PRINCIPLESOFPSYCHOLOGY

PRINCIPLESOF

PSYCHOLOGY

ASYSTEMATICTEXTINTHESCIENCEOFBEHAVIOR

FREDS.KELLERand

WILLIAMN.SCHOENFELDDEPARTMENTOFPSYCHOLOGY

COLUMBIAUNIVERSITY

Copyright©1950byAppleton-Century-Crofts,Inc.

Copyright©1995B.F.SkinnerFoundation,www.bfskinner.org

Allrightsreserved.Nopartofthisbookmaybereproducedinanyformorbyanymeans,withoutpermissioninwritingfromthepublisher.

ISBN1-46109-343-0

B.F.SkinnerReprintSerieseditedbyJulieS.Vargas

ACKNOWLEDGEMENTS

This reprinting ofPrinciples of Psychology reproduces the complete1950 edition of the original book and adds two forewords and a shorthistorical"Note"byFredKeller,thelatterwrittenforhisclassin1958–59.

Many people helped bring back this classic. I would like to thank themany members of the behavioral community who supported the projectfinancially, as well as with encouragement, and the board of theFoundation: Dr. Carl Cheney, Dr. E. A. Vargas, Dr. ScottWood, andespeciallyDr. JackMichaelwhosupervisedallof theactivities required inthereprintingprocess.

JULIES.VARGAS,PRESIDENTTheB.F.SkinnerFoundation

Cambridge,MassachusettsMarch23,1995

FOREWORDIK&S:THENANDNOW

K&S stands, of course, for Keller (Fred S.) and Schoenfeld (WilliamN.).But I hope the title of this foreworddoesnot leadyou, the reader, toexpectabiographythattellsyouintimatesecretsabouttheirlivesThen andNow.You perhaps know that K&S referred not only to Professors Kellerand Schoenfeld themselves but to their pioneering undergraduate text,which was first published officially in 1950. I have a few things to sayabout that book. But of course, in writing about their book I am alsowriting about them. Indeed, I shall sometimes refer to K&S asit andsometimes asthey.Manyof thebook's characteristics are theirs also.Theyaremaster teachers, in person and in print.They are also first-rate humanbeingsandthis,too,comesthroughintheirwriting.Iwasfortunateenoughtoknowthembothways, to receive inspirationfromthemin theirclasses,in theiroffices, in thecorridors,and in theputting togetherof thebook. Itwas quite evident to all of us who were in that position that it was notpossibletoseparatethebook,K&S,fromthepeople,K&S.

The importance ofK&S in establishing what are now called theexperimental and applied sciencesofBehaviorAnalysis iswell recognized.What is often not appreciated, however, is its current relevance. It tells usmuch not only about where we come from but also about where we aregoing—or shouldbegoing.When I reread thebook inpreparation for thisforeword,Idiscoveredtwothingsaboutmyself.First,Ifoundthatmuchofmyownbehavior that Ihadassumedwas theoutcomeof interactionswithmyowndata,assistedbymyowncreativethoughtprocesses,cameactuallyfromK&S; the behavior had become part of my repertoire, but I hadforgottenitssource.Second,IfoundthatK&S stillhadmuch to teachme;having reread it, I am wiser than I was before. I feel quite comfortablereferring to it in the present tense. I recommend it even to the mostsophisticated.Itwillrepayacarefulreading.

Simply to classifyK&S as an elementary text is to assume anequivalence relation that no longer exists. Introductory textbookscharacteristically confine themselves to presenting what is known or ispresumedtobeknown.Howcanabookbecalledanelementary textwhen,likeK&S, it does not just presentwhat is known but points outwhatwedonotknow,whatweneedtoknow,andsuggestshowwemightfindout?

Elementarypsychology textbooks these days are required to be eclectic.(Afriendofmineoncesaid,onlyhalfinjest,thattobeeclecticistostandwithone'sfeetplantedfirmlyinmidair.)Itisfearedthatthepresentationofa consistent point of view might narrow the vision of pupils andprospectivestudents.Bycontrast,K&Sdidsomething thathadneverbeendone before in psychology and has rarely been done since; they adopted a

systematic approach to their subject matter. Early in the book, they tellstudents that (a) they will have trouble with later chapters if they do notunderstand and make use of the facts and principles that are presentedearlier, and (b) their everyday language about psychology is imprecise andriddledwithuselesspreconceptions,andsotheyaregoingtohavetolearnanewlanguageanduseitconsistently.

Modern publishers and most teachers of the introductory psychologycourse would be terrified by a text that asked students not only to thinksystematically but also to learn a new language. ButK&S respects itsreaders. Taking it for granted that students are capable of understandingcomplex matters, K&S point out in a clear, precise, and readable fashionwhy it isworth one's trouble to learn this subjectmatter.Their style doesnotcondescend.

K&S also point out that they will be taking into account facts andprinciples that have arisen in the context of othertheoretical viewpoints:"Gooddataaregooddata,regardlessoftheory."Whatcriticshaveforgottenis that within their behavioral system, K&S discuss most of the matterscontained in the traditional general psychology course.Wherever possible,they systematize what is known, but they also discuss questions they donotyethaveanswersfor.Someofthosequestionsstillawaitanswers;somehave been answered but, like most good questions, lead us further intouncharted territory.K&S presentwhat is interesting and useful, nomatterwhat its source, and students are left not only with a practical way oflookingattheirownandothers'behaviorbutalsowithasetofmethodsforlooking into interesting puzzles that arewaiting to be solved.Rather thanconstrainingstudents,K&Sgivesthemspace,roomtogrow.

ThisapproachmakesK&Srichincontentandinspiration.Thereismuchi nK&S that is worth being reminded of—descriptions of interestingexperiments inmany areas, methodological details, problems that arise inconsideringwhywe seewhatwe see,whyneurosesdevelop,how the self-conceptarises—ingeneral,whatmakesussaywhatwesayanddowhatwedo.Iamgoingtopointoutjustafewexamples.LetuslookfirstatasmallpartoftheirdiscussionofPavlovianconditioning.

PavlovianandOperantConditioningMany of us have been tempted in recent years to downplay Pavlovian

research butK&S, upon being reread,makesmany aspects of that researchinteresting again. It is no longer clear that Pavlovian and operantconditioning take place independently of each other. K&S describe"intrinsically interesting"experiments inwhichphysiologicalchangeswerebrought about bywords that were spoken either by an experimenter orbythe subject. Subjects in these experiments became able, for example, tocommand theirownskin temperature to change.Although theexperimentswere done long before the field ofbiofeedback had arisen, K&S valued

them because of "their relation to the problem of 'controlling' bodilychanges." Was it just a coincidence that Ralph Hefferline, whose workhelped to initiate modern developments in biofeedback (for example,Hefferline, 1958; Hefferline & Bruno, 1971), had done his doctoraldissertationwithKellerandSchoenfeld(Hefferline,1950)duringtheperiodwhenK&Swasbeingwritten?

Concerning the separationofPavlovian andoperant conditioning,K&Shave this to say about the nature of "voluntary control": "This type ofconditioning fails to tell ushow the controlling word itself comes to bestrengthened" (italics added).AsK&S emphasize throughout, all behavioris related to environmental determinants. When, as with words, thosedeterminantsareself-produced,wemuststillfacetheproblemofexplaininghow thewords themselves come to be spoken.A proper understanding ofbiofeedback—voluntary control—and related phenomena requires anunderstanding of how respondent and operant conditioning interact.K&Sbringsthisproblemalive.

ConceptFormationThechapter,GeneralizationandDiscrimination,containsa longsection

onconceptformation.Atthebeginningofthissection,K&Saskaboutthedefinition of a concept. Starting with the question, "What is a 'concept,'"they go on to point out, "One does nothave a concept... rather, onedemonstrates conceptual behavior, by acting in a certainway."Then, afteraskingwhattypeofbehaviorthatis,theycomeupwith:"Whenagroupofobjects get the same response, when they form a class the members ofwhich are reacted to similarly, we speak of a concept." In their finalsentence, theygiveusthedefinitionthathascometobegenerallyacceptedeven by many who do not know the source of that definition:"Generalizationwithin classes and discriminationbetween classes—this istheessenceofconcepts."

Morebasicthanthedefinitionitselfistheirmethodofarrivingatit:notasking "What is a concept?" but instead, asking, "Whatmakes us say theword?"This tactic illustrateswhatWillardDay (1992) cited as the core ofradical behaviorism: "Skinner's account of the heart of radicalbehaviorism... rests fundamentally upon a viable conception of thefunctional analysis of verbal behavior" (p. 69). This conception mustinclude theverbalbehaviorofscientists themselves:"Themore[theradicalbehaviorist]canbringhisownverbalbehaviorunderthecontrolofwhathehasactuallyobserved,themoreproductiveanduseful it is likelytobe"(p.83).And so, insteadof looking for a thing called aconcept,K&Saskedaquestion aboutour own verbal behavior. They changed the definitionalproblem fromWhat is a concept? toWhat are the contingencies thatdeterminewhenwesaytheword"concept"?

Sadly, most of today's students remain unaware of an approach to

behavioral science that takes the verbal behavior of the scientist as itscentral datum. Radical behaviorists do not stand outside the behavioralstream, wisely commenting on psychological processes or states that thebehaviorissupposedtoreveal.Instead, theyask,"WhyamIsayingwhatIamsaying?Whereismyverbalbehaviorcomingfrom?"Thisiswhyradicalbehaviorists, contrary to the charge that is often brought against them, areable toconsiderconcepts, ideas,meanings, consciousness, andotherkindsof matters that are supposed to represent private events—events that aredirectly observable only by the person experiencing them. The questionthey ask is,What are the contingencies that determine when people saywhat theysay?AsSkinner (1957)pointedout, themeaningof such termsconsistsofadescriptionoftheconditionsunderwhichweusethem.

EquivalenceRelationsBecauseofmyowncurrentresearchinterests,Ihadalreadygonebackto

K&Stoseewhattheyhadtosayaboutthetopicofstimulusequivalence.Ifoundthat,too,underconceptsandmediatedgeneralization.

The authors begin with a definition of mediated generalization:"Generalizations are said to be mediated when they are based upon astimulus equivalencewhich results from training (Cofer& Foley, 1942)."TheygoontosummarizeanillustrativeexperimentbyRiess(1940).Riessfirstconditionedaresponse(changeinelectricalskinresistance)toeachofaset of visual words (style, freeze, surf, andurn), and then tested forgeneralizationof theskin response toagroupof synonyms(fashion, chill,wave,andvase)andtoagroupofhomonyms(stile, frieze,serf, andearn).Hefoundgreatgeneralizationtosynonymsoftheoriginalwords,evenmorethan to homonyms. "Whereas the generalization to the homonym[s]illustrates simply stimulus generalization [via auditory similarity], that tothe synonym[s] illustratesmediatedgeneralizationbasedupon theprevioustraining which produced the 'meaning' equivalence of these two words"(Keller&Schoenfeld,p.160).

And then, among the notes at the end of theGeneralization andDiscrimination Chapter, I found this: "The problem [of mediatedgeneralization]hasnotyetbeeninvestigated,however,tothedegreethatitsimportance would seem to justify" [italics added]. Prophetically, in thebook's final pages, where K&S have more to say about verbal behavior,they indicate the road along which a behavioral analysis of equivalencerelations is to take us: "The fact that adult speech bears relation to theenvironment in a more or less lawful manner is something to bescientifically explained, rather than taken for granted.How such acorrespondencearisesisacentralproblemforanalysis...."

K&S let thematter rest there. Iwas a graduate studentwhileK&S wasbeing written and tested and for a time after it was published, and Iremember being excited about the work on mediated generalization that

K&S summarized. But I was deeply involved in other work at the time(Sidman, 1989) and one could only do somuch. Not too long afterward,however, the problem did become the subject of intensive experimentalinvestigationandsophisticatedtheoreticalanalysis,largelyinthecontextofthe paired-associates technique rather than Pavlovian conditioning (see, forexample, Jenkins, 1963).By the timewebeganour studies in this area, aconsiderable literature had come into existence. Nevertheless, thesubsequent effort had advanced the topic little beyond the summary thatK&S offered their undergraduate readers, and this work eventually groundto a halt. It fell victim to limitations that are inherent in the experimentalpractices and theoretical orientation of methodological behaviorism (seeDay, 1992, pp. 61–70) and that still characterize much of experimentalpsychology.

My coworkers and I came into this field from entirely differentmethodological and conceptual directions. For one thing, our originalinterest was in reading comprehension (Sidman, 1971), not mediatedgeneralization,andwethereforedidnotfeelcompelledtocontinuewiththemediationmodel.Second,weaskedaboutthebehaviorofindividualsratherthan the averaged behavior of groups. In studies that preceded ours, thegrouping of data had prevented investigators from actually observinginstances of mediated generalization in the behavior of any individual.Evaluation of their data against the null hypothesis guaranteed thatexperimenterscouldneverexperiencethethrillofseeinganindividualrelatestimuli innewways,seeminglywithouteverhavingbeen taught todoso.Also, the grouping of data had caused positive and negative instances ofmediated transfer to cancel eachother out in the averages, therebyyieldingmassivenegative results in criticalexperiments (Jenkins, 1963).Therewasnotmuchreinforcementhereforexperimenters.

Our own data led us to talk aboutequivalence relations rather thanmediated generalization. Instead of assuming that stimulus equivalencerequired response mediation, we asked ourselves, "Under what conditionsdowe say things likeurnmeansvase, ortheword dog represents a dog,orthisshoppinglisttellsuswhattobuy?Ithinkwenowknowsomethingwedidnot knowbefore aboutwhywe say such things.The reinforcementthatcomesfromseeingthese"concepts"intheveryprocessofbeingformedby individual subjects—andpupils—haskeptusgoing in this research formore than 20 years.Andwe are only just now in the process of realizingthat the formation of equivalence relations is one of the functions ofreinforcement, with what used to be calledmediating responses, simplyjoiningdiscriminativestimuli,conditional stimuli, and reinforcingstimuliasmembersoftheequivalenceclassthatreinforcementestablishes(Sidman,1994).Butclearly,K&Shadseensomething that fewothers recognized inthis research area.When I eventually foundmyself involved, I also foundthatK&Shadpreparedmetoinvestigateit,astheyhadpointedout,to the

degreethatitsimportancewouldseemtojustify.

K&S'sRoleTodayWe should not lose sight, however, of the authors' introductory

statement: "This book is a new kind of introduction to psychology."Unfortunately, this statement remains true today; it is still a new kind ofintroduction. Itsbasic lessonhasneverbeenabsorbed intopsychology.Touse Skinner's apt term, "autonomous man" remains the conceptualfoundation of modern psychology, and students in the typical elementarycourseremainunawareoftheroleoforganism-environmentcontingenciesasdeterminersofhumanconduct.

RichardM.Elliott, inhisEditor's introduction to thebook, recognizedquite clearly what the authors had done: "I am sorry for the psychologistwhomissesthisout-of-the-ordinarytextbook...itwouldenhancehisvisionand build his morale to know that it has been possible already todemonstrate...somuchlawfulnessofbehavior...."

What is involved here is a new conception of human nature.Althoughthe authors do not explicitly discuss the philosophical underpinnings oftheir approach,K&S is in fact an example of radical behaviorism inpractice.WillardDay(1992), inhisroleasaphilosopher,madethecaseasfollows:

ThethreepropositionsintermsofwhichIdefinethe[radical]behavioristoutlook are these: First, behaviorism is at heart a concern with thecontingencies involved in behavioral control. Second, behaviorism... isopposed to something calledmentalism... Skinner's opposition towhat hecalls "autonomousman."Third, behaviorism involves at heart a particularconvictionwithrespect tosocialplanning,namelythat ifweare tosurviveasaspeciesweshouldbeginatoncetorestructureoursocialenvironment...so that it acts to produce people who have the behavioral equipmentnecessaryforusalltosurvive(p.179).

Psychology has not only not accepted this conception but has opposedit, often misrepresenting it and frequently distorting it into something itnever was (see Chiesa, 1994, for a keen analysis of the relations betweenpsychologyand radicalbehaviorism).Partlybecause the radicalbehavioristconception of human nature is rarely presented even for discussion to thethousands of students who go through the elementary psychology courseeach year, that conception has not reached the awareness of the generalpublic.K&S, in its original form, is still capable of providing studentswithwhatRichardM.Elliott called "... insights... [that]will beofuse toyou whether you become a psychologist, teacher, lawyer, salesman,philosopher, doctor, or just apersonwho feels theneed to seebeneath theseemingchancinessofhumanbehavior."

WhatIamsuggestinghereis that today'sstudentsdonotsomuchneedall thenewfactswehave learnedaboutbehaviorbut rather, just enough toarouse their interest in a viewpoint of humannature that canprovide themwith hope. Once again,Willard Day (1992) has said best what has to besaid: "With knowledge of contingencies one can see all too clearly theincalculabledamagewecontinuallydo toourselves, to thosewe love, andto thoseothers forwhomwewant toassumesomeresponsibilitywhenwebaseoursocialdecisionsonthemodelofautonomousman"(p.191).

Wehave tobe taught to seecontingencies.Oncewehave learned to seethem, the road is open to changing them.K&S teaches students thatbehavioral contingencies are real, and have to be taken into account ifbehavior is to be understood and if something is to be done about currentpractices.IamsuggestingthatanyrevisionofK&Swouldneedtoaddonlythe new facts that will help make that lesson more effective.We do notneedmorepsychologists,orevenmorebehavioranalysts.AsK&Stoldusin "A LastWord": "We need to hasten and train a generation of men ofgoodwill.Howthisistobedonemaybemankind'slastdesperatequestionofall."

MURRAYSIDMANSouthborough,Massachusetts

March1995

REFERENCES

CHIESA, M. (1994).Radical behaviorism: The philosophy and thescience.Boston:AuthorsCooperative.

COFER,C.N.,&FOLEY, J.P. (1942).Mediatedgeneralization and theinterpretation of verbal behavior: I. Prolegomena.PsychologicalReview,49,513–540.

CUMMING, W. W., & B ERRYMAN, R. (1965). The complexdiscriminated operant: Studies of matching-to-sample and relatedproblems. In D. I. Mostofsky (Ed.),Stimulus generalization (pp.284–330).Stanford,CA:StanfordUniversityPress.

DAY, W. (1992).Radical behaviorism: Willard Day on psychology andphilosophy.Reno,NV:ContextPress.

HEFFERLINE,R.F.(1950).Anexperimentalstudyofavoidance. GeneticPsychologyMonographs,42,231–334.

HEFFERLINE,R. F. (1958).The role of proprioception in the control ofbehavior.Transactions of the New YorkAcademy of Sciences , 20,739–764.

HEFFERLINE,R.F.,&BRUNO,L.J. J. (1971).Thepsychophysiologyofprivateevents.InA.Jacobs&L.B.Sachs(Eds.),The psychology

ofprivateevents(pp.163–192).NewYork:AcademicPress.JENKINS,J.J.(1963).Mediatedassociations:paradigmsandsituations.In

C. N. Cofer & B. S. Musgrave (Eds.),Verbal behavior andlearning: Problems and processes (pp. 210–245). New York:McGraw-Hill.

KELLER, F. S. (1968). "Good-bye, teacher...." Journal of AppliedBehaviorAnalysis,1,79–89.

MENZIES, R. (1937). Conditioned vasomotor responses in humansubjects.JournalofPsychology,4,75–120.

RIESS, B. F. (1940). Semantic conditioning involving the galvanic skinreflex.JournalofExperimentalPsychology,26,238–240.

SIDMAN,M.(1971).Readingandauditory-visualequivalences.JournalofSpeechandHearingResearch,14,5–13.

SIDMAN,M. (1989).Avoidance atColumbia. The BehaviorAnalyst , 12,191–195.

SIDMAN, M. (1994).Equivalence relations and behavior: A researchstory.Boston:AuthorsCooperative.

SKINNER,B. F. (1957).Verbal behavior . NewYork:Appleton-Century-Crofts.(CurrentlyavailablefromtheB.F.SkinnerFoundation.)

SKINNER, B. F. (1968).The technology of teaching. New York:Appleton-Century-Crofts.

SKINNER, B. F. (1969).Contingencies of reinforcement: A theoreticalanalysis.NewYork:Appleton-Century-Crofts.

STODDARD, L. T., B ROWN, J., HURLBERT, B., MANOLI, C., &MCILVANE,W.J.(1989).Teachingmoneyskills throughstimulusclass formation,exclusion,andcomponentmatchingmethods:Threecasestudies.ResearchinDevelopmentalDisabilities,10,413–439.

FOREWORDIIK&S:ASYSTEMATICAPPROACH

In 1950—and even today—Keller and Schoenfeld'sPrinciples ofPsychology: A Systematic Text in the Science of Behavior (K&S) was aradicallydifferentkindofintroductory-leveltextbookforpsychology.Inthetypical introductory textbook, each chapter covers one of the varioussubfields of psychology.Collectively, the chapters provide a broad surveyof topics, but in away that ismore patchwork than coherent.There is nooverarching framework to organize and integrate the chapters, no basis fortreatingsomematerialasbasicandothermaterialasderivativeoradvanced.Each chapter is self-contained, presenting its material in terms of thedistinctiveconceptual language typicalof thesubfield.The impressiononegets is that psychology is a loose federation of relatively independentsubfields, eachwith its own theoretical concerns and conceptual language,ratherthanaunifiedscientificdiscipline.

Textbooks that give high priority to the most up-to-date researchcontribute further to the sense that psychology is fragmented. Suchtextbookscanbeuseful,certainly, inprovidingasortofsnapshotorstatusreportonwhatresearchersataparticulartimefindmostinteresting.Butthematerial cited is not necessarily of lasting significance. It is often the casethat the "hot" topics of one period become passe in the next, and thepatterns of changing "hotness" do not, in retrospect, always seem likeprogress.Textbooksthatemphasizecurrentresearchinterestsforthesakeofbeing current can become dated quickly, and the field they describe canappearfaddish.K&Spresentsadifferentapproach,onemore typicalof textbooks in the

advanced sciences.Thebookprogresses in systematic fashion from simpleto complex. The goal is not to cover evenly all the topic areas ofpsychology but, instead, to give the student a set of serviceable principleswith which to address advanced topics and interpret everyday behavioralphenomena. "Our aim, throughout the text, is to show how the complexmaybeexplainedintermsofthesimple"(p.112).

The first several chapters present the elementary principles of behavioranalysis—the operant-respondent distinction, respondent (i.e., Pavlovian)conditioning, and the reinforcement, extinction, and differentiation ofoperantbehavior.Thenext fewchapterscovermorecomplexanalyticunits—the stimulus control of operant behavior (e.g., generalization anddiscrimination), chains, and secondary (or conditioned) reinforcement.Thefinal chapters address topics of even greater complexity—e.g.,motivation,emotion, anxiety, and social behavior including consciousness, the "Self,"and language.Thematerial in each chapter builds uponmaterial in earlierones so that there is a genuinely cumulative effect.The chapter on social

behavior, for example, is notmerely different from the earlier chapters onbasicprinciples,itistrulymoreadvancedinthesensethatthetreatmentoftopicsinthatchapterisbasedonprerequisitematerialcoveredearlier.

If one's primary aim is to teach fundamental principles, one is lessinclined tobegreatlyconcernedaboutwhethera textbookdoesordoesnotcover the very latest research findings.The more important questions areaboutwhether the principles covered in a textbook are, indeed, serviceableand whether they are presented effectively for the student.As teachers wehave a limited amount of time in our courses to teach difficult material.Introducing novices to the issues that engage the attention of experiencedresearchersprobablyisnotthebestwaytoteachthecoreprinciplesthataremost likely to beserviceable—even if the "cutting-edge" of research is onsomeofthebasicconceptsofthescience.

Teachers of the other sciences seem to appreciate this point. Myintroduction to physics, for example, was almost entirely devoted toclassical mechanics. I doubt that my instructor felt any shame aboutpresentingsuch"dated" topics insteadof,say, relativity theoryorquantummechanics, even though he undoubtedly knew that the principles he wasteaching were not viewed within the science as the most fundamental orgeneral.Ifhehadtriedtocoverthenewertopicsintheintroductorycourse,the other students and I probably would have come away with little ofvalue.IneverimaginedthatIwasgettinganintroductiontothecurrenthottopics in physics, and so I did not feel deceived or cheated. I assumed,correctly, that such topics were appropriate for advanced courses.What Iwanted and got was acquirable, given my entry-level skills, and certainlyusefulforgettingalongintheworld.

The intent inK&S was likewise to present principles that would helpstudents get along better in the world. The principles had to have beenestablished through rigorous experimentation. But they also had to behelpful in explainingmeaningful human behavior. "We... shall have littleinterestintheactivitiesofanimalsmarkedlydifferentfromman,orinthoseinfra-humanactivitieswhichthrownolightonhumanconduct"(p.2).K&S, in otherwords, should not be regarded as a 1950–vintage survey

ofthefieldthatisnowknownas"animallearning"(or,lessattractively,as"rat psychology"). It is understandable, however,why some readersmighthaveinterpreteditassuch.Muchofthematerialinthebook—especiallyinthe earlier chapters covering the elementary principles—is from researchconductedwithnonhumananimalsassubjects—particularlyrats.

The reason whyK&S stressed such research is important, and it hasnothing todowith interest in rat behaviorperse.The reason is thatK&Sregarded controlled and simplifiedprocedures as essential for establishingbasicrelationships."Inthediscoveryanddemonstrationofbasicprinciples,everythingdependsuponthekindofmethodthatweemploy"(p.53).Eventhe venerable rat-maze was considered too complex to reveal basic

relationships.

Alongwith this development [usingmazes to study learning] came thehope that an analysis of "learning" could be made with this usefulinstrument.Unfortunately, this hopewas not realized. It gradually becameclear thatmaze-learningwasanexceedinglycomplicatedaffair,and that themaze itselfwasnot thesimpledevice that ithadseemed tobe.Evenwhenattentionwas centered upon the behavior of a rat at a single choice point,the problems involved were too great for the kind of description mostdesiredbythescientist(p.58).

Keller and Schoenfeld were uncompromising in holding that complexhuman behavior arose from basic processes common across a range ofspecies. But performance in experimental procedureswas not necessarily asimplereflectionofthosebasicprocesses.Withhumans,languageandothercompetenciesacquiredinasocialcontextaddafurtherlevelofcomplexity.

In the case of human beings, the solution of problemsmay be speededup by a special set of conditions. It is too early for us to consider theseconditions here, butwemaynote that the possessionoflanguage is oftenof help in reducing the time required or the number of errorsmade in themasteryofcertaintasks....Insomeinstances,therateofimprovementissodramaticas toobscurethefact thatessentially thesamebasicprinciplesareinvolvedinverbalasinnon-verbalbehavior(pp.60–61).

Thus, research with nonhuman animals under contrived conditions wasemphasized because such researchwasmost suitable for establishing basicprinciples in relativelypureform.But theoverarchingconcernwithhumanbehavior is unmistakable. Indeed,K&S confronts head-on some of thethorniestandmostprofoundissuesinpsychology.Whatdowemeanwhenwespeakof"preparatoryset,"of"havingaconcept,"of"beinganxious,"of"beingmotivated,"of "beingconscious,"orof "beingself-aware"?What istheoriginof thephenomenatowhichthesetermsapply?Andwhatroledosuchphenomenaplay inhumanfunctioning?What is thenatureandoriginof language?What, indeed, is the nature of human nature?The answers tothese questions were to be found in the operation of a relatively smallnumberoffundamental,generalprocesses.Forexample:

Of great importance to the formation of personality is the fact thathumanbeingscandiscriminate theirownactions,appearance, feelings,andsuccessfulness. In the course of growing up, the child comes to "know"about himself; he becomes at least partially "aware" of his capacities andweaknesses, his likelihood of winning or losing in given situations, his

physical and social attractiveness, his characteristic reactions. This issometimesspokenofasthedevelopmentoremergenceofthe"Self,"awordthat ismeant todesignate theability tospeakof (be"aware"of)one'sownbehavior, or the ability to use one's own behavior as the SD for furtherbehavior, verbal or otherwise.... The "Self," in short, is the person, hisbody and behavior and characteristic interactions with the environment,taken as the discriminative objects of his own verbal behavior.They aremadediscriminativeforhimbyhissocialcommunity,asitteacheshimhislanguage....[A]personpossessingnoverbalbehaviorofanysortwouldnothavea"Self,"orany"consciousness."His reactions to theworldwouldbelikethoseofanyanimal....(pp.368–370).

Again,thegeneralityofthebasicprocessesisstressed.

Whatmenlearnindifferentsocietiesoughtnottoobscurethebasicfactsthatallmen learn in the samewayandare subject to the samedrives.Theprinciplesof reinforcement, extinction,discrimination, and the likeoperateuniversally, though the form of the response to be learned or the specialtypeofSDtobeobeyedmaybeselectedbythecommunity....Weneedtohold on to the idea of humanabsolutes as much as to that of culturalrelativity(p.365).

Giventheconcernwiththesekindsofissues,perhapsitisnotsurprisingthat graduates ofKeller and Schoenfeld's program at ColumbiaUniversityin the late 1940s and early 1950s developed a strong interest inpsychodynamic concepts and practice. Some of those individuals haveworkedcollaborativelywithpsychodynamically-orientedcliniciansandhavewritten sophisticated and sympathetic behavioral interpretations ofphenomena revealed in psychodynamic practice (Ferster, 1972; Hefferline,1962;seealsoDinsmoor,1989;Knapp,1986;Sidman,1989).

K&S'sCurrentUsefulness

A fair question to ask at this point is whether I would useK&S as atextbookinanundergraduateclass.AlthoughIobviouslyhavegreatrespectand affection forK&S, I must say, in all candor, that I would not.TheproblemisnotthatI thinkthematerial isout-of-date.Indeed,I thinkmostof the material is currently very useful. The problem is that the book,despite itsbeing exceptionallywellwritten, dealswithmaterial and issuesthat are just a bit too subtle, abstract, and sophisticated for many of theundergraduatestudentsIencounter.Arigorouslysystematicapproachmakesspecialdemandsthatmanystudentsareunpreparedtohandle.Imightwishitwereotherwise,butthatisthereality.Dinsmoor(1989)describedsimilarexperiencesusingK&Sathisuniversity.

Ido,however,urgeourgraduatestudents tostudyK&S,andI return toit frequently myself. It is an excellent introduction to Skinner'sexperimental program and the system of integrated principles derivedtherefrom (e.g., Skinner, 1938). Dinsmoor (1989) made the interestingobservation that he and other graduate students at Columbia around 1950foundithardtograspthebroadsignificanceofSkinner'sexperimentalworkuntilafter theyhadgonethroughK&S.Thematerial inSkinner'sBehaviorof Organisms initially seemed rather abstract and remote from everydayconcerns.K&Sbroughtthematerialtolifeandthuspreparedthestudentstoread Skinner'swritingswith deeper understanding.Whatwas true in 1950remains so today:K&S can be an excellent high-level introduction and asourceofmotivationforfurtherworkinbehavioranalysis.

There are some broad themes inK&S that are especially important forgraduate students to think about. Studentswill be reminded, for example,of why we conduct experimentsand what it means for results to bescientifically significant (as distinct from statistically significant). Theywillberemindedoftheimportanceofthinkingclearlyaboutthemeaningofpsychological concepts like "motivation," "emotion," "anxiety," and"consciousness."And they will be reminded of why a scientific systembasedonasmallnumberofempiricallyestablishedgeneralprinciples issohelpfulfordealingeffectivelywithawiderangeofcomplexphenomena.

Students who studyK&S may come to appreciate the irony in thecharge, sometimes heard, that behavior analysts ignore theory and favor"dust-bowl" empiricism instead. The irony comes from the fact that anoutstanding feature ofK&S is its attempt to organize the field ofpsychology in terms of a coherentsystem.The lesson to be drawn is thatthe supposedantitheoretical stanceofbehavioranalysts is amyth.Leadingbehavior analysts have neither advocated "dust-bowl" empiricism noropposed theory. Their opposition has been restricted to certain kinds oftheoryprevalentwithinpsychology(Skinner,1969,pp.vii–xii).

Also, it is hard to imagine a clearer,more insightful short introductionto Skinner's approach to verbal behavior than that presented inK&S (pp.376-400).ThismaterialwasderivedfromaseminaronverbalbehaviorthatSkinner offered at Columbia in 1947 and from mimeographed notes thatSkinnercirculatedafterdeliveringtheWilliamJamesLecturesatHarvardin1947. It thus predated the publication of Skinner's book,Verbal Behavior(Skinner, 1957). I regularly recommend the section inK&S to graduatestudents, and they consistently report finding it very effective (seeDinsmoor,1989forasimilarassessment).

I findK&S usefulmyself formany of the same reasons. It remindsmeof fundamental themes, issues, andconcerns that I tooeasily forget. I findit helpful, for example, to be reminded of the original questions aboutsignificant human functioning that led to particular research lines. It oftenhappensthatresearchlinesbecomeabitautistic.That is, theresultsofone

experiment raisequestions that prompt thenext experiment, and soon, sothat over time the direction of research becomes guided more byconsiderations internal to the line thanby theoriginal questions. It canberefreshing, in such cases, to step back a bit, take in the whole trajectory,and see howmuch progress has beenmade toward answering the originalquestions.K&Scanprovideexactlytherightkindofperspective.

Afinalcommentisinorderabouttheageofthematerialinthebook.Itis not hard to go though a 45-year-old book and identify places where anewerresearchfindingornewtermmightbepreferable.ButIamstruckfarless with the absence of certain newmaterial than with howmuch of thematerialinK&S remainsusefuland important. Indeed, I find it remarkablehow often I come across topics covered perceptively inK&S that arecurrently active areas of research and theoretical analysis. Such topicsinclude therelationbetweenverbalandnonverbalbehavior(i.e., theroleofinstructionsandrules),theconceptualstatusofmotivationandemotionandthe significance of the relevant operations, the significance of responsevariability for the selectionofbehavior,what itmeans to "havea concept"and the development of the relevant stimulus control relations, the role ofproprioceptive and other internal events in verbal and nonverbal behavior,the function of verbal behavior generally, and the analysis of culturalpractices.

Whatcananintroductory-leveltextbookpublishedin1950offerareadertoday? Ordinarily, the answer to such a question would be, "Not much."ButK&S is not the least bit ordinary. It is a book full of wisdom andinsight,containingmuchofvalue forbothnovicesandexperts in the fieldofbehavioranalysis.

It is, to be sure, important as an historical document, and it deservescarefulattentiononthataccountbyanyoneinterestedintherootsofmodernbehavioral psychology. The book's value is far more than historical,however. It presents the core principles of behavior analysis systematicallyand with unsurpassed clarity.And it shows how those principles can beapplied interpretively so as to make interesting cases of complex humanbehavior comprehensible. If you are new to the field, studyingK&S willgive you a deep understanding of the principles of behavior analysis andwhy they are important. If you are expert, youwill, I suspect, come awaywitha freshandbroadenedperspectiveon thoseprinciples, their roots,andtheirimplications.

RICHARDL.SHULLUniversityofNorthCarolinaatGreensboro

March1995

REFERENCES

DINSMOOR, J. E. (1989). Keller and Schoenfeld'sPrinciples ofPsychology.TheBehaviorAnalyst,12,213–219.

FERSTER,C.B.(1972).Anexperimentalanalysisofclinicalphenomena.PsychologicalRecord,22,1–16.

HEFFERLINE,R.F.(1962).Learningtheoryandclinicalpsychology—Aneventual symbiosis? In A. J. Bachrach (Ed.),Experimentalfoundations of clinical psychology (pp. 97–138). NewYork: BasicBooks.

KNAPP, T. J. (1986). Ralph Franklin Hefferline: The Gestalt therapistamong the Skinnerians or the Skinnerian among the Gestalttherapists?JournaloftheHistoryoftheBehavioralSciences,22,49–60.

SIDMAN,M. (1989).Avoidance atColumbia. The BehaviorAnalyst , 12,191–195.

SKINNER, B. F. (1938).The behavior of organisms: An experimentalanalysis. NewYork:Appleton-Century-Crofts. (Currently availablefromtheB.F.SkinnerFoundation.)

SKINNER,B. F. (1957).Verbal behavior . NewYork:Appleton-Century-Crofts.(CurrentlyavailablefromtheB.F.SkinnerFoundation.)

SKINNER, B. F. (1969).Contingencies of reinforcement: A theoreticalanalysis.NewYork:Appleton-Century-Crofts.

SUPPLEMENTARYNOTESbyF.S.Keller

Psychology1–2ColumbiaUniversity

1958–59

TobeusedinconjunctionwithPrinciplesofPsychology

Keller,F.S.&Schoenfeld,W.N.Appleton-Century-Crofts

1950

MATTERSOFHISTORY

TwoRoutestoModernPsychologyWhen Rene Descartes, in the middle of the 17th century, made his

famous distinction betweenbody andmind, he opened up two routes topsychologicalscienceasweknowittoday.Thefirstwasa"mental"one.Itled throughBritish philosophy andGerman sense-organphysiology to thefounding of experimental psychology and the early-20th-century school ofstructuralismintheUnitedStates.Thesecond,or"body"route,ledthrougha largeAnglo-German area of reflex physiology and the field of Russianconditioning to theAmerican school ofbehaviorism.These were not theonly routes tomodernpsychology, but theywere important ones and theydeserveasecondlook.Route One. After Descartes, the British "mental philosophers," from

JohnLocketoJohnStuartMill,tookthemindveryseriouslyasaprovinceforstudy.Eachoftheminspectedhisownmind,asbesthecould,andtriedtodescribewhatwasthere.Thefirstcontentstobenotedwere ideas.Theseideas included everything one could "think about," and they entered themindthroughone'scontactswiththeoutsideworld.Someweresimpleandunanalyzable,suchastheideasofwhiteness,hardness,orroundness.Otherswerecomplex,beingcompoundedfromsimpleones.Forexample,thethreesimpleideasjustmentionedmightbecombinedintothecomplexideaofaball. Ideaswerealsoassociatedwitheachother in succession,aswhenoneregularlyleadstoanother.Example:theideaofaballmightcalluptheideaof a child, and the idea of child might in turn bring to mind the idea ofmotherorfather,andsoon.

Later developments of this associationism led to a distinction betweentwo kinds of mental contents. First, there was theimpression (orsensation).Thiswastheimmediateeffectofcontactwiththeoutsideworld.Secondly, there was the idea proper.This was a kind of revival or faintcopyof the impression,occurringat some later timeand in the absenceofthe original outside circumstances. Laws of association were also putforwardtoaccountforthewaysinwhichideasappeared.Thus,anideawassaid to call up another because of thesimilarity of the two, or because oftheir earliercontiguity—their togetherness in either time or place.And soon.Weshallmeettheseconceptslater,intheseNotesorinyourtext.

During the 19th century, German scientists, mainly physiologists,carried this mental analysis from the armchair into the laboratory.Sensations, rather than ideas, were their basic elements—sensations ofsound, color, touch, taste, and so on; and they tried to show how thesesensationswere related to other things, such as events in the environment(stimuli),theworkingsofthesense-organs,andeventheactivityofnerves.By1861,abroadnewfieldofresearchwasmappedout,giventhenameofpsycho-physics,anddedicatedtothestudyofthewaysinwhichthemental

dependeduponthephysical.WhenWilhelmWundt (1832–1920), the founder of modern laboratory

psychology, published his first textbook of the new science, he leanedheavilyuponpsycho-physicsformaterial.Also,hewasinfluenced,directlyand indirectly, by the writings of the British philosophers. The newpsychology,hetellsus,willbebaseduponanintrospectiveexaminationofmind.Mindistobeanalyzedintoitselements.Theprincipalelementswillbesensations, eachwith its own peculiar quality and intensity.We are tostudythewaysinwhichthesesensationsarecompoundedorconnected.Wemustdeterminethelawsofsuchconnections—andof theassociationswithwhich theoldphilosopherswereconcerned.Finally,weshall lookinto the"bodily substrate of the mental life"—the physiological processes uponwhich the psychological ones arebased.Wundt had been a physiologist,andthenewpsychologywastobea"physiologicalpsychology."

Wundtwasascholar,asystematizer,apromoter,andatirelessworkerinevery branch of the new enterprise. In 1879, at Leipzig, he set up alaboratory of psychology in which all manner of research problems wereattacked.Mostoftheseproblemsweredrawnfromsense-organphysiology,butotherscamefromareasasdistantandfarapartasmedicine,astronomy,and education. He inaugurated a scientific journal for the publication ofresearch; hewrote voluminously on every subdivision of his field; and helecturedtostudentsfromallover theworld.At the turnof thecentury,hiswas the best-known psychology in existence, and the one with the mostscientific status. It had far outdistanced all of its European rivals andwasfastbecomingtheworld'sonlypsychology.HowitfaredonAmericansoil,youwillseelater.Route Two . As early as 1633, Descartes is said to have begun his

Treatise onMan .Among other things in this important book, he told ofthe way in which a large portion of human behavior might result, inmachine-like fashion, from environmental action. More than that, hedescribed,ingreatdetail, thebodilymechanismthroughwhichstimulationfrom the outside world might lead to muscular response. His descriptionwasbasedonfactandfancy, inaboutequalparts,but itcomesfairlyclosetomodernteaching.

External stimuli, says Descartes, may affect the sense-organs. Sense-organexcitationsetsupactivity innerves.Thenerves lead from thesense-organs to the"central cavern"orventriclesof thebrain, and from thebrainto themuscles.Thenerves are really little tubes,withdelicate threads thatmake up their "marrow." Excitation of a sense-organ acts upon these tinythreadsasabell-ringermighttugatabell-rope.Tuggingatthebell-rope,inturn, leads to the opening of little pores or valves in the ventricle walls.Within the ventricles are "animal spirits"—invisible, powerful, flame-likefluid or vapor from the heart.The opening of the ventricle valves permitsthesespirits to rush into the tubular motor nerves and then into the

muscles.Themuscles, swollenby the spirits, thenproduce theappropriatebodilymovements.

Descartes, it appears, was not quite certain that all human behaviordependedupon this sortof responsemechanismorwasevokedexclusivelyby environmental stimulation. Formost ofman's actions, yes; and foralltheactionsofanimals.Butman'sconductwassometimesgovernedbyhissoul (mind) and originated, not in the outer world, but in a little glandwithin the brain.This gland (called thepineal gland, from its pine-coneshape) was thought by Descartes to project into the fiery vapors of thecentralcavernand tobecapableofmovementat the soul'scommand.Thismovement,hesaid,wasin turnable todirect theflowofspirits into tubesotherthanthosenormallylinkedupwithsense-organ/bell-ropeaction.

Itwasthemiddleofthe19thcenturybeforenaturalscientistsrecognizedthe basically correct elements of Descartes' thinking—thatstimuli causeresponsesbywayofthelinkedactionofsense-organs,sensorynerves,brain(andspinalcord),motornerves,andmuscles.TheCartesianerrorshadtobedisposedoffirst.Ithadtobeshownthatmusclesdidnotswellwithanimalspirits; that thenerveswerenot little tubes,withbell-ropes inside; that theventricleswere less important than the graymatter of the brain and so on.Also, it had to be shown that muscles, detached from a living organism,would respond to direct stimulation or the stimulation of still-attachednerves—circumstancesinwhichnoanimalspiritscouldpossiblyflowfromthecentralcavern.And,finally, thewholeideaof reflexaction, understoodbyDescartes,hadtoberediscovered.

Theword"reflex"issaidtohavebeenfirstusedin1736byJeanAstruc,a French doctor who said that animal spirits from sensory nerves werereflectedfromthespinalcordorbraintocause"motioninthosenervetubeswhichhappen tobeplacedexactly in the lineof reflection."However, it isfrom the time of Marshall Hall, a great English physiologist of the nextcentury, thatwedate themodernuseof this term.Hall, in1832,describedfour different kinds of muscular reaction.We may pass by two of thesequickly,asof lesspsychological interest—themovementsofbreathingandthemovementsresultingfromdirectstimulationofthemusclesthemselves.A third and very important type was called "voluntary," and was said tostartspontaneously in thebrain, reaching themusclesbywayof thespinal"marrow"(cord)andmotornerves.Finally, therewas the"reflexfunction."This,saysHall,is"excitedbytheapplicationofappropriatestimuli,whichare not, however, applied immediately to the muscular or nervo-muscularfibre,but tocertainmembranousparts,whence the impression iscarried tothemedulla (spinalis), reflected, and reconducted to the part impressed, orconducted to a part remote from it, in which muscular contraction iseffected."

Hall'saccountofvoluntarymuscularreactionandthereflexfunctionwasbased,ingoodpart,uponhisobservingthebehaviorofalizard,freshlycut

into four segmentsandstimulatedwith forcepson the skinof the tailmostpart. In spiteofnopossible influence from theanimal'sbrain, the tailwasseento"moveandbecomecontorted"whentheforcepswereapplied.Theseobservations weremade almost 200 years after Descartes had sat down towritehisTreatise.Inthatperiodoftime,manyexperimentsonanimalshadbeen conducted and many claims had been tested—including some thatwere made by Descartes himself.Yet, if we ignore certain differences interminology, there is a striking degree of similarity between Hall'streatmentandthatofDescartes.Bothrecognizetheimportanceofstimuliineliciting behavior; both explain this stimulus-response connection byasserting that nervesmay conduct excitations from the sense-organs to thebrain or spinal cord and thence to the muscles; and both point to certainbehavior that isnot elicited—behavior which Descartes attributes to theactionofthesoulandHallspeaksofasvoluntary.

AfterMarshallHall, studies of reflexmovement came so fast and fromsomanydirectionsat thesametimethatwecannoteasilyholdtothetrail.Experiment followed experiment, in England, France, Germany, andelsewhere. Ideawasmetwith idea in every sphereof reflex research.Someworkersthoughtofthereflexasasimpleconnection—usuallyanunlearnedconnection—between a specific stimulus and a specific response, and triedto list asmanysuchconnectionsas theycould find, inanimals, inhumaninfants, and in human adults. Similar attempts to these had been madebeforeHall'stime,asinthelistofhuman"automaticmotions"madeupbyDavid Hartley, a British philosopher and physician, in 1749; and similarattempts are still beingmade, especially by students of infant behavior, ateithertheanimalorhumanlevel.

Another,largergroupofworkersturnedtheirattentiontothemechanismof reflex function rather than to the simple connection between two suchobservable events as a stimulus and a response.With the aidof dissectinginstruments, recording devices, compound microscopes, techniques oftissue-staining, andmethods of stimulating sense organs and nerves, theysought to teaseout the trainofevents that followedstimulationandendedwith response. Almost always they worked with animals, for obviousreasons. Usually, too, they did not work with intact organisms. Instead,they dealt with special "preparations"—animals in which one part of thenervoussystemwasremovedsurgicallyfromtheinfluenceofanother,asbycutting through the spinal cord or severing the connection between twobrainparts.

Thefloweringofallthesestudiesappearedinoneofthefirstgreatbooksof our century:The IntegrativeAction of the Nervous System (1906), byCharles S. Sherrington, an already famous British physiologist. Broughttogether in this book were countless facts of reflex action, gleaned byscientists down through the years since the time of Descartes.These factsweresiftedandarrangedinamosttellingmannertoshowthebasichookup

of thesimplest reflexfunction; toshowhowdifferent reflexes act together,at the same timeoroneafter another; and to show, ineverywaypossible,the kind of bodily changes that take place when a stimulus begets aresponse.

Sherrington's bookwas for physiologists, and its principal aimwas toreveal the hidden operation of those structures lyingbetween stimulus andresponse, comprising the so-called "reflex arc" (See K & S, p. 6).Yet itwas also a book for psychologists. It portrayed in detail the manner inwhichagoodmanyresponsesarecontrolledbystimuli,anditshowedhowsuch stimulus-response relations or "reflexes" (K & S, p. 6) could bestudied quantitatively by measuring some of their properties, such as"threshold," "latency," "response magnitude," and so on (K & S, 8–12).Youwillsee,lateron,thatpsychologistsmakegooduseofthesemeasures,and that they are indebted to Sherrington for a number of other things aswell(seeChaining,intheseNotes.)

Onemoremanandonemorebookmustbementionedinthispartofourhistorical introduction. The man is Ivan P. Pavlov, the great Russianphysiologist; and the book,whichwas translated intoEnglish in 1927, ishi sConditioned Reflexes. Together, man and book, they account for aseven-league step along the "body" route tomodern psychology.The stepwas achievedwhenPavlovdiscovered a newmethodof investigation—theconditioned-reflex method; when he and his pupils applied the methodexhaustively to the behavior of intact organisms, dogs in particular; andwhen, finally, he drew from a mass of experimental fact the clearformulationofsomeimportantprinciples.

Pavlov's method, his findings, and some of his more significantgeneralizations about behavior are treated in your text (K & S, 15–35),hence require no discussion here. Since we are on the topic of history,however,youmaybe interested inknowingwhatPavlovhad to sayaboutsomeofthosewhowentbefore.Hetellsusaboutthisinthefirstchapterofhis book. He says that he owes themost toDescartes, who"evolved theidea of the reflex. Starting from the assumption that animals behavedsimply as machines, he regarded every activity of the organism as anecessary reaction to some external stimulus, the connection between thestimulus and the response being made through a definite nervous path....This was the basis on which the study of the nervous systemwas firmlyestablished."

Pavlovalsoexpresseshis thanks to thephysiologistsof the18th,19th,and20thcenturies,onwhoseshouldershewasabletostand—especiallytoSherrington, forhis "classical investigationsof... the spinal reflexes."Andhe pays his respects in still other quarters: (1) to those psychologistswhowereledbythetheoryofevolutiontodevelopanobjectiveapproachtothereactions of various animal species; (2) to the American psychologist,Thorndike, for his method of investigating the intelligence of animals

experimentally; and (3) to the English philosopher, Herbert Spencer, foranalysisofinstinctintoachainofreflexes.Toeachoftheseweshallreturnlater, either in theseNotes or in your text. They would belong to thehistory of psychology even if Pavlov had not mentioned them in hisacknowledgments.

OtherRoutestoModernPsychologyIt would be oversimplifying matters to say that the above-described

routes fromDescartes to the present centurywere the only ones.Actually,therewereseveralothers.Somewerewhatyoumightcalldetours fromthemain lines, and others were separate roads entirely. A few of the moreimportantonesdeserveourattentionhere.Mental Acts and Functions . An essentially non-experimental route to

modern psychology may be said to have begun withAristotle (364–322B.C.),whodistinguishedbetweendifferent functionsof thehuman"soul."Forpracticalpurposes,however,weneednotgosofarbackinhistory.Wecan start with Franz Brentano (1838–1917), a German philosopher,theologian,andpsychologist inthedaysofWilhelmWundt. BrentanowasaprofoundstudentofAristotle'sworkandwasgreatlyinfluencedbyit.He,in turn, exercised considerable influence upon a number of German andBritishpsychologists—aninfluencethatwascarriedintoatleasttwoofthe20thcenturyschoolsofpsychology.

Brentano, likeWundt, believed that psychology should be based upontheobservationofmentalevents.Butwhenheexaminedhisownmindinanatural,commonsensefashion,hedidnotfindsensations,feelings,orothermental elements. Instead, he found what Aristotle had found—namely,mentalacts orfunctions.These acts were of threemain kinds. First therewereactsofideating—e.g.,seeing,hearingandimagining.Secondly,therewere acts ofjudging—e.g., rejecting, perceiving, and recalling. And,finally, there were acts of loving-hating, which included feeling, wishing,intending,andothers.

Brentano's classification of acts was a personal one and not easy toconfirm by any truly scientific method.You would therefore expect thatlater classification would differ from his, just as his had differed fromAristotle's.Thatisexactlywhathappened.OneofBrentano'smosteminentpupils,Stumpfbyname, cameout fortwo classesof acts,intellectual andemotional.Still later,AugustMesser, anotherGerman, returned toa three-group classification—thinking, feeling, andwilling. Messer also tried toshowthat theseactswererelatedtomentalcontents, suchas thesensationswith whichWundt and his followers had been concerned. For Brentano,youmightsay,theactofseeingwouldbemental,butthethingseenwouldnot. ForWundt, the thing seen, the sensation,wasmental, but the seeingwas not. Messer (and others of his time) accepted both the actand thecontent as mental. All three men, however, were agreed in one respect.

Psychologywastodealwithmental, rather thanphysicalphenomenaas itssubject matter. The route from Brentano to the present day was thus a"mind,"ratherthana"body,"route.

Brentano's influence finally was felt in other countries, especially inEngland, where it was prominent in the teachings of such pioneers inpsychology as JamesWard andGeorgeF.Stout, bothofwhomwereveryactive at the turn of the present century. Still another eminent Britisher,WilliamMcDougall,broughttheinfluencetoAmericainthe20s,underthebanner of "purposive" or "hormic" psychology. We shall take up thisviewpointlater,inourdiscussionofpsychologicalschools.Voluntary Behavior and Animal Learning . In 1832, as already

mentioned,MarshallHalldistinguishedbetweenvoluntaryandotherkindsof behavior—especially the "reflex function." Hall was not, however, thefirst to make such a distinction. Even Descartes had recognized thedifference between responses that follow stimuli and those which occur atthe behest of the "soul."He had even shownhow the soulmight overrulereflex action by moving the pineal gland and thus diverting the flow ofanimal spirits from their accustomed channels.At about the same time,Thomas Hobbes, the eminent British philosopher, was talking about twokindsofbodilymotioninman:vital,seeninsuchanactivityasbreathing;andvoluntary, as in human speech.A hundred years later than DescartesandHobbes, but still in advance ofMarshallHall,we haveDavidHartleytalking aboutautomatic motions, which originate in the stimulation ofsense-organs, andvoluntary actions, which commonly arise from "ideas."Andthesewerenottheonlyonestomakethedistinctionbetweenbehaviorthatwenowcallreflex,orrespondent,andthatwhichwecallvoluntary, oroperant(K&S,49–51).

Thestudyofreflexbehavior,aswehavealreadyseen,wentonataveryrapid pace following Marshall Hall's investigations. Voluntary behavior,however,didnotsoreadilygiveintoscientificattack.Advanceinthisareaseems to have awaited the fuller recognition of an important idea thatwasgradually emerging from the philosophies of the past.To this idea, let usnowturn.

The notion that man's behavior is determined byrewards andpunishments has been expressed atmany times since the dawnof recordedhistory.FromtheteachingsofEpicurusinthegardensofancientAthenstotheutilitarianphilosophyof JeremyBentham in the18thcentury, the ideawas voiced again and again that human conduct can be understood as asearchforpleasureortheavoidanceofpain.Butitwasnotuntilthemiddleof the19thcentury that the firstattemptsweremade toput thedoctrine ingoodscientificorder.

Herbert Spencer, British philosopher and psychologist, started the ballrolling in 1855 (K & S, p. 36) when he tried to explain how a hungryanimal might learn to get its food when the usual conditions of food-

getting had been changed and some new movement was required forsuccess. In essence, his explanation went like this: when the customaryresponseswillno longerwork, theremayoccurnervouschangesassociatedwithavarietyofmuscular reactions.Oneof these reactionswillultimatelybe successful in reaching the food. This success leads to "pleasurablesensations" and a flow of "nervous energy" into such channels that thenewfound solution to the problem will be given priority—that is, theappropriateresponsewillbemorelikelytoappearwhentheanimalisagainfacedwiththemodifiedstimulussituation.

Four years later,Alexander Bain, a Scottish psychologist, took up thesameprobleminabookentitledTheEmotionsandtheWill(1859).Oneofthe thingshe tried todo in thisbookwas to showhow thewilldevelops.He decided that it beginswithspontaneousmovements.Thesemovementsoccur in certain stimulus situations, of course, but in a chance, unguidedfashion.Those that turn out to be suitable or pleasing are singled out andgivensuchstrengththattheywilloccurthereafterwhenevertheorganismisplaced in that particular stimulus situation. SaysBain: "A few repetitionsof the fortuitous concurrence of pleasure and a certainmovementwill leadto the forgingof an acquiredconnection... so that after a time thepleasureor its idea shall evoke the propermovement at once."Themovementmaythenbecalledvoluntary.

Bainalsodescribesthewayinwhichvolitionmayarisefromasituationthat isdispleasing to theorganism.Herehesays:"Therepeatedconnectionof the feeling and this onemovement (at first accidentally stumbledupon)would end in a firm associationbetween the two; therewould be nomorefumbling and uncertainty; the random tentatives, arising throughspontaneity and the spasmodic writhing of pain, would give place to theone selected and appropriate movement and we should have a full-grownvolitionadaptedtothecase."

FromtheteachingsofBainandSpencer,youshouldnowbegintoseeapattern emerge. First, there is the idea ofaccidental orspontaneousmovement,occurringunderconditionsofhungerorconfinement.Secondly,this movement results in some form ofgratification orrelief, as throughremovalofhungerorescapefromconfinement.Thirdly,thisgratificationorrelief seems tostrengthen themovement in its connection with theprevailingcircumstances,makingitmorelikelytooccuronlateroccasions.Finally, this movement belongs to the class called voluntary, rather thanreflex.

At this point, you might suggest that what we need is a gooddemonstration.The argument seems plausible enough, but it is still at aprettyabstract level.Howaboutsomeconcreteobservations?This is, then,a good place to introduce the name and one of the studies of C. LloydMorgan,anotherBritisher,whobroughttheSpencer-Bainargumentashadeclosertoexperimentalinvestigation.

Morgan's contribution to the present theme may be illustrated by asingle quotation from his very influential book,Animal Life andIntelligence, published in 1896. It concerns his pet terrier,Toby, and theanalysisofoneofToby'saccomplishments.

Theway inwhichmydog learnt to lift the latchof thegardengateandthus let himself outwas in thiswise.The iron gate is held to by a latch,but swings open by its own weight if the latch be lifted.Whenever hewanted togoout the fox terrier raised the latchwith thebackof his head,and thusreleased the gate, which swung open. Now the question in anysuchcaseis:Howdidhelearnthetrick?Inthisparticularcasethequestioncan be answered, because he was carefully watched. When he was putoutside the door, he naturallywanted to get out into the road,where therewas much to tempt him—the chance of a run, other dogs to sniff at,possiblecatstobeworried.Hegazedeagerlyoutthroughtherailingsonthelow parapet wall...; and in due time chanced to gaze out under the latch,lifting it with his head. He withdrew his head and looked out elsewhere;butthegatehadswungopen.Herewasafortunateoccurrencearisingoutofthe natural tendencies of a dog. But the association between looking outjust there and the opengatewith a free passage into the road is somewhatindirect. The coalescence of mental processes in a conscious situationeffective for the guidance of behavior did not spring into being at once.Only after some ten or twelve experiences, in each of which the exit wasmorerapidlymade,withlessgazingoutatwrongplaces,hadthefoxterrierlearnt to go straight andwithout hesitation to the right spot.In this casethe lifting of the latch was unquestionably hit upon by accident, and thetrick was only rendered habitual by repeated association in the samesituation of the chance act and happy escape. Once firmly established,however, the behaviour remained constant throughout the remainder of thedog's life, some five or six years.And, I may add, I could not succeed,notwithstandingmuch expenditure of biscuits, in teaching him to lift thelatchmoreelegantlywithhismuzzleinsteadofthebackofhishead.

Inthisexample,wehaveamoreobjectivereportofanimalbehaviorthanwastheruleinthesecondhalfofthe19thcentury.Morgan,asweshallseeinamoment,wasextraordinarilysensitive to therequirementsofscientificreporting. Yet the account still leaves something to be desired. Forexample, one cannot help but wonder how Morgan could possibly knowthat the"guidance"ofhispet'sbehaviorwas lookedafterbya"coalescenceof mental processes in a conscious situation." Or one might argue thatMorgan's descriptionwas lacking in the sort of detail that wouldmake itpossible to repeat or extend his observations with some other dog thanToby.

Theseandrelatedmattersweresoon tobe takenupbyotherstudentsof

animalbehavior,includingayoungAmericannamedEdwardL.Thorndike,whosemore importantcontributionsare treatedbriefly inyour textbook(K&S,36–42).Thenextmoveaheadwas toproduce a situation like that inwhichLloydMorgan'sTobylearnedtoliftthelatchofthegardengate,andto replace the Spencer-Bain way of talking about reward and punishmentwith simpler and more objective statements. Voluntary behavior finallygave in toexperimentalattack,and thepleasure-painphilosophyofancientGreecefoundscientificrespectability.The Darwinian Influence. Psychology, like other biological sciences,

was greatly influenced byevolutionary doctrine, although probably notquite in the way that Charles Darwin himself would have expected.Thestory can be told quickly. Darwin believed in "mental," as well as"physical,"evolution.Hethoughtthatman'smindwasnodifferentinkindthan that of the animal, and he argued that some of the most cherished"human" qualities (a "moral sense," for example) could be recognized inthoseanimalsnearest tomaninthephylogeneticscale.Onlyaquantitativedifferenceseparatedthespeciespsychologically.

This view did not appeal to persons who felt that mind belonged toman, andman alone.On the other hand, therewere thosewho accepted itreadily. They saw the exciting possibility of a comparative psychology,through which the lines of mental evolution could be traced from onespeciestoanother.

So eager were some of Darwin's followers to demonstrate this form ofdescent, that they went to great excesses in reading human qualities intoanimal conduct. "Humanizing the brute" became a popular pastime.Talltalesof animalgenius flourishedandwereoften accepted aswell-groundedfact by men who should have known better. It was a period of"anecdotalism" and "anthropomorphism" for the less critical of Darwin'sdisciples. Thirdhand testimony from the pages of theDaily Times wassometimes given the status of a scientific paper; and the tendency to readhumanmentality into animal behavior—evenlower-animal behavior—waspracticallyunrestrained.

One of the more conservative of these post-Darwinians was LloydMorgan. Fortunately, he was also one of the most influential.When herecognized some of the dangers of ascribing to animals more humanmentality than the facts justified, he decided to do something about it. In1894,hecameoutwithwhatwenowknowas"LloydMorgan'sCanon,"anattempt to encourage parsimony in the interpretation of the animal mind.Thiscanon,orprinciple,wasstated inhisbook,Comparative Psychology,asfollows:

Innocasemayweinterpretanactionastheoutcomeoftheexerciseofahigher psychical faculty, if it can be interpreted as the outcome of theexerciseofonewhichstandslowerinthepsychologicalscale.

Morganhimself, aswehave seen inhis account ofToby's latch-lifting,was not averse to crediting animals with mental processes, at least of asimple, rudimentary sort.The net effect of his principle, however, was tobring theseprocessescloserandcloser to thevanishingpoint—thepointatwhich Descartes had left them when he described animals asautomata ormachines,withoutanymindwhatever.TheuseofMorgan'scanonsoonledtotheextremeviewonthepartofsomepsychologistsandbiologiststhatitwas unnecessary to assume the existence of mental processes anywhere inthe animal scale—that behavior could be adequately explained in terms ofphysical events. Before many years had passed, this view foundconsiderablefavor,andthestagewassetforastartlingnewdevelopment.Afewpsychologistswerepreparingtoasksuchquestionsasthese:Isitusefulto assumemental processes in explaining the behavior ofman? Is it validor useful to assume that mind exists in any organism?The viewpoint ofbehaviorismwasjustaroundthecorner.

"School"DaysinPsychologyItwould simplifymattersgreatly ifonecould say that allof theabove-

mentioned historical routes converged directly on one great center, called"modernpsychology."Unfortunately,thiscannotbedone.Otherthingshadt ohappen first, most important of which was a hectic period ofpsychological schools that began shortly after the turn of our century andlastedwellintothethirties,andperhapsevenbeyond.

Duringthisperiod,sayaround1925,ifyouhadtakenuppsychologyinanAmerican college, you would soon have met with one or more strongopinions as to the true nature of your science. Had you gone to Cornell,you would have been introduced to thestructuralistic viewpoint; atChicago you would have foundfunctionalism; and, at Johns Hopkins, itwouldhavebeenbehaviorism.IfyouhadgonetoHarvard,thetrendwouldhavebeenstructuralistic,butyoumightalsohavebeenexposed to lectureso n"gestalt," purposivistic, andbehavioristic psychology by eminentvisiting professors. At Columbia, you would have found eclecticism, amiddle-of-the-roadpolicythattoleratedalltheseviewsbutsponsorednone.And everywhere you would have heard reports of something called"psychoanalysis." This was a period of competing doctrines, vividleadership, and hot-blooded allegiances. It was a stormy, but healthy, andprobably unavoidable, stage in the development of a young science.Theveterans of thiswar between the schoolsmay be forgiven if, on occasion,theyreenacttheoldbattlescenesandrenewtheancientgrudges.Thoseweretruly exciting days, in comparison with which the current competitionbetween"learningtheories"seemspaleanddispirited.Structuralism.The first-establishedand foryears themost influentialof

the schools was structuralism. This school was headed, in the UnitedStates, by E. B.Titchener (1867–1927), at Cornell University.Titchener,

although British born, had been a pupil ofWundt's at Leipzig, and hissystem of psychology was essentially the one that Wundt started.Accordingtothissystem,mindwasthethingtobestudied,bythemethodofintrospection. Observers, carefully trained and well instructed, were toexamine their own experience, under special stimulus conditions, and toreportonthisexperienceinthegreatestpossibledetail.Thiswasessentiallythe procedure that German physiologists and psychologists had used forfiftyyearsormore in theirstudyof theway inwhichsense-organsreact toexternalinfluences.

The introspective or structural psychologist aimed to study mentalstructure by the method of introspection. He aimed to analyze experienceinto its basic elements; to study the compoundingof the elements; and torelateexperiencegenerally tocertainphysicalevents—especially tochangesin the stimulus world. In a typical introspective experiment, for example,anobservermightbeaskedtoreportonthechangesthatheexperiencedinatonal sensation when the experimenter changed the wavelength of thephysicalstimulus.

The structuralists were concerned with sensation, image, feeling,perception, idea, emotion, memory, association, imagination, attention,action, thought, anda fewother topics—all fromapurely subjectivepointof view.Behavior was something inwhich they had no interestwhatever,handing it over for study to the field of biology. "Action," for example,was of interest only as onefelt it in himself.The critics of structuralismwerelatertoarguethatanobserver'sreportofhisexperiencewasinitselfaformofbehavior,ofaverbalkind,butTitchenerandhisassociatesalwaysfeltthatthetrulypsychologicaleventwastheexperiencereported,ofwhichbehavior was no part. Fortunately for us, the observer's reports in suchexperimentsweremadeandrecordedwiththegreatestofcare,andtheywereusually related toknownandmeasurable stimulus conditions.Hence, evenif onedenies that experience canbe observed, hemay stillmakegooduseofthedatafromthestructuralist'sresearches.

Structuralismnolongerexists, inanyactivesense.Somehavesaid thatit diedwithTitchener, in 1927, and theremay be truth in this statement.While he was alive, Titchener certainly held undisputed sway over hispupils and co-workers, by virtue of his great scholarship, his skill inwritinganddebate,andhis fatherlyseverity.Uponhisdeath, therewasnoonewhocouldcommandthesamedegreeofrespectasa leaderorwasabletoenlistnewworkersforthecause.Itmaywellbethat,withoutahead,theschoolbrokeupsoonerthanwouldhaveotherwisebeenthecase.However,a more important reason for the decline of structuralism was the rise ofseveral opposing points of view, especially those of functionalism,behaviorism, and gestalt. Each of these schools attacked theTitchenerianposition;eachdrewstrengthfromadifferentlineofhistoricaldevelopment;eachhaditsowninspiringleader,orleaders;andeachmanagedtoattractits

ownshareofattentionandsupport.Functionalism. Of these rival viewpoints, functional psychology least

deservestobecalledaschool.Infact, it ishardlytobedistinguishedfromthemiddle-of-the-roadpositioninpsychologythatwasreferredtoearlieraseclecticism. In its history, its leadership, and its basic teachings,functionalism presents a picture of confusion. It never achieved a clearstatementofpsychology'ssubjectmatter,itsmethods,oritsprimarygoals.In fact, it isaquestionas towhowas,andwhowasnot,a functionalist;afewnames* are agreedupon.Others are not.Yet, there are certain signsorearmarksof the functionalist that seem to set himapart from themembersofotherschools;andsomeoftheimportantonesmaybenotedhere.

The functionalistwas influencedbymore thanonehistorical trend.Themajor influencewasBritish, but this included associationism,Darwinism,and the local version of Brentano's act psychology. In addition, therewasthe influence of English and German sense-organ physiology and thephysiologicalpsychology taughtbyWundtandbrought to thiscountrybyTitchener.Thiscombinationof factorsgave to functionalismabreadthandvariety of appeal that was lacking in structuralism, and in those systemsthatwere tocome later. Itwasnot,however, an entirely newcombination.Something of the sort had already beenwritten byWilliam James (1842–1910), the earliest dean ofAmerican psychology. James did not belong toany school, but hisPrinciples of Psychology (1890) is the book towhichyouwouldgotogetthebroadbackgroundandrichflavoroffunctionalism,withoutmuchappearanceofsystem.There,too,youwouldfindaccountsofhabit, emotion, theself, and other matters—classics in psychologicalliteraturewhichalonewouldmakeatriptothebookworthwhile.

The functionalist, as the name indicates,was interestedmore inmentalfunction than structure. He felt that a description of one's experience wasnot enough.Hewanted toknowwhatwentbefore the experienceandwhatcame after—what caused the experience and what the experienceaccomplished.IfhehadbeenforcedtosidewitheitherBrentanoorWundt,hewouldhavehadtopicktheformer.Yet,likeMesser(p.6)andothers,hewouldhavepreferredtoincludebothwithinpsychology.

The functionalist was willing to study behavior even when nothingmentalwas assumed to be involved.Hewas ready to broaden the field ofpsychology to include animal, child, and abnormal behavior, whereintrospectionsaredifficult,orimpossible,toobtain.

The functionalistwas a student of individual differences.Titchener andthemembers of his schoolwere concerned onlywithwhat youmight callthe-mind-in-general.Theyweresearchingforlawsthatheldforanynormal,adult, human being, anywhere in the world. Differences between humanbeingsweregivenvery little attention.They felt that suchdifferenceswereimportant mainly forapplied psychology—"mental testing," for example;and theyhadno sympathy forpurelypractical pursuits.Functionalism,on

the other hand, was favorably disposed, from the very start, to theapplication of science to daily affairs; and this meant the recognition ofhumandifferencesaswellassimilarities.

Finally, the functionalist was an evolutionist. He took Charles Darwinveryseriously.Mentalfunctionwasof interest tohimin largepartbecausehe felt that itwasan important factor in thesurvivaloforganisms—that ithelpedmantoadjustoradapttohisenvironment."Adaptivebehavior"wasaveryimportantfunctionalistconcept.Behaviorism. Structuralism, as noted earlier, was the school that stood

neartheendofthemindroutetomodernpsychology.Functionalismwasalittle further along, at the intersection of this route with several others,includingtheonesfromBrentanoandDarwin.Behaviorismcamestill later,atapointwherethebody route led into themain lineof traffic,andnot farawayfromthefunctionalists.Itwas,inasense,anextremedevelopmentofthe functionalist belief that psychology should include the study ofbehavior.Also, John B.Watson (1878–1958), the principal leader of thenew school, was himself trained by functionalists, at the University ofChicago.

Putting it baldly, you might say that structuralism argued for theexclusive study of mind; that functionalism argued for the study ofbehavior as well as mind; and that behaviorism argued for the exclusivestudy of behavior.Watson once said thatmental states, "like the so-calledphenomenaofspiritualism,arenotobjectivelyverifiableandforthatreasoncan never become data for science.... In all other sciences the facts ofobservation are objective, verifiable and can be reproduced by all trainedobservers....Psychology,on theotherhand,asascienceof 'consciousness'hasnosuchcommunityofdata.Itcannotsharethem,norcanothersciencesuse them.... Even if they existed, they would exist as isolated, unusable'mental'curiosities."

Behavior was treated conventionally by Watson as the activity ofmuscles and glands.With respect to the muscles (his main interest), hemadetheusualdistinctionbetweenthestriped orskeletalmuscles, suchasthose used in walking, talking, writing, and so on, and theunstriped orsmooth muscles, known to be involved in reactions of the stomach,bladder,bloodvessels, and the like.As for theglands,he recognizedboththe duct and ductless types as psychologically interesting. He knew thatsome of the former (e.g., the salivary glands) had been modified in theiractionby thePavlovianprocedureofconditioning;andhe felt thata studyofthelatter(theendocrines)shouldthrowmuchlightupontheproblemsofemotionandmotivation.

Watson tells us, also, that responsesmay be grouped in another, morepsychological, way. We may distinguish between "explicit," overt, orobservable responses and "implicit," covert, or non-observable responses.Talking, for example, isovert behavior, whereas thinking iscovert. (We

shall come back to this distinction later, in our discussion ofChaining.)Further, we may also distinguish between "habit" responses, acquiredduring one's lifetime, and "hereditary" or "innate" responses such as thosereflexeswithwhichaninfantisequippedatbirth.

Watson, like other behaviorists, was interested in theobjectiveobservation of behavior.This included, for him, the observation ofman'sconductinaneverydayworld,aswellashisreactionsinalaboratoryunderspecial conditions of stimulation. It included, too, the new and verypromisingconditioned reflex technique that Pavlov had developed, but ofwhich we still knew little inAmerica. It included the testing methods—intelligence tests, special ability tests, achievement tests, and so on.(Watson, like the functionalists, saw no vice in putting psychology topractical uses.)And, finally, it included the method of verbal report, theWatsonian substitute forintrospection.This report might be as simple asthe utterance of "cold" or "warm" by a subject who had been suitablyinstructed and stimulatedwith a cold orwarmobject.Or itmight refer toone'sownbehavior, aswhenaperson reports thathe iswriting somethingorthathisheartispounding.Inthefirstcase,Watsonsaysthatthemethodis on a par with the conditioned-reflex method, and that objectivepsychologycandealwithspeech reactions ("cold,""warm,"etc.)aswellasit can with other forms of motor behavior. In the second case, where thereport is of one's own behavior, he says the method is less reliable, butsometimes useful. It should, if possible, be supplemented or replaced bybetter techniques, suchas thoseofautomatic recording.Thus, rather thanaverbal report of a muscular cramp in one's leg, it would be better if weattached to the leg an instrument thatwould record directly any change inthetensionofthemuscles.

Psychology's goal, for Watson, was topredict andcontrol humanbehavior.Prediction, he says, has two aspects. We may observe someresponseandthentrytotellwhatbroughtitabout—topredictthesituationor the stimulus. Or, we may know the situation and try to tell what theresponse to it will be. In the first case, man has to be studied "in actionfrom birth to old age, in such away that the behaviorist, with reasonablecertaintywhenwatchingtheindividualbehavecantellwhatthesituationorstimulusisthatcallsouttheact."Inthesecondcase,wemust"experimentwithman'sbehaviorfrominfancytooldage(so)that,giventhesituationorstimulus,weareabletopredicttheprobableresponse."

As forcontrol,Watson says "Every scientist feels that he can makeprogressinhisfieldjusttotheextenttowhichhecangaincontroloverthematerial with which he works... the psychologist likewise, having chosenhumanbehaviorashismaterial,feelsthathemakesprogressonlyashecanmanipulateorcontrolit."

Watson expressed these ideas in 1919, in hisPsychology from theStandpoint of a Behaviorist.This is the book inwhich youwill find the

best statement of his position. Psychology is to study behavior.Observationswillbeobjective,asintheothersciences,andtheaimwillbepredictionandcontrol.Thedescriptionofbehaviorwillinvolveitsanalysisintoelementaryparts—thereflexresponsesofmusclesandglands.Behaviorwillbe related to theenvironment,but itwillalsobe related finally to thephysiological activity of the sense-organsand the nervous system.Ultimately,weshallattempt"to formulate, throughsystematicobservationandexperimentation,thegeneralizations,lawsandprincipleswhichunderlieman'sbehavior."

Other references toWatson and his views will be made later in theseNotes and inyour text. In1919,hispositionseemedpretty radical.Todayit seems less so,althoughnotwithoutdefects.Forexample, therewas toomuchdependenceonPavlov'sconditionedreflexinWatson'sexplanationofbehavior change. He failed completely to recognize that "voluntary" andreflex behavior must be measured in different ways and controlled bydifferent methods (K & S, p. 51). Basically, however, his programprospered.Heput themarkofobjectivismuponpsychology,apparently forall time. His system, as an organization of facts, has broken down.Anysystemmust, if science is to advance, because new facts replace old onesandrequireneworganization.Buttherearefewpsychologiststoday,ofanyreputation, who would talk seriously of mind, mental activity, or othernon-physical phenomena as the subject matter of our science. In thisrespect, we are all behaviorists, and Watson has triumphed. In otherrespects,asyouwillsee,hewasnotsosuccessful.Gestalt Psychology. If you were to make a survey of the hundreds of

experiments thatwere conducted in theUnitedStates during thepast year,you would find only a handful of references togestalt psychology or itsteachings.Yet,nomorethantwentyyearsago,thiswasoneoftheliveliestof the schools to brighten the American scene, and its influence hasprobably been much greater than your survey could hope to show. Theschool may no longer exist as such, but its contributions have benefited,andhavebeenabsorbedwithin,ouryoungscience.

"Gestalt" is a German word, variously translated as "form," "shape,""structure," or, most commonly, "configuration." The school itself,sometimes called "configurationism," originated in Germany. Its principalleaderswereMaxWertheimer(1880–1943),WolfgangKohler(1887–),andKurtKoffka(1886–1941).Allthreemenultimatelyadoptedthiscountryastheirhomelandandheldprominentpositionswithinouruniversities.

Gestalt psychologywas at first simply a rebellion against structuralismand its accepted doctrine of association. The rebellion started whenWertheimerandhis twopupils,KohlerandKoffka, found it impossible toanalyze "apparent movement" (such movement as we see in electric signswhen the individual lights flash in a certain succession) into the sensationelements described byWundt and his followers. Movement,Wertheimer

decided,wasinthiscaseaphenomenon(hecalleditthe"phiphenomenon")that defied further analysis. It was awhole ortotality that could not beprofitablydissectedassomesortofcompoundofelementsheldtogetherbyassociation. Itwasnot tobeexplained, either, as an "illusion"baseduponone'searlierexperiencewithobjectsthatreallymoved.

From this small beginning, a larger protest developed. Structuralisticanalysis was attacked on a broad front. It was treated as purposeless,artificial, and ruinous of the more natural "gestalten" of experience.Alsoattackedwas structuralistic associationism—the view thatmental elementswereassociated to formcompounds."Pastexperience" ingeneralwasruledoutalmostentirelyasanexplanatoryprinciple."IfIseethispatternoflinesas a cube in three dimensions of space," said the gestaltist, "it is notbecause I have had past experience withreal cubes in a three-dimensionalworld,but simplybecause the linesareorganized in thisparticularway.Adifferent organization would favor atwo-dimensional figure."And so on.There were arguments and counter-arguments, supported with experimentsfromeachofthewarringcamps.

But the gestaltists did not limit themselves to an attack uponstructuralism. They found similar flaws in behaviorist teaching. Theatomistic analysisofbehavior into little reflexunitswas just asbad, fromtheir viewpoint, as the analysis ofmind into sensations.Also, they foundthatthebehavioristwasusingtheconceptofconditioninginawaythatwasnomoretobedefendedthanthestructuralist'suseofassociation.

The structuralists and the behaviorists were not very well prepared toanswer the gestalt arguments against analysis into elements because theyhad not yet come to a very clear idea of what these elements were.Thestructuralistshad justaboutconcluded thatimages andfeelings were reallynodifferentfromsensationsintrospectively.Indeed,theywerealmostreadyto say that one did not evenobserve sensations.They had thus whittleddown theirdatum or object of observation almost to the vanishing point.The behaviorists were also in trouble withtheir units of analysis.Sometimesaresponseseemedtobenomorethanthetwitchofamuscle;atother times itmight be "building a house."A stimulusmight in one casebedescribedasarayoflight,focuseduponasinglespotintheretinaofthe

eye; in another, it might be a very complex pattern of light rays called,perhaps,bythename"dog."Watsonspokeofthesecomplexesas"stimulussituations" rather than "stimuli," adding that a "situation is, of course,uponfinalanalysis,resolvableintoacomplexgroupofstimuli."

Also, both structuralism and behaviorism had gone pretty far, on someoccasions, in using "association" and "conditioning" to explain why oneexperienced something or behaved in a certain way.Watson, for example,had suggested that practically all differences in human achievement wereduetodifferenthistoriesofconditioning.

Today,aswelookbackonthiswarbetweentheschools,wecanseethatthe structuralists and the behaviorists were vulnerable to attack. (Thefunctionalists' position was never clear enough to be fired upon.) But wecan also see that gestaltoverdid the objection to analysis andassociationism,eitherofthementalorbehavioralkind.Analysisisbasictoallscientificdescription,andthekindpracticedbythestructuralistsandthebehaviorists was as valid as that which the gestaltist accepted.Also, weknowthatone'spasthistoryisimportantindeterminingwhatonemaysee,hear,ordo.Eventhegestaltistscamefinally torecognizethesethings.Yettheir protest served a purpose. It pointed up weak spots in their rivals'armors, and it broadened our viewofwhatmight legitimately be includedassubjectmatterforscientificinvestigation.Someofthegestaltclaims,aswe'llseelater,cannotbeupheld.Otherscanbe,andformanintegralpartofmodernpsychology.HormicPsychology. "Purposive" or "hormic" (from theGreek,hormao,

meaning tourge orimpel) psychologywasnever as influential as either ofthe schools alreadymentioned.Yet,WilliamMcDougall (1871–1938), itsBritish-born leader, was one of the outstanding psychologists of thiscountry,andhisviewsarereflectedinseveralmoderndoctrines.

As noted earlier, McDougall was in the tradition ofact psychology.And, like somewhowent before, he recognized threemain classes of act.There wascognition (sensing or knowing),conation (willing or striving),andaffection(feelingor"emoting").Heputmoststress,however,upontheconativeaction,thepurposiveendeavoroforganismstoreachtheirgoals.

McDougall's views, as presented in hisOutline of Psychology (1923),hadmuch incommonwith thoseof functionalism.This isunderstandable,on two main counts. Act psychology is itself a kind of functionalism,dealingmorewithmental "activity"or "function" thanwith"content"; andMcDougallwas greatly influenced byWilliam James,whose teachings, asmentionedbefore,werefunctionalisticbefore therewasafunctionalschool.The similarity of the two schools will be especially apparent later, whenMcDougall's treatment of "instinct" (the motive power of conation) iscomparedwithHarveyCarr'saccountof"adaptivebehavior."

Inasense,too,McDougallwasabehavioristalthoughhespentmuchofhis time in fightingWatson's views. In theOutline, for example, he says

that behavior is not amatter of reflexes or reflex combinations, asWatsonstated.True behavior, he says, has six characteristics that reflex behaviorlacks. (1) It showsspontaneity. It need not be elicited by a stimulus,althoughitoftenis.(2)Itmaycontinueintheabsenceofanystimulusthatstarted it off.A squirrel, treed by a dog, may continue to scamper awaywhen the dog has passed out of sight or hearing. (3) It showsvariability."When an animal persists in the movements initiated by a sense-impression,itsmovementsarenotpredictableindetail."Therunningofthedog or of the squirrel, in our example,would never be exactly reproducedon successive encounters, although the sameend resultmightbe achieved.(4) The variedactivity ends when the aimed-for result is achieved. Thesquirrel ultimately ceases his flight and resumes his daily round ofoccupations. (5) True behavioroften shows preparatory movements. Thedogmay crouch before rushing his prey, or the squirrelmay sit up alertlybefore taking off. Finally, (6) true behavior shows the elimination ofuselessmovements with practice. On repeated runs, both the dog and thesquirrel improve in their efficiency—as by terminating the pursuit or theflightatanearlierpoint.

Regardless of the rightness or wrongness, or even the newness, ofMcDougall's argument, one thing is pretty clear. He approached afundamental distinction, to which you have already been introduced,between reflex and voluntarymovement.Thiswas something thatWatsonhadmissed, probably because of an understandable fear of admitting suchterms as "spontaneity" and "volition" into a natural-science description.Such termsoften seem to imply a stateof affairs inwhich eventsoccur inthe absence of any natural cause—a situation that is abhorrent to mostscientists.

A second noteworthy feature ofMcDougall's accountwas his emphasisontheserialnatureofbehaviorasweordinarilyobserveit.Hisexamplesofbehavior are never of single responses to single stimuli. Rather, they areacts in succession. Within the succession, McDougall thought he saw"purpose"—another term that raises scientific blood-pressure—butwemayaccept thefact of serial response without reading into it any hypothetical"goal-striving."

Still a third point. McDougall sawvariability in performance, even inhighly routine performance. To some psychologists, like Watson, whoaimedtopredicttheexactresponseofanorganismtoeachspecificstimulusin its environment, this sounded like an admission of defeat. ToMcDougall, itwasmerelyanacceptanceofobserved fact.Youwill see, indue course, that one may have variability and predictability at the sametime.Muchdependsonwhatwe try topredict—or,better,whatwedefineasaresponse.

Further intoMcDougall's psychologywe need not go at this time.Hisviewpoint reachedAmerica at a timewhen every effortwas beingmade to

developpsychologyasanaturalscience,modeling itafter theolder,better-established disciplines. Often he seemed to be pulling in the wrongdirection. He criticizedWatson and others whom he thought to be toonarrow. He thought he saw "mind" and "purpose" within or behindbehavior. He conducted an experiment that seemed for a time to supportLamarck'swell known theory of inheritance of acquired characteristics.Heevensponsoredsuchquestionableenterprisesasthoseof"psychicresearch."All thismade forsomuchunpopularity thathispositivecontributionsdidnotget theattention theydeserved.Yet,hewasamanofgreat ability andwide scholarship—in such diverse fields as those of social, abnormal, andcomparative psychology, not to mention the related fields of philosophy,ethics, and anthropology. His breadth of interest was exceeded onlyperhaps,by thatofhis idol,WilliamJames. In another land,or at anothertime,hemighthavefatheredamoresuccessfulschool.Psychoanalysis.Strictlyspeaking,psychoanalysisdoesnotbelonghere.

It was not, and is not, a "school" in the sense of theword as it has beenused in this section of yourNotes. Neither its founder, Sigmund Freud(1856–1939), nor any of his followers ever aimed to draw together all thefacts of psychologywithin a coordinatedwhole like that of structuralism,behaviorism, gestalt, or other known systems. Only in the sense of theirdevotion to amethod of therapy or atheory of personality can theexponents of psychoanalysis be said to have formed a school.Although itgrew up alongside these other systems, it was not a part of any academictradition. It had its roots in clinical observation and curative technique,rather thanclassroomdiscussionor laboratoryexperiment.Also in spiteofexercising considerable influence upon academic teaching and research, ithas never in turn been seriously affected by developments within thosefields.

How long this will continue is a question.There are signs of a closerrelationbetweenthelaboratoryandthedoctor'sofficeofthefuture.Modernstudies of the effect of reward and punishment upon the behavior of manand thehigher animals oftenpoint to conclusions that are similar to thoseofpsychoanalyticdoctrines.Someofthesestudieswereundertakenwiththeexpress purpose of finding experimental parallels or analogues ofpsychoanalytic concepts such as those of "regression," "repression,""fixation,"and soon.Othersaimedonlyat the systematicdevelopmentofbehavior science, but led to ideas that were later seen to have apsychoanalytic counterpart.Thorndike's "law of effect," for example, wasderived from the experimental investigation of animal learning, but isclearlycomparabletoFreud's"pleasureprinciple."Stillothersprobablyhadatwo-foldpurpose,beingundertakentoadvanceourbasicknowledgeatthesame time that they illustrated the operation of an analytic "mechanism."Someoftherecentstudiesofavoidancebehaviorwouldseemtofallwithinthiscategory.

There will be no attempt, in theseNotes, to present an outline ofpsychoanalysis.A review thatwould be suitable in size for inclusion herewould probably tell you little that you didn't know already or that youcouldn'tfindinmoreattractiveformelsewhere.*Nowhere in theworldhasFreudian theorybeenbetter receivedormorewidelypublicized than in theUnitedStates.Majorelementsof thedoctrinewill,however,beconsideredin numerous connections, especially as they may be related to laboratoryfindingsorprinciplesofbehaviordrawnfromthem.Youwillfind,insomeinstances, that analytic concepts fit readily within the rapidly risingstructureofmodern"learningtheory."Inothercases,theywillnot,andyouwill have to decide for yourself the place they should fill within ourscience.

LearningTheoriesSchoolsofpsychologyno longerexist.The fightinghasdiedout inall

but a fewout-of-the-wayplaces, andonly a handful of die-hards remain todefend the old positions.Psychologies have given way topsychology—apsychology that has some of the features of all its predecessors, minusmany of the defects. This psychology is essentially behavioristic in itscoloration. Its methods are objective and it shows a general reluctance totreat any form of the "mental" either "act" or "content," "sensation" or"gestalt"—asitssubjectmatter. Inmostofourcollegesanduniversities, itis a psychology of stimulus and response (or stimulus, organism, andresponse)inwhichthereislittleornoreferencetononphysicalmatters.

Yet, thispsychologydiffersfromitsWatsonianancestor inanumberofways. It deals with voluntary (operant) as well as reflex (respondent)behavior. It does not limit itself toPavlovian conditioning as the soledevice for modifying behavior. It encourages sensory studies that wereformerly restricted to the laboratory of structural and gestalt psychology.And so on.These are not the only differences, but perhaps theywill helpyouseewhymostacademicpsychologiststodaywouldnotcallthemselvesbehavioristsevenwhenthebehavioroforganismsistheirsoleconcern.

Within this studyofbehavior, there is still room for disagreement.Forexample,wehaveseen,during thepast25yearsorso, thedevelopmentofdifferentlearning theories. Starting out with the modest aim ofsystematizing the facts in a single psychological area (learning), thesetheories have become almost as broad in scope as were the schools theyhave replaced.Theyhaveattempted togiveanorderly, economicalaccountof behavior in general, including all the principal areas of psychologicalresearch.

Among these attempts, we have the so-calledcontiguity theory,advocated formanyyearsbyEdwinR.Guthrie (1886– ), at theUniversityof Washington. This theory, the best statement of which is found inGuthrie'sPsychology of Human Learning (rev. ed., 1952), was based

originally upon the principle of Pavlovian conditioning and derivedmuchof its support from Guthrie's analysis of the behavior of cats in a specialkindofpuzzlebox. Itsbasic tenet is that "acombinationof stimuliwhichhasaccompaniedamovementwillon its recurrence tend tobefollowedbythat movement."The theory has not led to much laboratory investigationand has not foundmuch support from experimental psychologists.Yet, ithas not been proved "wrong" so much as incomplete and less appealingthanrivalformulations.

Thesign-learning theory,originatedbyEdwardC.Tolman(1886– ),attheUniversityofCalifornia,haditsmainsourceinexperimentswithwhiteratsinmazes,carriedoutbyTolmanandhispupils.TheimportantbookisTolman'sPurposive Behavior in Animals and Men (1932), but this hasbeensupplementedbyanumberofimportantfollow-upsinboththeory andexperiment. It has been a productive theory experimentally—more thanGuthrie's,andlessthanthosetobementionednext.Ithasbeendescribedaskind of "purposivistic-gestalt behaviorism"—a theory that recognized, andattempted to meet, the principal objections toWatson's system. Its mostdistinguishing feature has been its emphasis upon "intervening variables."Thesearespecialpsychologicalprocesses,suchas"cognitions,""purposes,"and "hypotheses," which are assumed to exist between stimuli andresponses,andaresupposedlyrequiredif theresponsesaretobeadequatelyexplained.They are not physiological processes, and they are not mentalprocesses. They are thought of by Tolman simply as "processes whichinterconnectbetweentheinitiatingcausesofbehavior,ontheonehand,andthe final resulting behavior itself, on the other."They are conceptual, youmightsay,ratherthanphysicalormental.AndmuchoftheargumentaboutTolman's theory has centered about the need for assuming their existence.Tolman,ofcourse,arguedthattheywereessential;hiscriticsdisagreed.Reinforcement theory is without doubt the reigning viewpoint today

among those experimentalists in psychologywho feel the need for puttingthe facts of behavior into some kind of order. There are, however, twobrands of this theory as presently taught. One has been associatedmainlywith thenameofClarkL.Hull (1894–1952)atYaleUniversity; theother,with that of B.F. Skinner (1904– ), at Harvard. The two brands wereindependentlydevelopedanddifferinseveralimportantrespects.Theygivedifferentmeanings tocertainbasic terms, even "reinforcement" itself.Theystress the importance of different measures of behavior and differentexperimental designs. They differ in their willingness to assume theexistence of "intervening variables" between stimulus and response.Theyrepresentdifferentphilosophiesof research.Yet, theyoften speak the samelanguageandleanuponeachother'sfindings.Commonly,too,theypredictt hesame sort of experimental results. The differences in the way theyexplain these results are, therefore, seldom of any practical importance.Today it appears that the two are at least cousins in their systematic bias,

and perhaps they're even closer. Certainly, they resemble each othermuchmorethantheyresembleanyrivalviewpoints.

TheseNotes, and your text, are written from the standpoint ofreinforcement theory. The adopted version is the one outlined first bySkinnerinhisBehaviorofOrganisms (1938),andelaboratedmost recentlyin hisScience and Human Behavior (1953). Yet, you will also findreferencesinyourreadingtostudiesthatwerecarriedoutundertheguidanceofHullorhisassociates,andwithintheframeworkofhissystem.Youwilleven find an occasional description of some "Hullian" concept. In somedegree, that is, you will be introduced to the other branch of the family.Later on,whenyouhavemasteredone formof the theory, youwill easilygrasptheother.

InBehalfofSystemYears ago, in hisBeginner's Psychology (1915), Titchener, the

structuralist, gave the following advice to his readers. "Psychology," hesaid, "has only recently turned to scientific methods; and when the timecame for it to take its place among the sciences, there was naturallydifference of opinion regarding the standpoint it should assume, theprocedure it should follow, themodel it should seek to copy.When suchdifferencesofopinionareobtained, thebestway tobeginyour study is tomasteronesystemthoroughly;yourideasarethusmadeconsistentandyourknowledge receives an orderly arrangement; then, as you read further, youcan use this system as a touchstone whereby to test new ideas and toarrangenewknowledge;and if thenewideasseempreferable to theold,orif theold frameworkbreaksdownunder thenewknowledge, you can alterthe system accordingly. If you begin, on the contrary, by studying anumber of worksabreast, you are liable to become confused.And it isbetter to be wrong thanmuddled; for truth, as Bacon said, emerges morequicklyfromerrorthanfromconfusion."

Thiswaswisecounsel in1915,andevenwisercounsel today,whenwehavesomuchnewmaterial, suchawide rangeof facts, tocontendwith inour science. It is important, both practically and theoretically, that someorganization be given to your knowledge.This is especially true if yourfirst course in psychology is also to be your last one, asmust be the casefor many students.The person who takes away from his first coursenothing more than a large body of disconnected and sketchily examineditems of fact, method, or theory has only a superficial and temporaryadvantageover thepersonwhonever attended the course. If, on theotherhand,hehasbeengivenasystematicorientation,heshouldhavelearnedtothink psychologically, with respect to everyday affairs as well as those oftheclassroomorlaboratory.Hisapproachtonewproblemsshouldbemoredirectandincisive;hisestimatesofthesignificanceofnewfindingsshouldbemoretelling;andheshouldbelessdisturbedbytheclaimsofquacks,or

blindedbythedustofpopularfancy.As a beginner in psychology, you need not fear that your study of a

system of behavior will either restrict your movement or narrow yourvision.The opposite is more nearly true.A thorough understanding of asystem will increase both the quality and quantity of your productiveactivity,anditwillopenupmoreexcitingrealmsofthinkinganddiscoverythan you have ever known. In brief, it will give you apower andperspectivethatnootherapproachtopsychologycouldpossiblyprovide.

EDITOR'SINTRODUCTION

Psychologists havebeen ardent professionals, an eager, easily convertedlot. No wonder the cry is often heard among them "There is news in theland of Babel," meaning "Here now isthe psychology!" So it has comeabout that there are mechanisms of automatic defense against the assertedexclusivenessand thepropagandaofbehavior theorists.Chief among themis undoubtedly negative adaptation or, as the authors of this book wouldhaveit,withgreaterillumination,"absenceofreinforcement."

Igrant theserviceablenessofsuchdefenses inpreservingcommonsenseandhealthyskepticism,yetIamsorryforthepsychologistwhomissesthisout-of-the-ordinary textbook. He may be one whose own work lies farafield.But nomatterwhat thatworkmaybe, itwould enhance his visionand build his morale to know that it has been possible already todemonstrate, operationally and therefore beyond challenge, so muchlawfulness of behavior on the single assumption that all the features oflearnedbehaviorarebuttheroutes,straightroutesanddetours,downwhichan organism has been baited. Hemight quibble over the excessive use ofratsandbalkattheextrapolationstohigherbehavior,buthecouldnotdenymassivefactsthatstick.

Iespeciallycongratulateyou,thethoughtfulstudent,whosefirstorearlyexposuretopsychologyisthroughthisbook.Itsuseasatextisaguaranteethat you have an instructorwho knows that the basis of every science liesnot in talkandproofbysay-so,but inexperimentalmethods.Atbestyouare going to learn psychological science by your own sciencing, in alaboratory.Ifcircumstancesdenythatprivilege,yourinstructorwillstillseeto it that you get the next best by perfectly feasible demonstrations in theclassroom.Finally,ifthisbookarousesinyouthetinglingenthusiasmthatinanearlier form ithasplainlyevoked inmanystudents,youareonyourway to insights of the greatest value.Theywill be of use to youwhetheryoubecomeapsychologist, teacher, lawyer,salesman,philosopher,doctor,orjustapersonwhofeelstheneedtoseebeneaththeseemingchancinessofhumanbehavior.

RICHARDM.ELLIOTT

PREFACE

Thisbookisanewkindofintroductiontopsychology.Itisdifferentinthat it represents for the first timeapointofview that is coming toguidethe thinking and research of an active group of psychologists in thiscountry.Themembersofthisgrouparemainlyexperimentalists,laboratoryworkers, who spend much of their time in observing and measuring thebehavior of organisms–rats, dogs, guinea-pigs, apes, pigeons, and, ofcourse, human beings. They are unflaggingly on the lookout forfundamental principles of behavior–principles that hold true for the whiterat aswell as thecollege student, for thedog in laboratoryharness aswellas thepatienton thepsychoanalyst'scouch, for the tribal savageaswellasthe sophisticatedproductofourownculture.Already theyhavediscoveredsomeof theseprinciplesandhavebrought themtogether in thebeginningsofscientifictheory.Otherprinciplesare,atpresent,onlysuspected,andthesearchgoesonat anever fasterpace. In thisbook,we try to tell about theones of which we are certain; we describe some of the research they arebasedon;andwepointoutthewayinwhichtheymaybeorganizedtogiveameaningfulpictureofhumanconduct.Wehopethatsomethingofinterestand use, perhaps even something of adventure, will be found in ouraccount.

This has not been an easy book to write. Our aim, at the outset, wasclearenough:wewantedtoconstructanelementarytext inpsychologythatwould be suitable for our ownuse and, hopefully, for the use of a rapidlygrowing number of like-minded teachers.We had felt, for some time, theneedforabook thatwould integrateclassicalandcontemporary thought ina way that would adequately represent the dominant theoretical trend oftoday. But when, at last, we undertook to write it ourselves, we soonbecameawareofthedifficultiesinvolved.Wehadnomodelstoworkfrom,no tradition to follow, at the undergraduate level of exposition. Withrespect to the content aswell as the formofour text,wehad to relyuponourownjudgment,againandagain—oftenwithmisgivings.

Wefound,too,thatseveralreadershadtobekeptbeforeusaswewrote.There was the beginning student, the object of our greatest concern, whomight never go beyond an introductory course. There was the advancedundergraduate, even thegraduate student,whohadnotyet beenacquaintedwithourwayofthinking.And,finally,therewereourcolleagues,watchingus to see howwellwewould be able to carry reinforcement theory to thestudent.

Now,withourbookinthehandsofthepublisher,wearewellawarethatit reflects our struggle with these problems, rather than their completesolution. No one could think otherwise.We will have to be content, atpresent,withsomethinglessthantheloafweintendedtoputonyourtable.Ontheotherhand,wethinkthereissomethingnutritiousinit.Howwellit

hasbeenkneaded,andhowwellbaked,thereadermustjudge.

Allbooksareindebtedtomorepeople,formorethings,thantheauthorscan remember.At this juncture,we recallmost clearlyRichardM.Elliott,whoseconfidence inourprojectnever falteredandwhowasmostgenerouswith his reinforcements;ThomasW. Reese and Kenneth MacCor-quodalewho,withgentlepainstaking,rootedouterrorsandaddedtheir thoughtstoourownatmanypoints;andMurraySidmanandJamesA.Dinsmoorwho,withgreatkindness,tookupthedoubleburdenofillustrationandindexing.Morethanbyanyotherman,wehavebeenguidedandinspiredbytheworkand thought of Burrhus F. Skinner,who led us into the field of behaviortheory, and whose influence has so perfused this field as to be lost to apropercontemporaryperspective.We are aware, too, of howmuchweowetoourownstudents–thosemenandwomen,workingalongsideus—whosevitalityandcreativityhavebeenarecurrentsourceofourownenthusiasm.

F.S.K.W.N.S.

Pour atteindre à la verité, il faut une fois dans saviesedéfairedetoutes lesopinionsquel'onareçueset reconstruire de nouveau, et dès le fondement,touslessystèmesdesesconnaissances.

RenéDescartes

CONTENTS

Editor'sintroductionPreface

CHAPTER

1.PSYCHOLOGYANDTHEREFLEXPreliminaryThesubjectmatterofpsychologyStimulusandresponseThereflexTheresponsemechanismThereflexarcandthereflexExtensionofthereflexconceptSomereflexpropertiesThethresholdorlimenLatencyStimulusintensityandresponsemagnitudeReflexstrengthAfewremarks

2.RESPONDENTCONDITIONINGPavlovandtheconditionedreflexTemporalfactorsinconditioningCompoundstimuliTheextensionofPavlovianresearchRespondentbehaviorHigher-orderconditioningPavlov'splaceinpsychology

3.OPERANTCONDITIONINGThorndikeandthelawofeffectSkinnerandoperantconditioningThelawofoperantconditioningOperantconditioningandthelawofeffectTherunwaytechniqueThemazetechniqueQuicklearningPositiveandnegativereinforcementOperant-respondentoverlapTheimportanceofoperantconditioning

4.EXTINCTIONANDRECONDITIONING

TheadaptabilityofbehaviorRespondentextinctionSpontaneousrecoveryOperantextinctionResistancetoextinctionasameasureofstrengthWhenisaresponseextinguished?SpontaneousrecoveryofaTypeRresponseExtinctionandforgettingRegression:aby-productofextinctionOne-trialextinctionPeriodicreconditioningP-RatfixedintervalsResponserateandP-RintervalExtinctionafterperiodicreconditioningFixed-ratioperiodicreconditioningThe'ceiling'offixed-ratiorespondingExtinctionfollowingfixed-ratioreinforcementTheeffectofaperiodicreinforcementSuperstition:anexperimentalexamplePunishmentandextinction

5.GENERALIZATIONANDDISCRIMINATIONStimulusgeneralizationStimulusdiscriminationFormingarespondentdiscriminationGeneralizationanddiscriminationinoperantconditioning:the

'discriminativestimulus'TheeffectofprevioustrainingAhumandiscriminationanalyzedAbolishingadiscriminationExtinguishingadiscriminativeoperantThemeaningof"similarity"GeneralizationgradientsThestudyofdiscriminativecapacityDiscriminationand"experimentalneurosis"DiscriminativereactiontimeReactiontimeandoperantlatencyLatencyandthediscriminativeprocessWorkadayreactiontimesMultiplediscriminations'Higherunits'inperceptionConceptformationGeneralizationanddiscriminationineducation

6.RESPONSEVARIABILITYANDDIFFERENTIATIONVariabilityversusstereotypyResponseinductionThewhyofvariabilityVariabilityandnegativereinforcementNegativereinforcementandthe"lawofleasteffort"MotivationandvariabilityExtinction,reconditioning,andvariabilityThedifferentiationofresponseResuméDifferentiationanddiscrimination

7.CHAININGTheprincipleofchainingChainsandnewunitsHomogeneousandheterogeneouschainsChaininganddelayedreinforcementExteroceptiveandproprioceptiveSD'sinchainingCovertresponseCovertresponseincodelearningCovertresponseandreactiontimeThecontexttheoryofmeaningTheassociationexperiment'Complexes'and'guilt'Mazesandnonsensesyllables

8.SECONDARYREINFORCEMENTApauseforreviewAnewfunctionofthestimulusSecondaryreinforcementofTypeSreactionsSecondaryreinforcementofTypeRreactionsSDandSrMoreaboutchainingThegeneralityofSr

SomeparametersofSr"Tokenrewards"and"sub-goals"SecondarynegativereinforcementSecondaryreinforcementand"feelings"SecondaryreinforcementinsocialbehaviorTakingstockandlookingahead

9.MOTIVATION

AnewlineofinquiryTheneedfortheconceptofmotivationThenatureofdrivesDriveasan"internalstate"DiscoveringdrivesThemeasurementofdrivesTwoclassesofdrivesSomemisconceptionsaboutdriveSomerepresentativedrives:ActivitydriveHungerdriveThesub-hungersSexdriveTheaversivedrivesAnxietyandavoidanceTheconflictofdrives

10.EMOTIONApreliminaryquestionSomeobservationsEmotionanddriveEmotionasaninferredstateThreetheoriesResponsepatternsandemotionSpecificemotionsEmotionalconditioningObservingourownemotionsEmotionandfeeling

11.SOCIALBEHAVIORIntroductionSocialbehaviorinanimalsSampleanalysesofsocialinteractionHumanbehaviorandtheculturalenvironmentPersonalityandhistoryConsistencyandintegrationofthepersonalitySocialmotives–arethereany?Verbalbehavior:IntroductionThenatureofabehavioraltheoryoflanguageTheemergenceofverbalbehaviorinchildhoodTwofunctionsoflanguage:themandandthetactThespeaker-hearerrelationDistortionsofthetactrelationOthercontrollingfactorsinspeech

FurthercharacteristicsofverbalbehaviorConsciousnessandunderstanding

ALASTWORDReferencesandauthorindex

1

PSYCHOLOGYANDTHEREFLEX

NATUREWITH its myriad phenomena assumes a unified aspect only inthe rarest cases; in the majority of instances it exhibits a thoroughlycomposite character...; it is accordingly one of the duties of science toconceivephenomenaasmadeupofsetsofpartialphenomena,andatfirsttostudythesepartialphenomenaintheirpurity.

P.Volkmann,ErkenntnistheoretischeGrundzügederNaturwissenschaft,1896

PreliminaryThepurposeof this text is threefold: (1) toacquaintyouwithanumber

of well-established psychological principles; (2) to show you how theseprinciples are related, one to the other; and (3) to suggest how you mayapply them in the analysis of everyday human activity. In addition, wehope that youwill draw from it somenotion ofwhere the frontiers of ourscience lie,ofwhat importantproblemsremain tobesolved,and,perhaps,anideaofhowyoucanhelpinfurtheringtheirsolution.

With certain matters of theory and practice commonly considered inelementary texts,weshallnotbeconcerned.Youwill find in thisbooknochapters on the workings of the human nervous system, the structure andfunctionofoursenseorgans,or themeasurementof intelligence.Theseareinteresting matters for advanced study, but they do not rightly belongwithin a text devoted to fundamental problems and basic principles ofpsychology.

For the general reader, there are excellent surveys and samplings of thevariousfieldsandproblemsofpsychology,whichprovideabird's-eyeviewof the science. For the advanced student, there are many books that dealexclusively with this or that area of specialized interest.The present textbelongs within neither category. It aims simply to furnish the beginningstudent with a solid foundation and a trustworthy framework for thinkingabout, or dealingwith, humannature as he finds it in his daily life.And,needless to say, it should also prepare him for further, more intensive,studyinourfield.

Asyoumakeyourway,stepbystep,throughthisbook,youwillnoticecertain outstanding characteristics of our approach. These may well bementionedinadvance,asapreparationfor thingstocome.Ourapproachisbiological, experimental, and systematic.Biological, in that our basicprincipleswilloftenbedrawnfrom thestudyofanimalbehavior, andwillbefoundtoapplyatvariousevolutionarylevels;experimental, inthattheseprincipleswillbederived,notfromcasualobservationoruntestedopinion,

but from laboratory studies inwhich the important factors are isolated andvariedinsuchamanneras topermitscientific lawfulness tobediscovered;andsystematic, in that the interrelationofexperimental factswillbeoneofourmajorconcerns.

TheSubjectMatterofPsychologyTentativelywemaydefinepsychologyas the science of the behavior of

organisms. Such a simple statement, however, is both incomplete andmisleading. Psychologists are not, as a rule, equally interested in thebehavior ofall organisms. Their attention is usually focussed primarilyupon the human being and a few of his near-relations in the animalkingdom —for example, the rat, the cat, the dog, the monkey, and thechimpanzee. The comparative psychologist makes it his business toexaminedifferencesandsimilaritiesofbehaviorthroughouttheevolutionaryscale.We, however, shall have little interest in the activities of animalsmarkedly different from man, or in those infra-human activities whichthrownolightonhumanconduct.

Moreover, the psychologist studies behaviorin its relation toenvironment.Behavioralonewouldhardlyconstitutethesubjectmatterofascience. Imagine, for a moment, the senselessness of a motion-picturerecordofanorganism'sbehavior frombirth todeath,withevery indicationoftheworldinwhichitlivedcarefullyblottedout!Onlywhenwebegintorelate behavioral to environmental happenings does the possibility of ascientificpsychologymakeitsappearance.

StimulusandResponseBut we must go still further. Behavior and environment are unwieldy

terms,toobroadintheirmeaningtobeveryuseful.Assoonasweattempttodealwitheither,we findourselvesaskingWhat sort of behavior?Whataspectofenvironment?Thisisbutanotherwayofsayingthat,wheneverwetry to describe either the behavior or the environment of an organism, weare forced to break it down into parts.Analysis is essential to description,inourscienceaswellasothers.

Throughanalysis,psychologistshavearrivedat theconceptsofstimulusandresponse. A stimulus may be provisionally defined as "a part, orchange inapart,of theenvironment,"anda responsemaybedefinedas"apart,orchange inapart,ofbehavior."Weshall recognize,however, that astimulus cannot be defined independently of a response.An environmentalevent becomes a stimulus by virtue of the fact that it is followed by aresponse. Activities of our muscles and glands (the so-called bodilyeffectors) make up our responses.These terms provide the specificity wedesire and make a study of the environment-behavior relation feasible bygivingus things to observe and tomeasure.Stimuli and responses are thebasicunitsofourdescriptionsandprovidethestartingpointforascienceof

behavior.Wewouldnot go far astray in asserting thatmodernpsychologyitselfisessentiallyastimulus-responsepsychology.

TheReflexPhysiologists have supplied us with a very useful wordwith which to

designate any specific stimulus-response relationship.The word is reflex,and, in thechapters tofollow,weshallemployit frequently.Forexample,weshallspeakof thesalivaryreflexwhenwe refer to the fact that food, inthemouthofahungryman,quicklyevokesaflowofsaliva.Weshallevenextendtheuseofthetermtodenoteresponsesforwhichrelatedstimuliarenotclearlyobservable.Thus,weshallgive thename reflex to the responsemade by a child in clapping his hands, even when we have no way ofknowingexactlywhat stimuli are responsible for the act.Butmoreof thislater.

TheResponseMechanismBetween the stimulus and the response there are obviously bodily

happenings. It is well recognized today that stimuli affect sense organs(calledreceptors)andthatsenseorgans,whenexcited,causenerveimpulsestobetransmittedovernervepathwaystothebrainorspinalcordandthenceto the muscles and glands (effectors).This sequence of events results inthose responses which we, as psychologists, are interested in relating tostimuli.The structures involved—the receptors, the nervous system, andthe effectors— make up what is commonly known as theresponsemechanism.

Thedetailed functionof the responsemechanism isofprincipalconcerntothephysiologist.Hetellsusthatthereceptorsorsenseorgansareofthreemaintypes:

1.Exteroceptors, small structures within the eye, the ear, or the skin,whichareexcitedbystimuliexternaltotheorganism.

2 .Interoceptors, tiny organs which lie mainly within the alimentarytract,andwhichareexcitedbyinternalstimuli.

3 .Proprioceptors, which are located within the muscles,joints, andtendons, as well as the inner ear, and are excited, that is, stimulated, bymovementsofthebodyoritsparts.

In line with this distinction, stimuli themselves are often described asexteroceptive, interoceptive, or proprioceptive. This gives us greaterspecificitythanisconveyedmerelybythewordenvironment.

We are also told that the nervous systemhas subdivisions.One part ofit,thesomaticnervoussystem,servesprimarilytotransmitnerveimpulsesfromthesenseorgans to thosemusclesof thebodywhichareemployed inwalking,talking,writing,andthelike—responseswhicharehighlyspecificand of great importance inmaking changes in our environment.The otherpart, theautonomic nervous system, is essentially an extension of the

somatic nervous system, but serves mainly for the conduction of nerveimpulsestotheglandsandtheso-called"smooth"musclesofourbodies—forexample,thedigestiveglandsandthemusclesoftheblood-vesselwalls,the stomach, and the iris of the eye.The action of part of this autonomicsystem is extremely diffuse and ismost conspicuously responsible for thewidespread internal changes which occur in all of us when under strongemotion.

TheReflexArcandtheReflexAnatomists and physiologists have analyzed in detail the structure and

function of the response mechanism. Their interest in filling the gapbetween stimulus and response has led them to the conception of areflexarc, in order to distinguish it from the observed relation of stimulus toresponsewhichwehavecalledthereflex.Thesimplestconceivablechainofstructuresbetweenourtwoend-terms,stimulusandresponse,ispresumablyone that includes (1) a receptor element (cells or cell groups in the sense-organ tissues), (2) a sensory or afferent nerve element (nerve cells orneurons)forconductingimpulsestosomenervecenterineitherthebrain orthe spinal cord, (3) amotor or efferent nerve element for conduction fromnerve center to effector, and (4) the effector itself, amuscular or glandularelement(muscleorglandcells)thataccountsforthefinalresponse.

This chain or arc of structural elements—receptor cells, sensory andmotor nerve cells, and effector cells—has sometimes been called a reflex,and some theorists have maintained that it is the organic basis of thestimulus-responserelation.Weneednothereexaminetheevidenceforsuchan argument, which has often been disputed, but it is important that wemake a clear distinction between the reflex as arelation actually observedand the reflex as a hypotheticalmechanism. Confusion will be avoided ifweusereflexintheformercase,andreflexarcinthelatter.

The newcomer to psychologywill gain little from further discussion ofphysiologicalmatters.Itistruethatpsychologistshavelongbeeninterestedin relatingbehavior to receptor, effector, andnervous-system function; andphysiologicalpsychology istodayathrivingfieldofstudyfortheadvancedstudent.Butasfarasgeneralprinciplesofbehaviorareconcerned, thisareaofinvestigationhasnotasyetbeenveryhelpful.It is, infact, thecasethatour knowledge of the stimulus-response relation has more often precededthanfollowedourknowledgeofitsphysiologicalcounterpart.

ExtensionoftheReflexConceptThe reflex, as well as the reflex-arc, concept grew up in physiological

science. Muscular movement, originally attributed to supernatural forcesand later ascribed to a power residingwithin themuscles themselves,wasgraduallyrecognizedasduetotheactionofsuccessivelyexcitedpartsoftheresponse mechanism.Also, the importance of an initiating stimulus was

demonstrated, and itbecameclear thatmanyanimal reactionswereadirectand almost inevitable result of stimulus presentation.Observation of suchphenomena as the writhings of a decapitated lizard in response to mildlyinjurious stimulation of its skin led to the search for stimuli to otherreactionsofoperatedanimals,andsoondisclosedaconsiderablenumberofcomparable stimulus-response relations. The advocates of scientificdeterminismwere thusencouraged toextend their researches into the realmof normal, intact animals and, eventually, to man himself. It becameobvious that, regardless of the physiological activities involved, stimuliand responses were often associated in a definite and openly observablecause-effect sequence. Lists of animal and human reflexes were compiled,and psychologists, as well as physiologists, turned to the task of furtherexplorationandstudy.

By 1919, it was possible for JohnB.Watson, anAmerican pioneer inthe objective, natural-science approach to psychology, to suggest that theidentification,enumeration,andclassificationofstimulus-responserelationswas the principal concern of our science. In effect, he proposed acataloguingof reflexesasbasic to thepredictionandcontrolofhumanandinfra-humanbehavior.Inthefurtheranceofsuchaprogram,Watsonhimselfmade extensive observations on the 'unlearned' reactions of newly-bornhumaninfantstovarioussimpleformsofstimulation.

Watson's proposal is now deemed impracticable. Even if we assumedthat the entire reflex repertory of a given organism might some day bedetermined(whichisveryunlikelyinviewoftheeffectsoftrainingandtheextreme difficulty in identifying the stimuli for many common forms ofresponse), we would still be unable to formulate general principles of thesort thatwerequire.Weneedadynamic, rather thanastatic,pictureof thebehavior of organisms.To describe process, not to inventory elements, isourmajorconcern.Amoreproductiveapproachhasbeenone that takes forgrantedthebasicallyreflexcharacterofourbehavior,selectsafewexamplesforobservation,andexaminestheirdynamicpropertiesinsomedetail.Thisistheapproachthatwillbetakeninthepresenttext.

SomeReflexPropertiesWebeginouraccountwithafewfactsthataresosimpleandubiquitous

that theymight easily be overlooked.Yet they actually have the status offundamental principles.They are readily demonstrable in any experimentalsituation where the intensity of a stimulus can be controlled and themagnitude of a response can be measured, and they may also be crudelyobservedunderthenon-quantitativecircumstancesofeverydaylife.

Take,asanexample,the scratchreflexofaspinaldog in the laboratoryof a physiologist. In such an animal, a severed spinal cord permitsmovements of the hind legswhich are free from any influence stimulatingthe fore part of the body.When a touch stimulus is applied to the dog's

hind ribs, a rhythmic flexion and extensionof the right hind legwill takeplace.Thisresponseresemblesanormaldog'sreactiontothebiteofaflea,but since it is uncomplicated by the effect of any stimuli concurrentlyexciting the dog's eyes or ears, it has a somewhat artificial appearance.Nevertheless,itiseasilyelicitedandveryusefulinillustratingpropertiesofreflexactioningeneral.

TheThresholdorLimenOne of the first facts that may be demonstrated with such a reduced

animalisthatastimulusmustbeofacertainintensitybeforeitcanelicitaresponse. The intensity just sufficient to excite is called a liminal orthreshold intensity. Lesser intensities are called below-threshold or sub-liminal since, when applied singly, they never evoke a reaction.Correspondingly, greater intensities are called above-threshold orsupraliminal.

Alloftheelicitedresponsesoforganisms(eitheroperatedorintact)showthis dependence upon stimulus level, and great effort has been expendedthroughoutthepastcenturyindeterminingtheliminalintensitiesofvisual,tactual,auditory, and other stimuli. Indeed, an entire field of research,psychophysics,hasbeenorientedaboutsuchmeasurements.Workersinthisfieldhavedealtnotonlywithabsolute thresholds, thekinddefinedabove,but also withdifference thresholds, in which one studies the organism'scapacitytodetectdifferencesintheintensityofstimulithatareattheoutsetwellaboveabsolute-thresholdvalue.

We shall come back to thresholds later, inChapter 5.At present, it isenough to note that many common observations point to their existence.Whenweraiseawatchtooureartohearthetick,whenwefindthefirststarin the evening sky,orwhenweobserve thedimmingof theatre lights,weare dealing with the fact of absolute or relative intensity limens. Otherinstanceswillquicklyoccurtoanyonewhothinksaboutthematter.

LatencyA short interval of time elapses between the application of a stimulus

and thearousalofa response.This interval is called the latency orlatentperiod. In the scratch reflexof the spinaldog, the intervalmay range from140 ms. (milliseconds) to 500 ms. (half a second), depending upon theintensityof the stimulusemployed. In sucha reflex, strong stimuli reducethe latent period and weak ones lengthen it. Other reflexes may show adifferent range of latency values, but for any single reflex of this sortconditionsmaybesocontrolledas toproducea fairlyconstantvalue.Thisconstancyis importantsinceitgivesusameasureof theeffectofchangingtheconditions.

Although this relation is quite dependable in such cases as that of thespinal dog's scratch reflex, there is another type of stimulus-response

connectionwhere changes in stimulus intensity are not so clearly reflectedinalterationsof latency.When thedriverofacar responds toagreen lightbymaking themovementswhich set his car inmotion, the latency of theresponsedoesnotappear tobedirectly related to theintensityof the light.He starts up as quickly for a weak light as for a strong one.The processinvolved in such a case requires, as we shall see, its own explanatorytreatment.

We shall return to thematter of latency inChapter 5,whenwe discusssome experimental studies ofreaction time—a term thatmay already havesuggesteditselftoyou.Weshallthenseethatlatencyandreactiontimearenot unrelated, the latter being a variation of the former. For the present,however,thedistinctionshouldnotbeprematurelypressed.

StimulusIntensityandResponseMagnitudeIfthescratchreflexofourspinaldogiselicitedbyabarelysupra-liminal

stimulus,thelatency,asstatedabove,willberelativelylong.Thiswillnotbe the only effect observed.The flexion of the leg may be limited to asingle short excursion or, at most, to two or three. If, now, the stimulusintensityisraisedtoahighervalue,therewillensueanincreaseinboththeamplitude and number of scratching movements. Within limits, themagnitudeoftheelicitedresponseisdependentupon,thatis,isafunctionof,theintensityofthestimulus.

Asinthecaseoflatency,however,theresponsemagnitudeisnotalwaysseentobearsuchaneatrelationtothestimulusvariable.Inasnake-infestedcave,afaintlyheardrattlemayevokeasvigorousareactionasthereportofa revolver; in a tropical jungle, the hum of a mosquito may sometimeselicit as strong an emotional response as the roar of a near-by plane. Suchobservationspointtoothercontrollingfactors,andthelaboratorybearsthisout.Adifferent, but nonetheless lawful, relationwill occupy our attentionlater.

ReflexStrengthThere arepropertiesof reflex actionother than those just described, and

theytoomaybeobservedin thebehaviorofaspinalanimal.Weneednot,however, consider them in this text, sincewe are here concernedwithbareessentials, andwe already have enough facts to take the next step in ourthinking.

FIG.1.Schematictaperecordsofthreereflexproperties.Theheightofaspikeonthestimuluslineindicatesstimulusintensity;thatontheresponseline, responsemagnitude.The distance between the stimulus and responsespikes on the constant-speed tapegives the latency.The time line ismadeby an automatic marker which ticks off any desired unit of time. Thebroken threshold tapeshows that the stimulationswereverywidely spacedtopreventpossiblesummationofstimuluseffects.

Returning to the scratch reflex again, let us ask ourselves a question.Whathappenstothreshold,latency,andresponsemagnitudewhenweapplya stimulus not once but many times and in fairly quick succession?Amoment's thought may suggest the answer.The reflex undergoes fatigue.This shows itself on the observational level. The latency increases, themagnitude of successive 'scratchings' becomes less and less, and thestimulus intensity required for threshold excitation heightens appreciably.Moreover,thesechangestakeplacetogether.

Since other procedures or "experimental operations" than that ofrepeating stimulations will also produce concurrent changes in reflexproperties, itwill be useful to have a single term to denote such changes.Reflex strength has been suggested for this purpose and will be adoptedhere.Weshall, then,speakofa strong reflexwhenthere isarelatively lowthreshold of stimulation, a relatively short latency, and a relatively largemagnitudeofresponse.Conversely,aweakreflexwillbeconsideredasonewith a high threshold, a long latency, and a smallmagnitude of response.Frequently,we shall takeasour indicatorormeasureof strengtheither thelatencyor the responsemagnitudealone,withoutbothering toconsider therelatedvaluesofotherproperties.

Ithasalreadybeenmentioned,inpassing,thatthetermreflexwilloftenbeappliedinthisbooktoresponseswhicharenotobservablyrelatedtoanystimuli. In such cases,where concepts like latency and threshold can havenomeaning(sincetheydependuponastimulus-responserelation),weshallnevertheless speak ofreflex strength. Frequency of occurrence will thenbecomeaveryimportantmeasureofstrength.UntilwecometoChapter 3,however,thismatterneedsnofurtherdiscussion.

AFewRemarksSo much by way of preliminaries. This is a short chapter, but you

shouldnowbeabletosaysomethingabout(1)thegeneralaimofthistext;(2) the subjectmatter of psychology; (3) the concept of the reflex; (4) thenature of a few basic reflex properties; and (5) the meaning ofreflexstrength.All this isessential toyourunderstandingofwhatwillbe treatedin the pages to come. Each step in our expositionwill depend uponwhathasgonebefore.Unlessyouhaveunderstoodtheearlierfactsandprinciples,youwill almost surely have troublewith the later. For this reason it will

oftenbehelpfulforyoutogooveragainaparagraph,asection,orachapterpreviouslyread,inordertostrengthenyourfoundationforthetopicathandorthetopicstocome.

Itwill alsobewise, in reading the coming chapters, to divest yourself,as far as possible, from preconceptions concerning psychology. Oureveryday language is shot through with purportedly psychological termsand concepts.Most of these are lacking in scientific significance, becausethey are either poorly defined or unrelated to anything else, and theywillfind no place in our account. You are asked here, as you would be inphysics,chemistry,orbiology, to learnanewlanguage.Thepowerof thislanguagecanbeappreciatedonlybytheonewhoappliesitconsistentlyandrigorously, avoiding the contamination of ill-defined or long-discardedterms.

NOTESInthebackofthisbook,alongwiththesubjectindex,isanalphabetical

list of some of the more important books and articles to which we havereferred in preparing our chapters. We have adopted the practice ofmentioning, in the text, authors' names and the dates of their publications(e.g., Hilgard and Marquis, 1940).The complete reference is provided inour list—thus "HILGARD, E. R., and MARQUIS, D. G. (1940).Conditioning and learning. NewYork,Appleton-Century-Crofts, xi, 429pp."After each reference you will find, in italics, the page numbers onwhichthecontributioniscitedinthistext.

Theintroductiontopsychologyprovidedbythisbookissomethingofadeparturefromtheusualapproach,andyoumayatsometimebeinterestedincomparingitwithothertexts.ThreeverypopularelementaryaccountsarethosebyMunn(1946),Ruch(1948),andWoodworthandMarquis (1947).Two otherwell-known books, byGarrett (1941) and byCrafts, Schneirla,Robinson, and Gilbert (1938), are often used as collateral reading inbeginners'courses.Anyoneofthesewillgiveyouagoodideaoftherangeof topics towhichstudentsareusuallyexposed in their firstcourse,but innone of themwill you find a serious attempt to show how the topics areinterrelated or integratedwithin a unifiedwhole. Only a few authors havemadeaneffortsimilartoours(e.g.,Muenzinger,1942).

Thesystematicpositionportrayedinthepresenttextisbestdescribedasreinforcementtheory,whichis thedominantviewpointinmodernbehaviortheory.Ourownaccount leansheavieston theworkofB.F.Skinner (Thebehavior of organisms, 1938), but is not uninfluenced by otherexpositions–as found, for example, in the writings of E. L. Thorndike(1911,1932)andtherecentteachingsofC.L.Hull(1943).

The studentwho does notmind reading something of amore technicalstripe than we offer may go to Skinner's book. There, in the first twochapters, he will find further discussion of such topics as the data of

psychology,theneedforanalysis,theconceptofthereflex,thestructureofpsychological theory, and related matters. The viewpoint in Hull'sPrinciplesofbehavior(1943)providesastimulatingcontrast.

Our systematic position has not kept us from looking at facts in othertheoretical contexts.Gooddata aregooddata, regardless of theory, andwehavebeenhappy todrawupon the experimental findingsof our colleagueswheneveritseemedprofitable,regardlessof theviewpoint thatguidedtheirresearch.Therefore,ifyouareonewholikesto"gototheoriginal,"donotbesurprised ifyouoccasionally findanarticlewritten in termswithwhichyou are unfamiliar. If you can push aside the interpretations anddiscussions,andattendtotheexperimentalproceduresandresults,youwillsee why we thought the article worth consideration. Later on, if yourinterest in psychology continues, you will be better able to evaluate theexplanationsanddebatedpoints.

2

RESPONDENTCONDITIONING

TAKEAhungryhorse;expose theductof theparotidglandon thesideofthe jaw;nothingwillcomeout—thegland isat rest.Now,showthehorsesome hay, or better, make some movement indicating that his meal iscoming.Immediatelyasteadyflowofsalivawillstart...

ClaudeBernard.LaScienceExpérimentale,1878

PavlovandtheConditionedReflexIn the history of science, it often happens that the facts, principles, or

methodsofonefieldareputtoaccountinthedevelopmentofanother.Thisis especially true of a young science. In psychology, the boundaries ofwhich have only recently been established, borrowings from other fieldshave been numerous. Since 1879, whenWilhelmWundt set up the firstpsychological laboratory inLeipzig,Germany,our sciencehasoftendrawnuponhereldersister,physiology.Anexampleof thiswascited in Chapter1: the concept of the reflex actually emerged as a result of purelyphysiologicalinterestinthefunctioningoforganisms.

Wemustnowacknowledgeanotherdebt,againtophysiology,foroneofour keystone concepts.This is theprinciple of the conditioned reflex, firstclearly stated by the Russian physiologist, Ivan Petrovich Pavlov (1849-1936), as the outcome of investigations begun in Petrograd (nowLeningrad)duringtheclosingyearsofthelastcentury.

In1904,Pavlov received theNobelPrize inmedicine forhis studiesofthe digestive activity of dogs.During these studies, he noticed somethingthatsuggestedtohimanexperimentalsolutionofsomeof theproblemsofbrain function.Drawing about him a number of co-workers, he soonlaunched a large-scale program of research—a program that took up theremaining years of his life and won him grateful recognition frombiological scientists throughout the world. Psychologists, however, haveprofited more from this research through the light it shed uponbehaviorthanthroughthespeculationsPavlovadvancedconcerningbrainfunction.

Pavlov's basic observations were simple. If food or certain dilute acidsareputinthemouthofahungrydog,aflowofsalivafromtheappropriateglandswillsoonbegin.This is the salivaryreflex, long known to exist invarious animals, including man. But this is not all. Pavlov noted, likeothers before him, that the animalwould also salivatewhen food had notyet reached the mouth: foodseen or foodsmelled would elicit the sameresponse.Also, the dogwould salivatemerely upon the appearance of themanwhousuallybroughthisfood.

ForPavlov, theseobservations raised importantexperimentalquestions.

Howdid it happen that themere sight of the personwho fed the dogwasenough to evoke a salivary secretion? Surely, this was not an innate orinborn stimulus-response relation, typical of all dogs andasuneducable asthe scratch reflex of a spinal animal. On the contrary, it seemed obviousthat the effect of such a pre-food stimulus could be understood only intermsoftheindividualexperienceoftheorganism.Somehow,anoriginallyineffective stimulus for salivary response must have taken on newsignificanceforthisanimal;itmusthavecometosignalize theapproachoffood.Also, it seemed toprepare the animal for the food by starting thedigestiveprocess.

This led Pavlov to develop an experimental method for studying theacquisitionofnewstimulus-responseconnections. Inpractice, thismethodrequiresnosmalldegreeoflaboratorycontrolandtechnicalskill,butitmaybe outlined rather simply. First, a normal dog is familiarized with theexperimental situation until he shows no disturbance whenplaced in aharness and left alone in a room especially designed to cut off unwantedoutsidestimuli.Asmallopeningorfistulaismadeinthedog'scheeknearthe duct of one of the salivary glands.When the fistula is healed, a glassfunneliscarefullycementedtotheoutsideofthecheeksothatitwilldrawoff the salivawhenever the gland is activated. From the funnel, the salivathen flows into a glass container or falls, drop by drop, upon a lightlybalancedrecordingplatform.Themagnitudeofresponsestovariousstimulican bemeasured by the total volume or the number of drops secreted in agiven unit of time.The experimenter,who sits in an adjoining room, canmake his measurements, apply what stimuli he desires (including food),andobservethedog'sbehaviorthroughawindow.

FIG. 2. Representation of a Pavlovian situation for conditioning thesalivaryresponseinadog.(AfterDashiell,1949.)

Wheneverything is ready, thedog is exposed,on successiveoccasions,to a pair of stimuli.One stimulus, say a small portion of powdered food,initially elicits a flowof saliva each time that it appears and the dog eats.The other, say a tone, has no such effect, but may cause some otherbehavior,perhapsatwitchingoftheearsoraturningoftheheadtowardthesource of sound. The combination of the two stimuli is presented atirregular intervals over a period of days, always at a timewhen the dog ishungry.Thepurpose, of course, is to determinewhether one stimulus (thetone)willacquirethepowerofelicitingthesameresponseastheother(thefood). So, after a certain number of pairings, the originally ineffectivestimulus(tone)ispresentedalone,toseeifitwillproducesalivation.

TableIgivesdatafromanexperimentbyAnrep(1920),oneofPavlov'spupils;anditshowswhathappenswhensuchanexperiment iscarriedout.In this study, a tone of 637.5 cycles per second was sounded for a five-second stimulation period; two or three seconds later the dog was givenbiscuitpowder.At intervalsof five to thirty-fiveminutes, thispairingwasrepeated. In sixteen days, fifty such combinations were presented and sixtests were made with tone alone. The test tone was of thirty seconds'

duration, andAnrepmeasured responsemagnitudeby thenumberofdropsof saliva that were secreted in this period. In addition, he recorded thelatenciesoftheresponse,inseconds.

TableIACQUISITIONOFACONDITIONEDSALIVARYREFLEX

(Anrep,1920)

Fromthistableyoucanseethat theamountofsalivationinresponsetotone-alone increased fromazerovalue, after a singlecombination, to sixtydrops in the test interval following the thirtieth combination.Alongwiththis increase in responsemagnitude, therewas a decrease in the latency ofthe response to tone, fromeighteen to two seconds.Littlechange occurredas a result of further pairings of tone with food, showing that the tone-salivationtie-upwaswellestablishedbythethirtiethcombination.

ExperimentslikethisledPavlovtoformulateanewprinciple:Ifanycasualstimulihappen,onceorafewtimes,toaccompanystimuli

whichelicitdefiniteinbornreflexes,theformerstimulibeginofthemselvestoproduce theeffectof these inbornreflexes....Wedesignate the twosortsof reflexes, and the stimuli by which they are elicited, asunconditioned(inborn) andconditioned (acquired) respectively. (Pavlov, 1923, translatedby£.B.HoltinAnimaldriveandthelearningprocess,1931,p.24).

A schematic picture or paradigm of Pavlovian "conditioning" may behelpfulatthispoint.

In thisparadigm, threereflexesare represented.Thefood-salivationand thetone-ear-twitching reflexesare"unconditioned"; the tone-salivation reflex is"conditioned." The lettersS andR refer, of course, to stimulus andresponse.The use of r ismerely to show that the ear-pricking response totone is of no great importance in this conditioning process, it may evendisappear during the repeated application of the tonal stimulus. Theimportant response, and the one that ismeasured, is the one belonging tothesalivaryreflex.

Early in their studies, Pavlov and his students found that this sort ofconditioningcouldoccuronlywhen the food-salivation reflexwasstrongerthan the reflex elicited by the "casual" stimulus. For example, an intenseelectric shock (rather than a tone, a light, or a touch)would not become aconditionedstimulusforsalivationbecauseitproducedaviolentemotionalupset in the animal.This ledPavlov to say that a conditioned reflexmustalways be based upon an unconditioned reflex thatwas "biologicallymoreimportant"or 'physiologically stronger." The stronger of the twounconditioned reflexes is the one that strengthens orreinforces the newstimulus-response relation. The stimulus of the stronger unconditionedreflexisoftencalledthe"reinforcingstimulus."

Pavlov's principle has been restated, by Skinner (1938), in a way thathighlightstheimportanceofthereinforcingstimulus,andpointsupthefactthat a new reflex is formed by combining elements of the two that werealreadypresentintheorganism'srepertory.

The approximately simultaneous presentation of two stimuli, one ofwhich (the "reinforcing" stimulus) belongs to a reflex existing at themomentatsomestrength,mayproduceanincreaseinthestrengthofathirdreflex composed of the response of the reinforcing reflex and the otherstimulus.(Skinner,Thebehavioroforganisms,1938.p.18.)

TemporalFactorsinConditioningIntheabovestatement,asinPavlov's,acloserelationofthetwostimuli

in time is specified.Onestimulus is to"accompany"orbe"approximatelysimultaneouswith"theother.Wearetemptedtoaskfurtherquestionsaboutthis relation. Does conditioning proceed more rapidly with simultaneousthan with successive stimulus presentations? If successive stimulation iseffective, which of the two stimuli should come first for best results? Isconditioningstillpossiblewhenconsiderabletimeelapsesbetweenthetwo?

Answers to these questions have been sought by several investigatorsand we now know that a strict simultaneity of the two stimuli isunnecessary for the rapid development of a conditioned reflex; and that aclose successionof stimuli, onebeingpresented twoor three seconds aftertheother, isprobablythemosteffectivearrangementofall.Weknow,too,that a conditioned reflex is set uponlywithverygreat difficulty, if at all,

when the conditioned stimulusfollows the unconditioned, even by afraction of a second. In terms of our tone-food example, the tone shouldprecede the food (as it did in Anrep's experiment) if the conditioningprocedureistotakeeffect.

As to how far in advance of the unconditioned stimulus the other onemay come, research does not yet give a final answer.What evidence wehavemakesitseemlikelythatalimitwouldsoonbereached.TwotypesofPavlovian procedure bear upon this problem. In one, the salivary responseis first conditioned to a sound or some other stimulus by the method of"simultaneous" presentation. Then, as the pairing continues, theunconditioned stimulus isnotprovideduntil the conditioned stimulushasbeen steadily present for a given period of time, say three minutes.Eventually, under such circumstances, a "delayed" conditioned reflex maybe established: the animal will respond with salivation only after theconditionedstimulushasbeenpresentfortwominutesormoreofthethree-minute interval. One is led to say that he can now "tell the time" withconsiderableaccuracy.

The second type of procedure is similar to the first, but with oneimportant difference: the conditioned stimulus isnot maintainedcontinuously during the interval of delay, but is presented onlyat thebeginning of the interval.As in the case of the delayed reflex, however,long-continued pairings of this sort will bring about a temporaldiscrimination: the dog will not salivate until the time for reinforcementapproaches.Pavlovcalled thisa"trace"conditionedreflex,arguing that theimmediate cause of salivationwas some trace of the conditioned stimulusthathadbeenleftinthenervoussystemoftheanimal.

Related to these two procedures, because of the time discriminationshown, are the following observations, also made in Pavlov's laboratory.(1)A dog was fed regularly at thirty-minute intervals.When this routinehad been well established, food was withheld at the usual feeding time.Salivationwasneverthelessnoted tooccur at approximately the endof thethirty-minute period—the timewhen the foodwould ordinarily have beenprovided.InPavlov'sterms,a"timereflex"wasformed.(2)Inanothercase,the same thirty-minute interval between feedings was used, but the foodwas always presented to the accompaniment of a metronome beat.Afterrepeatedpairingsofmetronomeandfood,salivationwasconditionedtothesound,asyouwouldexpect,butitwasalsodependentuponthetime-since-feeding. If the metronomewas sounded alone early in the period betweenfeedings, no salivation would occur; if it came slightly later, a smallmagnitude of response might be produced; and, as the end of the periodapproached, theeffectwouldbecorrespondinglygreater.Finally,with longtraining,salivation-to-metronomewaselicitableonlyat theveryendof thebetween-feeding interval; the response was conditioned, so to speak, tometronome-plus-thirty-minutes.

FIG. 3.Time relations in respondent conditioning. CS = conditionedstimulus;US=unconditionedstimulus.

These rather astonishing results tell us that the dog can make anextremely delicate time discrimination, but they do not bear upon thequestion of the maximal possible delay between the conditioned and theunconditioned stimulus. The experiments on the delayed and the traceconditioned reflexaremore to thepoint.The fact that a three-minute lapsebetween stimuli results in conditioning only aftermany pairings probablyindicates that the limit was almost reached. Certainly, under the usualconditions of Pavlovian experimentation, inwhich the pairings of stimulido not come at regular intervals, we would not expect to train a dog tosalivatetoatonethatprecededfoodbyhalfaday!

CompoundStimuliIn Pavlovian conditioning, a relation is established between a response

and some stimulus that accompanies the reinforcing stimulus.Why, then,youmayask,doestheresponselinkitselfexclusivelytothetone,thelight,orthetouchprovidedbytheexperimenter;aretherenototherstimuliinthesituation which regularly accompany the presentation of food?This is asimple enough question, but the answer is complex, having at least twomajor aspects. First, these"other" stimuli may be present not only whenreinforcement is given, but also under conditions ofnon-reinforcement, inwhich their eliciting power would be expected to dissipate itself (inaccordance with the principle ofextinction, to be treated inChapter 4).Secondly, a number of experiments fromPavlov's laboratory have pointedtothefactthatwhencertaincompoundsofstimuli,suchaslight-plus-soundor sound-plus-touch,are regularlypairedwith food, it ispossible thatonlyone member of the compound will become a conditioned stimulus. Forexample, Palladin conditioned salivation to a combination of touch-plus-temperature. Then he tested separately the elicitive function of the twocomponents.The tactual stimuluswas found to elicit as strong a responseas the compound, but the thermal stimulus was without the least effect(Pavlov,1927).

Such findings have opened up a brand-new field of research inconditioning, but one into which we need not enter here. We shall,however,returntotheprobleminanotherconnection,whenweconsiderthestimuluscontrolofanothertypeofconditionedresponse(Chapter8).

TheExtensionofPavlovianResearchPavlov and his collaborators studied many other aspects of salivary

conditioning than theoneswehavementioned.Someof this research isofinterest only to the specialist in this province, andwemay ignore it here.

Wecannot,however,leavethebasicprinciplewithoutsomegeneralremarksabout its extension, its significance, and the influence it has had uponpsychologicalthought.

Weknow today that the principlemay be demonstrated in the behaviorofmanymore animals than the dog.Hardly a species has been studied inwhichconditioningcannotbeestablished.Evenone-celledorganismsseemto display similar changeability. Special experimental conditions may beneeded.Thus rats, guinea pigs, and other small animals require apparatusand techniques that are clearly unsuitablefor humanbeings.But the broadgeneralityoftheprincipleisnottobequestioned.

Extension of the procedure has also involved the use of reflexes otherthan the salivary response to food (or acid) as the "biologically stronger,"reinforcingreflex.Since1916,anumberofinvestigators,mostlyAmerican,have shown that the constriction of the pupil of the human eye, whichresults naturally from stimulationwith strong light, can be conditioned tothesoundofabell,orsomeotherstimulus.Othershavedemonstratedthatchanges in the electrical resistance of the skin (through sweat secretion),elicited by such stimuli as a mild electric shock or a fairly loud buzzersound, may readily serve as a basis for new reflexes. Still others haveworked with such reinforcing reflexes as blood-vessel constriction inresponse to stimulationwith cold objects, changes in pulse beat resultingfromelectricshockorskininjuries,andsoon.

Many agents have been used as conditioned stimuli, within the sense-fieldsofsight,sound,touch,taste,andsmell—evenproprioceptivestimuli,arousedbymovementsofthelegs,hands,orarms,havebeenemployed.Inseveral experiments, responses have been conditioned towords, eitherspokenbytheexperimenterorbythesubject.Acoupleoftheseexperimentsmaybedescribedbrieflybecauseoftheirintrinsicinterestandtheirrelationtotheproblemof'controlling'bodilychanges.

Hudgins (1933) seems to have been the first to condition a response toself-initiatedverbalstimuli.Heused,ashisbasicunconditionedreflex,onewith which Cason (1922) had already worked in human experimentation:the constriction of the pupil of the eye in response to bright light. In arather complicated sequence of stimulus pairings and combinings, he wasable to condition this pupillary reflex to (1) the sound of a bell; (2) avigorous contraction of the hand and fingers; (3) the wordcontract, asspoken by the experimenter; (4)contract, when spoken by the subjecthimself; (5)contract, whenwhispered by the subject; and, finally, (6) thesubject's silent, or sub-vocal,contract.Which is to say that, through aconditioning procedure, the subject came to control his own pupillarycontraction—afeatthatisordinarilythoughtimpossibleforhumanbeings.

FIG.4.Anacquisitioncurveshowing themagnitudeof theconditionedgalvanic skin response after varying numbers of reinforcements. The CSwas a tone; theUS for this respondentwas an electric shock to thewrist.(AfterHull,1943,basedondataofHovland,1937d.)

Menzies (1937) conditioned changes in the skin temperature of humansubjects by a very ingenious technique. Unconditioned stimulation wasapplied by immersing a subject's hand in a beaker of ice-water, and themeasured response was the elicited change in temperature of the subject'sotherhand.(Ithadbeenknown,since1858,thatafallinthetemperatureofonehandisregularlyaccompaniedbyasimilarchangeinthetemperatureofthe other.)With this stimulation was paired, in various parts ofMenzies'experiment:(a) thesoundofanelectricbellorabuzzer;(b)avisual patternof illuminated crosses;(c) verbal stimuli—the meaningless wordprochaska, spokenaloudby theexperimenter and repeated inawhisperbythe subject, or merely whispered by the subject alone; and(d) theproprioceptivestimulationprovidedbyextensionofthearm,clenchingofafist, or holding the head in a thrown-back position. Conditioning was

effectivelyestablished,infromninetothirty-sixpairings,fortwelveofthefourteen subjects. (In one of the two 'failures,' conditioningwas doubtful;in the other it did not take place, presumably because the unconditionedstimulus itselfwasnotalwayseffective.) Itwassetup toverbal stimuliasreadily when the subject whispered the word to himself as when thewhisperingwascombinedwith theexperimenter's spokenword.Moreover,inthreesubjectswhohadbeenconditionedtorespondtothevisualpattern,the temperature change could be induced by asking them to "recall" or"think about" the stimulus! In short,Menzies showed convincingly that aconditionedthermalchangecouldbesetupeasilyinhissubjectswithallofthestimulithathetried,bothexteroceptiveandproprioceptive.

Suchexperimentsastheseraisesomeimportantquestionsconcerningthenature of "voluntary control," but this is not the place for theirconsideration.At thispoint, it isprobablyenough to say that theproblemwillnotbesolvedon thebasisofPavlovianconditioningalone,since thistype of conditioning fails to tell us how the controllingword(contract oranyother)itselfcomestobestrengthened.

Physiologists tellus thatall the reflexes thus farmentionedarealike inone important respect: they depend uponautonomic nervous systemfunction.They involve the actionof glands and smoothmuscles (e.g., thesecretionofsweatandthecontractionofblood-vessels).Sincetheactionofsucheffectorsisoftenassociatedwithstatesofemotion(in"fear,"thesalivadriesup, thesweatpoursout, theskincools, thepupilsof theeyesdilate,etc.), it will come as no surprise foryou to learn, in later chapters, thatthesestatesmaybeconditionedinPavlovianfashion.

Yet, a few reflexes do not require autonomic function and mayapparently be reinforcing. Foot-withdrawal at the onset of electric shock isneitherasmooth-musclenoraglandularresponse,butithasbeenusedasabasic unconditioned reflex by Russian and American workers in manyexperiments. Bechterev (1857-1927), one of Pavlov's contemporaries, wasthe first to work extensively with this response and, in this country,throughthestudiesofLiddellandothers,ithasbecomeacommonformofconditioningprocedure.

The experiments of Liddell and his collaborators, using sheep assubjects,areofparticular interest inshowingtherelativecomplexityof thebehavior involved in such "motor" conditioning situations. A commontechnique in their studies is one in which the sound of a metronome ispairedwith an electric shock to the animal's left foreleg.At first, only theshock will elicit a flexion of the leg muscles, but, after a few stimuluscombinations, thebeatingof themetronome is in itself sufficient toevokethe response. (Along with this effect there may also be measured aconditioned change in skin resistance and breathing rate.)This is, to allappearances,asimpleandstraightforwardcaseofPavlovianconditioning.Itmay be shown, however, that seemingly minor alterations in the

experimental procedure are enough to produce dramatic changes in thesubject'sbehavior.Inoneexperiment,whenshockregularlyfollowedafive-second series of metronome beats, an increase in the daily number ofstimulus pairings from ten to twenty resulted in an entirely unpredictedchange in the sheep's behavior. Formerly a steady and tractable animal, hesuddenlybegan toshowdistinctly"neurotic"symptoms.He resistedbeingled into the laboratory; once in the experimental harness, hebecamequiet,but only for as long as the experimenter remained in the room;when leftalone, and before any stimuli had been applied, he began tomake forelegmovementsas if inexpectationof theshock;the effect of the shock,whenactually given, was to quiet the animal for a minute or more but, as thetime for the next stimulation approached, the frequency of the legmovementsincreased.Inspiteofareinstatementof theearlierexperimentalconditions, this deviation from normal behavior became daily morepronounced, and was alleviated only by a long vacation in pasture(AndersonandLiddell,1935).

Observations of this sort, on other animals and over a period of years,have raised important questions concerning the origin, development, andcureofneuroticbehavior—questionsalreadyraisedbyPavlovinhisstudiesof the discriminative capacities of dogs (seeChapter 5), and questions towhichLiddellhimselfhasgivenmuchattention(Liddell,1938).Theyalsosuggest the presence, in an apparently "pure" Pavlovian set-up, of factorswhich have not as yet been fully identified. Certainly the results of suchexperiments as Lid-dell's are strikingly at oddswith those obtained in theconditioningofsalivaryandotherautonomicfunctions.

The latter suggestion is supported by several recent demonstrations ofpseudo-conditioning, inwhichmotor responsesnot unlike foot-withdrawalhave been employed, and in which "conditioning" occurred without anypairing of stimuli. For example,Reinwald (1938) has observed thatwhiterats, after jumping and running in response to a few electric shocks, willreact similarly to a tone that was initially without observable effect upontheir behavior. Had this effect resulted from a succession of tone-shockcombinations, it could easily have been mistaken for true conditioning.Certain strong unconditioned stimuli may, apparently, be so generallydisturbing as to render an organism sensitive to influences which, underothercircumstances,wouldnothavebeenfelt.

Indirect evidence of the complexity of supposedly simple extensions ofPavlov's technique to responses that involve action of the somatic, ratherthanautonomic,nervoussystemwillbepresentedinthefollowingchapter.You will see that it is possible, in some such instances, to point to theoperationof another basic principle of conditioning—one that clearlyapplies tomostofour everydayactions and isoften found in combinationwith the one now under discussion.You will also be able to understandwhyonetheorist(Skinner,1938)hassuggestedthatPavlovianconditioning

islimitedexclusivelytoautonomicresponses.

RespondentBehaviorWhatever the strengths or limitations of the Pavlovian principle, one

pointstandsoutclearly:thistypeofconditioningalwaysdependsupontheelicitation of response. Food elicits salivation; strong light elicits apupillary constriction; shock elicits foot-withdrawal; and so on. Theunconditioned stimulus is observable, and the basic reflex occurs with aregularity and automaticity comparable to the reaction of a spinal dog.Also, as with a spinal reflex, strengthmay bemeasured in terms of suchpropertiesaslatencyandresponsemagnitude.

The namerespondent has been given to stimulus-elicited behavior inordertocontrastitwithbehaviorforwhichnostimulicanbeidentified.Wehaveadoptedthistermandwilluseitinthesucceedingpagesofthisbook.By introducing it here,we justify the title of the present chapter and pavethe way for later discussion. Since all the reflexes thus far mentionedinvolve the action of identifiable eliciting stimuli,wemayuserespondentas the equivalent ofPavlovian conditioning, and we may speak ofarespondent when referring to a specific instance of such conditioned orunconditionedbehavior.

Higher-OrderConditioningIt was reported from Pavlov's laboratory, early in the twenties of this

century, that a conditioned reflex, once set up, might serve as theunconditioned-reflex basis of another; and a distinctionwasmade betweenprimary andsecondary, orhigher-order, conditioning. Frolov, one ofPavlov's coworkers, conditioned salivation to both the sound of a buzzerand the beat of a metronome.When these two first-order conditioningswere well established, he used them in building a second-order reflex—salivation in response to avisual stimulus, a black square. Great cautionhad tobeexercised inpresenting thestimuli:an intervalof fifteensecondshad to elapse between the black square and the sound of the 'reinforcing'metronomebeatornoconditioningwaspossible.Also,thesecondaryreflexneverbecameverystrong:thelatencywasgreatandtheresponsemagnitudewas small. But some effect was discernible, in spite of the fact that theblacksquarewasneverpaireddirectlywiththeoriginalfoodstimulus.

In another experiment, Foursikov used the foot-withdrawal response toelectricshockashisbasicreflexandwasabletoobtainresults thatpointedtothepossibilityofthird-orderconditioning.Thewithdrawalresponsewasfirstconditionedtoatactualstimulus,thentothesoundofbubblingwater,and,finally,toatoneof760cyclespersecond,witheachnewreflexbasedexclusively upon the preceding one. This is schematized in the threeparadigmsshownonthispage,whereRomannumeralsI,II,andIIIindicatethe successively conditioned reflexes.Again, however, the effect required

highly controlled experimental conditions, was rather unstable, and grewlessastheorderwenthigher.Also,prolongedattemptsbyFoursikovtosetupafourth-orderreflexwereentirelywithoutsuccess.

It is possible that the factsdonot clearlyprove the existenceofhigher-orderconditioning.Conceivably,thefindingsareduetootherfactorsinthesituationthanstimuluscombination.InFoursikov'sstudy,onemightpointto thesensitizingeffectofelectricshockand thesimilarityof thesoundoftone to that of bubbling water for at least some of the effects of theconditioning procedure. Be this as it may, the influence of higher-orderconditioning could hardly be expected to play much of a part in theeveryday behavior of organisms, where conditions are seldom wellcontrolled.

Pavlov'sPlaceinPsychologyThe principle of respondent conditioning, firmly established on an

experimental footing, had many repercussions in psychology. It appealedespecially to the objectivists in the field as awelcome replacement for theolder, subjective "associationof ideas"—a legacy fromBritishphilosophy.Men like JohnB.Watson saw in the concept at least a partial explanationof the fact that many stimulus-response relations, not discoverable ininfancy,arepresentinadultlife.Ignoringtheproblemwhichthisraisedforanyone who sought to identify, in any adult,all the stimuli for hisresponses, they seized upon the principle to show that everyone's behaviorrepertory is the final product of countless stimulus substitutions.

Overwhelmedbythevisionofanatural-scienceexplanationofbehaviorthathadpreviouslybeenattributedto 'psychic'or 'mental'influence,theyforgotfor a time that they were at the beginning, rather than the end of theirlabors.

The apparent demonstration of higher-order conditioning gave addedimpetus to thismovement.Overlooking thedifficulties involved in suchademonstration, they accepted the experimental findings with alacrity asevidence of the all-embracing power of Pavlov's formulation. If the merecombination of stimuli, even if remote from the one that was initiallyreinforcing, sufficed to setupnewstimulus-responseconnections, theverycitadel of subjectivity—the "highermental processes" of imagination andthought—mightsoonbestormed.

Pavlovhimself, althoughnot unaware of the behavioral implications ofhis work, was more interested in the light he thought it shed upon thefunctionsof thebrain.Conditioning, forhim,dependedupon the rigorouscontrolofexperimentalvariables—timeofstimuluspresentation,numberofreinforcements,strengthofthebasicreflex,andotherfactors—allofwhichweretobestudiedindetailbylaboratorymethods.Whereverhelooked,hesawproblems, the analysis of which required research, andmore research.On the other hand, his most ardent psychological admirers saw onlysolutions, answers to age-old questions. When these early enthusiastsrecognizedanyscientificproblem, itwasmerely theoldoneof identifyingthe stimulus components of every environmental situation and describingthe responses associated therewith.And such a problem does not readilygivewaytoexperimentalattack.

Nowadaysweviewthematterinadifferentway.Modernpsychologists,althoughlessinterestedinthephysiologicalimplicationsoftheirstudiesinthis field, tend to lean in Pavlov's direction. That is, they havewholeheartedlyadoptedhisexperimentalattitudeandingeneralarewaryofextending the principle into territory not already cleared by laboratoryresearch.Gradually,theyhavetaughtustoseethelimitationsaswellasthestrengthofPavlov'swork.

Respondent conditioning is now a well-accepted principle of behavior.Pavlovwoulddeserve aplace in thehistoryofpsychology, if fornootherreason. Fortunately for us, his work did not stop at this point.When weconsider, in the chapters to come, such concepts as those of "extinction,""generalization,"and"discrimination,"weshallagainhaveoccasion topayhomage to this Russian genius.He did not give us a complete system ofbehavior. In fact, we shall see that other, non-Pavlovian, principles haveactuallybecomemore important in thedevelopmentofsuchasystem.Buthecarriedusagreatstepforward in thepathweweredestined to followinthe scientific study of animal and human conduct. In retrospect, it isinteresting to consider that aphysiologist should have been theman to dosomuch in promoting our enterprise.We are in no position toweigh his

contributions within his chosen field; we can say very little about thedegreetowhichheclearedupthemysteryofbrainaction;buthisworkwillstandformanygenerationsasalandmarkintheanalysisofbehavior.

NOTESThebestsinglesource-bookforEnglish-speakingstudentsofrespondent

conditioning is Pavlov'sConditioned reflexes: an investigation of thephysiologicalactivityof thecerebralcortex,a translationfromtheRussianby G.V.Anrep, published in London by the Oxford University Press in1927. Another book, a collection of Pavlov's Lectures on ConditionedReflexes, translated by W. H. Gantt, was published in New York byInternational Publishers in 1928. Besides the lectures themselves, thisvolumeincludesashortbiographyofPavlov(byGantt)andabibliographyof nearly two hundred different articles which emanated from Pavlov'slaboratorybetween1903and1926.Before1927,however,psychologistsinthis country had a very incomplete picture of Pavlov's work, and hisinfluencewasfeltonlygradually.

Ithasbeenpointedoutrecently,byHilgardandMarquis(1940),thatthework of Vladimir M. Bechterev actually aroused more interest in thiscountry than did that of Pavlov himself. Bechterev, asmentioned in yourchapter, dealt with responses like foot-withdrawal to electric shock. Sincehis publications were often in German, and occasionally in French, theyweremoreaccessibletoAmericanreaders.HavingbeenastudentatLeipzigunder Wundt, he retained a strong interest in psychology and anacquaintance with its problems; and he used human as well as animalsubjects in his experiments. His most important book,Objectivepsychology, was translated into French and German in 1913, and histeachings are now available in English(General principles of humanreflexology,1932).Since1927,however,withtheappearanceinEnglishofPavlov's monumental work. Bechterev's influence has graduallydisappeared. Modifications of his basic technique are still used in manylaboratories in theUnitedStates (atYale,Cornell,Rochester, and Indiana,to mention but a few), but researchers have quite generally adopted theterminologyandsystematicconceptsofPavlov.

Among the names associated with the early development of interest inPavlov (and Bechterev) in this country are those of R.M.Yerkes and S.Morgulis (1909), J. B. Watson (1916), K. S. Lashley (1916), W. H.Burnham(1917),F.Mateer(1918),S.SmithandE.R.Guthrie(1921),H.Cason(1922),andF.H.Allport (1924).Youwill findanexcellent reviewofthisdevelopment,togetherwithasummaryofsomeimportantPavlovianconcepts, in the first two chapters ofConditioning and learning (HilgardandMarquis,1940).

Agoodreviewofearlystudiesinstimuluscompounding(inRussiaandelsewhere) is available in apaperbyG.H.S.Razran (1939c).This is one

ofmanyreviewsbyRazran,whosescholarlyinterestsandacquaintancewiththeRussianlanguagehavepermittedhimtorenderinvaluableservicetohisAmerican colleagues. Except for the translations of Pavlov's books, ourprincipal contact with Russian research has been made through Razran'sefforts.

3

OPERANTCONDITIONING

SUPPOSE,now, that in puttingout its head to seizeprey scarcelywithinreach,acreaturehasrepeatedlyfailed.Supposethatalongwiththegroupofmotor actions approximately adapted to seize prey at this distance,... aslightforwardmovementofthebody[iscausedonsomeoccasion].Successwill occur instead of failure.... On recurrence of the circumstances, thesemuscular movements that were followed by success are likely to berepeated:whatwas at first an accidental combination ofmotionswill nowbeacombinationhavingconsiderableprobability.

HerbertSpencer,PrinciplesofPsychology,1855

ThorndikeandtheLawofEffectIn 1898, five years before the termconditionedreflex appeared in print,

animportantpsychologicalmonographwaspublishedintheUnitedStates.Its title was "Animal Intelligence: An Experimental Study of theAssociative Processes in Animals," and it was written as a doctoraldissertationbyEdwardLeeThorndike(1874-1949)atColumbiaUniversity.This study was important for two main reasons: (1) it introduced anexperimental method of investigating the problem-solving behavior ofanimals; and (2) it attempted to account for such behavior in terms ofassociations (stimulus-response connections) that were strengthened bytheirresults.

InThorndike'sexperiments,cats,dogs,andchicksweresubjects,but,inthepresentaccount,weshalllimitourselvestohisstudieswithcats,whichmay be taken as typical of his work.With these animals, fifteen differentforms of "problem box" were used as apparatus, representing as manydifferent problems for solution. Most of the boxes were crate-like affairs,about twenty inches long, fifteen incheswide, and twelve inches high.Atthe topofeachwasa trap-door throughwhichacatmightbedropped intothebox,andononesidewasadoorthroughwhichhemightescapeandgettoasmallbitoffood(meatorfish)ontheflooroutside.Thedoorwasheldin a closed position by a bolt or a bar, but could be opened fromwithinwhensomerelease-mechanism—alatch,awireloop,alever,orsomeothersimpledevice—wasproperlymanipulatedbythecat.

Theexperimentalprocedureforanygiventaskwasroughlyasfollows.Ahungry catwas dropped into the box and left there until, in the course ofhisactivity,hehappened tooperate theappropriate release-mechanism—forexample, until he pulled a loop or depressed a lever that opened the door.As soon as he left the box and ate the morsel of food that awaited himoutside,hewas takenupby theexperimenter andputback in thebox, the

doorofwhichhadagainbeenlocked.Afterasecondescapeandfeeding,theprocedurewasrepeated;andsoon.

For each problem,Thorndike noted the time required by the animal toescape from the box on each successive trial.Figure 5 shows, graphically,the number of seconds needed by one cat on each of twenty-four trials tomakealoop-pullingescaperesponse.

This curve,which is fairly representative of results obtainedwith othercats and other problems, helps us to understand what took place in thesestudies.First,we see anoverall reduction in the timeper trial required fortheanimaltogetoutofthebox.Thenumberofsecondsneededforthefirstescapewas160; for the twenty-fourth, itwasonly seven.The amount andthe rapidity of the dropwas greater for some problems and some animalsthanitwasforothers;andtherewasawidevariationinthenumberoftrialsrequiredbefore theescape-timebecameminimal.Secondly, it appears that,in spite of the general decrease of time as the number of trials increased,there was considerable irregularity. Setbacks in the animal's progress werefairly common.Thus escape from the box on the second trial required butthirtyseconds,whereas,ontrial three,whichfollowedimmediately,ninetyseconds were needed. In some of Thorndike's other experiments, theseirregularities were even more pronounced and continued for many trialsbeforeaconsistent,quicksolutiondeveloped.

FIG. 5.Time taken to escape by one ofThorndike's cats on successivetrialsinaproblembox.(AfterThorndike,1898.)

While gathering the time records,Thorndike did not fail to observe thebehavior of his subjects. He noted, for example, that a cat, when firstconfrontedwithhisproblem,commonlymadevigorousattempts toescapeconfinement."It tries tosqueezethroughanyopening; itclawsandbitesatthe bars orwire; it thrusts its paws out through any opening and claws ateverything it reaches;... For eight or tenminutes itwill claw andbite andsqueezeincessantly."(Psychol.Monogr.,1898.)

In the course of such activity, nearly all the cats hit upon the responsethatopenedthedoorandgaveaccesstofood.Whenreturnedtotheboxforthe second test, the struggling recommenced, andcontinueduntil a secondsolution wasachieved. Gradually (in some cases, rather suddenly), withsuccessive confinements and escapes, the amount of useless activitydiminished and the cat's behavior became clearly oriented toward therelease-mechanism. Finally, a well-directed and stereotyped mode ofresponsedeveloped:the"problem"wassolved.

According toThorndike, the solution of such a problem by cats andotheranimalsinvolvedtheformationofanassociationbetweensomeaspectof the stimulus-situation, such as thewire loop or thewooden lever,withthe specific movement that led to door-opening. Further, he argued, thestimulus-response relation that finally appeared was obviously influencedby theoutcomeof thismovement.Thepleasureexperiencedby theanimalingettingoutoftheboxandtothefoodservedtostampintheconnectionbetweenstimulusandresponsethatledtothepleasure.Bythesametoken,stimulus-responseconnections thatdidnot lead to apleasurable after-effectwerenotstrengthened,andtendedtodropout.

Here was the first approximation to a basic principle of behavior.Thirteenyearslater,whenThorndikerepublishedhismonographaspartofabook onAnimalintelligence (1911), the same ideawas presented formallyastheLawofEffect:

Of several responses made to the same situation, those which areaccompanied or closely followed by satisfaction to the animal will, otherthings being equal, be more firmly connected with the situation, so that,when it recurs, they will be more likely to recur; those which areaccompanied or closely followed by discomfort to the animal will, otherthingsbeingequal, have their connectionswith the situationweakened, sothat, when it recurs, they will be less likely to occur. The greater thesatisfactionordiscomfort,thegreaterthestrengtheningorweakeningofthebond. (Thorndike, E. L.,Animal intelligence: experimental studies. NewYork,Macmillan,1911,p.244.)

By this time, Thorndike had joined the staff of Teachers College atColumbia University and had embarked upon a long career of productiveresearch in psychology and education.Aman of strong practical interests,powerful motivation, and striking originality, his yearly output of workwas enormous in bulk and varied in content. From his early experimentswith animals, he was led to problems of human learning, classroomprocedures, dictionary construction, intelligence testing, and vocationalguidance. Wherever he turned, he never lost sight of his fundamentalprinciple and, from time to time, he brought forth specific evidence of itsoperation.

The law of effectwas notThorndike's only answer to theway learningtakesplace inanimals andhumanbeings.Early inhis thinking (1898),herecognized alaw of exercise according to which connections arestrengthened through mere repetition and weakened through disuse. By1932, however, he was led by his own research to renounce his formerpositionand toargueagainst exerciseasa factorworking independentlyofeffect.Also, in 1913, he proposed a law of readiness, but this was littlemore than a guess, in termsof "conductionunits," as to thephysiologicalconditionsunderlyingtheoperationofhisbasicprinciple,andneverplayedavery important role inhis research.Somewhatmore interestingandmoreempirically grounded were five subsidiary laws which were intended tosupplement theprimaryones.Oneof these, that ofassociative shifting, istheThorndikiancounterpartofPavlov'sprinciple,andanother (response byanalogy) bears some resemblance to the principle ofgeneralization withwhich we shall deal inChapter 5. In the main, however, he left thesesubordinate laws in a relatively undeveloped state and we need not givethemelaboratetreatmenthere.

The law of effect was, then, Thorndike's major contribution, and thebeginnerinpsychologygenerallyacceptshisformulationwithoutquestion."Trial and error,with accidental success" seems to describe satisfactorily agreatdealofproblem-solvingbehaviorasheknows it inhiseveryday life.He is usually quite ready to believe that many of his own reactions tostimulussituationsarefirmlyimplantedbecauseof theireffects. In fact,hemay tell you that, long before he heard ofThorndike, he had assumed theoperationofsomesuchgeneralprincipleandcouldhavetoldyouaboutitifyouhadaskedhim.Howelse,hemaysay,doesonefindhiswayabout inastrangecity,'learntheropes'ofanewoccupation,solveaChinesepuzzle,ormasteranycomplicatedskill,exceptthroughtheeffectofthesuccessthatattendshistrial-and-errorbehavior?

Youmay, therefore, be surprised to hear thatThorndike's principlewaschallenged by psychologists in many quarters. How can it be, said onegroup of critics, that "pleasure" or "satisfaction," which aremental states,exerciseaninfluenceuponsuchclearlyphysicalphenomenaas responses tostimuli?How,saidothers,cantheresultsofanactionhaveanyeffectupon

theactionitselfwhentheactionisoveranddonewithbeforetheresultsareapparent—what sort of cause-effect relation is this, inwhich the effect hastofunctionasthecause?Stillothers,lessconcernedwiththephilosophyorlogic ofThorndike's position, argued that his formulation was lacking ingenerality and not always supported by fact. Observations were broughtforward toshow(1) that trial-and-errorwas typicalonlyofaveryrestrictedform of problem solving—one in which "insight" into the situation wasprohibited by the very conditions of the experiment; (2) that, even withThorndike's cats, the solution of a problem was not always "hit-or-miss"andgraduallyachieved,butwas, at least in somecases,practically reachedontheveryfirst trial;and(3)thatlearningispossiblewithoutany effect—aswhenrats,givenachance to runamazeprior to the introductionofanyfood reward, showedby their speedof learning in later, rewarded runs thattheyhadprofitedfromtheir'unrewarded'explorations.

In view of later developments, such arguments, and a great deal of theresearchthattheyfostered,weresomewhatbesidethepoint.Theyservedtoobscure, rather than clarify, a fundamental principle; to delay, rather thanhasten, an important line of investigation. A full appreciation ofThorndike's contribution did not come until thirty-odd years after his firstmonograph was published, when the principle was re-affirmed and placedclearlywithinalargerbodyoftheory.

SkinnerandOperantConditioningIn 1930, there was published, in the Proceedings of the National

AcademyofScience,ashortpaperontheeatingbehaviorofwhiterats.TheauthorwasB.F.Skinner,thenagraduatestudentinpsychologyatHarvardUniversity.Skinnerdescribed,inhispaper,anexperimentalmethodthat,inslightly modified form, has become a fixture in modern experimentalresearch.Itinvolvedsimply(1)adeviceforgivingasmallpelletoffoodtoahungrywhiterateachtimethattheanimalpushedopentheswingingdoorof a food-bin at one end of his experimental chamber; and (2) a recordingmechanismthatcausedaverticalmovementofapenuponthepaper-coveredsurfaceofaslowlyrevolvingcylinder(akymographdrum)whenevertheratopenedthedoorofthebintoobtainthepellet.Theverticalpenmarksweremadeat rightangles to thedrummovementandwerecumulative—that is,eachupwarddistancethatthepenmovedwasadded,byaratchetdevice,totheprecedingone.Since a shortperiodof timewas required for the eatingof eachpellet before thenext door-opening tookplace, and since thedrumcontinued to revolve steadily during this period, eachmarkwas displacedslightlytotherightoftheprecedingone.Thisprovidedastep-wiserecord,of the sort shown inFigure6.The vertical lines (of equal length) indicatesuccessivedoor-opening responses and thehorizontal lines (not necessarilyequal in length) indicate the time elapsing between responses. Since thepelletswereofaconstantsizeandweight(about1/15ofagram),aneating

rateisrepresented.The step-wise effect is very obvious inFigure 6 because of the size of

the time-units and response-units we have selected. If the units had beensmaller, the effect would have been less pronounced.Figure 7 is copiedfrom an actual record obtained in Skinner's experiment. In this case, thesteps are so close together that they are imperceptible in our copy and afairlysmoothcurveofeatingresults.

FIG. 6. Illustrating the construction of a cumulative record of bar-pressingresponses.

FIG. 7. A cumulative record of a rat's food-getting responses (bar-pressingforpellets)duringadailyeatingperiod.(AfterSkinner,1938.)

Such a curve is characteristically obtained when a rat is fed daily withpellets at a regular time, and is deprived of food during the interveningperiods. It shows that, under such a regimen, the animal begins byresponding at a relatively high rate and gradually slows down as he nearsthe point of satiation. The curve is, to use the mathematician's term,negativelyaccelerated.

Thecurvedoesnot, of course, representall eatingbehavior in rats.Hadtheanimalsbeenpermittedtoliveintheexperimentalboxandeatwhenever

their hunger led them to do so, they would probably have eaten at asomewhat slower but very constant rate on each occasion:straight-line,rather than negatively accelerated, curves would have resulted. Theimportant fact is theorderliness of the behavior that appears under aspecifiedsetofexperimentalconditions.Thesearchfororderlyrelationshipsis characteristic of all science and,with thediscoveryof a single example,oneisoftenledtohuntformore.

The only behavior required of the rats in Skinner's experiment was thesimpleandfairlynaturalactofpushingopenadoor to reachfood.Howorwhen the animallearned to perform this act was not determined, but asecond experiment, reported by Skinner in 1932, dealtwith this question.Usingamodificationofhisearlierapparatus,hetriedtofindoutthewayinwhichanewact,onenotpreviouslyrelatedtofood-getting,mightcometobesorelated.Theactchosenwasthatofpressingdownward,withaforceofabouttengrams,asmalllever.Thislever,orbar,wassituatedatoneendofa response chamber in a sound-resistant and light-proof experimental box(seeFigure 8). Its downwardmovement caused the ejection of a pellet offood, fromamagazine in an adjoiningchamber, into a smallmetal cuportray.With every activation of the food-magazine by the bar depression, arecordwasmadeon akymographdrumoutside the experimental box.Therecordwascumulative,asinthestudyofeatingrate.

Theexperimentalprocedureinvolved(1)apreliminaryacclimatizationofthe hungry animal to the response chamber, with a supply of food in thetray, until hemoved and ate freely in the situation; (2) further sessions inwhichtheratwasaccustomedtoeatingpellets,whentheyweredischarged,oneatatime,fromthefood-magazinebytheexperimenter;and(3)traininginthebar-pressingresponse.Stage3,theimportantoneinthisexperiment,wasconductedasfollows.

FIG. 8. An early model of Skinner's bar-pressing apparatus. (FromSkinner,1938.)

After twenty-four hours' deprivation of food, the rat was placed in theresponse compartment of the box.The barwas present, the food-magazinewasfilledwithpellets,andawatersupplywasobtainablefromasmalltubenear the tray, but the tray itself was empty.When the rat approached thetray,ashehadlearnedtodoinstages1and2,andfoundnofood,hesoonengaged inexploratorybehaviorofonesortoranotherwithin the chamber.In tenor fifteenminutes,oftensooner, thisexploration led toadepressionof thebar,usuallyaccomplishedbyoneorbothof the rat's forepawsasheraised himself to sniff at the wall above the tray or at the slots throughwhichthebarenteredthechamber.Thebar-depressionwasaccompaniedbytheclickofthefood-magazine(towhichtheratwasaccustomedinstage2)andtheejectionofapelletintothetray.Atthesametime,theresponsewasrecordedonthekymographdrumoutsidethebox.

FIG.9.Some typical cumulative responsecurvesobtained fromhungryrats on the day of conditioning a bar-pressing response for the first time.Eachresponsewasreinforcedwithapelletoffood.Noticethatconditioningis commonly "instantaneous," and that the response rate is usually steady.(AfterSkinner,1938.)

A second bar-pressing response usually followed soon after, in somecases immediately after, the first pellet had been seized and eaten, and theanimal quickly developed a maximal rate of pressing and eating. Thesample records inFigure 9 illustrate this clearly.The two lower curves inthis figure show the immediate development of a maximal response rate;theuppercurvesshowaslightpositiveaccelerationat theirbeginning,withafewresponsesoccurringbeforeaconstanteatingrateappears.

It isclearfromthesecurves that theratsquickly learnedtopress thebarwhen food resulted from the act. Indeed, if wewere unaware of thepreliminarystagesoftheexperiment—theacclimatizingoftheanimaltotheapparatus and the training to approach the tray when the magazine wasoperated—we might conclude from some of the records that the rat hadalreadysolvedthebar-pressingproblemwhenitwasfirstpresented.Exceptin those animals that produced positively accelerated curves, the learningprocess was practically instantaneous.And in no case was there anythingcomparable to the gradual and irregular progress that typified the behaviorofThorndike'scats.

Skinner calledhis study an experiment "On theRate ofFormationof aConditioned Reflex." It was obvious to him, however, that the rat's

behavior could not adequately be described in terms of the conventionalPavlovianparadigm.Anumberofinsurmountablebarriersstooddirectlyinthe path of such an analysis.A practical exercisewill convince you of thedifficulties in applying Pavlov's principle to the bar-pressing situation.Construct foryourself theparadigm,after themodel inChapter 2, labelingeachS andR appropriately to showhow the reflex isestablished.Whatarethe two reflexes with which you begin? Where is the third reflex, theconditioned one? How does the 'conditioned' stimulus come to act as asubstitute for the 'unconditioned' in eliciting the response to the latter?Besure that you limit yourself toobservable, rather than purely hypothetical,stimuliandresponses.

As an outcome of his own struggle with this problem, Skinnerproposed, in 1935, and again in 1937, that we recognizetwo types ofconditioning:Type S andType R.Type S is no more than the classicalPavlovian conditioning, in which reinforcement is always related to thepresentation of astimulus—for example, food is given when a tone issounded.Type R, which is represented in the acquisition of bar-pressingbehavior, involvesa relationbetween reinforcementanda specificresponse—thus,foodisgivenwhenthebarispressed.

A paradigm suitable forType R conditioning is shown below, alongwiththefamiliarTypeSschema.Acomparisonofthetwowillhelpyoutounderstandsomeofthewaysinwhichthetwotypesdiffer.

Type S conditioning, as we have seen, involves the elicitation of aresponse (salivation) by an identifiable conditioned stimulus (tone) that isunder the experimenter's control. In Type R conditioning, the specificstimulus that initially evokes the response (bar-pressing)cannot beidentified.This is indicated by the small s of theType R paradigm.Weneed not assume that bar-pressing has no cause or that it can bear norelation to environmental stimuli;we shall see, inChapter 5, that stimulimay"settheoccasion"forthisresponse.But,forallpracticalpurposes,theresponsejustoccurs,isinitiallyemitted,withoutrelationtoanyspecifiablestimulusagency.

TypeSconditioninginvolvesstimulussubstitutionandtheformationof

anewreflex.Thetone,inourexample,comestoactasasubstituteforfoodin eliciting salivation; and tone-salivation is the new reflex. In Type Rconditioning, however, there is merely the strengthening of a reflex thatalreadyexistsintheorganism'srepertory.Bar-pressing,forinstance,occurswithsome frequency prior to any reinforcementwith food.At any rate, nosubstitutionisinvolvedandnonewstimulus-responserelationisformed.

Type S conditioningprepares theorganism for reinforcement.The tonecomes to elicit salivation in advance of the food, paving the way, as itwere, for its digestion. Type R conditioning procures orproduces thereinforcement—bar-pressing provides the rat with a food pellet.The twoprocessesmay take place concurrently, although our observation is usuallylimitedtooneofthem.PavlovhimselfnotedthatTypeSconditioningwascommonly accompanied by "motor reactions" of head-turning and the likewhichwewouldnowascribetothedevelopmentofaTypeRconditioning.More recently,another investigator (Brogden,1939b)hasshown thatwhendogs are reinforced with food for making a leg movement (Type Rconditioning) in the presence of a tone, there is the simultaneousdevelopmentofconditioned(TypeS)salivation.

With respect to this last distinction, a simple example may not besuperfluous.Ahungryboy,homefromschool, ismetat thedoorwith theodor of freshly baked cookies. In accordance with his history ofType Sconditioning, his 'mouthwaters,' preparing him forwhatmay follow.Butnothing will follow unless he has been conditioned inType R fashion tomakehiswaytothekitchenandexhibittheverbalorotherbehaviorwhich,inthepast,hasbeenproductiveofcookies.

TheLawofOperantConditioningThese differences between Type S and Type R conditioning are

associated with a broader distinction, mentioned briefly in the precedingchapter,between two fundamentalclassesofbehavior.Oneof these,whichincludesall thoseresponses,conditionedorunconditioned, thatareelicitedby known stimuli, we calledrespondent.The other class, comprising allthose responses that areemittedmore or less independently of identifiablestimuli, we may now calloperant. The spontaneous movements of aninfant organism, humanor otherwise, aremainlyof the latter type; and soare the "voluntary" acts of human beings. In fact,most of our behavior inthe routine affairs of everyday life is clearly operant, in that itoperates oracts upon the environment to produce the satisfaction of our basic needs.Respondentbehaviorismuchlesscommonlyobservedandseldom,itever,operatesupontheenvironmenttoproduceanything.

Operantbehaviorisconditionedprimarily,ifnotexclusively,inTypeRfashion; respondent behavior is usually conditioned in Type S fashion.Hence,whenitssuitsourconvenience,wemayspeakofTypeRasoperantconditioning, in the same way that we speak of Type S as respondent.

Again, just aswe refer to any single example of respondent behavior asarespondent, we shall refer to each example of operant behavior asanoperant.

We have seen, inChapter 2, that thestrength of a respondent iscommonlymeasuredintermsoflatencyandresponsemagnitude.Neitherofthese measures is satisfactory in determining the strength of an operant.Latency can have no meaning in the absence of an identifiable stimulusfrom which to measure theS-R interval; and the magnitude of an operantresponse does not change during conditioning in the orderly manner thattypifiestherespondent.Theamountofsalivasecretedbyadoginresponseto a tone may increase gradually with successive tone-food combinations,but the force of a bar-pressing response may be as great on its firstappearanceas it ison its fifty-first, and itmay fluctuate throughouta longseriesofemissions.

Our best measure of operant strength isfrequency of occurrence. Anoperant is strong when emitted often within a given period of time; it isweakwhenemittedrarely.Wehave,inasense,assumedthisalready,inthecase of bar-pressing: a steady, high rate of responding implied a strongresponse-tendency,whereasa slow,uneven rate impliedaweakone. In thecaseofarespondent,frequencyisauselessmeasure—infact,nomeasureatall—since the response rate is determined solely by the rate at which theelicitingstimulusispresentedtotheorganism.

"If the occurrence of an operant is followed by the presentation of areinforcing stimulus, the strength is increased." (Skinner, 1938.)We cannow begin to grasp the significance of this statement of the principle ofType R conditioning. Bar-pressing is an operant. It occurs with a certainlow frequencyprior to any experimental procedures thatwemayapply. Itsstrength is increased when it is followed by reinforcement. Increasedstrength means merely that it occurs with higher frequency than it didbefore.

TableIICOMPARISONBETWEENTYPESANDTYPERCONDITIONING

TypeS TypeRParadigm: Paradigm:

Responseiselicited. Responseisemitted.

Stimulussubstitution. Nosubstitutionofstimuli.Formationofnewreflex.

Strengtheningofreflexalreadyinrepertory.

"Preparation"byconditionedstimulusfortheunconditionedreinforcementthatfollows.Theresponsedoesnotmanipulatetheenvironment.

Response"procures"thereinforcement.Theresponse"operates"ontheenvironment.

Commonly,ifnotalways,ismediatedbytheautonomicnervoussystem,involvingsmoothmusclesandglands.

Mediatedbysomaticnervoussystem,involvingskeletalmuscles.

Usuallymeasuredintermsofreflexlatencyormagnitude.

Usuallymeasuredintermsofreflexrate;sometimes,latency.

OperantConditioningandtheLawofEffectWhen you compare the work of Skinner with that ofThorndike, you

may be impressed by the numerous dissimilarities. The two men useddifferent species of animals, different apparatus, and different experimentalprocedures. Yet, when you consider that both situations required amanipulation of some environmental object; when you note that thepresentation of food was in each case contingent upon this manipulation;andwhen you compare the principle ofTypeR conditioningwith the lawof effect, you may notice a striking agreement. Both formulationsemphasize the influence of the outcome of a response upon its strength:Thorndike calls it "satisfaction" and Skinner speaks of a "reinforcingstimulus."Skinner's formula seems to be thenarrowerof the two, since itcontains no equivalent ofThorndike's "discomfort," yetThorndikehimselflatercametodiscounttheweakeningeffectofdiscomfort—apointtowhichweshallreturnlater.

Thereisstillanothersimilarity.InspiteofThorndike'semphasisuponaconnectionorbondbetweensituationandresponse,it isperfectlyclearthathedoesnot refer to thePavlovian typeofconnection.Hewouldhavebeenthe last to suggest that the loop-pulling or othermanipulative behavior ofhis cats was elicited in the sameway that food elicits salivation, a shockelicitsfoot-withdrawal,oracinderintheeyeelicitstears.

This last point requires some elaboration. Throughout much of thepresent chapter, we have underlined the fact that a large proportion of anorganism's behavior is emitted rather than elicited, and is conditioned inType R rather thanType S fashion.We may have led you to think thatresponses like loop-pulling and bar-pressing can have no relationwhateverto stimuli. If such was your impression, it should be corrected. Operantbehavior,howeverspontaneous in its initialoccurrence,verysoonbecomesassociatedwithstimuli.Thecatthathaslearnedtopull theloop,ortheratthathaslearnedtopressthebar,reactstostimuliorstimuluscombinations,evenifwecannotspecifythemcompletely.Intheabsenceofalooporbar,theanimalseldompawstheair.Butthesestimuliarenoteliciting:theyarenot related to their responses in the sameway that salivation is related tofood or foot-withdrawal is related to electric shock. It is just that theresponsesaremorelikelytooccurinthepresenceofsuchobjects.Touseanexpressionquotedearlier,thesestimulisettheoccasionforresponses.Lateron,whenweconsider thismatter inmoredetail,we shall refer to themasdiscriminativestimuli.

TheRunwayTechniqueIn the discovery and demonstration of basic principles, everything

depends upon the kind ofmethod thatwe employ.We seek, first, to findlawful relationships between known variables. This is the aim of allscientificendeavor.But,atthesametime,wehopethatourfindingscanberelated to each other, within an integrated whole, and that ourgeneralizationswillapply to lessrestrictedexperimentalsituations than theonefromwhichtheywereinitiallydrawn.Intheserespectsthebar-pressingtechnique has proved especially useful, as youwill see again and again inthepages tocome. It is,however,bynomeans theonlymethod towhichexperimental psychologists have appealed in their attempts to analyze thebehavioroforganisms.

A relatively simple means of studying operant behavior, and one inwhichType R conditioning is readily apparent, is the runwaymethod. Inthisprocedure,as recentlyemployedbyGrahamandGagné(1940),hungryrats are used as subjects. After several periods of acclimatization toexperimentalconditions, theanimal isplaced ina starting-box, the slidingdoor of which opens to an elevated wooden pathway, three feet long andthree-quarters of an inchwide.At the other end of the pathway is a food-box, identical in its dimensions with the starting-box. The rat's task is

merelythatofrunningfromthestarting-boxtothefood-boxwhenthedoorof the former is opened by the experimenter.Upon entering the food-box,thedoorofwhich is thenclosedbehindhim,he is reinforcedwithabitoffood.Whenthefoodiseaten,theboxesarecarefullyinterchangedand,aftera pause for equalizing the between-run intervals and baiting the new food-box,theprocedureisrepeated.

FIG.10.AversionoftheGraham-Gagnéapparatus.(AfterRaben,1949.)

Progress in this task ismeasured in termsof the timeelapsingbetweenthe opening of the starting-box door and the rat's passing of a point fourinches along the runway. This time interval was found by Graham and

Gagné to decrease on successive runs. For a group of 21 animals, theaverage (geometricalmean)valueobtainedon the first runwas71seconds.Thesecondtrialrequiredonly17secondsand,bythefifteenthrun,thelowvalueof2.8secondswasreached.ThesedataarerepresentedinFigure11.

It isobvious, in this study, thatoperantconditioningoccurred, throughthe reinforcement of the running response with food.Yet the situation isdifferent in certain important respects from the one used in bar-pressingstudies. For example, in the Graham-Gagne method, the experimenter, aswellasthesubject,determinesthefrequencyoftherunningresponse.Onlyat certain intervals is the response made possible.This is essentially thetrial-by-trial procedure employed by Thorndike in his problem-boxexperiments; andThorndike, too, used a time-measure of progress. It isinteresting to note that, although the curve ofFigure 11 is an averagedrecord of 21 rats, it resembles closely that ofFigure 5 (here), obtained byThorndikewithasinglecat.

Another point of difference between the runway and bar-pressingtechniques lies in the fact that the runway requiresconditioningof aseriesor chain of responses (seeChapter 7) that take more time and arepresumably greater in number than in the case of bar-pressing.Consequently, the reinforcement of the bar-pressing response is moreimmediate and direct than the reinforcement for leaving the starting-boxwhen the door is opened. It is probably for this reason that the wholerunwayperformanceislearnedmoregraduallythanisbar-pressing.

TheMazeTechniqueYouhaveprobablyrecognizedthat thebar-pressing technique isactually

a simplified form of the problem-box method, in which the trial-wiseprocedure is replaced by the more useful "free operant" arrangement—thatis, the animal determines his own response rate. Similarly, the runwaymethod may be considered as the final stage of another important line ofdevelopment in the investigation of operant behavior. In 1901, whenThorndikewasbusywithhisproblem-solvingstudiesatColumbia,W.S.Small,atClarkUniversity,wasexploringthetrial-and-errorbehaviorofratswith a device that was soon to enjoy tremendous vogue amongAmericanpsychologists. Even today, when the popularity of this device has greatlydecreased, the beginning student of psychology generally expects anyreferencetoratstobefollowedbytalkaboutmazes.

Interest in mazes or labyrinths is understandable. The early Greekequivalentof"labyrinth"wasapplied to intricateundergroundpassagesandnetworks of chambers, which never fail to stir the imagination. From thegreat Egyptian labyrinth, described by Herodotus as containing 3,000chambers, to the decorative gardenmazes of eighteenth-century France andEngland, and the amusement-parkmazes ofmodern times,men have beenintrigued by such devices. It is not strange that Small, in his search for a

FIG. 11.A plot of average startingtimes on successive trials by a groupof rats on the Graham-Gagnéapparatus. (From data provided by

taskthemasteryofwhichwouldbeslow and measurable for animalsthat spend most of their lives indark andwindingpassages, shouldhave hit upon this form ofapparatus.

ThemazethatSmallbuiltwasacrude, wire-mesh affair, with asawdust-covered floor. In design(seeFigure 12), it was modeledafterthefamoushedgemazeonthegrounds of Hampton Court PalaceinEngland.Likemostmazes usedsince, it comprised a series ofstraight-aways, turns, choice-points, and blind alleys, with areward at the end of the route.Progress can be shown by thereduction of running-time onsuccessive trials, or in the numberofblind-alleyentrances.

Small's own studies were notextensive, but suggested severallinesofresearch.Henotedthatratsdecreased their running-time anderrorswith continued practice; thatthey adopted a short-cut providedby the pattern (at the fourth choicepoint); that lessened hungerbrought greater variability ofbehavior; and that they seemed todepend less upon sight and smellthan upon touch or movement(proprioceptive) cues. All thesematters were subjected to laterstudy by other investigators, whoimproved and standardized thetraining procedure and developednew and more reliable apparatus(seeFigure 13). Maze units wereequalizedinlengthandincreasedinnumber; one-way doors wereintroduced to prevent retracing ofpaths; extra-maze distractions were

GrahamandGagné,1940.) eliminated; motivation was morerigorouslycontrolled;andsoon.Suchstudieswereundertakentodeterminetheinfluenceuponmaze-learningscoresofsuchfactorsasage,sex,previousmaze experience, sense-organ participation, distribution of practice, andbrain destruction. Maze performance was, thus, the dependent variable orindexreflectingtheeffectofwhateverindependentvariabletheexperimenterchose.

FIG.12.DesignoftheHamptonCourtmazeusedbySmall.TheanimalstartsatE,andfindsfoodatF.(FromSmall,1901.)

FIG. 14. A T-maze with a singlechoice point.The animal starts atE,andreinforcement is locatedatoneoftheend-boxes(G).

FIG. 13.A typical pattern of theWarner-Warden maze. (From C. J.Warden,T.N.Jenkins,andL.H.Warner,Comparativepsychology,Vol. I.Copyright1935byTheRonaldPressCompany.)

Along with this developmentcame the hope that an analysis of"learning"couldbemadewith thisuseful instrument. Unfortunately,this hope was not realized. Itgradually became clear that maze-learning was an exceedinglycomplicated affair, and that themaze itself was not the simpledevice that it had seemed to be.Even when attention was centeredupon the behavior of a rat at asingle choice point, the problemsinvolved were too great for thekindofdescriptionmostdesiredbythescientist.

We shall return to the problemof maze-learning in a laterchapter,after you have become acquaintedwith the concepts required for itsexplanation. For the present, wemay limit ourselves to thestatement that, for aclose analysisof behavior, the maze is usefulonly when reduced to its very

simplestform—thatis,whenitbecomesnomorethanarunwayorasingleT.

QuickLearning

It was noted, earlier in this chapter, that the bar-pressing response of awhiteratmaybeconditionedwithasinglereinforcement;andtwocasesofthis were pictured in the cumulative response curves ofFigure 9.This isnot an unusual finding. Suppose you were to use a modification of theSkinner apparatus (seeFigure 8) in which the animal's living-cage istransformed into a response chamber by themere insertionof a bar.Undersuch circumstances, with hungry animals, and with only the briefest ofacclimatizing and pre-training periods, "one-trial learning" is commonlyobserved.Seldomdoesconditioningfailtooccurwithinafewminutesafterthe bar is first presented. The rat, one might say, "gets the point"immediately,oftenreachingasteadyresponse-rateafterthedischargeofbutonepelletintohistray.

This kind of behavior has sometimes been likened to the sudden,"insightful"achievementsofanimalshigherthantheratinthephylogeneticscale.For example,WolfgangKöhler (1925)presentedchimpanzeeswith avariety of fruit-getting problems—stacking boxes, using sticks, takingdetours,andthe like.Henoticed,anddescribedvividly,manyinstances inwhich the apes, after a preliminary survey of the situation, attained theirobjectivesquicklyandconclusively, inalmost-humanfashion,withoutanydiscernible "trial and error." Such results are in sharp contrast with theslow, stepwise progress of rats in mazes and cats in problem-boxes, andKöhlerargued thathisownanimals,byvirtueof theexperimenter's choiceof problem situation,were permitted to show their true intelligence or useof"insight."

Todaywecansaythatthesuddenorone-trialmasteryofanyproblemisduetoone,orboth,oftwofactors:(1)thesimilarityof theproblemtoonethatwas solvedonanearlieroccasion,or (2) thesimplicity of theproblemitself.Köhler's findings are probably attributable, in large part, to the firstof thesefactors.He tooksamplingsofbehavior, like those taken inhumanintelligence tests; and the success of his animalsmay be ascribed to theirpast history of conditioning (see our treatment of similarity inChapter 5).Rapid acquisition of the bar-pressing response, on the other hand, istraceable primarily to the relatively uncomplicated nature of the task. Bar-pressing has considerable unconditioned strength to begin with: it occurswithmore than zero frequency prior to any reinforcement thatwe provide.Moreover, the specific movements involved are very few in number, andthis is a good reason for calling a problem "simple."We could easilycomplicatemattersinthebar-pressingsituation,saybydemandingagreaterforceofresponse,orbyplacingthefoodtrayatsomedistancefromthebar,sothattheanimalwouldhavetoleavethevicinityofthetraytoobtainthenext reinforcement. By requiring this additional activity, we couldundoubtedlylengthenthelearningtime,unlesstherathadhadexperienceingettingfoodundersimilarcircumstances.

In the case of human beings, the solution of problemsmay be speeded

up by a special set of conditions. It is too early for us to consider theseconditions here, butwemaynote that the possessionoflanguage is oftenof help in reducing the time required or the number of errorsmade in themastery of certain tasks. Thus, the person who 'verbalizes' his choice-responseswhile finding his way through a complicatedmazewill make aperformance record definitely superior to that of a person who does notemploy such aids (Warden, 1924). Apparently, verbal sequences can bememorizedbythehumanlearnerfasterthanthepurelymanualorlocomotorpattern.Theseverbalsequencesarisefromhismovementsandthencometodirect the chain in a discriminative fashion. In someinstances, the rate ofimprovement issodramaticas toobscure thefact thatessentially thesamebasicprinciplesareinvolvedinverbalasinnon-verbalbehavior.

PositiveandNegativeReinforcementThorndike, in his 1911 statement of the law of effect, spoke of the

strengthening effect of "satisfaction" upon the bond between situation andresponse.Today,avoidingcontroversyaboutthenatureof"satisfaction,"wewouldsaythatthefoodhegavetohiscatsforopeningaproblem-boxdoorwaspositively reinforcing. On the observational level, this would meanexactlywhatThorndikemeant—that the effect of the foodwas to increasethe frequency of the response that produced it.We know, too, that water,forathirstyanimal,wouldhavehadasimilareffect.Foodandwaterbelongtoaclassofpositivereinforcers.

This is not all that Thorndike said. He spoke also of the weakeningeffect of "discomfort"upon situation-response connections.Certain stimuli(electric shocks, loud sounds, strong lights, etc.) serve todecrease thefrequency of responses in the wake of which they follow. Nowadays, wecallthemnegativereinforcers,buttheyarenotbestdefinedintermsoftheirweakeningfunction.By1932,Thorndikehimselfarguedthat"rewards"and"punishments"arenotopposed toeachother in themanner impliedbyhisearlierformulation;andweshallofferevidence,inthenextchapter,toshowthattheweakeningeffectofnegativelyreinforcingstimuliisnotpermanent.

Another, and probably a better,way of handling thematter is to definepositive reinforcers as those stimuli which strengthen responses whenpresented(e.g.,foodstrengthensbar-pressingorloop-pullingbehavior),andnegative rein-forcers as those which strengthen when they areremoved.Experimentally, a number of responses have been conditioned in animalsentirely on the basis of escape from, or reduction of, certain stimulusconditions. Mowrer (1940)showed, for example, that a panel-pushingresponsecouldberapidlysetupinwhiteratswhenitwasreinforcedbytheremoval of electric-shock stimulation; and Keller (1942) obtained similarresults when he conditioned bar-pressing in rats by the simple device ofturning off a bright light for sixty seconds whenever the response wasemitted in itspresence.At theeveryday level, too,weoftensee the results

ofthiskindofstrengthening.Wemoveoutoftheheatofthenoondaysun;weclosethewindowthatshutsouttheroaroftraffic;wetakeofftheshoesthatpinchourfeet;andwekindlethefirethatwillwarmourhands.Ineachcase, we perform an act that has previously been strengthened because itproducedthecessationofa"noxious"or"annoying"stimulus.

Wehave, then, twowaysofdefiningnegative reinforcers: the firstis interms of the weakening effect they have when presented; the second is interms of the strengthening effect of their removal. The effect is uponoperant behavior; an operant is weakened in one case and strengthened inthe other. Yet, it should be noted that the same operant cannotsimultaneously undergo both changes. A strong light, applied brieflywhenever a bar-pressing response occurs,will depress the frequency of thepressing(Schoenfeld,1947);ifthesamelightiscontinuouslyapplieduntila bar-pressing response occurs, and is then immediately extinguished, theresponsewillbestrengthened.Buttheresponsecannotproduceandremoveastimulusatoneandthesametime.

Operant-RespondentOverlapCoincidental with their effect upon operant behavior, negatively

reinforcing stimuli may exercise another function: they may serve aseliciting stimuli forrespondent behavior.Two cases arise: (1) elicitationmayaccompany theweakening functionof a negative reinforcer, aswhen astrong electric shock inhibits bar-pressing and, at the same time, elicitsfoot-withdrawal,squealing,blood-pressurechanges,andsoforth;and(2) itmayaccompany thestrengthening function,aswhen a shock elicits a foot-withdrawal,which is thenstrengthenedoperant-wiseby theshock-removal.Thesecondcaseisrepresentedintheaccompanyingparadigm.

Another combination of elicitation and operant strengthening is seen inan experiment reported by Konorski and Miller (1937). Using an electricshock just intense enough to elicit a leg-raising response in a dog, theygave food to the animal after each elicitation.The scheme, then,wouldbethis:

Here the shock elicits flexion and the flexion 'produces' food. We arethereforepreparedforthefindingthat"afterafewreinforcementstheanimalstarts to raise its leg independentlyof electrical shock—as soonas it findsitself in the given experimental situation." (Konorski andMiller, 1937, p.266)Such cases of overlapmaybehard to understandunless youperceivethat, with the use of shock, one elicits responses of those very muscles

whichareusedinoperantbehavior.A question often occurs in connectionwith operant-respondent overlap:

since a skeletal-muscle or 'motor' respondent, such as foot-withdrawal toshock, may be conditioned in a Type R manner, cannot an autonomicrespondent,likesalivationorthegalvanicskinresponse,bestrengthenedbythe same procedure?Very little information exists on this point, but theanswer is probably No. One may apparently strengthen an operant whichwill in turn produce the condi tioned stimulus for a respondent, as in theHudgins and Menzies experiments (here-here), but we do not therebyliberate the respondent from its dependenceuponan eliciting stimulus.Asfor the reverse effect, it is not clear, at this time, that amotor respondentcanbeconditionedbyaTypeSprocedure.Thepossibilityof sensitization(here) or some kind of operant reinforcement, such as shock-removal, hasseldombeeneliminatedinexperimentsdesignedtotestthismatter.

Related to the latter point are some unpublished observations byReinwald (1941).A dog was first trained, by aType R procedure, to liequietlyonhisrightside,withhisleftlegsuspendedbyaharnessinsuchawayas topermit itsunimpededmovementwhenever thepatellar tendonofthe leg was tapped lightly with a small hammer.Then, in an attempt toconditionthispatellarreflexorkneejerk,tapstothetendonwerecombinedwith visual stimulation—a moderately intense light just in front of theanimal'seyes.TheprocedurethusconformstothePavlovianparadigm:

Onethousandcombinationsoftheconditionedandunconditionedstimuluswerepresentedtothedogoveraperiodofmanyweeks,withthefollowingresult:no evidence of Type S conditioning was obtained . Although thetendon tap regularlyelicited theknee jerk, the lightnever came to exercisetheleasteffectupontheresponse.Resultsofthissortshouldnotbehastilygeneralized, since the patellar-reflex findingsmay not be typical, but theydosuggestacriticalscrutinyoftheallegedcasesofrespondentconditioningwhensomaticor"motor"responsesareinvolved.

TheImportanceofOperantConditioningThe principle of operant conditioning may be seen everywhere in the

multifariousactivitiesofhumanbeingsfrombirthuntildeath.Alone,orincombination with the Pavlovian principle, it is involved in all thestrengtheningsofbehaviorwithwhichwe shallbeconcerned in thisbook.It ispresent inourmostdelicatediscriminationsandoursubtlestskills; in

our earliest crudehabits and thehighest refinements of creative thought. Itaccounts, in large part, for our abnormal 'fixations' as well as our normal'adjustments'; for our parades of power and our shows of weakness; forcooperation no less than competition. It may be seen in our friendlyrelationswith,andourwithdrawalsfrom,ourfellows;inourexpressionsofbigotryandtoleration;inourvirtuesaswellasourvices.

We do not expect you to accept this appraisal without question orreservation at this time.We have scarcely begun our analysis of behavior.Only a few experiments have been cited, most of them drawn from thelaboratoriesofanimalresearch.Otherprinciplesandotherfindingshaveyetto be considered, and these in turn must be related to the ones alreadytreated. Nevertheless, so basic and so far-reaching is the law of operantconditioning that, even now, you should be able to find, in your ownexperience,many illustrationsof its action.Later on, aswedealwith evermore complex problems, you will be more, rather than less, aware of itsexplanatorypower.

NOTESWe have ignored, in our discussion of Thorndike's work, his many

studies of connection-formation in human beings, where the mereannouncementofRightorWrongby theexperimenterwasusedasa rewardorpunishmentforsomeresponse,usuallyverbal,onthepartofthesubject.This omissionwill be understood after you have seen, inChapter 8, howwords and other stimuli, not initially reinforcing, may come to exercisesuchaneffectuponoperantbehavior.Also, inour treatmentof the runwayand themaze,nomentionhasbeenmadeof theway inwhichresults fromthese devices have exemplified, even clarified, the operation of otherprinciples than theone towhich thischapterisdevoted.This, too,willbecorrectedaswegoalong.

The way in which we have treated the concept of "insight" deserves aword of qualification. Köhler's fascinating reports of chimpanzee behavior(The mentality of apes, 1925) were once widely regarded as convincingdemonstrations of the inadequacy of conditioned reflex theories. HisUmweg ordetour method of experimentation, in which animals wererequired to use crude tools (sticks, strings, etc.) orroundabout approachesto their food-objectives,washailedby some as thebest of all instrumentsfor analyzing 'intelligent' behavior.Today, however, we see that Köhler'smethod leftmuch to be desired, since it involved no study of the relationbetween known variables, as presupposed in scientific inquiry. Hisobservations did little more thansuggest problems for further research.Conditioningprinciplesmaybeexpectedtoapplytothemaswellastoanyotherinstanceofeverydayproblem-solving.

The distinction between operant and respondent conditioning, althoughanticipated in the writings ofmany psychologists, was not sharply drawn

until 1935, in an important paper by Skinner. Since then, it has beenendorsed and discussed by a number of authorities in this area of research(Schlosberg, 1937; Hilgard, 1937; Razran, 1939a; Hilgard and Marquis,1940; andMowrer 1947).The terms classical andinstrumental, employedbyHilgardandMarquis,areratherwidelyusedequivalents,respectively,ofTypeSandTypeRconditioning.

Hilgard and Marquis have described four categories of instrumental(operant) conditioning. (1)Reward training, in which responses arestrengthened through the presentation of positively reinforcing stimuli; (2)escapetraining, inwhich they are strengthened through the termination orreduction of "noxious" (i.e., negatively reinforcing) stimuli; (3)avoidancetraining, which is accomplished when "the learned reaction prevents theappearance of a noxious stimulus"; and (4)secondary reward training, inwhich strengthening results frompresenting stimuliwhichhavepreviouslyaccompanied positive reinforcements.You will note that we have alreadyconsidered the first two of these categories;avoidance behavior andsecondaryrewardtrainingwillbetreatedinChapters8and9.

Some psychologists prefer not to use the termconditioning inconnection with the strengthening of operant responses, bar-pressing orotherwise. They speak of trial-and-error, law-of-effect , or, simply, effectlearning.Weneednotobjecttotheseterms,aslongasthereferenceisclear,but we think it more appropriate to adopt the notion of two types ofconditioning: (1) thePavloviancase, inwhicha reinforcingstimulus(e.g.,food) is provided in connection with the presentation of astimulus (e.g.,tone); and (2) theThorndikian case, in which the reinforcing stimulus iscontingent upon aresponse.The important thing, for us, is that there aretworeinforcementcontingencies—onewithanSandonewithanR.

In connection with the concept of positive reinforcement, it has beenargued (e.g., by Hilgard andMarquis, 1940) that our quoted principle ofoperant conditioning onhere is circular.We use reinforcing stimulus toexplain the strengthening of an operant; but wedefine a reinforcingstimulus in termsof its strengthening effect.The fact of thematter is thatsome stimuli strengthen the responses they follow, and others do not.Tothe former alone do we apply the termreinforcing. A less debatablewording of our principle might run as follows:There are stimuli whichhave the power to strengthen the operant responses that produce them.This strengtheningmay be termed 'operant conditioning,' and the stimulimay be referred to a class called 'reinforcing.'This is the essence of ourquotedprinciple,andthecircularityissuperficial.

4

EXTINCTIONANDRECONDITIONING

NEVERsufferanexceptiontooccur....Eachlapseislikethelettingfallofa ball of string which one is carefully winding up; a single slip undoesmorethanagreatmanyturnswillwindagain.

WilliamJames,onthemakingandbreakingofhabits,PrinciplesofPsychology,1890

TheAdaptabilityofBehaviorSo long as life endures, a creature's behavior is a clay to bemolded by

circumstances,whimsicalorplanned.Actsaddedtoit,andotheractswhichfall out, are the means by which it is shaped. Like the two hands of anartisan, busily dabbing and gouging, are the two processes,reinforcementandextinction.

Reinforcementistheindispensableconditionforstrengtheningreactions.But,asweknow,itseffectisexercisedinthepresenceofallofthestimuliexistent at the time it occurs. Some of these stimuli (such as the day'stemperature,passingodors,andmomentaryillumination)maybeirrelevantinthattheyarenottheonesnecessarilycorrelatedwithreinforcement.Theymay,onlateroccasions,bequitedifferentwhilethereinforcementcontinuesto be associated with only one stimulus. If irrelevant stimuli wereperpetually to arouse the response, we would have a picture of sheerbiological inefficiency: energy spent uselessly, time lost, and impairedchances for survival. The adaptability of behavior to critical stimulidependsonthepossibilityofdiminishingtheresponsetonon-criticalones.Such a decline in reaction strength follows the withholding ofreinforcement.This is called extinction, and is theprocesswehavenow toexamine.

RespondentExtinctionJust as a Type S reaction is strengthened by the presence of the

unconditionedstimulus,so it isweakenedbyitsabsence.Supposewehalttheconcomitantpresentationofconditionedandunconditionedstimuli,butcontinue to present the conditioned stimulus.When this is done, and wemeasurethemagnitudeoftheresponse,wefindthatonsuccessiveteststheresponse to theconditionedstimulusdecreasesandeventually reacheszero.This is what Pavlov calledexperimental extinction, and the principlemaybestatedasfollows:

If a conditioned reflexofTypeS is elicitedwithout thepresentationofthereinforcingstimulus,itsstrengthdecreases.

Aresponse, then, is said tobe extinguishedwhen, as a consequenceof itsdissociation from reinforcement, the conditioned stimulus has lost itselicitingpower.

The table below shows the phenomenon as reported from Pavlov'slaboratory.Adogwasconditioned to salivateat the sightofmeat-powder,throughmanytrialsinwhichhewasshownthepowderandthenallowedtoeat some.Extinctionwas then carried out,with thirty-second showings ofpowder which were never followed by eating. In a few trials, theconditionedstimulus lost itspower toelicit salivation.Youwillnote thatsome extinction occurred with each unreinforced stimulation, and that thedrop in response went rapidly at first, and then more slowly. Thisprogression is typical of the variousType S connectionswhich have beenset up and broken down in many laboratories since Pavlov's earlyexperiments.

TableIIIEXTINCTIONOFACONDITIONEDSALIVARYREFLEX

(DatafromPavlov,1927)

Successiveunreinforcedstimulations

Numberofcc.ofsalivasecretedineachthirty-second

period1 1.02 .63 .34 .15 .06 .0

FIG. 15. Average extinction curve of the conditioned galvanic skinresponseobtained from20human subjects.Theordinate is in per cent, sothat for no preceding extinction trials (i.e., on the last conditioning trial)the amplitude of the GSR is given the value of 100 per cent. TheunconditionedeffectoftheCS(tone)accountsforthecurve'sfailuretodropto a zero level. (From Hull, 1943, based on data supplied by C. I.Hovland.)

The actual speed of extinction depends on several factors. (1) Fully-conditioned reactions extinguish more slowly than those based on only afew reinforcements, andover-conditioned reactionsmore slowly than thosejustbroughttofullstrength.(2)Extinctionseemstotakefewertrialswhen

unreinforced elicitations come close together than when distributed over alonger time. (3) Higher-order conditioned respondents are very susceptibleto extinction: a few presentations of the stimulus without reinforcementfrom the preceding conditioned stimulus, and the reaction is at zero. (Youwill recall that higher-order Type S reflexes are difficult to establish,unreliablewhen obtained, and usually quiteweak.Their inability to resistextinctionisanaddedreasonfordiscountingtheir importancein the lifeofanorganism.)

SpontaneousRecoveryExtinction is not necessarily permanent or complete when the response

hasoncereachedzeromagnitude.Whentheanimalisagainbroughtintothelaboratory, the response to the conditioned stimulus usually reappears insome strength, albeit below the former maximum. This phenomenon iscalledspontaneousrecoveryand,althoughitsoriginisnotwellunderstood,ithasregularlyoccurredinbothTypeSandTypeRbehavior.Forexample,thedogwhosesalivaryextinctionisshowninTableIIIwas testedagainbyvisual presentation of meat powder only two hours after this record wasobtained. His response then measured .15 cc. of saliva. Greater recoveriesthan this have been found with other responses and under othercircumstances.Thespontaneousrecoveryofhigher-orderTypeSreactionsisnegligible.

Extinction following spontaneous recovery is faster than originalextinction. If there is a second recovery, it is less in magnitude than thefirst and is extinguished more rapidly.As a rule, only a few extinctionsessions are needed to overcome the diminishing recoveries and to reducetheresponsestrengthtoastablezero.

OperantExtinctionConditioned operants are extinguished by severing the relation between

theactandtheeffect.Assuccessiveresponsesfailtoproducereinforcement,the recurrenceof the response is lessand less likely.TheprincipleofTypeRextinctionmaybeputasfollows:

The strength of a conditioned operant may be diminished bywithholdingitsreinforcement.

As in thecaseofoperant conditioning, theprincipalmeasureofoperantextinction is the frequencyof the response in time.The loss in strength isseeninafallenrateofemission.ThisisportrayedclearlyinthecumulativeresponsecurveofFigure16.Asresponsescomemoreandmoreslowly,thecumulativecurvebendsoverandtakesonacharacteristicshape.

FIG. 16. Typical cumulative response curve for extinction of bar-pressingbyawhiteratfollowingabout100reinforcements.(AfterSkinner,1938.)

Theextinctioncurve fora responsehitherto regularly reinforced (that is,with a reinforcement foreach emission) is usually, if not always, ratheruneven.Itbeginswithasteeperslope(higherresponserate)thanthatduringregular reinforcement, partly because responses are no longer separated byeating time, and partly because the animal is apt to attack vigorously thenow-unrewarding bar.Thereafter, the curve ismarked bywavelike changesin rate which distort it in detail but still permit the drawing of a smooth'envelope' to describe the over-all trend.These bursts and depressions ofresponse might be called emotional in character, the counterpart of morecomplicated frustrations and aggressions seen inman. Consider, forexample, the frustration-aggression pattern in the behavior of a child whostruggleswithknottedshoe-lacesoranunyieldingdoor,or theeffectofanunresponsivelistenerincoldlydousingourdinner-partystories.

ResistancetoExtinctionasaMeasureofStrengthA strong habit is one that tends to persist after the reinforcement has

been discontinued,whereas aweak one succumbsmore quickly.Taken byitself, emission rate under regular reinforcement is not a good indicator ofoperant strength.With bar-pressing for food, for example, rate is greatlyaffectedby such incidental things as the size andhardness of thepellets—whichdeterminechewing time.On theotherhand, resistance to extinctionserves to disclose quite well what strength an act has acquired during atrainingperiod.

But how to measure this resistance?We can use the total number ofresponses emitted in extinction, or the number required to reach some

arbitraryextinction criterion such as the first five-minute or ten-minuteinterval in which no responding occurs. The number of responses in acomplete extinction is generally more satisfactory, since any short-timecriterion excludes possibly important data contained in the rest of theextinction curve.On the other hand, extinction criteria have the advantageof being time-saving and experimentally convenient, and are thereforefrequentlyemployed.

Resistance to extinction after regular reinforcement is governed by anumberoffactors,amongwhicharethefollowing:

1 .The number of reinforcements applied to the response. Williams(1938) trained five groups of rats to press a bar for food.The animals ineachgroupwereallowedagivennumberofreinforcementsbeforeextinctionwasbegun,thelow-numbergroupreceivingbutfivereinforcements,andthehigh-number group, 90. InFigure 17 is plotted the mean number ofunreinforcedresponsesmadebyeachgroupuptoacriterionoffiveminutesofnoresponse.Resistance toextinction isseen to increasewithnumberofreinforcements up to a point beyond which additional reinforcementsproduce very little increment in strength. (In mathematical terms, anasymptoteofstrengthisapproached.)

FIG. 17.Curve showing the relation betweennumber of reinforcements

andnumberofextinctionresponses.Thenumberofresponsesmadeistakenasthemeasureofstrengthofconditioning.(AfterWilliams,1938.)

2.Theamountofreinforcementgivenforeachresponse.Thisfactorhasrecently been investigated by Zeaman (1949). Four groups of rats weregiven one run a day on a GrahamGagné runway for twenty days, with adifferent amount of food reinforcement for each group (.2, .4, .8, and 1.6grams).Zeamanwasabletoshowthattheamountofreinforcementreceivedduring trainingwas related to thehighest speedof starting (here) that eachgroupultimatelyattained.Thegroupgivenasmallamountoffooddidnotreachas lowastarting-timelevelas thegroupgivena largeamount.Theseresults accord with those from an early study by Grindley (1929) whichindicated that chickswill traverse a runway faster for six grains of rice pertrial than they will for onegrain. Similarly, Wolfe and Kaplon (1941)foundthatchickslearnedamazefasterwhenawholekernelofpopcornwasgiven for each run than they did when only a quarter of a kernel wasprovided.At the end of the training period inZeaman's experiment,whenthe starting-times had decreased to a minimal value for each group, therunning response was extinguished. Although extinction began with thegroups atdifferent levels, and although thenumberofnon-reinforced trialswas insufficient for a complete test, Zeaman's analysis suggests that anincrease in amount of reinforcement goes with an increased resistance toextinction.

FIG.18.Finalstrengthachievedbytheconditionedsalivaryreflexinthedog as related to the amount of food reinforcement used in theconditioning. One dog was used with four different CS's, each CS beingreinforced by a different amount of food.The training schedule involvedpresentation of all four stimuli in random order, but each stimulus wasalwaysassociatedwith itsownamountof food.Thecurveshows the limitofresponsestrengthreachedforeachstimulus,andindicatesthatrespondentas well as operant conditioning is influenced by the amount ofreinforcementemployed.(FromHull,1943;datasuppliedbyW.H.Gantt.)

3.Thedegreeofmotivationpresentduringextinctionandconditioning.Youwill see, in Chapter9, that resistance to extinction depends upon themotivation (hunger, thirst,etc.) present at the time. Suppose a group ofanimals, equally hungry, are conditioned with the same number ofreinforcements. If they are then extinguished under different degrees ofhunger, the hungrier ones make more responses than do the less hungry(pages265-266).That is,resistancetoextinctionisgreaterwhenextinctionis carried out under highmotivation than under low (Perin, 1942).Oddlyenough, however, when groups of animals areconditioned under differentdegreesof(hunger)motivationandextinguishedunderthesamedegree, theextinction responsesdonot reflect thedifferences in trainingconditions.Acertainlowdegreeofmotivationisrequiredforanyconditioning,butadded

incrementsdonotseemtogiveaddedstrength(Strassburger,1950).

TableIVTHEEFFECTOFVARYINGDRIVELEVELATTIMEOF

CONDITIONINGUPONTHESTRENGTHOFCONDITIONING

(DatafromStrassburger,1950)

The table summarizes the experimental design and findings. Each entryin the table represents one group of animals conditioned in bar-pressingunder the indicated length of food deprivation and with the indicatednumberof reinforcements; thenumerical valueof the entrygives themeannumberofbar-pressing responsesmade in the first hourof extinction.Tentotwelveanimalswereusedineachgroup.Extinctionwascarriedoutatthesamedrivelevel(23hoursofdeprivation)forallgroups.Noneofthevaluesin any row is significantly different from the others in that row bystatistical test.Thus, for agivennumberof reinforcements, thedrive levelat time of conditioning did not affect the strength of the response (asmeasuredbyresistancetoextinction).Note,however,thatthereisarelationshown between the number of reinforcements and resistance to extinction.ThislatterfindingcorroboratesthatinFigure17.here.

WhenIsaResponseExtinguished?Anoperantmust exist in some strengthbefore it canbe conditioned; it

mustbeemittedonceinawhileatleast,inorderforustoreinforceit.Thisunconditioned rate of emission may be called theoperant level for thatresponse, and it appears as part of the general activity of the organism. Itdeterminesthequicknesswithwhicharesponsecanbereinforced:iftheactcomes infrequently, there isa longwait; if the first reinforcementdoesnottakehold,conditioningisinconvenientlydelayed.

From the fact of operant level, it follows thatan extinguished responsewill not reach a zero rate, but will return to one that existed beforereinforcement. Thus, the cumulative response curves for bar-pressing

extinction in the rat will approach or reach a slope that approximates theone that existed prior to conditioning. For this reason, experimentswhichaim to compare numbers of responses in extinction after different kinds oramountsof trainingmust take accountof theoperant level.Unconditionedemission ratesmust be determined for all animals before the experimentalfactor is introduced. The groups may be equated for prior level, or acorrectionmaybeappliedtothefinaldata.

SpontaneousRecoveryofaTypeRResponseSuppose that, on agivenday, after conditioning, a periodof extinction

has reduced bar-pressing to a low rate. The next day's session, with allconditions the same,will give another, but smaller, extinction curve.Thisobservation has been made in a number of studies. Youtz (1938), forexample, extinguished bar-pressing in rats to a criterion of no response intwentyminutes.Oneday later, ina secondextinction session,heobtainedasmuchas55percentoftheresponsesemittedinthefirst.Theamountofrecovery diminishes, however, with successive extinctions until none isdiscernibleapartfromthatwhicharisesfromgeneralactivity.

Further consideration of this spontaneous recovery suggests that thenumber of responses 'recovered' does not add up to more than would beexpectedfromanextensionoftheoriginalextinctioncurve;inotherwords,iftheextinctioncurvehadbeencontinuedwithoutinterruption,thenumberofresponseswouldhavebeenaboutthesameasthatobtainedfromaspacedsuccession of shorter extinction periods.This may, or may not, be true;thereissomeevidenceto thecontrary(Ellson,1939;Keller,1940).Itdoesappear,however,thatthelongertheintervalbetweenoriginalextinctionanda later one, the greater the accumulation of unexpended responses, and thegreatertheapparentrecovery.Thiseffectmaybequitepronounced,evenforintervalsmeasured inminutes, if extinction is interruptedvery soon in thefirst session.At that time, responses are coming out at a good clip and,consequently,manywill pile up in a short space of time.An example ofthisisaffordedbyEllson's(1938)studyinwhichextinctionwascarriedouttoafive-minuteperiodofnoresponding.(Suchacriterionisreachedbeforethe process is very far advanced, and many responses are still to come.)Subsequently, four groups of his animalswere given recovery intervals of5.5, 25, 65, and 185 minutes, after which extinction was resumed to thesame five-minute criterion.The average numbers of responses accumulatedintherespectivegroupswere7.6,14.4,19.5,and24.4.

FIG.19.Aschematiccurveillustratinghowspontaneousrecoveryafteralapse of 8 hours might approach the extinction curve (dashed line) thatwould have been expected if extinction had not been interrupted. (AfterSkinner,1938.)

ExtinctionandForgettingThe beginner in psychology often takesextinction to be synonymous

withforgetting.Acloseexaminationofforgettingshowsus,however, thatmattersarenotquiteassimpleasthis.Letuslook,briefly,intothehistoryof the problem. Back in 1885, Hermann von Ebbinghaus published avolume entitledUeberdasGedächtnis or, inEnglish,OnMemory. In thisbook he proposed methods for the quantitative study of human verballearningandtheretentionof learnedmaterial, togetherwitha largebodyofdatahehadpainstakinglyamassedinstudyinghisownability to learnandrelearnseriesofnonsensesyllables(rop, fim,zeb, etc.).We shall consider,inChapter7,theactualprocessofseriallearning,andwillhaveoccasiontoreferagain toEbbinghaus. In thepresentconnection,wemaynote thatoneof theproblems towhichhe addressed himselfwas that of forgetting.Thecurve inFigure 20 shows how he found the amount of retention ofnonsensesyllables todependupon theamountof time thatpassesbetweeninitial learningofa listand later relearning toagivendegreeofperfection.It indicates that the major loss in remembering occurs quickly afterlearning,butsomesmallresiduecanbeexpectedtosurviveforalongtime.

FIG. 20. Curve of nonsense-syllable retention obtained by Ebbinghauswith the "savings method." This method takes the difference betweenoriginal time taken to learn and relearning time after some period, andcomputes the "per cent time saved" (here called "per cent retained") bydividing the time difference by original learning time. Each periodrepresents the learning and relearning of a different list of nonsensesyllables. Ebbinghaus served as his own subject. The first three pointsplotted are for intervals of .33, 1.00, and 8.8 hours, respectively. (AfterEbbinghaus,1885.)

Despite Ebbinghaus' results and the supporting evidence later providedby other investigators, objections soon arose. We now know that his'forgettingcurve' isnotbroadlyapplicable toother thannonsensematerial.Manyobservershavepointed to the recallof supposedlydeadmemories indreams, hypnosis, and reveries. Psychiatrists have convinced us that,withsuitableprompting,personsmayrecovermemories thathadat firstseemedhopelesslylost.Experimentalists,too,haveaddedevidencewhichindicatesthat 'meaningful' material, such as prose and poetry, does not fade withtimeinthemannersuggestedbytheEbbinghauscurve.

Scientifically,themeaningofforgettingshouldprobablyberestrictedtotheweakeningofresponsewhichresultsexclusivelyfromthe lapseof timebetween conditioning and some later test of strength. Simple conditionedresponses, operant or respondent, lend themselves to the study of thisphenomenon.After the response is established, any lengthof timemaybeallowed to elapse before it is tested again and compared with its formerstrength. If forgetting takes place during the interval, the loss in strengthshould be reflected in, say, a decreased resistance to extinction of the

response.The upshot of several studies has been that themere disuse of aconditioned responseyields very little diminution in strength over periodsof months or years.Figure 21 presents averaged extinction curves of bar-pressing in two groups of rats.The upper curve was obtained on the dayfollowing the one in which 100 responses had been reinforced; the lowercurve was obtained forty-five days after initial conditioning and the samenumberof reinforcements.Experimental extinction is, apparently, farmoreeffectiveinweakeningaresponsethanisaperiodofpassivedisuse.Today,manyinvestigatorsbelievethat,givenidealcontroloveracreature'sactivityduring the interval between learning and testing, a conditioned responsewouldshownoweakeningatall.

FIG.21.Averagecumulativeresponsecurves forbar-pressingextinctionin twogroupsof ratsextinguishedafterdifferentperiodsof timefollowingoriginal conditioning. Only slight loss of strength results from the merelapseoftime.(FromSkinner,1938.)

The seeming dilemma is not insoluble.The crucial difference betweentheEbbinghausresultsandtheconditioned-reflexresults lies in the typeofmaterials employed. Nonsense syllables, in long lists, memorizedindividually in a fixed serial order, are subject tomuchmutual confusionandblocking.Theintervalsbetweenlearningandrelearning inEbbinghaus'work were undoubtedly filled with verbal behavior—of which nonsensesyllables are special cases.These syllables probably cannot be isolated orprotectedfromblendingwithotherspeech,andthiswouldinvolve,intime,a loss of their identity.This is probably related to the fact thatwe 'forget'foreignlanguagesandourschool-daymathematics,butnothowtoswimorrideabicycle.

Ifweasknowwhether there is sucha thingas forgetting, theanswer is

probably No.A simple and distinct act probably does not expire throughdisuse; an intricate set of reactions may suffer, if not rehearsed, frominternal attrition. If, in theinterval, other things are learned or practisedwhich conflict with the original responses, 'forgetting' will be the result.The everyday use of the wordforgetting does not, of course, distinguishbetween lapse of time and other factors that may account for a loss inretention.

Regression:aBy-productofExtinctionOccasionally we see an adult behave, as we put it, "like a child"; or a

youngsterwho,seeinghisparents'attentionturntoanewinfant,mayonceagainwet himself, insist on being hand-fed, and refuse to sleep in a darkroom or in his own bed. These and other instances have seemed topsychiatriststoillustrateincommonamechanismwhichtheydescribeasaregression or going-back. They consider regression to be a retreat fromone'smorerecentlyacquiredbehavior to thatofanearlierperiod.Fromthecritical writings on this subject, we may extract a central idea: If presentbehavior isnotcapableofgetting reinforcement,one reverts toolder formsofresponsewhichwereonceeffective.Anindividualwhosebesteffortsfailto reach a solution of his difficulties is thrown back upon other resources,andtheseresourcesaretheonesheonceusedsuccessfullybutoutgrewwithmaturationorsocialtraining.

Clinicalobservationsarealwayscomplexanditisdangeroustoentertainover-simplifiedexplanationsofthem.Nevertheless,thenotionofregressionhas led toa search forapossibleprototypewhichwouldgive somedegreeof validity to the idea. Mowrer (1940), in the study cited earlier (here),believed he found such a prototype. In his experiment, rats wereconditioned to push a panel in order to turn off a shock coming from thecage floor. The first response to the shock, before panel-pushing waslearned,was a tiptoeposturingor dance,which reduced the severity of theshock.Withfurther training, thisresponsediminishedinfavorof themoreeffective response to the panel. In a later extinction series, when panel-pushingwasitselfnolongerreinforcedbyshock-removal,theanimalswentback to the earliermincing steps which had been their partially satisfyingmode of coping with the noxious stimulus. Such findings as these havebeen reported by Hull (1934) and a number of other investigators (e.g.,Masserman,1946);andweshallhaveoccasiontoreturntothismatterlaterinconnectionwiththeproblemof'chaining'(Chapter7.).

One-TrialExtinctionExtinction,aswehavebeendescribingit,isaprettyslowprocess,butit

may have occurred to you that many of our own responses show arapiddecrement when reinforcement ceases—that you, for example, would verysoonstopmakingaresponsethatnolongergotresults.Withtheexception

ofthefeeble-minded(Mitrano,1939)ortheveryyoung,nohumanbeingislikely towastemanycoinsonavendingmachine that isoutoforder; andfewofuswillblindlypersist in turningakeythatnolongeropensadoor,orscratchingmatchesthatneverlight.

It is undoubtedly true that extinction of an operant maybe very rapid,evencomplete ina singleemission (howmanyofusput. twonickels inacoin box when the first is not returned?), but this statement is deceptivebecause it ignores the fact that immediate extinction is not a property oforiginal behavior.We learn the signs of failure by virtue of long training,and these signs come to govern the number of responses made inextinction. Wediscriminate, that is, between reinforcing and non-reinforcingsituations(seeChapter5).

An analogue of such behavior can be found in lower organisms.At thebeginning of an experimental hour, a rat is given a set number of regularreinforcements, say ten or twenty, for bar-pressing and then extinction iscarriedout for the rest of thehour.The sameprocedure is followedon thenext day, and the next, for as long as wewish.The number of responsesmade in extinction on successive days decreasesmarkedly, indicating thattheveryfailuretoobtainthereinforcementbeginstoactasasignaltoceaseresponding (BullockandFischer,1950). It isbasically thisprocess,whichhuman beings improve upon over the rat by direct verbal instruction, thatenables a response to be dropped at once. To speak of it withoutqualification as one-trial extinction is incorrect. It would be equallyerroneous to do so in the case of a conditioned Type S reaction likesalivation which, in successive extinctions following reconditioning, maygettoapointwhereasinglenon-reinforcedelicitationisenoughtodroptheresponsetozeromagnitude(Pavlov,1927).

PeriodicReconditioningThe interspersal of conditioning and extinction that we have just been

discussing leads us into an extremely interesting area. The regularreinforcement of response is not the world's rule; and some importantproperties of behavior should arise from periods of extinction that arebroken into by occasional reinforcements. Laboratory research has, in fact,disclosedthatthisisthecase;andwemaynowexaminesomeoftheresultsobtainedbythesystematicrepetitionofconditioningandextinction.

One suchprocedure,whichhasproved tobeavaluable research tool, isbased upon the use of single reinforcements that are periodically providedthroughouttheexperimentalsession.Theschedulemaybeoftwokinds:inone, the reinforcements are separatedby a fixedtimeinterval; in the other,theyareseparatedbyafixednumberofunreinforcedresponses.Letustakethemupinorder.

PeriodicReconditioningatFixedIntervals

To illustrate the first procedure, we take a hungry rat and ourrepresentative bar-pressing response, and we choose three minutes as abetween-reinforcement interval. The animal's first response to the bar isrewarded with a food-pellet, but none of the responses in the next threeminutes is permitted to be effective. The very first response after theinterval isreinforced, and the responses in thenext threeminutes againgounreinforced. And so on, for as many cycles of reinforcement and non-reinforcement as we desire. (The interval of extinction responses is notprecisely 'fixed,'ofcourse, since the ratmaynotalways respondatexactlythetimewhenthefood-pelletbecomesavailable,butexcessivedelayisveryrarebecauseof thesteadyhigh rateof respondingwhichusuallydevelops.)If we are patient and do not cut the procedure off too soon, we get aninteresting set of changes.The cumulative response curve passes throughthreestages.

FIG.22

1.At first, the intervalsofunreinforced respondingare likely tocontainlittle curves which bend over, and which are actually small extinctioncurves. Each reinforcement brings out some responses, but notmany, andtheydwindle during the interval.Figure22-A reproduces a recordobtainedundertheseconditions.Youwillobservethatthelittlecurvessumupsoastogiveapositivelyacceleratedtrendtotheover-allcurve.

2.When reconditioning has been continued for a while, the positiveacceleration comes to an end, and the rateswithin the intervals fuse into asteady stream of responding.The temporally spaced reinforcements sufficeto maintain a consistent strength of response which accounts for thestraight-line appearance of the cumulative record.The beginning of such arateisseeninthelaterportionofFigure22-A,andthroughoutFigure22-B

.LEGENDFORFIG.22

A.The first stage of P-R (periodic reconditioning) training for a whiterat. The vertical strokes in the curve mark the reinforcements. Note theextinction-like shape of the cumulative response curve in the first fewperiods,andthelatersuggestionofasteadyrate.B. The practically linear cumulative response record obtained as the

second stage in P-R responding.The P-R intervalwas threeminutes, andthisportionoftherecordoccurredafterabout60reinforcements.

C. The third stage of P-R responding, showing the presence of atemporaldiscrimination.This record is fromthesameratas inB, andwasobtainedafter17daysoftraining(onehour,or20reinforcements,perday).(AfterSkinner,1938.)

3.After many reconditioning sessions, the response curves during thethree-minute intervals take on a scalloped appearance opposite to that instage 1. Right after a reinforcement, the animal delays his response for atime,but,as thesecondsslipaway,hestartsupagain,slowlyat first,andthenfaster,until theinterval isoverandthenextreinforcement isreceived.Thisshowsthatatemporaldiscriminationhasbeen formed.The responsesjust after eating a pellet are weakened because they are never reinforced,whereas later responses are strengthenedbecause the reinforcement is givenlater.The animal comes to "tell the time" at which responding is morelikely to succeed. Since, in this case, reconditioning isperiodic, somedegreeoftemporaldiscriminationisunavoidableinthelongrun.Thecurvei nFigure 22-C, obtained after seventeen one-hour sessions of periodicreconditioning,showsthisclearly.Youwillnotetheessentialsimilarityofthisfindingtothatdescribedinourtreatmentofthetracereflex inChapter

2. In the course of trace-reflex formation, salivation to the conditionedstimulus is increasingly delayed until it comes just before thereinforcement.

The close examination of curves, like that you have just made, is anexample of honoring small details, an occupation that has paid richdividendsinthehistoryofscience.Letus, then,consideranotherexample.(Andletus,forconvenience,adoptthepracticeofusing"P-R"for"periodicreconditioning.")Figure 23 presents a telescoped P-R curve for the sameanimal that provided the curves forFigures22-B and22-C. It gives us anover-allviewof response rateduring twenty-four experimentalhours.Withthe details washed out, another trend emerges. (This can best be seen byraisingyourbookinthehorizontalplaneandsightingalongthecurvefromitsnearend.)Thecurvebendsoverslightlywithcontinued training,as thenumber of unreinforced responses falls off.This is not unlike the drop inresponse frequency mentioned earlier in connection with "one-trial"extinction, but it is much less marked. Apparently the temporaldiscriminationwhichcontrolsthenumberofresponsesis,inthiscase,slowtoform.Onemightsaythatitisharderto'tellwhen'thenextreinforcementis coming than to 'tell that' there will be no more. Even singlereinforcements,properly spaced,keep the responsestrengthhigh fora longtime.

FIG. 23. Cumulative response curve under three-minute P-R over anextended period of time. Note the slow over-all negative acceleration thatprobably shows the control over responding gradually assumed by thetemporal discrimination.Figures 22-B and22-C are enlarged portions ofthiscurve.(AfterSkinner,1938.)

ResponseRateandP-RIntervalIntheaboveillustration,afixedintervalofthreeminuteswasused.You

may wonder what the effect would be if the length of this interval werechanged.The answer to this query has been given tentatively by Skinner(1938).Within limits, the greater the interval length the lower the rate ofresponse;andchangingtheintervalforanindividualanimalwillchangetheslope of his curve. Since the mere specification of slope does not tell uswhat interval gave rise to it, we need a term for the slope-intervalcombination.Skinnerhassuppliedthisinhisextinctionratio,whichreferstotheratioofunreinforcedtoreinforced responsesemitted.Thisratiomaybe obtained by taking the number of responses made in an hour anddividing it by the number of intervals within the hour. For example,supposea rat,onP-Rat three-minute intervals,hasmade346responses inone hour. Since there are twenty intervals in the hour, his extinction ratiowillbe346/20,or17.3responsesperreinforcement.Theratio tellsushow"willing" the animal is to make unreinforced responses when thereinforcementsarespacedastheyare—thatis,whateffectagivenspacingofreinforcementhasuponthestrengthoftheresponse.

For the present, two things only need to be noted about the extinctionratio.(1)Itssizedoesnotseemtovarywiththelengthoftheinterval.Thedataonthispointaremeagre,buttheyindicatethatslowerratesoverlongerintervals give the same ratio as faster rates over shorter intervals. (2) Ratsthatdifferwithrespect toratiosize,differ in thesamewaywhenresistancetoextinctionismeasured.

Relative toour secondpoint are findingsobtainedby120 students in aGeneral Psychology course at Columbia College in 1946. Each studentworked with a single rat and, in one class experiment, a two-minute P-Rwasemployedtodeterminetheextinctionratioforeachanimal,afterwhichthe bar-pressing response was extinguished.The 120 rats were ranked forsize of extinction ratio and for number of responses during one hour ofextinction.The coefficient of correlation between the two rankswas +.72,indicatingthatifaratgivesalargenumberofunreinforcedresponsesduringP-Rhewillalsogivealargenumberinextinction.Thishadbeenpredictedby the class, and the results bore out the prediction, but the findingneedstobecheckedwithamorerigorouslycontrolledexperiment.

ExtinctionAfterPeriodicReconditioning

One thing is very certain: P-R increases the resistance of a response toextinction.This finding comesoutof experiments like the following.Onegroup of hungry rats is given twenty regular reinforcements with food forbar-pressing, and the response is then extinguished.A second group alsoreceives twenty reinforcements before extinction, but these reinforcementsare spaced out in a three-minute P-R schedule. Suppose that extinction inboth cases is carried out for four hours, one hour per day on successivedays.Whenalltheextinctionresponsesofthetwogroupsareaddedup,thesecondgroupcomesoutfaraheadinthetotalnumbermade.

This simple experiment leads to some additional comparisons.As wesaw earlier, extinction following regular reinforcement starts with a sharpinitial spurt in rate, exceeding that which prevailed during regularreinforcement. Afterwards, the extinction curve is marked by wave-likedepressions and accelerations in responding. In contrast, periodicreconditioning gives rise to an extinction curvewhich does not noticeablyexceedthepreviousP-Rrate.Ifwecomparedthetwoextinctionsforonlyabrief period after reinforcement had ceased, regular reinforcement wouldappear tohavegeneratedagreater resistance toextinction thanP-R.Suchaconclusion, however, would be unjustified, since a continuation ofextinctionwouldrevealthattherespondingafterP-Rgoesonunabatedforaconsiderable time,whereas that after regular reinforcement soon tapers off.Here,asinothercircumstances,itiswisetowaituntilallthereturnsarein.

Not only is extinction after P-R more resistant, but the curve is alsosmoother than that after regular reinforcement. The vacillations betweenaggressive attacks on the bar and depressions in responding, which comeafter regular reinforcement is discontinued, do not occur somarkedly afterP-R.The greater smoothness of extinction suggests that P-R increases acreature's"frustrationtolerance."IntrinsictoP-Ristherecurrenceofperiodsof non-reinforcement during which "frustration" is repeatedly experiencedandovercomeby continued responding.Wemight expect the same sort ofresult in training children. Occasional reinforcement gives stability tobehavior,andpersistenceinthefaceoffailure.Skinnerhasarguedthat thisstability,aswellas the increasedresistance toextinctionfollowingP-R, isa significant property of our normal behavior, and responsible for"equanimity in aworld inwhich the contingency of reinforcing stimuli isnecessarily uncertain." He points out that our behavior "would be clumsyand inefficient if the strength of an operant were to oscillate from oneextreme to anotherwith the presence or absence of its reinforcement"(Thebehavioroforganisms,1938,p.138.)

FIG.24A.CumulativeresponsecurveofextinctionimmediatelyfollowingP.R.

Notethesmoothnessofthiscurveascomparedwithextinctionafterregular

reinforcement(FIG.16).

B. Cumulative response curve of extinction following P-R. Comparedwithextinctionafterregularreinforcement,respondinggoesonunabatedforaconsiderabletime.(AfterSkinner,1938.)

Fixed-RatioPeriodicReconditioningInadditiontowithholdingreinforcementforafixedintervaloftime,we

can make it wait upon the emission of a certainnumber of unreinforcedresponses—that is, we can establish afixed ratio of unreinforced toreinforced responses.This procedure brings to mind cases in which a settask must be carried out before we reap a reward. If we use it carefullywithin the laboratory, it should enable us to make some interestingdiscoveriesaboutourownbehavior.

Theexperimentalschemeisasfollows.AratisfirsttrainedunderP-Rata fixed interval of, say, fiveminutes, until his response rate is stabilized.Supposewefindthathisextinctionratiois10:1(tenunreinforcedresponsesfor each reinforced one), and we then decide to change over to fixed-ratioreinforcement,making food contingent upon a givennumber of responses.Threealternativesareopen tous:wecan set a fixed ratio (1) less than, (2)equalto,or(3)greaterthantheextinctionratio.

If a reinforcement is given forless than ten responses, itwill comesooner than usualwithin each interval—that is, itwill take him less thanfive minutes to get his pellet. If it comes sooner, there will be morefrequent reinforcement, and thiswill result in an increased rate of responsejustasitdoeswithshorterfixed-intervalP-R(here).Thenewrate, in turn,will bring reinforcement even more quickly, and the process will repeatitselfuntila limitinghighrateof response isachieved.Themore thefixedratiofallsshortoftheextinctionratio,thefastertheaccelerationinrateandthe earlier the final rate-limit will be reached.Figure 25 shows thisschematically for four fixed ratios below our assumed extinction ratio of10:1.

FIG. 25. Hypothetical curves showing acceleration of the response rateunder several fixed ratios smaller than the10:1 extinction ratio representedbythestraight-lineP-Rcurve.(FromSkinner,1938.)

Consider,now,thehypotheticaleffectofusingafixedratiowhichisthesame as the extinction ratio—that is, 10:1. Presumably, the frequency ofreinforcementwill continue togiveus the same slopeof curve as the five-minute P-R with which we started out, since the fixed ratio calls for thesame number of unreinforced responses as would be emitted, on theaverage, in the five-minute interval.This turns out to be the case, at leastfor awhile.After a time, however, there is an acceleration like thatwhichoccurs when the fixed ratio is less than the extinction ratio.A rat maymaintain his fixed-interval slope for as long as three experimental hours,whereupon hemay (sometimes suddenly) assume a new and higher rate ofresponding. What could have 'informed' him that reinforcement wascontingent upon the number of responses rather than the interval, and thathe might rather respond rapidly? The answer lies in the fact that theextinction ratio varies from time to time.Ten-to-onewas an average ratio;atonetimeoranotheritwascertaintobemorethanthat—sufficientlysoto

trip off the circular accelerative process. Such variations are especiallycommon in an animal's natural habitat, and are accentuatedby fluctuationsin his level of motivation which, in turn, determines the size of theextinctionratio(seeChapter9).

Whenthefixedratioisgreaterthantheextinctionratio,werequiremorethan ten responses before we reinforce. This means that reinforcementscome, on the average, less often than every five minutes. The rate ofresponding should, therefore, go down. But a lower response rate at fixedratio decreases the reinforcement frequency still further.We have a circularinteraction in which response rate continually decelerates and finallyextinguishes.The rapidity of the extinction will depend upon how muchthefixedratioexceedstheextinctionratio.Althoughreinforcementswillbereceivedeverynowandthen,theresponserate,cannotbesustained.Evenifitweresustained,ahighfixedratiomightinvolvemorelossofenergythanthe single reinforcements could offset, and the animal would waste awaybecauseofthisunfavorablebalanceoffoodintakeandenergyoutput.

The running response in an apparatus like that of Graham and Gagné(here)alsorevealsthebasicpropertiesoffixed-ratioreinforcement.Sincetherunwayisoffixedlength,wecanthinkofitinaroughwayasdemandingacertainnumberofstepsfortraversal.Thedistanceisshort,sothefixedratioofunreinforcedstepsiscomparativelysmall.Thedecreasedrunningtimeonsuccessivetrialsrepresentsanaccelerationinrateofstep-responding,thatis,the rat emits his running steps with increasing speed up to a limit.Although, unlike bar-pressing, the animal moves into a new stimulus-situation with each running response, the parallel between the twoproceduresisstriking.

The'Ceiling'ofFixed-RatioRespondingIn the example just considered, where the fixed ratio exceeded the

extinction ratio, we said that the response extinguishes. This happenswhenever the fixed ratio issuddenly set at a valuefarabove the extinctionratioandthesameeffectwouldbeachievedwiththerunwayapparatusifwewere suddenly to lengthen it to, say, a thousand feet.But there is anotherprocedure, of a more gradual sort, by which we can establish very highfixed-ratiovalues.Takingourratwiththeassumedextinctionratioof10:1,let us begin with a fixed-ratio schedule of 8:1. The response rate willaccelerate, as described above. Once the rate is up, consider the situationthat exists.What is being reinforced?The responding, of course, but alsothehigh rate itself is now feeling the effect. Every time a reinforcementcomes it impingesuponresponsescomingoutatacertain rate, so thatnotonly the response is kept strong, but the rate, too, is being conditioned.Theanimal"learns"thatbar-pressingisrewardedwhenitisfast.ThisisanaspectofTypeRconditioningwhichwenowmeetforthefirsttime,anditmaypayyoutostopforamomenttothinkitthrough.

Withour animal now responding at a high rate,wemay raise the fixedratio toanew level,perhapsof fourteen responses.Bridging the smallgapfromeighttofourteenisnothard,becausetheanimalisalreadyconditionedto a high response rate, and the exact number of responses is somewhatelastic.Hetendstokeepupahighpressingspeeduntilthereinforcement isforthcoming. After the new ratio is strongly practised, it is raised oncemore;aftera time,we increase itagain. If the increasesarenot toogreatatany one step, a ratio can be reached far beyond the starting point. In oneexperiment(seeFigure26), itwasactuallypossiblebyaprocessofgradualapproachandthoroughtrainingtoreachandsustainaratioof192responsesper reinforcement, without extinction! (Skinner, 1938.)Figure 26 alsoreveals in each curve the acceleration of response during periods of non-reinforcement.This acceleration resembles that shown in theP-R curve onhereandhasasimilarorigin.Actually,atemporaldiscrimination,althoughof no "use" to the rat in fixed-ratio P-R, is nevertheless formed, since theexperimental arrangement never permits one reinforcement to followimmediately after another. Or, think of it this way: receipt of a pellet isnever followed immediately by another reinforcement, so the animal holdsoff further responding for a while.But, when responding is resumed, thespeed picks up again because reinforcement is less delayed by a quickcompletionof the fixed ratio thana slowone. In thevernacular,wewouldsaythat"theratbecomesmoreeagerfor thefoodpelletashegetscloser towinningit."

FIG.26.Reinforcementat several fixed ratios.Thecumulative responsecurves are for one rat that was graduallyworked up to the ratio of 192:1.Thehorizontallinesindicatereinforcements.(AfterSkinner,1938.)

Thisphenomenon is not alien toour everyday experience.Facedwith amonotonous,routinejobofdefinitelength,likemowingalawn,weareaptto hasten our work as we approach its end at which there are the waitingrewards of a cool drink and soft chair. Indeed, it is so natural a thing thatwe can spontaneously recognize its operation in other persons. Imagine aman who, after a long enforced separation, is returning to his family andhome.Ashisobjectivecomesintoview,hispacequickensuntil,unabletohold back any longer, he breaks into a run. Or, imagine a couple who,under the stressof strongemotion, approacheachother fromadistance, atfirst slowly, butwith gathering speed, until the last few steps before theirembrace are covered at a run. Sensitive cinema directors have used suchwordless scenes with great effect, knowing that they could count uponaudiences'understandingandsympathy.

Themethod of extending the fixed ratio by delaying reinforcement and

gradually advancing the size of the ratio, has been employed with rats inother than bar-pressing situations. A caged rat may be taught to pull astring to which a piece of food, lying outside the cage, is attached (e.g.,Tolman,1937).Assurenessandspeeddevelopintheshortpulls,thestringmay be lengthened by slow degrees. Eventually, the string may reach alength well beyond that which the animal could initially have beenconditioned to pull, but he hauls away heartily and persistently on eachoccasion.Or,aratmaybetaughttodigsandfromatubethroughwhichhehas previously run unimpeded to obtain food (Stone, 1937). Once theresponse iswellestablished,moreandmoresandmaybe let into the tube.Iftheincreaseisnottooprecipitous,atrulyenormousamountofsandmaybe removed by the rat between reinforcements.The extension of the fixedratio by a judicious manipulation of reinforcement bears a resemblance tothetypeofresponsetrainingcalled"differentiation"(seeChapter6).

ExtinctionFollowingFixed-RatioReinforcementFixed-ratio reinforcement leads to extinction curves which are different

from those following either regular or periodic reinforcement (fixedinterval).Thecurvebegins steeply, andmaintainsamaximal rateuntil theresponses,quiteabruptly,cometoanend. Figure27showsthisforoneratthathadbeenworkingatahighratio,andforoneatalowratio.Thereasonfor the difference between these and themore usual extinction curvesmaynot at first be obvious. You will remember, however, that fixed-ratioreinforcement strengthens not only the individual response, but also therateofresponse.Whenareinforcementcomes,therateisusuallymaximal.A high rate of responding is usually followed by a reinforcement and,adopting a term from the following chapter, we can say that the animal"discriminates"hisown rateof responding. Inhisworld, theoccurrenceoffastrespondingmarksthetimewhenreinforcementcanbe 'expected,'henceitsetstheoccasionformoreresponding.Whenextinctionhasbegunandhereaches the endofhis fixed ratio, nopellet is discharged,yet the situationis thenormaloneforreinforcement.Consequently,hekeepsonrespondingrapidly because, in his history, reinforcement has come on such anoccasion.Each new responsemakes the occasion stillmore appropriate forfurther responding,with the result that the responsescontinue tocomeoutat a sustained maximal rate.This cannot go on indefinitely, because pasttraininghasgiventheresponseonlyalimitedstrengthwithwhichtoresistextinction.When thebreakcomes, itcomessuddenly.A temptinganalogyisthatwhichlikenstheresponsetendencytoareservoirthatisdrainingoffat full flush under high constant pressure right up to the point of totaldepletion.

In contrastwith this process, extinction following fixed-intervalP-Rorregular reinforcement shows little influence of a special "discriminativevalue" attached to the rate of response. The effect of reinforcement is

primarily to strengthen theindividual response, irrespective ofrate, while non-reinforcementdecreases that strength. In fixed-ratio reinforcement, each non-reinforcementaddstothe'occasion'for more responding and hastenstheappearanceofthenextresponse.These different consequences ofnon-reinforcement arise, of course,from the opposed trainingprocedures. Since reinforcement isconnectedwithdifferent features ofresponding,wecanexpectdifferentextinctioncharacteristics.

TheEffectofAperiodicReinforcement

Outside of the laboratory,regular reinforcement is by nomeans the rule, but neither isstrictlyperiodicreinforcement.Itishardly to be expected that aschedule of any fixed interval orany fixed number of responseswould be scrupulously honored byan environment so crowded withdifferent events.Wemaywell ask,t h e n ,whether the results ofaperiodic reinforcement are thesame as those ofperiodic orregularreinforcement.

Following Skinner's (1933)early studies of P-R with rats, anumberofotherscientists triedoutmodified forms of intermittentreinforcementwithotherorganismsand other responses. Brogden(1939a), using dogs as subjects,conditioned salivation, or foot-withdrawal from shock, withregular reinforcement.Subsequently, the animals weregiven training with aperiodic

FIG. 27. Cumulative responseextinction curves of two rats afterfixed-ratio P-R. (After Skinner,1938.)

reinforcement. He found thatresponse strength was maintainedat ahigh level evenwhennomorethan 40 per cent of the responses

werereinforced.Apparently theeffectof theaperiodicprocedurewasstrongenough to offset that of the relatively small number of reinforcementsprovided.

Humphreys' (1939) experiments in this field are of especial interestbecause he used human subjects. He studied the conditioning of an eye-wink response and, later, the galvanic skin response (change in electricalresistanceoftheskin).Inthefirstexperiment,alightofshortdurationwasfollowed, after 400 milliseconds, by a puff of air to the eyeball.Conditioning was assumed to be present when the eye-closure, originallyelicited by the air-puff, occurred regularly in response to the light and inadvanceofthepuff.Onegroupofsubjectswasregularlyreinforced,theair-puff following each presentation of the light on ninety-six occasions.Another group, given "partial reinforcement," received the air-puff after thelightonly50percentof the time—that is,on forty-eightof theninety-sixoccasions. In spite of this difference in the number of reinforcements, theconditioningwassetupasreadilyinonegroupasintheother.

In conditioning the galvanic skin response, Humphreys paired a tonewith a mild electric shock (the unconditioned stimulus for the change inskin resistance) under two reinforcement schedules. Members of a regular(100percent)reinforcementgroupweregivensixteencombinationsoftoneand shock: andmembers of an aperiodic (50 per cent) reinforcement groupwere given only eight combinations, randomly interspersed with eightpresentations of tone alone. Test trials, with tone alone, followed eachtraining period.Again, the results showed the two schedules to be aboutequallyeffectiveinmaintainingresponsestrength.

Extinction trials were given at the conclusion of each of theseexperimentsand,inbothcases,agreaterresistancetoextinctionwasshownbytheaperiodicallyreinforcedresponses.Theeffectwasmorestrikingwiththe conditioned eye-wink, but was clearly present with the galvanic skinresponse—afindingconfirmed ina laterstudy(Humphreys,1940).This isprobablywhat youwould expect, since aperiodicwould seem to resembleperiodic more than regular reinforcement, but these experiments do notentirelysatisfyourcuriosity.Oneofthemdealtexclusivelywithrespondentconditioning, and the other involved operant-respondent overlap. Neitherwas based upon the emission of straightforward operant responses such asmakeupthebulkofoureverydaybehavior.

Whenweturntotherelativelyunrestrictedor 'free'situationprovidedinthebar-pressingtypeofexperiment,wefindonlyafewpublishedstudiestodate (Humphreys, 1943; Mowrer and Jones, 1945; Jenkins and Clayton,1949; Skinner, 1950).The effect of such a reinforcement schedule can be

described, however, with some assurance. If bar-pressing is reinforced atintervals of four, three, two, and one minutes, together with somereinforcements that come immediately after a reinforcement (a 'zerointerval'),withtheseintervalsappearinginrandomorder,averysteadyrateof respondingwill quickly develop. Even aftermany days of training, thestraight-line character of cumulative-response curves will be maintained,sincenotimediscriminationispresenttogivetherecordsascallopedeffectlike that observed in fixed-interval or fixed-ratio P-R. Also, a greatresistance to extinction will be built up by such a procedure—possibly agreaterresistancethanthatwhichresultsfromP-Ratfixedintervalsorfixedratio.Theextinctioncurveitselfwillshowlittle,ifany,initialacceleration,and its course will be marked by only minor fluctuations in rate. Undercertainconditions,aswhen theorganismhasbeenaccustomed to relativelylong intervalsofnon-reinforcement inhis trainingsessions, therateduringa large part of the extinction curvewill be indistinguishable from the ratethatprecededextinction.

Casual observation suggests, of course, that the operant behavior ofhuman beings in their daily affairs is greatly affected by aperiodicreinforcement.Inveryfewspheresofhumanactivityisreinforcementeitherregular or strictly periodic, and, in certain cases, the effect of thisaperiodicity is dramatically impressive. The chronic gambler, whoseinfrequent winnings do not keep him from trying his luck again; the"punch-drunk" pugilist who stumbles into the ring long after his fisticpowerhaswaned; even theoft-disappointed farmerwho tills his landoncemore in the hope of a bumper crop—do not these all suggest thestrengtheningeffectofoccasional,unpredictablyspacedrewards?

WiththefactsofP-Rinhand,youshouldbeabletomakesomecriticaldeductions about educational procedures which strive to control behavior.Youshould, forexample, seehowonewouldgoabout teachingachild tobe persistent in the face of failure. One would make sure, in training forskill, for confidence at work, or for willingness to persist in socialactivities,thatthechildisguaranteedsomemeasureofsuccessandapproval—regularlyat first,but lateronlyoccasionally, so thathewillnotgiveupinthefaceofsetbacks.Inthecaseofemotion,whereTypeSreactionsplayan important role, you should be wary of using fear stimuli, like talk ofbogeymen,tomakethechildobedient.Onceconditioned,fearsaredifficultto extinguish; and great care must be taken thereafter to avoid evenaccidental reconditioning by stimuliwhichwould have gotten little noticeprior to conditioning.Also, if someunwantedbehavior in a child is tobeextinguished, even one surrender by the parent, in amoment of fatigue orembarrassment,mightrenewthebehaviorstrengthmorethanitwouldhaveifextinctionhadnotbeenundertaken in the firstplace.Toonewhowouldbreak a habit, we might paraphraseWilliam James and say: Never sufferreinforcements to occur during the extinction process, else you may

inordinatelystrengthentheveryresponseitwasyouraimtoweaken!

Superstition:AnExperimentalExampleSuperstitions may be considered as beliefs or practices based upon

assumed"if... then" relationswhichareeither falseorundemonstrable.If ablack cat crosses my path,then I shall have a bad day; if I take a turnaround the table,my luck at cardswill change; if there is a solar eclipse,disaster will befall the tribe; rain on this daymeans rain each day for thenextforty;anewborninfantmustnotbepraisedaloud,lestspiritsoverhearand harm him through jealousy. Cause-effect connections, or sequences ofevents, are thought to operate which have no basis in fact. In groping tounderstandthetroublesandjoysoflife,menofteninventexplanatoryforcesor causeswhich they hope can then be dealtwith by placation or counter-measures.Superstitionsarefrequentlybasedonanxietyaboutwhatadimly-graspednatural environment holds in store for us, and it is this emotionalquality which secures their unreasoning acceptance and unshakable grip.Somesuperstitionsarewidelyheld,asocialheritageofbeliefstaughtusbyour elders, but others are private convictions arising,we think, fromvalidpersonalexperiences.

One class of superstitions is that in which the supposed cause-effectrelation involves some act of our own as the agent in producing a givenenvironmental effect. This effect may be the procurement of a positivereinforcerortheavertingofanegative.IfIwishdownawell,mywishwillcome true; if I carry a rabbit's foot, no harm can befallme; if I pierce themanikin'sheart,myenemywilldie;ifIperformthisritualinthemorning,I insure against failure during the day. Such instances spring from thetemporal contiguity of a responseand a reinforcement in the individual'sown history or in his observation of others. As you are well aware, areinforcement will strengthenany act which precedes it, and this is trueevenwhenthecontingencyisaccidental.Thecorrelationmaybe 'false,'butthis will not side-track the conditioning. So far as the organism 'knows,'his reaction was indeed the effecting agent. With strengthening, theresponse appears more frequently, and the probability of anothercoincidental reinforcement increases. Thereafter, only occasionalcontingenciesareneededtokeepupthe'magical'responseatsomestrength.

This situation has been duplicated experimentally in a simple manner(Skinner, 1948a).The method of establishing a 'ritualistic superstition' isessentially that of periodic reinforcement, except that the reinforcement isnotmade to depend upon the emission of any arbitrarily selected operant.Into ahungrypigeon's cagea smallportionof food ispresented at fifteen-secondintervals.Thepresentationisgovernedautomaticallybyaclockand,once the mechanism has been started, requires no attention from theexperimenter,whomayevenretirefromthescene,leavingtheanimaltoitsowndevices.Anhourorsolater,ifhereturnstoobservetheresult,hewill

findsome interestingbehavior.Thebirdmaybewalkingsystematically incircles; hemay be tossing his head repeatedly, or swinging it from left torightinapendularmotion;hemaybemakingregularbrushingmovementswith his beak in the direction of the floor, or poking it into the uppercorners of the cage. He has been caught, as it were, by the accident ofreinforcement, supplied by an unwitting environment, at a chance-point inhis on-going behavior. He has acquired a 'personal superstition.' Theresponse is conditioned fasterwith shorter between-reinforcement intervals,when the chance of extinction and the emission of alternative responses islessened. Once the response is established, however, this interval may belengthened to, say, one or two minutes without changing the behavior.Moreover,agreatresistancetoextinctionmaydevelop.Inoneanimal,morethan10,000emissionsoftheresponse(aside-to-sidehop)occurredbeforeafifteen-minuteperiodofnon-respondingappeared.

PunishmentandExtinctionAlmost any discussion of the control of behavior eventually runs head-

on into thequestionof thepartplayedby "punishment."Atvarious timesand in various places, men have based their ideas of formal education,discipline, and social training upon the premise that punishment affectsbehavior inamanner theopposite to thatof reward. Inotherwords, it hasbeen supposedbymany to "stampout" behavior just as reward "stamps itin." On the other hand, throughout the years, isolated observers andthinkers have been skeptical about the long-term effectiveness of this kindof behavioral control.As a rule, thesemen did not deny that punishmentinfluenced behavior, but they questioned the permanence, as well as thedesirability,ofsuchinfluence.

Sigmund Freud, the psychoanalyst, was one whose clinical experienceled him to the latter view.He believed that one's early development is ofgreat importance in shaping his adult personality; and he argued that,during this development and later, many desires and acts (mainly sexual)mustbekeptfromovertexpressionifoneistobecomeasociallyacceptablehumanbeing.Overtexpression ispreventedbypunishmentor the threatofpunishment, but the wishes and tendencies toward the forbidden behaviorare not thereby done awaywith.They remain alive and forceful, althoughone may be unaware of their existence within him. Freud described thisbanishment of acts from overt expression as "repression" to the"subconscious."

Thorndike, on the basis of human learning experiments, also came todoubt the efficacy of punishment as a means of permanently weakeningbehavior. In 1931, hemodified his famous 'law of effect' so as to excludethe idea that annoyerswere the opposite of satisfiers in their effect uponresponse, asserting that "Annoyers do not act on learning, in general, byweakeningwhateverconnectiontheyfollow."Again,in1932,hestatedthat

"a satisfying after-effectwhich belongs to a connection can be relied on tostrengthen the connection... but there is no evidence that an annoyer takesaway strength... in any way comparable to the way in which a satisfyingafter-effectaddsstrength."

Neither of these men can be said to have solved the problem ofpunishment to the satisfaction of the laboratory scientist. Thorndike'sexperimentswereconcernedalmostentirelywiththeeffectofaspokenrightorwrong upon the strength of verbal connections such as those formedwhen one memorizes a series of word-number associations; and Freud'sview was part of an elaborate conceptual system which laid no claim toexperimental support—either in its origins or its implications. The netresultof their teachings, asyoumightguess,wasnot theblindacceptanceofdoctrine,butaquickeningandabroadeningoflaboratoryresearch;anditis out of such research that the true status of punishmentmaybe expectedtoemerge.

Wecanbestapproachthismatterherebyreferringbacktoourdiscussionof positive and negative reinforcement. It was pointed out, inChapter 3,thatcertainstimuli,suchasloudsoundsorelectricshocks,belongwithinaclass ofnegative reinforcers: when presented, they weaken operantbehavior;whenremoved, theystrengthenit;andtheyalsopossesselicitingpower. Let us examine, more closely than before, theweakening effect ofsuchstimuli.

Negativereinforcementcanbemadecontingentuponaresponse—thatis,theresponsecanbringiton.Thisisanimportanttypeof"punishment,"asthewordiscommonlyused.Thecontingencymaybeprovidedbynature(achild's finger is always burned when thrust into a flame) or through thebehavior of another organism (thehandmaybe slappedwhen the finger isplaced in the mouth). In either instance, the childis "punished," butpsychologists have, naturally, been more interested in the social than thenon-socialcase.

Punishmentmaybeapplied toanoperantunder twogeneralconditions:while the response is(a) undergoing positive reinforcement (regular orotherwise) to see if itwill nullify the effect of positive reinforcement; and(b)whenitisbeingextinguished,toseeifitwillsubtractfromthestrengthalreadypossessedby the response.Bothcasesare interesting,but the latterhasmorepractical and theoretical significance, since it is closely related tothe age-oldquestionof thepermanencyofpunishment's effects.Webegin,therefore, by asking whether punishment, applied to an operant duringextinction,willhavea short-ora long-term influenceupon the strengthoftheresponse.

This problem was first attacked, in an exploratory way, by Skinner(1938), inconnectionwithhisstudiesofoperantbehavior in thewhite rat.Two groups of animals were given periodic reconditioning for three days.Extinctionwasthencarriedoutintwoperiods,oftwohourseach.Withone

group, all responsesmade in the first tenminutes of the first periodwerepunished;with theothergroup,extinction tookplace in theusualway,noresponsesbeingpunished.Thepunishmentforthefirstgroupconsistedofaslapadministeredbythebaritselfwheneveritwasdepressed.Thisslapwasgiven to the animals' forepaws by a sharp return kick of the bar providedwhen an electric hammer struck against the bar shafts outside the responsechamber.Figure28showstheaverageextinctioncurvesforthetwogroups.The remarkable thing in these curves is that the effect of the slaps insuppressing the response was only temporary; the two groups eventuallycameout even in the total number of responses emitted during extinction.Apparently punishment did not subtract from the over-all strength of theresponse, and Skinner concluded that the experiment gave "no evidencewhatsoever for a process of negative conditioning directly the opposite ofpositiveconditioning."

FIG. 28. The effect of punishment on extinction. Experimenter'scomment:"Thetwocurvesarefromgroupsoffourratseach,withthesameexperimentalhistory.All responsesmadebyonegroupduring the first tenminutesofextinctionwereslapped.Therateisdepressedforsometimebuteventuallycompleterecoveryismade."(FromSkinner,1938.)

That more remained to be learned about the effects of punishment wasindicatedbyasecondexperimentinwhichfourrats,afterPRtraining,weregiven five extinction periods of one hour each. During the first two daysand forty minutes of the third, no responses were slapped. For the lasttwenty minutes of the third day, and the entire hour of the fourth, all

responses were slapped. On the fifth day, unpunished extinction wasresumed.Theaveragecurveforthefouranimalsisgivenin Figure29. It isplain that therewasno recovery from thepunishmentduring the fifthday.Itappearsthataprolongedperiodofslappinghadamorelastingeffectthanthe short one employed in the first experiment.We cannot say, however,thattheeffectwaspermanent,sincenofurtherextinctiontestsweregiven.

A recent extended and significant research dealingwith the punishmentof operant behavior is that of Estes (1944).This investigator, taking offfromSkinner'sobservations,concernedhimselfprimarilywith theeffectofelectric shock upon bar-pressing rate. In a long series of experiments, heshowed(1)that,aftertwoorthreeone-hoursessionsoffixed-intervalP-R,ashort period of mild punishment during the early part of extinction willdepress the response rate temporarily, but willnot decrease the number ofresponses required for subsequent complete extinction; (2) that a longperiodofmildpunishmentoraperiodofseverepunishment(eithershortorlong)willproduceanappreciabledecreaseinthenumber of later responses,butwillnotdecreasethetimerequiredforcompleteextinction;and(3)that,whenagreater amountofP-Rprecedes the extinctionperiodduringwhichpunishmentisgivenforeachresponse,therewillbeconsiderablylesseffectuponthenumberofresponsesfinallyemitted.

FIG. 29.The effect of extended punishment on the average extinctioncurve of four rats, the extinction being carried out after P-R at fixedintervals.(AfterSkinner,1938.)

Theseresultstellusthattheeffectofpunishingthebar-pressingresponseisnotpermanent,at least in termsof thetime required for extinctionwhenpunishment has been stopped; and that, if the response is strongly

conditioned in advance, it may not even reduce thenumber to be emittedlater.ButEstes'nextfindingisstillmoredramatic.(4)Theeffectofelectricshock was no greater when bar-pressing brought it on than when it wasdelivered independently of the response.When shock was given only attimes when the animal was not pressing the bar, there was the sameimmediate depression of rate and the same slow recovery as when theresponse itself was punished. In a word, the effect was nonspecific. ToEstes, this suggested that the important relationship was not between thepunishment and the response, but between the punishment and the generalsituationinwhichitoccurred.

According to this analysis, shock is an unconditioned stimulus for a"changed state of the organism," involving emotional respondent changesand accompanied by a decreased frequency of bar-pressing. (See thefollowingscheme.)

Since the shock is delivered in a "general situation"—the rat's cageenvironment—this situation becomes, for all practical purposes, aconditionedstimulusforthebehavioralchangesevokedbytheshockitself.Moreover, if this analysis is correct, it would follow that the merepresentationofthe 'conditionedsituation,'unaccompaniedbyshock,wouldleadtoextinctionoftheeffect.

In support of this hypothesis, Estes found that when bar-pressing waspunishedintheusualwayduringashortperiodofextinction,andthiswasfollowed by a two-hour period in which the rat was left in the responsechamber,withnobarpresent andno furtherpunishment, the effect of theshock was almost entirely absent when the bar was re-introduced andextinctionwasresumed.Leavingtherat in thechamberwithout thebar ledto a nearly complete dissipation of the emotional upset caused by theshock.Except for a smalldepressionat the start, the final extinctioncurvewas innodiscerniblewaydifferent fromthatprovidedbyanimals thathadneverbeenshocked.

The just-mentioned small remainder of shock-effectwas not ignored byEstes,whoexplained it in the followingway.Punishment hadbeengiventothebar-pressingresponse;anditseffecthadpresumablybeenconditionedt oall aspects ofthe stimulus situation which were present when theresponsewasmade.Extinctionof theeffecthadoccurred in thepresenceofmost of these conditioned stimuli, but not all.The bar itself was absent

during this extinction, and so was the response of pressing. Hence, thestimuli provided by the bar and the pressing movements themselves hadnot had a chance to lose their conditioned-stimulus status. Consequently,when the bar was returned to the chamber for the final extinction of theoperant, thebar-stimuli and the (proprioceptive) pressing-stimuliwere stillabletoexerciseasmallbutmeasurableinfluenceupontheresponserate.

Estes' "changed state of the organism" was, as he said, "of the sortcommonly called 'emotional,' " andwe shall havemore to say about thiseffect of punishment later,whenwe attack thegeneral problemof emotion(Chapter10).Already,however,wecanseesomeinterestingparallelsofhisrats' behavior, in the field of everyday human conduct. Consider, forexample, the not-uncommon experience of bringing down upon ourselves,through something we have said, the strong disapproval of others.(Disapproval itself may be negatively reinforcing—Chapter 8.) The ill-chosen words have been uttered and the damage is done; we are coveredwith embarrassment and confusion at the outcome.We are 'emotionallyupset' and we seek escape from our predicament (as the rat often tries toclimboutofhischamber)byonemeansoranother.Ifourattemptsfail,weremain to face the music, but for some time after thefaux pas, and evenwhen disapproval is no longer in evidence, our behaviormay be seriouslydepressed (just as the rat needs time to recover from his shock). If we doescape from the embarrassing situation, we need only to enter it again inorder to re-experience the upset that it caused; indeed, the merereinstatement of the fateful words, at some other time or in some othercircumstances,mayrearousetheoldemotioninwhatseemstobefullforce.Repeatedexposure to thesituationorutteranceof thepreviously 'punished'words will, in time, cause the disturbanceto vanish just as, in the rat'scase, repeated returns to the response chamber and further pressings of thebar will lead to an extinction of the effect and a recovery of operantstrength.

Themajor conclusion tobedrawn from these studies is, of course, thattheeffectsofpunishmentarelikelytobeimpermanent.Whentheemotionaldisturbance resulting therefromhas disappeared, the punished actmay stillneeda longperiodofextinctionbefore it iseradicated from theorganism'srepertory. If this is true,and if itapplies tomenaswellas rats,wecannotbutwonderwhy theuseofpunishmenthasbeensowidespread throughouttheages:whythistruthreceivedsolittlerecognition.Amoment'sthought,however, will suggest two answers. First, if one disregards the restrictiveand biologically uneconomical effects of punishment, he may use it indepressing"wrong"behaviorandtherebypavethewayforthestrengtheningof "right" behavior. This is a technique still commonly met with ineducational practice andoften supported by experimental studies in humanand animal learning (Bunch, 1928, 1935; Dodson, 1932;Warden, 1927;Gilbert, 1936, 1937; and others).Various researchers have shown that a

combination of reward and mild punishment will reduce the time or theerrorsinvolvedinthesolutionofproblems.

The second reason is not often mentioned: the use of punishment ispositively reinforcing to theuser.There is no gainsaying that punishmenthas its advantages in the control of others. Given the requisite physicalstrength, or the symbol thereof,we can always force others into a state ofsubmission—at least temporarily.Whateverannoyance theyhaveprovidedforus is therebyeliminated,andwe arepositively reinforced.On thebasisof this fact alone, it is easy to see why we live in a world wherepunishmentorthreatofpunishmentistherule;wherepowerorthesignsofpowerareconsideredall-importantinachievingsocialcontrol.

ConcludingRemarksTwo new principles, those of operant and respondent extinction, have

beenintroducedinthepresentchapter,andwehaveshownhowtheyoperatein conjunction with operant and respondentconditioning to build upreflexes of exceptional strength; extinction, as well as conditioning, isinvolved in the procedures of periodic and aperiodic reconditioning.Wehave alsopresented evidence to show that theweakening effect of negativereinforcement (punishment), when added to that of non-reinforcement, isprobably temporary. By now you should begin to appreciate the kind ofanalysisofbehavior thatweareattempting.Ouraim, throughout this text,istoshowhowthecomplexmaybeexplainedintermsofthesimple.Thiswill be even more apparent as we move along. Not many new principleswill be added in later chapters; but these will be required for yourunderstanding of a great many aspects of human nature which, for theaverageperson,areoftenshroudedindarkness.

NOTESRelated to the effect of different amounts of reinforcement upon

extinction (here) is the effect, reported by Jenkins and Clayton (1949), ofdifferent amounts upon the rate of response when aperiodically reinforced.Pigeons,trainedtopeckataresponsekeyforfood,weregiveneithertwoorfivesecondsofeatingtimeatirregularlyspacedintervals.Themeannumberof pecking responses in half-hour sessions was found to be 1205 for thetwo-second amount and 1557 for the five-second amount. Unpublishedobservations of white rats in the Columbia laboratory indicate that acomparable effectmay be obtained under aperiodic reinforcementwith oneversustwopelletsoffood.

In discussing the relation between response strength and the amount ofmotivationpresentduringconditioning,wementionedonlyStrass-burger's(1950) study, which we felt to be the most conclusive treatment of thismatter to date.We neglected, perhaps unfairly, to describe (1) the initialattack upon this problembyFinan (1940),who also used the bar-pressing

technique;(2)alaterstudybyMacDuff(1946),whousedthemaze;and(3)a still later study by Reynolds (1949) with the runway method. Finanfound that conditioning under twelve hours of food deprivation providedthe greatest response strength in his rats,with twenty-four-, forty-eight-,and one-hour deprivations decreasingly effective. MacDuff's order ofeffectiveness,frommosttoleast,wasforty-eight-,twenty-four-,andtwelve-hourdeprivations.Reynolds,whohas certainobjections toboth theFinanandMacDuff studies, concludes, from his own data and his predecessors',thatwecannotyetsaywhetherresponsestrengthis,orisnot,afunctionofmotivationalstrengthatthetimeofconditioning.

Characteristics of unconditioned or operant-level bar-pressing (here) byrats when motivated by hunger have been described by Hefferline (1950),and have also been investigated by Schoenfeld, Antonitis, and Bersh(1950b) who obtained measures of such activity under both hunger andthirst.

Inthetext,wesaidthatforgettingiscausedbywhathappensduring thepassage of time to interfere with, block, or extinguish the material-to-be-recalled.Amoredetailed statement, and some relevantdata,maybe foundinGuthrie(1935)whosetheoreticalviewpointisneverthelessdifferentfromtheonewehaveadopted.

In our discussion of aperiodic reconditioning, we cited experiments byBrogden(1939a)andHumphreys(1939)inwhichavoidanceresponseswereconditioned. Regular reinforcement was, in both cases, presumablycompared with aperiodic reinforcement in its effects.This is, as youmayhaverecognized,anover-simplificationofwhattookplace.Anticipatingourlater treatment of avoidance behavior (Chapter 9),wemay say that neitherthe shock nor the puff of air constituted the reinforcement in theseexperiments.Rather, the leg-flexionandtheeye-winkwerestrengthenedbytheremoval of stimuli which had been regularly or irregularly associatedwith the shock or the air-puff.The periodicity or aperiodicity involved insuch studies is of a different sort from that which is involved when bar-pressingisreinforcedwithfood.

Many investigations, not mentioned in this chapter, have been carriedout for the purpose of determining whether punishment is an aid or ahindrance in themasteryof complex learningproblems.Electric shockhasbeen the favorite punishing agent, and, for obvious reasons, animals havebeen theprincipal subjects.Mazesanddiscriminationboxes (here, Chapter5) have been the most popular forms of apparatus; and the shock hascommonly been applied in connection with an animal's choice-pointbehavior—for example, a rat may be punished before, after, or at themoment ofmaking a right or awrong turn in a visual-discrimination test(Muenzinger,etal.,1934,1935,1936,1938).Withrespect to thesolutionof such problems, it now appears that there is no single answer to thequestion of how punishment affects learning. This will be quite

understandable when you see that learning, in these situations, involvesmuch more than mere conditioning, either operant or respondent. Inaddition, and still more important, is the fact that such a stimulus aselectric shock may have any one of several functions. It may depress theresponse that it follows, or its removal maystrengthen—as you alreadyknow.But itmayalsobe informative, acting as a discriminative stimulusfor a specific response (seeChapter 5). Under certain conditions, it mayevenbepositivelyreinforcingwhenapplied(Chapter8).Anyorallof thesefunctions may be exercised in the maze or the discrimination box,depending upon the degree of shock employed and the kind of timerelationship with stimuli and responses that may be provided. This hasbeenmadeespeciallyclear in thestudiesconductedbyMuenzingerandhisco-workers.

The strongest opposition to the use of punishment in education comesfromThorndike's many studies of human learning.Thorndike argued thatthe effect ofWrong or other mild 'annoyers' is often to strengthen, ratherthanweaken,theresponsestheyfollow.Admittingthatintensepunishmentmight sometimes have a suppressive effect upon behavior, he pointed outthat itwasalwaysadangerousweapon,beingmosteffectivewithsensitivepersons, who need it least. At best, Thorndike argued, punishment issuccessfully applied onlywhen it leads one "to shift to the right behaviorandenjoy it,or tohavesuchexpectations fromthewrongbehavior thatheismore comfortable to avoid it than to enter upon it." (The full import ofthe last part of this quotation will be clearer when you have read thediscussionofavoidancebehaviorinChapter9ofthistext.)

5

GENERALIZATIONANDDISCRIMINATION

INSTANCES of this kind are so plentiful everywhere, that if I add onemore, it isonly for thepleasantoddnessof it. It isofayounggentleman,who,having learnt todance,and that togreatperfection, therehappened tostandanoldtrunkintheroomwherehelearnt.Theideaofthisremarkablepieceofhouseholdstuffhadsomixed itselfwith the turnsandstepsofallhisdances,thatthoughinthatchamberhecoulddanceexcellentlywell,yetit was only whilst the trunk was there; nor could he performwell in anyotherplace,unlessthatorsomesuchothertrunkhaditsduepositionintheroom.

JohnLocke,AnEssayConcerningHumanUnderstanding,1690

StimulusGeneralizationThischapterbeginswithasinglefactorcharacteristicofbehavior:When

an organism is conditioned to respond to one stimulus, itwill respond inthe same way to certain others.We call thisgeneralization and, as weproceed, we shall find that it helps us to explain a great deal of behaviorthatseems,atfirstglance,tobeverycomplicated.

The existence of stimulus generalization was discovered early in thehistory of psychology. Before 1910, Pavlov and his co-workers hadobserved it and reportedupon it in thesewords: "If a toneof 1000d.v. isestablished as a conditioned stimulus, many other tones spontaneouslyacquiresimilarproperties....Thesameisobservedwithstimulationofotherreceptor organs. This spontaneous development... we have termed...generalization of stimuli...".The fact is now so well established in bothrespondentandoperantbehaviorthatwemaystateitasaprinciple.

An increase or decrease in the strength of one reflex, throughreinforcement or extinction, is accompanied by a similar butsmaller increaseordecrease in the strengthof other reflexes thathavestimuluspropertiesincommonwiththefirst.

Alittlethoughtwillconvinceyouthatthisprincipleisofimportancetoanyorganismin itsdaily life.Ourenvironment is inperpetual flux,and itisveryunlikely thatanystimulusever recurs in identical form.Thevisualstimuli suppliedbya running rabbit to apursuing fox,orby the faceof afriend as you see it from moment to moment, are subject to countlessvariations in pattern, movement, brightness, and so forth, yet the foxcontinuesitschase,andyoudonotfeelyourselfconfrontedbyaprocessionofstrangers.Intheeverchangingenvironment,thegeneralizationofstimuli

givesstabilityandconsistencytoourbehavior.

StimulusDiscriminationOn the other hand, the behavior of organisms would be equally

ineffective and unadaptive if it could never get over the barrier ofgeneralization.Anorganismmustbeabletoresponddifferentlytodifferentobjects in its environment; and common observation tells us that it does.The fox pursues the rabbit, but not the hound; and we distinguish onefriend'sfacefromanother's.Behaviorcanshowaspecificitywithrespecttostimuli,andwhenthisspecificityisdevelopedinthefaceofgeneralization,wespeakofstimulusdiscrimination.

An organism manifests a discrimination when it comes to respond toone, but not to the other, of two previously generalizing stimuli.To becalleddiscriminative,theresponsetothefirstmustbemaintainedwhiletheresponse to the second isweakened. In the laboratoryand in everyday life,thedevelopmentof thisdifferencedependsupon the reinforcement, or lackof reinforcement, that attends responses.The basic principlemay, then, bestatedasfollows:

A reflex strengthened through generalization may be separatelyextinguished, so that an organism responds to one stimulus andfailstorespond,orrespondslessstrongly,toanother.

In contrast with generalization, the process of discrimination gives ourbehavioritsspecificity,variety,andflexibility.

FormingaRespondentDiscrimination"Generalization"and"discrimination"areanaturalpair,liketheopposite

polesofamagnet.Pavlov's(1927)wayofshowinghowdiscriminationsareformed is embodied in the "method of contrasts," which combines theprocedures of reinforcement and extinction. A hungry dog is firstconditioned to salivate to a 1000-cycle tone (or some other) by the usualType S technique. Such a conditioning, as we already know, will begeneralized—thedogwill salivate, to a lesser degree,whenother tones arepresented. One stimulus, say the 1000-cycle tone, is then chosen as'positive' (to be followed by reinforcement) and another, say the 900-cycletone,asNegative'(nevertobefollowedbyreinforcement).Thetwostimuliare thenpresented, inhaphazardorder,onmanyoccasions.Theoutcome isa difference in the strength of the two reflexes—in other words, adiscriminationisformed.

An indispensable part of this procedure of selective reinforcement or"contrasts"liesintherandomalternationofthestimulitobediscriminated.Toomany successive presentations of either tone alonewill not overcomegeneralization. Frequent interchange of 'positive' and 'negative' stimuli is

essentialifadiscriminativeresponseistobeestablished.The formation of a discrimination is, then, a double process.Through

generalization,eachdirectreinforcementofstimulusAaddstotheelicitingpowerof stimulusB; each extinctionof stimulusB subtracts a little fromthepowerofA.Further reinforcementsgivemorepower toA thanB; andfurther extinctions takemore fromB thanA.As the stimuli draw apart intheirpower,adiscriminationisinthemaking.Thegradualaccumulationofdifferencesinthestrengthofthetworeflexesistheheartoftheprocess.

GeneralizationandDiscriminationinOperantConditioning:The"DiscriminativeStimulus"

We now return to our representative operant, bar-pressing.You saw, inChapter3, that this responsewas emitted, not forced; and that its strengthcouldbe increased through reinforcement.You saw, too, in the caseof therunwayoperant, that a connection developedbetween the runningbehaviorand the opening of the starting-box door.This door-opening stimulationwas called "discriminative," rather than "eliciting," and we are now in apositiontomakethisdistinctionclearer.

Suppose,inthebar-pressingsituation,thatwetaketwostimuli,abrightlight and a dim light, and arrange an experiment inwhich they are turnedon alternately within the rat's response compartment. The rat has beenconditioned previously to press the bar for food. He is hungry again and,when placed in the box, begins to respond immediately. This time,however, theconditionsaredifferent.Althoughheisfree torespondatanytime,wereinforcetheresponseonlywhenthebrightlightison.

In such a case, neither stimuluselicits bar-pressing. The bright lightmerelysetstheoccasiononwhichreinforcementwillfollowif the responseis made; the dim light, on the other hand, provides the cue fornotresponding and sets the occasion for non-reinforcement. Each is adiscriminative stimulus.Adopting conventional notation, we may refer tothebright light, in thepresenceofwhichreinforcementoccurs,asSD (ess-dee) and thedim light, in thepresenceofwhich there is no reinforcement,asSΔ(ess-delta).

Letuscontinuetheprocedure.Periodically, thebright light(SD) comeson and stays on until the animal presses the bar.At the response, food isdeliveredandthelightgoesoff.Forfiveminutes,thelightisdim(SΔ) andnoresponseduringthattimeisreinforced.Then,SDappearsagain,thenextresponse is reinforced, and the cycle is repeated—for asmany times aswedesire.

As in the caseof respondents, generalizationoccurs.A reinforcement inthepresenceofSDincreasesitseffectivenessasacue,but italsoincreases,

to a lesser degree, the effectiveness of SΔ; an unreinforced response to SΔ

decreases its effectiveness and weakens slightly that of SD. Continuedalternation of SD and SΔ causes them to draw apart in their evocativepower.

FIG. 30.The formation of a light-dark discrimination by one animalfollowingtwosessionsofP-Ratfixedintervals.(AfterSkinner,1938.)

Here, again,extinction is the hallmark of discrimination–responding toSΔ extinguishes while responding to SD is maintained. A cumulativeresponse curve should therefore reveal, by its shape, the formation of thediscrimination. That is, it should resemble a curve of extinction. It isbecauseextinction,asopposed toconditioning, is the importantprocess indiscrimination,thatourexperimentprovidedsuchrelativelylongperiodsofSΔ.

Figure30 presents such a curve,made by a single rat, during a ten-dayperiodofexperimentation inwhich lightwas theSDanddarknesswas theSΔ. The discrimination, which appears as a relatively slow change in

response rate under SΔ, followed two days of periodic reconditioning (inthedark) at five-minute intervals.Onehour of trainingwasgivenper day,andapproximatelysixhourswererequiredfor theamountofresponding insuccessivefive-minuteperiodstoreachaminimalvalue.RespondingunderSΔneverdisappearsentirely,andtheSDresponsesareofcourseincludedinthe cumulation;hence, the curvenever reaches anoperant level comparabletothatofasimpleextinctioncurve.

TheEffectofPreviousTrainingDoes the length of time needed to form a discrimination depend upon

the amount or the kind of an organism's previous training? In answeringthisquestion,our lineof reasoningwouldbeas follows:Adiscriminationinvolves extinction; resistance to extinction varies with the kind andamountof training,hence the timerequiredforadiscrimination todevelopwill depend upon previous training. InFigure 31, we have an averagediscrimination curve for four rats. Before the discrimination began, theseanimals had received fifty regularreinforcements, followed by a period ofextinction.Thestrengthoftheresponsewasthereforemuchlessthaninthecase of the rat (seeFigure 30) that had previously received two days ofperiodic reinforcement when the discrimination training was introduced.Figure 31 shows that the discriminative process for this group waspracticallycompletebytheendofthesecondday.

FIG. 31.Average cumulative response curve for four rats showing the

formationofalight-darkdiscriminationfollowing50regularreinforcementsandoneperiodofextinction.(AfterSkinner,1938.)

An interesting result,not soeasilypredicted,ariseswhenweattempt toteach a discriminationwithoutany previous training in bar-pressing.Here,onthefirstoccasionwhenthebarisaccessible,theSDispresent.Thefirstresponse is reinforced, theSDis replacedbySΔ for a five-minute interval,and thecycle is then repeated.Asamplecurve fora singleanimal isgiveninFigure32. It shows that the generalization to SΔ is almostnil, so thatthereisnothingtoextinguish.Weconcludethat,undersuitableconditions,a discrimination may be formed immediately.The basic requirements arethatneither reflex is strengthened inadvanceof the trainingprocedure, andthatthefirstreinforcementsoccurinthepresenceofSD.

FIG. 32. The formation of a discrimination without previousconditioning. The vertical strokes over the cumulative response curveindicatereinforcementsinthepresenceoftheSD.(AfterSkinner,1938.)

AHumanDiscriminationAnalyzedWhen human rather than animal subjects are taught to make

discriminations,itisnotalwayseasytoidentifythechangesthatoccur,butan experiment by Hilgard, Campbell, and Sears (1938) offers convincingevidence that the same fundamental processes are involved in each case.Avisual stimulus (a light in the left one of two small windows) waspresented on sixty occasions to fourteen subjects, each presentation beingfollowedby a puff of air to the eyeball (the eliciting stimulus for the eye-wink reflex).After this training,when the subjectswere responding to thelight in advance of the air-puff about 75 per cent of the time, anotherstimulus (a light in the right-handwindow)was introduced and randomlyalternatedwith the first during sixty trials of discrimination training.Theleft-handlightwaspositive(alwaysfollowedbytheair-puff)andtheright-handlightwasnegative(neverfollowedbytheair-puff).

Figure33 shows the change in strength of the response to positive andnegativestimuliduringthediscriminationprocedure.Youwillnotethatthe

FIG. 33. Progress in forming adiscrimination by human subjects.The conditioned response employedwas the eye-wink; the stimuli to bediscriminated were visual. (AfterHilgard,Campbell,andSears,1938.)

responsestoSDandSΔdrawapartgradually in their frequency, andthat the major change is one ofextinction—aweakeningunderSΔ.Except for a slight drop in the SDcurve at the very beginning of thediscrimination trials (possibly dueto a generalization from theextinction of response to SΔ), theeye-wink occurs with the samefrequency as at the end of theconditioningtrials.

AbolishingaDiscriminationIf we wish to abolish a

discrimination, we must restoreresponding to SΔ.Two proceduresare available. (1)Wemay keepSΔpresent continuously and reinforceeveryresponse,thus reconditioningthe extinguished response. (2)Wemay return to periodicreconditioning at the same intervalemployed in discriminationtrainingbutwithouttheuseofSD.This kind of abolishment gets theresponse back to its original P-Rrate.

ExtinguishingaDiscriminativeOperantAbolition of a discrimination is different from extinguishing the

discriminative response. When we abolish, SD and SΔ are equalizedthroughreinforcementofboth(responsesunderSΔarereconditioned);whenwe extinguish, the equalization of SD and SΔ is accomplished by theweakening of the response toSD.To illustrate the latter, suppose thatwehave a well-formed discrimination, in which SD regularly evokes theresponse and SΔ seldom does. The response to SD is, then, ourdiscriminative response. It may be extinguished, like any other, by

withholding all reinforcement in the presence of SD. Butwhat kind of anextinctioncurvewouldyou lookfor?Youmightexpect,at first thought,agreatresistanceoftheresponsetoextinction,sinceSDrespondinghasbeenreinforced, say, at five-minute intervals, and this is like the procedure ofperiodic reconditioning.Onsecond thought,youwill see that the responsehasalways been reinforced under SD—the only non-reinforcement wasunderSΔ.Thecurveof extinction toSD should therefore resemble that ofextinctionafterregularreinforcement—whichitdoes.

TheMeaningof"Similarity"British philosophers of the seventeenth and eighteenth centuries made

muchof two "laws" throughwhich "ideas"were associated:contiguity andsimilarity. The influence of these philosophers is still apparent in oureveryday speech, aswhenwe say that one idea calls up another because oftheir likeness or because theywere once associated in time or place, but amodernobjectivepsychologvdoesnot find the conceptof the "associationof ideas"veryuseful.Wedofind,however, that acontiguityofstimuli, orofstimulus and response, is essential toType S conditioning or operantdiscrimination.Canwegiveanobjectivemeaningto"similarity''aswell?

If youwill pause to consider thematter, youwill see that "similarity"and generalization are the same thing. In everyday affairs, we talk as ifstimulicouldbesimilarinthemselves,butactuallytheirsimilaritydependsuponourownbehavior;theyaresimilarwhen,andonlywhen,wemakethesamesortofresponse to them.Similaritydoesnotreside instimulialone,anymorethanin"ideas."

Stimuli, it is true, may have common physical properties and, in aphysicalsense,are therefore"similar."Butwhenpeoplesay that thingsaresimilartheymeanthattheytendtoreacttotheminthesameway.Theyarereally reporting this tendency with the wordsThey are similar. This isquite different from the physical similarity which is often, though notnecessarily,presentwhentheresponsesaresimilar.

RelatedtothispointisanexperimentbyPlotkin(1943)onthelearningof International Morse Code by college students. Before their instructionbegan,thesestudentswerepresentedwithpairsofcodesignalsandaskedtorate them on the degree of their "similarity." Later, during the actual codetraining, he found that the confusions of signals with each other(generalizations) were directly related to their previously estimated"similarity."Forexample,thesignalsforthelettersC(—.—.)andY(—.——),whichwerecommonlygeneralizedduringtraining,hadalreadybeenjudgedtobeverymuch"alike";whereas,thesignalsforA(.—)andO(———),whichhadbeen rated as dissimilar,werenever confusedwith eachother. The identity of "similarity" and generalization was quite clearly

indicated, and givesweight toHull's (1943) statement that "the common-sense notion of similarity and difference is based upon the presence orabsenceof...generalization."

When two or more stimuli are found to generalize, so that a responseconditionedtooneismadetotheothersofthegroup,wemayofcoursetryto identify the property that they have in common. It is sometimesdifficult,orevenimpossible,todothis.Stimuliareknowntoacquiretheirfunctional equivalence from several sources. We may, for example,generalize two visual stimuli on the basis of wave-length, energy, size,shape,orlocation.Also,iftheyarecompositestimuli,theymaygeneralizethroughthe identityof theirparts.Apparently, too, theymaygeneralizeonthebasisof theemotionalorotherresponses that theyarousewithinus(asin the caseof "mediated"generalization, to bementioned later).These andother factors play a part, singly or in combination, in determining thedegree towhich stimuli areequivalent inevoking response.Youmaywellimagine that theproblemof identifying theminanygivencase isone thatchallengesthebesteffortsofspecialistsinthefieldofdiscrimination.

The concept of generalization has been extended to more complexproblemsofhumanbehavior.Psychiatristsandclinicalpsychologistsoftenencounter striking instances of respondent generalization, in whichemotional upsets originally connected with a single event in a patient'shistorycometobeevokedbystimulicommonlymetinthedailyroutineofliving.Inonecase,aphobiathatpreventedawomanfromlookingintotheeyesofanyone,whetherfriendorstranger,wastracedtoasingleexperience.Some years before, when surreptitiously opening the drawer of a chest inthe home of a blind aunt, shewas confrontedwith a pair of staring glasseyes.Thepanicinducedlaterbythesightofanyeyeballstemmedfromthis"trivial"but"similar"incident.

Up to this point we have emphasized thestrengthening action ofgeneralization. But the opposite effect may also occur. Experiments showthat reflexes, strengthened through generalization, may be weakened whenthe originally conditioned reflex is extinguished. Also, if a number ofdifferentstimuli,sayatone,atouch,andasound,areattachedtothesameresponse through conditioning, the extinction of any one of them willdecreasethestrengthoftherest.

Into the details of such experiments we need not go, but it may bementioned that psychiatrists, again, have apparently utilized this principle.Anumberof their curative techniques, aswhen theyprobe repeatedlywithverbal stimuli into the lives of their patients, clearly provide for theextinctionofemotionalrespondents.Throughgeneralization, theemotionalresponses to theseupsettingstimuliwhenencounteredoutsideof theclinicare also extinguished to somedegree in spite of obviousdifferences in thestimulussituation.

GeneralizationGradientsGeneralizing stimuli may be classed in terms of the sense-organs

involved—visual stimuli, forexample, involve the functionof receptors intheeye.Withineachof these 'sensedepartments,' stimulimayformseries,like the series of pitches, colors, loudnesses, brightnesses, warmths, anddistances apart on the skin surface.Now, ifwe inquire how, after any onemember of a series has been made a conditioned stimulus, the effects ofgeneralizationspreadoutovertherest,wediscoveraninterestingfact:thereis agradation of effect, depending upon the physical proximity of thestimulitotheoneconditioned.

Pavlov(1927)notedthisearlyinhisstudiesofgeneralization:"Ifatoneof 1000 d.v. is established as a conditioned stimulus, many other tonesspontaneously acquire similar properties,such properties diminishingproportionally to the intervals of these tones from the one of 1000 d.v."(The italicizing is our own.) This gradation effect has led to the termgeneralizationgradient.

Inaddition toagradient forstrengthening,Pavlovalso reportedone forextinction,obtainedfromthefollowingexperiment.Alongthehindlegofadog, five spots were selected, the first at the paw and the others spacedthree, nine, fifteen, and twenty-two centimeters from the first. By directreinforcement, the four spots were equalized in their effectiveness asconditioned stimuli for salivation, while the pawspot was extinguished.When this state was reached, the paw spot was given three moreunreinforced stimulations, and the other four spots were separately tested.The generalized extinction was greatest for the three-centimeter spot anddecreasedprogressivelytothetwenty-two-centimeterspot.

FIG. 34.Average generalization gradient for tonal pitch obtained from10 human subjects. The conditioned response was the galvanic skinresponse, labeledPGR(psychogalvanic reflex)on theordinate; theUSwaselectric shock.The responsewas firstconditioned toa toneof1967c.p.s.,and generalization was tested to tones of 1000, 468, and 153 c.p.s.Theabscissa is logarithmic inscale tokeep thefiguredowninsize. (DatafromHovland,1937a.)

Both conditioning and extinction generalization gradients have beenfound for human respondents byAmerican investigators (Bass and Hull,1934;Hovland,1937a).Moreover,gradientsfordiscriminativestimulihavebeen demonstrated in studies of operant behavior. Frick (1948) put five

groups of rats on a discriminative training schedule in which SD and SΔ

were alternately present for two and a halfminutes each. SD and SΔwerelightsofdifferentbrightness,andeachgroupofanimalswastrainedwithadifferentpair,asshownbelow:

FIG. 35.Average cumulative curves for SΔ responding by the severalgroupsinFrick'sexperiment.ThecurveforSΔislinearanddoesnotdifferfrom the P-R rate before discrimination training began (i.e., the differencebetweenSDandSΔwasnotdiscriminable).Theseparationofthecurvesisan indication of the varying degrees of generalization. The curves arecompositesofanumberofexperimentalperiods.(AfterFrick,1948.)

Frick's expectation was that the ease of forming the discriminationwould depend on the difference between SD and SΔ; and his results bore

FIG. 36. Average generalizationgradient of tactual stimuli varying indistance from a conditioned spot onthe skin of human subjects. (AfterGibson,1939.)

out the hypothesis. Both therapidity with which responding toSΔ extinguished and thecompletenessoftheextinctionwereincreased as the difference betweenSD and SΔ grew larger. SimilarresultshavebeensecuredbyRaben(1949), who used the Graham-Gagné runway and a techniquedevelopedbyVerplanck(1942).

Evidence for generalizationgradientswith human subjects anddiscriminative stimuli wascollected, in 1939, by Gibson. Inone experiment, the SD was avibratory stimulus, applied to thesubject's back. After a verbalresponse (a spoken nonsensesyllable,dut) had been connectedwith an SD at one spot on theskin,thevibrationwaspresentedatspotsfour,eight,andtwelveinchesdistant in a straight line down thesubject's back. Each subject wasinstructed to respond with thesyllableonly when the stimuluswas felt at the initial position.Generalization was measured interms of the percentage ofdutresponsestothestimulusateachofthe other positions. The averagepercentage values obtained for thefour points were, respectively,ninety-eight (at the 'conditioned'spot), twenty-five, fourteen, andnine, indicating a steep butcontinuousgradient(FIG.36).

TheStudyofDiscriminativeCapacity

A proper understanding ofdiscrimination must develop along two lines: first, an analysis of the

processofdiscriminationandthefactorsinfluencingthisprocess;secondly,the investigation of the discriminativecapacities of organisms.We havealreadyconsidered theformer,andwenowfocusourattentionbrieflyuponthe latter. Again we shall deal with matters of general validity, notattemptingtomakeaninventoryofallthesensorycapacitiesthroughouttheanimal scale. or to catalogue the changes in capacity accompanying theindividual development of an organism. Our emphasis will be upon thebehavioral aspects of the problem, the experimental methods used, and afewofthemoresignificantconcepts.

1 .Capacity and the threshold. It is possible to alter an organism'senvironmentinawaythathasnoeffectuponitsbehavior.Itiseasytopickaphysicalchangetowhichanorganismcannotbebroughttorespondatall,orwecanchoosetwostimuliwhichitwillnotdiscriminatenomatterhowlongwecontinueselective reinforcement.Thus,neitherahumanbeingnoradogwouldeverrespondtoa touchontheskinasfaintasadustparticle,norcouldtheyeverdiscriminateatoneof1000cyclesfromoneof1000.01cycles.Theproblemofcapacityarisesfromtheseelementaryfacts.

The measurement of sensory capacity reduces to two determinations,bothofwhichwerementionedinChapter1:a. The smallest value of a stimulus to which a response can be made.

Such would be the softest sound, the faintest light, the lowest pitch, thelightest touch. In discriminative terms, it would be the least stimuluswhich canbediscriminated fromnoneatall, so that a response cannot beconditionedtoalesservalue.AsnotedinChapter1,theseleastvalueshavebeencalled"absolutethresholds"or"absolutelimens."b. The leastdifference between two stimuli, each abovethe absolute

threshold, thatcanbediscriminated.Twostimulimaybesoclose togetherthat, behaviorally, they are identical; the difference between them mustreach or exceed a certain value before a response to one but not the othermay be established. An extension of the metaphor of "threshold" lednaturally to the designation "difference threshold" or "difference limen."Bothtypesofthresholdarestatedintermsofsomephysicalmeasureofthestimulus,whether energy, frequency, pressure, temperature, or some other.Stimulus values may be thought of as "supra-liminal," just "liminal," or"subliminal"forbothtypesofthreshold.Nevertheless, itshouldbekeptinmind that the meaning of threshold lies in an organism'sbehavior.Wedefine in physical terms those stimuli which are or are not adequate fordiscriminative responding. The behavioral data are always prior andnecessary to the concept of capacity; and stimuli which are notdiscriminated even after arduous training areheld to be below capacitylevel.Wespeakasif stimulicanorcannotbediscriminatedbecauseof thelimitations of capacity, but we intend that capacity be defined by thebehavioralevidenceofstimulusdiscriminability.

2 .The 'psychophysical' methods. Historically, the study of human

sensory capacities was the first to get under way in psychology with anadequatesetofmethods.By1860,GustavFechner,anoutstandingGermanscientist of his day, had formulated several experimental procedureswhicharestillthebackboneofhumansensoryresearch.Designedtotestthelimitsof discriminative capacity, these methods aim to determine both absoluteand difference thresholds. Fechner himself thought of them as ameans ofdetermining the relation between "mental sensations" and the "physicalworld" (hence "psychophysics"), butwe see them today, not as a solutionto a philosophical problem, but as important contributions to themeasurementofsensorycapacities.

Thepsychophysicalmethodsdiffer in their complexity, their usefulnessunder given circumstances, and in the typeof threshold they measure.BecauseFechner realized thatasubject'ssensitivityvaries frommoment tomoment,themethodseachprovideformeasurementstoberepeatedasoftenasdesiredinordertosecurereliableaverageestimatesofthreshold.Theyarealike, too, in presenting to a subject either single or paired stimuli.Thesubject is asked to report whether he can perceive the stimulus at all, orwhetherhecan tell that twostimuliaredifferent.Repeatedstimuluspresentationsandstatisticaltreatmentofresponsesfinallyyieldanestimateofhisthresholdand theconsistencyofhis responses,bothofwhicharemeasuresofhisaccuracyorcapacity.

The psychophysical methods are unquestionably endorsed by commonsense. They are characteristically human in their reliance upon verbalbehavior.They do not ask how such behavior is acquired and howwordscometobediscriminativeresponses.Nothingseemsmorereasonableonthesurfacethantoaskapersonwhetherhecanorcannotperceiveastimulusora stimulus difference. In this case, however, as in many others ofunanalyzed human intercourse, common sense does not reveal the actualcomplexity of the experiment or the tacit assumptions upon which weproceed.Thevery feasibilityof themethods, inwhich 'instructions' to thesubjectareindispensable,dependsuponalonghistoryofverbalbehaviorinboth experimenter and subject which circumvents the need for training anew discriminative response such as would be necessary with lowerorganisms. The former "asks" the latter to "observe," "perceive," "payattention," "judge," "cooperate," and so on; and the subject's onlyreinforcementis"socialapproval."Truly,thereismaterialforanalysishere.Yet, despite their naivete, the psychophysical methods give a very finemeasureofhumansenseacuity.Therelianceuponverbalbehaviorisamplyjustifiedbythelawfulnessanddependabilityoftheresults.

With human or animal subjects, and with respondent or operantmethods, threshold determinations are subject to the influence of manyimportantvariables.Theseexperimentalsituationsincludefarmorethanthestimuli to be discriminated or the receptor organ being tested.Variableswithintheorganismandintheexperimentalsituationmaycountheavilyin

determining results. Hence, it is always premature to conclude that theabsoluteordifference thresholdobtained isa finaland immutablevalue.Afailure to discriminate with onemethod does not mean that the organismwould fail with a different method. Human subjects, for example, mayimprove their thresholds when given information on their performanceduring the experiment. Only recently, too, the use of a conditionedrespondenthasrevealedforthefirsttimetheexistenceofcolorvisionintherabbit,whenothermethods had failed (Brown, 1936). Such instanceswillundoubtedlyoccurinthefuture.

3 .Some examples of human acuity. Studies of human receptorsensitivityhave turnedup somedramatic facts.Our sensesare reallymuchmoreacutethanwecommonlyimagine.Indeed,theyfrequentlyprovetobemoreresponsive tosmallenvironmentalenergiesandchanges thanmostofthe modern physical instruments which we regard with awe. Under goodexperimentalconditions,apersonwithnormaleyescantelltwocolorsapartwhentheydifferbyas littleas twoor threemillimicrons inwave-length(amillimicron is a millionth of a millimeter); the brightest light whichpermits ordinary good vision may be 2.00 x 108 times as bright as theabsolute thresholdvalue; andawire3/32of an inch thickcanbe seenat adistanceofhalfamile.

Hearing is initiatedby the impactof soundwavesagainst theear-drum.It has been computed that the faintest sound we can hear arises fromimpacts which do not greatly exceed in force the casual collisions of airmolecules against the drum. With a little more sensitivity, we couldactually "hear" the molecular movement of the air. As for pitch, somesubjects can,with100per cent certainty, discriminate two toneswhich, atthe200-cyclelevel,areseparatedbyonlyone-eighthtoone-halfofacycle.

The sense of smell is also extremely acute, despite popularmisconceptions. One sniff suffices to tell the presence of a substance likemercaptanwhen the amount ofmaterial involved ismuch less than canbepicked up with a spectroscope. For odorous gases injected into the nasalpassages, a rise in pressure of a fewmillimeters of mercury is enough tomake some subjects unfailingly distinguish the smell where before theywerecompletelyinsensitive.

Insomeofoursensing,weareexcelledby loweranimals, like thedog;in others, we stack up quite well against our biological contemporaries.This is true, however, only for acuitiesmeasured under rigid experimentalcontrols. In everyday life, our senses function under crude conditions andwedonotusuallyhaveoccasion to fallbackupon theirextremecapacities.Forthesereasons,wedonotlearnourfullpotentialitiesuntilthelaboratorytellsusitsstory.

4.Othermethods.Since the capacitiesof infra-humanorganisms cannotbe gauged byway of verbal responses,we resort to the basic procedure offormingadiscrimination.Forexample,usingarespondentlikethesalivary

reflex,webeginwith twovisual stimuli and themethodof contrasts.The"positive" stimulus is a luminous circle, the "negative" an ellipse. Theellipsehasaminoraxisequaltothecircle'sdiameter,butthemajoraxisistwice that length. The discrimination between the two stimuli is easilyformed.Then,bysmall steps,webring theellipsenearerandnearer to theshapeof thecircle,reinforcingregularly theresponseto the latterandneverreinforcing the response to the former.At some point in the progression,the major-minor axis ratio of the ellipse will be such that the animal nolongerrespondstoitanydifferentlythantothecircle—that is,wecometohisdifferencethreshold.Pavlov(1927)foundthisratiotobeabout9:8.Bya similar procedure of selective reinforcement, we may determine theabsolute threshold (e.g., for a sound) by seeing how small a stimulusmagnitudecanbemadeintoaconditionedstimulus.

The basic experiment foroperant discrimination may also be used indetermining thresholds.The difference limen is obtained by bringing SD

andSΔtothepointwhere,withaslightlysmallerdifferencebetweenthem,the cumulative response curve shows exactly the same properties as theordinary P-R curve. That is, extinction to SΔ does not occur and thereinforcementofonlySD responses sets up a periodic cycle.The absolutethreshold is determinedwhenwe find the least value of SD which, whenpairedwithanSΔofzeromagnitude,stillpermitsadiscrimination—or,toturnitaround,thegreatestvalueofSDwhich,withSΔequaltozero,givesonly a P-R curve. (This method, as well as the preceding one forrespondents,tellsushowthedifficultyofformingadiscriminationchangesasthestimuliapproachliminalvalues.)

FIG. 37. One of Elder's subjects, earphone adjusted over left ear, inposition for tests.The experimenter, who is behind the screen, raises theinclined sliding door at the ape's right, exposing the telegraph key.Whenthe key is correctly pressed by the subject, his reward is deliveredautomaticallythroughthechuteattheleft.(AfterElder,1934.)

Variations of the operant procedure have been used in studying thediscriminativecapacitiesoforganismshigherinthescalethantherat.Elder(1934),forexample,determinedtheabsolutethresholdfortonalintensityinseveral chimpanzees by having them press a telegraph key whenever the"positive" stimulus sounded in their ear-phones. A "ready signal" (theopeningofadoorwhichgaveaccess to thekey)precededexposures toSD

and SΔ (silence). The SD was presented until a response occurred, afterwhichthedoortothereactionkeywasclosedandtheapewasgivenabitoffruit; SΔ was presented for short periods only—eight seconds of silencefollowing the signal, or until the subject responded "falsely"—afterwhichthe door to the key was closed again.An equal number of SD and SΔpresentationswereprovided, inrandomorder,andthresholddeterminationswere based on the percentage of correct responses at different intensitylevels.The threshold values for seven tones, ranging in frequency between128and8192cyclesper second, showed that,on thewhole, theapeswereequalintheirsensitivitytohumanbeingsofacomparableagegroup.

Amorecomplicatedformofoperant techniquecomesunder theheading

FIG. 38. The Lashley jumpingapparatus.s—animal's stand;n—net;f.p.—food platform. (Reprinted fromCrafts, Schneirla, Robinson, andGi l b er t ,Recent experiments inpsychology,as redrawnfromLashley,1930. Copyright, 1938. Courtesy ofMcGraw-HillBookCompany.)

of "choice-reaction." There areseveralvariationsofthistechnique,but none is intended to give, nordoes it give, a picture of thediscriminativeprocess.Thestimulito be distinguished aresimultaneously presented, and oneobserves merely the all-or-noneevent on each trial inwhich theresponse is made either to SD orSΔ. A growing preponderance ofresponses to SD as against SΔindicates an emergingdiscrimination. The technique isadequate to show whether adiscriminationispossible,hencetosupply a measure of threshold.Two rather widely employedexamples of the choice-reactionmethod are the "jumpingapparatus" (Lashley, 1930) and the"T-box" (Yerkes and Watson,1911). In the first, a hungry rat istaught to jump from a ledgethrough either of two identicalloose curtains, behind which itlands on a platform and isreinforced with food. Afterconditioning, the curtains arechanged so that they bear differentstimuli, such as a circleversus asquare. A jump to the arbitrarily'correct' stimulus carries himthrough the curtain to food; an'incorrect' jumpdasheshimagainsta door which is closed behind thecurtain and drops him into a net.The stimuli may be equated in allrespects (brightness, color, etc.)save the oneto be discriminated

(shape), and are frequently interchanged in position to avoid thestrengthening of response to irrelevant cues or the growth of a 'positionhabit.'Themethod could, of course, be usedwithout employing negative

reinforcement for the 'wrong' response but simply omitting positivereinforcement.

FIG. 39.An apparatus for teaching a visual discrimination.The animalstarts atS; d is the choice point; g andg' are electric grids;a anda' aredoors to the food alleysf andf' ;L is the light box; the stimuli to bediscriminated are the differently illuminated round windows. (Reprintedfrom Crafts, Schneirla, Robinson, and Gilbert,Recent experiments inpsychology, as redrawn fromLashley,1929.Copyright, 1938.CourtesyofMcGraw-HillBookCompany.)

TheT-boxisessentiallyabifurcatingalleyinwhichtheanimalstartsatthe foot of the shaft,moves down, and turns right or left into one of twoend-boxes,onebeingcorrectinthatitcontainsfood,andtheotherincorrectinthatitleadstonoreinforcementortonegativereinforcementlikeelectricshock.The correct turn is signalled to the animal by the SD in use, and,again, care is taken to avoid mere position habits.As in the case of thejumping technique, somecriterionof successive correct choices is takenasthemeasureofdiscrimination.

Both the T-box and the jumping technique may be employed instudying the effect of various factors (brain operations, drug injections,motivation, past history, etc.) upondiscriminative capacity, but theT-box

hashadthegreaterusefulness.Itpermitsthetestingofmoresensefunctionsandawidervarietyofanimals.Inaddition, itmaybenotedthat theT-boxis essentially a one-choice maze (a simpleT-maze) and is suitable for atleast the exploratory investigation of still other influences upon operantbehavior.

Discriminationand"ExperimentalNeurosis"Much attention has been paid in recent years to aberrations of behavior

whichappearwhenanimalsareforcedtotheirdiscriminativelimits.Pavlov(1927)observed, in theellipseexperimentmentionedabove, thatwhenhisdog reached the 9:8 ratio of axes, an attempt to push this discriminationcompletely disrupted the dog's behavior. He became violent, bit theapparatus, whined, and barked; in the investigator's opinion, he presented"all the symptomsof acuteneurosis."OneofPavlov's students obtained asimilareffect in a pitch-threshold experiment, a visiting psychiatristagreeingthat theanimalwas"neurotic."These induceddisorganizationsarenowcalled"experimentalneurosis,"andhavesincebeenestablishedinrats,cats,sheep,pigs,andevenhumanbeings.Theypointtoacloserintegrationofpsychopathologywiththetheoriesandlawsofgeneralbehavior.

Experimental neurosismay be obtained in both respondent and operantdiscriminative conditioning. In the latter, the disruption of behavior isaccentuated if, in addition to going unreinforced, the responses to SΔ arepunished. Forcing a response in the jumping apparatus by goading theanimal with a blast of air or pushing him off the ledge, may have thisresult. When negative is combined with positive reinforcement in thisfashion, theamountof"conflict" increasesas the threshold isnearedor thedifficulty is increased.There is no choice but to respond, and a "wrong"responseiscostlyintermsofpunishment.

The symptoms of experimental neurosis include refusal to work,excitement,cowering,disturbancesofbreathingandheartbeat, irregularitiesofactivityandrestcycles, rigidity, trembling,convulsions,andspasmodicmuscular twitchings ('tics'). Oftimes, the animal will seem normal whenoutsidetheexperimentalsituation,buttheabnormalitieseruptassoonasitis put back. In some instances, the disorganization is carried over into theanimal's routine environment (Lid-dell, 1938), and may slowly disappearwith prolonged vacation, only to emerge with full force when theexperiment is resumed. Once set up, the neurosismay not only affect thediscriminationofthresholdstimuli,butanevencruderdiscrimination(e.g.,anellipseratioof2:1)maybedisturbed.

Human beings cannot, for obvious social reasons, be subjected toextreme experimental neurosis. In Pavlov's laboratory, however, smalldegrees of upset were produced in a child of six by requiring adiscriminative motor response to metronome beats. Rates of 144 and 92

beats per minute were easily discriminated, but after narrowing down thedifferenceto144and120perminute, the childbecame surly, disobedient,and uncooperative. Other children, organically impaired by encephalitis orcretinism,were alsobrought to thepoint ofmildbehavioral disruptionbybeingrequiredtomakeoverlyfinediscriminations.Suchmanifestationsarenot unlike those sometimes observed when human adults are repeatedlycompelled to attempt delicate discriminations in psychophysicalexperimentsor in certainoccupational training.Studentsof radio code, forexample, have been known to show deviations from normal behaviorwhich, in extreme cases, warranted a psychiatric classification of "codeneurosis."

It is still too early to judge the implications of this line of research forthe whole range of human maladjustments. To many psychologists, itseems a good beginning.The precipitating causes of human abnormalityoftenappearsimilarinprincipletotheexperimentalobservations.Theworkis being actively pressed inmany laboratories, by psychiatrists as well aspsychologists. Phobias, compulsions, anxieties, and other long-recognizeddisordersarefastbecomingtheconcernoftheexperimentalistaswellasthetherapist and the clinical investigator.We may look forward with someoptimism to a day of mutual aid and understanding between workers inthesetwohistoricallyunrelatedfields.

DiscriminativeReactionTimeIn our study of discrimination, we are interested in how a response

depends upon stimulus conditions. When we come to analyze thisdependency, severalmeasures of response are possible. For example, therei sresponse frequency, in the now familiar bar-pressing situation, whenanimalsareinvolved.Inpsychophysicalexperiments,weusethefrequency-of-detection of stimuli or stimulus differences.Latency of response affordsanothermeasure—onewhichhashistoricallybeencalledreactiontime.

How fast canaperson react to adiscriminative stimulus?This questionis ordinarily answered in the laboratory by the following procedure.Weinstruct a subject to respond as quickly as he can; we give him a 'readysignal' so thathecanget 'set'; andwedeliver the stimulus.Wemayusealight, a sound, a touch, or any other stimulus we wish; and the subject'sresponsemaybepressinga telegraphkey,callingoutaword,oranyotheroperantthatwecanconvenientlyrecord.Thetimebetweenthestimulusandtheresponseishis"reactiontime."

The study of reaction timewas not launched by psychologists. Duringthe nineteenth century, before the advent of modern registering devices,astronomerswere vitally concernedwith thismatter. In attempting to notetheexactmomentatwhicha starpassed themeridian, theyhad toobserveits movement and respond immediately when it crossed a hairline in theobjectiveofa telescope.Over thespaceofmanyyears, theystruggledwith

the problem of cutting down discrepancies in the observations of differentpersons;and theybequeathed tous theexpression,"personalequation," forindividual differences in reaction time among various observers.Also, H.L. F von Helmholtz, a brilliant investigator in physics, physiology, andother fields, attempted, in 1850, to determine the speed of nerve-impulseconduction in human beings, with a reaction-time method modeled afterone he had employed on a nerve-muscle preparation excised from a frog'sleg.A subject's skin was stimulated at points differently distant from thebrain, and to each stimulus hewas instructed to respond as quickly as hecould.Bydividing thedifference inestimated lengthofnervefor twoskinspots by the difference in the two reaction times, Helmholtz arrived at anapproximatespeedofconduction.Hisfinalfigureofaboutsixtymeterspersecond was surprisingly close to currently accepted speeds (about sixty-eightmeters per second), althoughhis datawere sovariable that hewouldnottrustthem.

It was not until 1868, however, that Donders, another physiologist,pointed out that severaltypes of reaction time might be investigated. Hesaw that increasingly complex stimulus and response alternatives shouldhavetheireffectuponreactiontime,andhesetdownwhathethoughtwerethethreetypicalexperiments.

1 .TheA-reaction .The subject makes only one specified response, saypressing a key, and a single stimulus is used each time.This is the so-calledsimple reaction time. It has been used as an index of the effect ofdrugsandfatigue; incomparingresponsespeedswithstimulibelonging tothe different senses; in determining the importance of stimulus intensity;andunderotherexperimentalconditions.

2 .The B-reaction. Two stimuli are employed, each with its ownappropriateresponse.Forexample,asubjectmaybeaskedtorespondwithhis right handwhen shown the color green, andwith his leftwhen shownred. On successive trials, the stimuli are interchanged at random. Thesubject, on eachoccasion, discriminateswhich is beingpresented, reactingwith the appropriate hand. This is sometimes called the "disjunctive"reaction time, because of its either-or character—that is, the subjectrespondsinonewayortheother.

3.TheC-reaction.This is anotherdisjunctive reaction time, sometimesknownasthe"discrimination"reactiontime.Twostimuliareused,andthesubjectmustdiscriminateby responding toone,butnot to theother.Thisis essentially the discriminative situation that we described earlier inconnectionwithbar-pressingandElder's (1934)studyofauditoryacuity inthechimpanzee.

If you think this over, you will probably conclude that the order ofincreasing complexity inDonders' types is notA-BC, butA-C-B.TheC-reaction adds a discriminative factor to theA-reaction; and the B-reactionadds a response factor to the C-reaction.The average reaction-time values

for the three cases come out as you might expect: A is fastest, C isintermediate,andBisslowest.

Other investigators went further. It seemed reasonable that two thingsshould hold. Inmodern terminology,wewould say that (1) reaction timemight be used as a measure of degree of stimulus generalization; and (2)reaction time should increase with the number of different responsesbrought into play. The correctness of both assumptions has been wellestablished.Inoneseriesofstudies,Henmon(1906)studiedtheB-reaction,usingpairsofcolors, tones,and lines.Membersof thepairsweredifferentinvaryingdegree.Onepairoflines,forexample,mightbenearlythesamein length, whereas another might be quite different. The subject wasinstructed to respond to a specified color or line of the pairwith the righthandifthestimuluswasontheright,andwiththelefthandifthestimuluswas on the left.With the tones, a right-hand reactionwas called for if thesecond of the two tones was higher in pitch, and a left-hand reaction iflower.Thekindofresultsobtainedaretypifiedbythoseshownbelow.

The importanceof increasing thenumberof responses (better, stimulus-response units) has been demonstrated in several ways. One experimenter(Merkel, 1885) used a separate stimulus and reaction key for each finger.ThestimuliwerefiveArabicandfiveRomannumerals,visuallypresented,and each was associated with the reaction of a specific finger. In thesimplest situation, only one stimulus and only one fingermovement wasemployed; in the most complex situation any of the ten stimuli mightappear and the appropriate reaction had to be made.The average reactiontimesinthetendifferentcaseswereasfollows:

Merkel's experiment, and possibly Henmon's, have a bearing uponanother problem, that ofresponse generalization ("induction"), to bediscussed in the following chapter. The evidence is clear, however, thatdiscriminative difficulty, arising from stimulus generalization, is reflectedinreactiontime.Thegreaterthegeneralization,theslowerthereactiontime.

ReactionTimeandOperantLatencyEvery reaction time is an operant latency. More specifically, it is the

minimallatencyofanoperant response toadiscriminative stimulus.As inthe case of threshold determinations, the measurement of minimalhumanlatencies is enormously facilitated by previous conditioning to verbaldiscriminative stimuli.The subject is instructed, he understands, and hecooperatesbydoinghisbest.Hehaslongbeenreinforcedforrapidresponsewhen such requests or their equivalents form a part of the discriminative-stimulus compound. It is even possible that, finding himself in alaboratory, face to facewith an experimenter, and so on, hemay not evenrequire"instructions" torespondwithallspeed.He"setshimself,"wesay,forthetask;he"assumes"thatitisthequickresponsethatiswanted.

The animal usually behaves otherwise. Reduced latencies are ordinarilyobtained only after considerable training. But the basic procedure is thesame. The reinforcement of a response is made contingent upon theoccurrence of that responsewithin a given time after SD is presented.Westartwiththeanimal'snormalrangeoflatenciesandgethimtospeedupbywithholding reinforcement for the slowest responses. By graduallyeliminating the longer latencies through extinction, while continuing to

reinforce the shorter ones,we ultimately reach a latency that isminimal—beyondwhich selective reinforcement is no longer effective.We have thendetermined his reaction time for this SD.The process may have requiredseveralhours,distributedoverseveraldays,whereasahumanreaction timecould have been determined in a single experimental session, but theoutcome is the same; in each case we have achieved a minimal latencythroughselectivereinforcement.

LatencyandtheDiscriminativeProcess

The normal operant latency of response to an SD in the bar-pressingsituation may fluctuate considerably from one stimulus presentation toanother, depending at least inpart upon the animal's location andongoingactivityat the timeofstimulusonset.Onemight,however,expect that, informingadiscrimination,therewouldbeachangeinlatencyofresponsetoSD and SΔ, the former decreasing, the latter increasing, as trainingprogresses. If this were so, we would have another way of analyzing theprocessbesidesthatprovidedbythecumulativeresponsecurve.

Unfortunately, the findings on this score have not been very helpful.Whileachange in latenciesdoesoccur in thebar-pressingexperiment, it isquite small, it comes in very early, and it stops long before thediscrimination is firmly established. In other studies, where a runningoperant is used, the situation is apparently different. In teaching adiscrimination with the runway apparatus, the end-box contains areinforcementonlyifanSD(e.g.,lightovertherunway)ispresent,andSΔ

(nolight)runsarenotreinforced.OnsuccessiveSDtrials,theanimalstartsout more and more quickly after the release door is opened; whereasinterspersed SΔ trialsgive progressively slower starting times, increasedvacillation, and more instances in which the animal does not leave thestartingboxatall.Thetechniqueisatrial-by-trialaffair, likethatoftheT-box and the jumping apparatus, with no free operant to be observed inprocessofchange;but,ifwetreatstartingtimesasoperantlatencies,wedoobserve a trendof timeswhich is somehow related to the formationof thediscrimination. The problem of bringing a latency measure into closerharmonywitharatemeasureisyettobesettled.

WorkadayReactionTimesIn everydaybehavior,most of our responses arenot emittedunder such

favorable conditions as those of the laboratory; and the demand for high-speedreactions ismadeonlyoccasionally,as inathletics,militarycombat,and the control of suchmachines as the airplane and the automobile. Forseveral reasons, the values reached in human reaction-time experiments are

seldom approximated, even in these pursuits.Warning signals are oftenlacking; one may not know from what direction the stimulus is coming;one is usually engaged in doing something else when the response issuddenly demanded; the response may involve the action of large, ratherthansmall,musclegroups;thestimulusmaybeveryweak,orsostrongastocause"freezing";andsoon.Thus,agroupoffootballplayersinuniformmayaverageasmuchas400millisecondsingettingoffascrimmagelineatan auditory signal (Miles, 1931); and the reaction time of automobiledriversmay rise to several secondswhen an impending accident requires ashiftofthefootfromgaspedaltobrake.

Individual differences in reaction time have long been recognized, andattempts have often been made to use these differences as a basis forselectingmentobetrainedinspecialskills.Ifonecouldpredictthesuccessof radio operators or fighter-plane pilots from their performance on a shortspeed-of-response test, then one would pick only the fastest men forschooling and thereby save the effort and expense of training those menwhowouldultimately"washout"orbecomecombatcasualties.Whenusedfor such purposes, however, reaction-time tests have proved of indifferentvalue, probably because other factors (discriminative capacity, motivation,adequate training procedures, etc.) are also important for success in theseoccupations.

MultipleDiscriminationsAlthough an analysis of discrimination necessarily begins with simple

cases, our behavior is usually guided by stimulus combinations that arequite complex. Many SD's may sometimes operate insuccession and wemaybeunabletorespondtoallofthemproperly.Atothertimes,theymayoperatesimultaneouslyandourresponseistothecompoundratherthananysingle element.Again, a responsemaybemade tooneor agroupofSD'swhich are constantly changing. In any or all of these cases, the basicprocesses of discrimination remain as we have outlined them, but theirworkingoutoftenpresentsdifficultproblems.

1 .Successive discrete SD's. We can illustrate the first type ofdiscriminative complexity bywhat is encounteredwhen one tries learningtoreceiveInternationalMorseCode.Aseveryoneknows,thiscodeconsistsof short and long sounds ("dots" and "dashes") which are combined invariousways to formdistinctivepatterns.Abeginnerneeds to learn thirty-sixofthesepatternsorsignals—twenty-sixforthealphabet,plustenforthenumbers0-9.Hisjobistobecomeabletodiscriminateeachwellenoughsothathecanwritedown,or"copy,"theappropriatelettersornumbersasthesignalsaresoundedoneatatimeandinrandomorder.Atfirst,manyofthesignals soundalike, and the student puts down thewrong character.Theseare obviously cases of generalization.A table of errorsmade for a class of

beginners in code will show that these generalizations fall within certaincategoriesdependinguponthedot-dashcompositionofthesignals(Spragg,1943; Keller and Schoenfeld, 1944).The problem for the learner is verylargelyadiscriminativeone,sincehisresponses—thewritingorprintingofletters and numbers—are usually well established ("differentiated") longbeforehehearscode.

Anexampleof thesortofgeneralizationthatoccurs incodemasteryhasalreadybeengivenearlierinthischapterwhenweconsideredtheconceptof"similarity."InTableVbelowaregivenseveralother instancesof thekindofconfusionthatmakestroubleforbeginners.Thesearetypical,butnottheonly ones that occur. Most of the signals actually generalize withseveralothers.Thus, the signal for Pmay generalize not onlywith the one for J,butalsowiththeonesforF,L,C,(—.—.),Q,X,andperhapsahalfdozenmoreintheearlystagesoftraining.Moreover,thisgeneralizationisreducedby training less readily for some signals than for others.Long afterW,P,andF are discriminated, 5 and6maygive trouble by generalizingwithHand B respectively, 4 (....—) will continue to generalize withV (...—),andsoon.

TableVGENERALIZATIONSCOMMONLYMADEBYBEGINNERSINLEARNINGTORECEIVEINTERNATIONALMORSECODE

Aswithalldiscriminations, the speedof learningcodecanbe increasedby making reinforcement as immediate as possible for the correctdiscriminative response. (For more on delayed reinforcement,208-209).This fact isutilizedin the "code-voice"methodof teachingnowemployedbytheU.S.ArmySignalCorps(TM11-459).Aftereachsignalissounded,there is a pause of two or three seconds during which the student writesdown, if he can, the character that was represented.Then the character is

identified vocally by the instructor and the next signal is presented.Eventually, under this form of tuition, the student accurately anticipateseach announcement, and the use of the reinforcing "voice" may bediscontinued.Figure40showsthesortofprogress thatmaybeexpectedofcollegestudentswhengivenonehourdailyofthiskindoftraining.

FIG. 40. Progress curve for a group of beginning students learning toreceive International Morse Code by the "code-voice" method. The dataplotted are the average per cent correct responses on successive training"runs"of100randomizedsignalseach.Priortothefirstrun,the36signalswere identifiedonce,and this isone reason thecurvedoesnot startatzeropercentcorrect.(AfterKeller,Christo,andSchoenfeld,1946.)

2.Compound SD's.This category of multiple discriminations includessome cases which seem so natural to us thatwe do not often think to

inquire about them, and other cases which are the basis of some amusingand unusual effects. Let us consider, as our first example, that ofdepthperception.We accept unquestioningly the fact that we can see objects indepthor, better, thatwe react appropriately to the third-dimensional aspectof objects. Under ordinary circumstances, we reach for common objectswithgreataccuracyandassurance,andwejudge,withconsiderablesuccess,thedistanceofthosebeyondourgrasp.Wepickupthepencilorbookwithasurereach;weunhesitatinglyaverthatthehouseisnearertous,orfartheraway, than the barn; and we even estimate, in feet or yards, the distancefrom 'here' to 'there.' Moreover, objects themselves appear solid.All thisdespitethewell-knownfactthatimagesofobjectsupontheretinaoftheeyeare,asinaphotograph,inonlytwodimensions.

Thehistoryofresearchtellsusthatmanycuescontributetotheadequacyof theseadjustments, andhelpsus toappreciate theextremecomplexityoftheSD'sweemploy.Forconvenience,wemaysummarizethesecuesundertwo headings: those which are effective even when only one eye is used,andthosewhichdependupontheoperationofbotheyessimultaneously.

1.Monocularcues includesuchSD's as arise from(a) theinterpositionof objects (the nearer object hides, in part, the object that is farther away);(b)size andperspective (the farawayobject is smaller than theonenear-by,and the continuous change inobject sizewithdistance is perspective); and(c) thedistribution of light and shadow (concavity and convexity, whicharethird-dimensionalcharacteristicsofobjects,commonlydependuponthiscue).These and one or two others have long been recognized by painterswhosoughttorepresentdepthuponcanvas.

2.Animportantbinocular cue is that provided by the fact that our twoeyes,beingapart,cannotbestimulatedinexactlythesamewaybyasinglesolid object.The depth provided by this disparity may be demonstrateddramaticallywith a stereoscope.This device, a fixture in the old-fashionedparlor,permits twoslightlydifferentbi-dimensionalviewstostimulate thetwoeyesseparately,witharesultanttri-dimensionaleffectthatmaybeverystriking.'Stereoscopicvision'providesuswithoneofourmostsubtlecues.

Notallofthesecues(andwehaveomittedseveralothers)needtooperateat one time to guide our adaptive reactions, although many of them arecommonly available. Numerous experiments have shown that one or twomayfunction,oftenveryeffectively, in theabsenceof the rest.The lossofaneye,whichautomaticallyeliminatesallbinocularcues,doesnot therebyrender the loser helpless; it merely makes him more dependent upon thecuesthatremain.

Some of the depth SD's are apparently learned in the earlymovementsandexperiencesofthechild.Ashereachesandcrawls,forexample,hesoonlearns the importanceof interposition,perspective, and themovement cuesprovided by themuscles of the eyes. Some are so obscure that, evenwith

coaching, they are not easily appreciated. Yet the effectiveness of ourmovements inspace isclearlydependentupon them,and thesemovementsare continually being reinforced by their outcome.After awhile, we thinknothingof reachingforobjectsandfinding thatourhandhasgone just theright distance and in just the right direction to make the contact; or wethrowaballwithconsiderableaccuracywithoutonce stopping toattend tothediscriminativebasisoftheact.

Another instanceof theoperationof compoundSD's is seen in the factthatagivenobject,underchangingstimulusconditions,seemstoretainitspropercharacter.Amanstandingonthesidewalkacrossthestreetseemsastall as a man should be, although his image on the retina of your eye ismuch smaller than it would be if he were standing close by. Thisphenomenon is referred to as "size constancy"; but other object propertiesthan size lead you to behave in a 'constant'way. In a similar fashion, youcan judge with great accuracy theshape of an object despite the changingstimulation it supplies as it moves about in your environment.You haveno difficulty in seeing a dish as round, although its projection on yourretina is usually elliptical.Thebrightness of objects provides still anotherexample.Yourhandkerchieflookswhitewhetheryouseeitinthesunlight,intheclassroom,orinadimcorridor.Theimportantpointaboutallformsof constancy is simply that, in responding to the physical properties ofobjects,yourresponsesaretheproductofthediscriminativestimuliarisingnot only from that object but also from asmuch of the environment as isavailable. If you were asked to judge the size of a line of light in acompletelydarkroom,youwouldfindthetaskdifficult—youwouldasktoknow,oryouwouldtrytoguess,itsdistancefromyoubeforeyouventuredanestimate.Understrictexperimentalcontrol,thedishcanbemadetolooklike an ellipse, by cutting off all cues of its tilt with respect to otherenvironmentalobjectsandwithrespecttoyourownperson.Andbrightnessconstancy can be destroyed by a similar reduction in the number of cues(such as prevailing illumination of the environment) upon which youordinarily depend, so that the handkerchief will look black under specialconditions.

"Redintegration" is another case in point. It is easy to show thatwhenSD compounds are the basis of a response it is possible to obtain thatresponse to aportion of the compound. The recognition of a work ofpictorial art may be based on only one detail, or the identification of asymphonymay bemadewith only a few notes carrying a simplemelodicline.Redintegrationisacaseofgeneralizationthroughpartialidentity.Itisnot,aswasoncethought,aseparateprincipleofdiscrimination,butistobethoughtofintermsofgeneralizationandelementsofcompoundedSD's.

3 .Changing SD's. Our third category of discriminative complexity isthat inwhichSD'sare inacontinuousstateofchange,andafewexampleswill suffice to show its prevalence.To a tennis player, a bounding ball

givesaninfiniteandcontinuousseriesofstimuli,buthisresponsetothemmustbe sure and quick. To the sportsman who brings down a duck orshattersaclaypigeon,or theskilledmachinistwhofollows themovementof thecutting toolonhis lathe, theproblemofdiscrimination isagainoneofchange.Forafighterpilotorgunner, thesameis true.Adevicecalledapursuit-meter has been employed in selection tests for gunners and pilots.This apparatus tests the accuracy with which a candidate can follow amoving target through an erratic course. In general, men who do notperformwellwith such an apparatus are likely to fail in flight or gunneryschool, hencemay be eliminated in advance in order to save expense anddisappointment.

"HigherUnits"inPerception

In the laboratory,we seek to isolate theSD'swhich enter into complexdiscriminations like the above, although we know that in ordinaryexperience they are not so isolated. Butwewould also like to know howthey become integrated. In learningMorse code, for example, the studentprogresses to a pointwherehe is no longer responding to discrete signals,but is hearing anumberof them together.An indicationof this is the factthat, with plain-language transmissions, he can "copy behind" withconsiderableease,theextentofhislagbeinganindicationofhowmuchhecan grasp at one time—that is, how big hisunit is.A burst of signalssounds to him like aword and he does not pause to break it up intoseparate letters. In early studiesof code learning (Bryan andHarter, 1899),itwasobserved thatbeforeastudentcouldpassbeyondacertainword-per-minute proficiency he might spend a good deal of time during which hemade no apparent progress. If he persisted, he could resume his advance,this time going on to a new level of mastery in which he could handlegroups of signals or short words. Periods of no progress were called"plateaus" and it was at first thought that these were necessary stages inwhich the learner consolidated smaller units or elements into higher units.Todaywe believe that such intervalsofnoprogress canbe largelyavoidedby carefully and systematically combining signals andgivinghimpracticeon higher units, rather than allowing the student to form themhimself onthe basis of accidental combination and by dint of mere perseverance.Higherperceptualunitsinvisioncanalsobestudiedinconnectionwiththeso-calledspanofapprehension.Flashforone-tenthofasecondanumberofletters on a screen before a student who has been instructed to call themback. If he is tried on one, two, three, ormore,we can quickly determinethe largest number that he will apprehend without error, and how hisaccuracy diminisheswhenwe exceed this number. Supposewe find that agiven subject can unfailingly report correctly asmany as six letters. Ifwechangeoverandflashshortwordson thescreen,wefind thathecanreport

about as many words as he did letters. If these words were three-letterwords,thenheisreportingeighteenletterswhereashecouldreportonlysixbefore. Obviously, in the case of the words, the letters are no longerfunctioningasseparateelements,but this time thewordsareelements.Wehave here another instance of the combination, through training, of simpleSD'sintocompoundsorhigherunits.

ConceptFormationWhat is a "concept"? This is another term which has come into

psychology from popular speech, carrying with it many differentconnotations.Weshallhavetobecarefulinusingit,rememberingthatitisonlyanameforakindofbehavior.Strictly speaking,onedoesnothave aconcept, just as one does nothave extinction—rather, one demonstratesconceptualbehavior,byacting inacertainway.Ouranalysis should reallystart with a different question:What type of behavior is it that we call"conceptual"?And theanswer is thatwhenagroupofobjectsget thesameresponse, when they form a class the members of which are reacted tosimilarly,wespeakofaconcept.Achild'sconceptof"horse"maybesuchthathisfirstsightofacow,amule,oracamelmayallresultinhissaying"Giddap" or "Whoa," or simply "Horsie."A group of eventsmay also beresponded to in the same way and thus form a concept, such as "war."Classes of objects or events, differently responded to, develop differentconcepts. "But," you may say, "this is only generalization anddiscrimination all over again"—and so it is. Generalizationwithin classesanddiscriminationbetweenclasses—thisistheessenceofconcepts.

1.Thegrowthofconcepts. It is importantthatwemaintainanobjectiveattitude toward concepts, that we see them in terms of behavior.A goodwaytodothisistonotehowsomeconceptsdevelopinhumanchildren.Atbirth,theworldofthechildmaybe,asWilliamJamessaid,nomorethana"booming,buzzingconfusion,"butverysoonherespondsindifferentwaystodifferentpartsofhisenvironment.Forexample,attheageofaboutthreemonths, he seems to show the rudiments of a 'social smile'—that is, hesmiles at other human beings.We like to think, at this stage, that he'knows us,' and parents fondly believe that they alone can evoke theexpression.Atestwillquicklyshow,however,thatthechildisrespondingtothemovingornoise-makingaspectsofparentalbehavior;movinganimalsandrattlingtoyswillalsosetoffthereaction.Wemay,ifwewish,thinkofthisasaprimitiveconceptualdistinctionbetweenmovingandnon-movingobjects. Later on, the child may reserve his smile for members of thefamily, even crying at the approach of strangers, but many months oflearningmustprecedetheadvanceofhisdiscriminationtothisstage.

An interesting set of studies byPiaget (1929) traced the changes in thechild's concept of livingversus non-living objects. This was done by

asking children of various ages what they thought was alive and why,whether objects feel hurt when kicked or become sad when locked in acloset, and so forth. It turned out, as you would expect, that the criteriaused for defining living and non-living objects change withage and theaccumulationofexperience.Atonetimeallmovingobjectsarecalledalive.Lateron,adistinctionismadebetweenthingsthatmovewhenpushedandthings thatmove 'by themselves' (brooks, clouds, trees in thewind).Onlygradually, and never completely, do they approach in their conceptualbehavior the 'realism' so highly prized by modern man. Since our ownmovementsof locomotion, respiration,speech,and the likeareperhaps thefirst,andcertainlythemostimportant,tooccasiontheresponse"living,"itisnotstrangethatwefindthebrook 'running,' thewind 'sighing,' thetrees'whispering,' and the table 'groaning' under its load of viands. 'Animism,'whichisopposedto 'realism'andwhichissaidtoinvolvetheascriptionofthepropertiesofanimate things to those thatare inanimate, is readily seento be another case of generalization. Rigorous discriminative training isrequired before a concept of living objects can be formed which excludesthisprimitiveextension.

Through education, formal and informal, our concepts are altered andenlarged.Where oncewe thought that trout, eels, sharks, andwhaleswereall "fish," we learn in school that these organisms fall into differentcategories, and our concept of "fish" is radically changed. Similarly, ourconcept of airplane once included heavier-than-air craft with propellers,wings,engines,fuselage,andtail.Todaywearemodifyingthiscategorytoincludepilotless planes that are jet-propelled andwithout the conventionalengine.Apparently the presence of wings is the indispensable part of ourconcept of airplane, since wingless ships, like rockets, do not evoke thename. In these cases, and in others, the difficulty of classifying oftenbecomes acute. For example,we knowwhat "fighting" is, butwemay beunabletosaywhetherpuppies,engagedinrough-and-tumble,are"fighting"or"playing."

2 .Experiments on concept formation. In the laboratory, studies ofconcept formation fall roughly within two major classifications. As anexample of the first, we have an experiment by Hull (1920) in whichsubjects were taught to respond with nonsense words(li, ta, yer, etc.) todifferent Chinese characters. Six lists, of twelve characters each, werememorized successively by each subject.The seventy-two characters weredifferent in total composition,but eachof the twelvecharactersof the firstlisthadacomponent(a 'radical')whichwaspresent inoneof thecharactersof each later list, and the same responsewordwas requiredwhenever thisradical appeared.Thus, the word li was applied to six different characters(oneineachlist),allofwhichpossessedthesameradical.Thepointoftheexperiment was to see whether the learning of successive lists would befacilitatedasthesubjectwasgivenmoreandmoreexperiencewiththebasic

radicals.Theresultsshowedthattheprocessofmasterywasindeedspeededupaseachnewlistwaslearned.Thesubjectscameto'formconcepts'inthesense that they generalized on the basis of the key radicals in the differentcharacters.Moreover, some subjectswhowere ablewithvery fewerrors torespond to the characters of the sixth list were unable, when asked, toidentify the radicals to which they were correctly responding. Evidently,this kind of conceptual behavior may be established without a subject'sbeingabletostateinwordsjustwhatheisdoing.Thisismostinformativewhenweconsider that incommonspeechwenormally tend to identify the'possession' of a concept with one's ability to verbalize it, even if onlyapproximately.

In the secondmajor typeof experiment on concept formation, a subjectmay be shown a number of objects andinstructed to find the commoncharacteristicwhichestablishesacategory.Thus,Smoke(1932)studiedtheformation of ten concepts each of which concerned a certain type ofgeometricaldesignhavinganonsensename.A"pog," forexample, alwayscontained a circle within a rectangle, although a series of "pogs" mightdifferinthesizeorsomeotheraspectofthecircleorrectangle.Thesubjectexaminedone"pog"afteranotherwhentheywerepresentedandadvisedtheexperimenterwhen hewas ready to define the class.Hewasthen asked tostatewhata "pog"was, todraw twoexamples,and to select froma listofsixteen test figures thosewhich fulfilled the "pog" requirements.Althoughthe ease of the generalizations depended upon the kind of geometricaldesign employed, the subjects were usually able to single out thosecontainingthesamebasiccomponents.Moreover,inagreementwithHull'sfindings, theywere sometimes unable todefine a class satisfactorilywhentheyhadalready'passed'theothertests.

3.Non-verbalizedconcepts.Inviewofthefactthatasubjectmaybehaveconceptuallywithoutbeingabletotellusthebasisofhisdiscriminationorgeneralization, it is natural to ask whether concepts are peculiarly human.From observation and theoretical analysis, we are led tentatively toconclude that lower animals exhibit fundamentally the same behavior.Consider a hunting dog, being trained to pursue rabbits. Early in histraining he may fail to discriminate adequately and find himselfembarrassed in an encounter with a polecat or a hedgehog.When we say,later, that he is properly trained, we mean, among other things, that hisconcept of rabbit will exclude these other creatures. In the laboratory, wemay train animals, aswell as young children, in concept formation,whenthere is no possibility of verbal definition.A child, for example, may betaught to reach toward a triangular figure for a piece of candy, whilereachingtoacirclegoesunrewarded.Later,ifwesubstitutedifferentshapesof triangles and different sizes of circles, his response remains correct.Animals may also be taught to generalize on the basis of triangularity(Fields, 1932). We approached this matter before, in our treatment of

"similarity" and stimulus equivalence, and nowwe can see that equivalentstimuliiswhatwemeanwhenwespeakofaconcept.

The"transposition"experimentprovidesuswithanotherinstanceofthissort (Spence, 1936; Jackson, 1939). A child or an animal is trained toapproach the brighter of two discs. After thorough conditioning, thebrighterdisc is removedand in itsplaceanother,dimmer,one is added sothat the same ratio of brightness exists between them as previously didbetween the earlier pair.The organism will respond to the disc which isnow the brighter one, although it was formerly the dimmer. Thisexperimenthasbeenspokenofasanexampleoftheconcept"brighterthan,"but the essential fact is not altered. We are here dealing with thegeneralizationofstimuluspatterns.

It iscurious tonote theresistance thatmaybeshownto thenotion thatthetermconceptneednotbe limited tomatterscapableofbeingverbalizedor found only in the behavior of human adults.We seem to have here aproblem in our own behavior.We have formed a concept of conceptualbehavior which is based upon such factors as the age of the subject, hisability toverbalize,and the fact thathe ishuman. It is true thatourverbalbehavior as adults becomes very complex and that, in philosophy, forexample,wemaywrestlewith the identifyingcharacteristicsofsuchverbalconcepts as justice, virtue, and wisdom. In such highly sophisticateddiscourses,wehaveanexampleof theattempt toreachagreementas to thecriteria of the just, the virtuous, and thewise—to enumerate the essentialSD's for evoking these words. Disagreements arise through differences inindividualdiscriminativehistories.Theproblemisnotunlikethatfacedbywriters on semantics; if we wish to know whether two persons 'have inmind' the same thingwhen theyuse the sameword, theanswerwill lie intheoverlapofdiscriminativestimuliwhichevokethewordfromeach.Thefact that two persons using the same word may agree on some of itsmeaningsanddisagreeonothersshouldnotbesurprising—theyhave,afterall,nothad identical training,and theconditionswhichevoke theword inonemaydisagreeinpartwiththosewhichevokeitintheother.

Wehavebeendealingwithconcept formationasa resultantof stimulusgeneralization and discrimination.This is all right but, especially in thecaseof adult humanbeings,it isnot thewhole story.An important sectorofadultbehaviorisverbalinnature,andtheanalysisofthisbehaviormustprecede a fuller understanding of conceptual activity. In the concludingchapterofthisbook,abriefsketchwillbeofferedofapromisingtheoryofverbal behavior that is in accord with the general principles you havelearned.Afteryouhave read that sketch,youmightprofitably turnback toreviewwhatissaidhere.

4.Conceptsandmediatedgeneralization.Generalizations are said to bemediated when they are based upon a stimulus equivalence which resultsfrom training (Cofer and Foley, 1942). Some concept formation is an

example of such equivalence.The words vase andurn have few stimuluspropertiesincommon.Ifweweretoconditionarespondentoranoperantinayoungchildtothesoundofoneofthesewords,therewouldbeverylittlegeneralization to the sound of the other. An adult, however, who haslearned that these words are almost synonymous and may be usedinterchangeably, is likely to show considerable generalization. Followingthe lead of Razran (1939b), who first explored this area, Riess (1940)conditionedthegalvanicskinreflexinitiallytoonestimuluswordandthentested for generalization to twootherwords, oneofwhichwas a synonymand one of which was a homonym of the first. For the stimulus wordsthemselves(style, freeze, surf, andurn), therewas an average gain of 346per cent in magnitude of skin response through conditioning. Thesynonyms(fashion, chill, wave, andvase) gained 141 per cent throughgeneralization;andthehomonyms(stile,frieze,serf, andearn)gained94.5per cent.Whereas the generalization to the homonym illustrates simplystimulus generalization, that to the synonym illustrates mediatedgeneralization based upon the previous training which produced the'meaning'equivalenceofthesetwowords.(Weshallhavemoretosayabout'meaning' inChapter 7.) Riess (1946) was also able to show that, forchildren below twelve years of age, there was more homonym thansynonym generalization—a findingthat accords well with the generallyacceptedbeliefthatthemeaningvalueofwordsincreaseswithage.

TableVIMEDIATEDGENERALIZATIONINTHECONDITIONINGOFTHE

GALVANICSKINRESPONSETOVERBALSTIMULIINSUBJECTSOFVARYINGAGE

(DatafromRiess,1946)

In the experiment, the GSR (measured in microamperes) was firstconditionedtoaword;afterwhichgeneralizationwastestedtoahomonym,anantonym,andasynonymof thatword.Exposureof thestimuluswordswasvisual,andofthree-secondduration.Thetableistobereadasfollows:The 21 children aged seven years and nine months (average) showed anincreaseinGSRtothewordafterconditioningof222percentoftheirpre-conditioning GSR to that word; the degree of generalization between theword and its homonym is shown by a GSR to the homonym of 159 percentof thehomonym'spre-conditioningGSRvalue;between thewordandits antonym, 139 per cent; between the word and its synonym, 129 percent. At this age, stimulus (auditory) similarity between the word andhomonym outweighs the 'meaning' similarity: but, by the age of 14, theorder of generalization is reversed, with mediated generalization through'meaning'beingthelargest.

GeneralizationandDiscriminationinEducationThe process of education is greatly concerned with generalization and

discrimination. It ispossible that,at thebeginningof life,all stimulimaygeneralize to produce mass and profuse responses in the infant. As hematures and learns to discriminate objects in his environment, thosegeneralizationswhichpersevereduringhispre-schoolyearswillprobablybeadequateformostofhisgrossadjustments,buttheymustbebrokendownlater in the interests of his educational progress.The new generalizationsanddiscriminationswillinturnundergochangeashisschoolingcontinues.Fromhisscienceteachers,forexample,hewilllearnthatfallingstonesandfallingpaperobey the same law, that fishbreathe in awayvery similar to

ours, thatanalleycatanda lionhavemuch incommon. In fact, fromonepointofview,thewholebusinessofsciencewillbeseenasthearrangementof nature's facts into new categories,with a stress upon the important butnot obvious similarities and a disregard for the obvious but unimportantdissimilarities. Even changes in the fundamental theories of science are ofthis nature.Thehistoryof great discoveries is oneof the reorganizationoffacts into new classifications on the basis of related properties. Planetarymovementsaretiedintoonegenerallawwiththemotionoffallingbodies.Electric currents are tied to the behavior of ions in chemical solutions.Nerve impulses are shown to be electrical phenomena.The behavior ofallorganisms follows the same basic laws.On simple and complex levels ofanimal and human behavior, the operation of generalization anddiscrimination are among the most important phenomena with which wedeal.TheArmy dog that growls at men in strange uniform but greets itsown soldiers; the child who learns to tell "Daddy" from "Mommy" (andmay then call all men "Daddy"); the student who learns that bats andwhalesarebothmammals,notbirdand fish;andapsychologistwhocitesthree apparently different instances of behavior as examples of the samebasiclaws—allofthesearedoingverymuchthesamething.

NOTESA very readable source on respondent generalization and discrimination

maybefoundinPavlov's(1927)ownbook,especiallyinChaptersVII andVIII. The statement of the principles as we have given them (tor bothoperantsandrespondents)havebeendrawn,however,fromSkinner(1938).

If you are interested in the technical details of the psychophysicalmethods,Woodworth (1938) offers a thorough account.A comparison ofhuman sensory capacities with those of animals at various phylogeneticlevels has traditionally been dealt with by the comparative psychologist.Fornumerous illustrations in this area, youmayconsultWarden, Jenkins,andWarner(1935-1940).

The early history of the "personal equation" and the study of reactiontimes has some entertaining features, well brought out by Boring (1929)andWoodworth(1938).

Amongthefirst torecognize the importanceofasystematicattackuponthe problem of "mediated generalization" were Cofer and Foley, in 1942andlater.Theproblemhasnotyetbeeninvestigated,however,tothedegreethatitsimportancewouldseemtojustify.

6

RESPONSEVARIABILITYANDDIFFERENTIATION

AfterE.R.GuthrieandG.P.Horton,CatsinaPuzzleBox,1946

IntroductionInthischapter,weshalldealexclusivelywithoperantbehavior.Weshall

concentrate upon the behavior itself, rather than the stimuli that set theoccasion for its emission. Indeed, we shall deal with changes in behaviorthat take placeirrespective of theSD situation.Ourmajor concernwill bewith three related matters. First, we shall consider the fact, perhapsobvious, that operant responses may differ on successive emissions, evenwhenpositive reinforcement is regularlyapplied.Secondly,weshall try torelate thisvariability to the influenceofnegative reinforcement that resultsdirectly from the emission of the operants themselves—an influence thatmay accompany the positive reinforcement. Finally, it will be shown thatdifferent strengths may be given to different variations of a response byapplying positive reinforcement in a selective manner. In treating thesematters, our task will not be simple because the data areincomplete andstill a source of conflicting opinion; but the problems are important andcannotbeby-passed,eveninatextforbeginners.

VariabilityversusStereotypySupposethat,inthesituationwhereawhiteratobtainsfoodbypressing

a bar, the first response to be reinforced is that of bar-biting, rather than aleaning, pressing, or climbing response.Wewould expect,wouldwe not,that this responsewould be strengthened: that it would bemore likely toappear at a later time—that it would gain a definite advantage over otherpossiblemodesofbardepression?Wewouldexpect,too,ifitdidrecurand

wasagainreinforced,thatitsprobabilityoffurtherrecurrencewouldincreaseevenmore.Eventually,wewouldlookforaconsiderableautomatizationorstereotyping of the rat's bar-pressing behavior.We would expect a single,fairlyrestrictedmodeofresponsetodevelop.

Althoughthedegreeofautomatizationactuallyachievedinthissituationis not as great as the above paragraph suggests, observation indicates that,withinlimits,somethingofthesortactuallydoestakeplaceregularlyinthecase of the rat, and there are supporting data from experimental studies ofother organisms. Guthrie and Horton (1946) observed and photographedapproximately 800 solutions of an escape-to-food problem by 52 cats,obtaining camera records of all the "escape postures" used by thirteen oftheir subjects. Solving the problem required that the cat's responses bringabout a slight inclination of an upright pole within an otherwiseunfurnished response-chamber. A small deviation of the pole from thevertical position was enough to open a door in the glass front of thechamber,whereupontheanimalwasfreetoobtainabitofsalmonfromthetop of a table a few inches away from a point of exit.Contact of any sortwith the release device occurred "inadvertently" on the animal's first trial,after a period of sniffing, clawing, or pawing at the glass door,and aftervariousturnsaboutthechamber.ThiswasapparentlyaroughequivalentofthebehaviordescribedbyThorndikemanyyearsearlier—notethequotationonhere.Thefinal,andeffectiveresponseoftheseriestookmanyaccidentalforms.Onecatbrushed thepolewithhis flank;onesteppedon thebaseofthepolewithhishind foot;anotherbacked into thepole;andsoon.Afterescape and feeding, the animal was again permitted to enter the response-chamber froma starting-box.On thisoccasion, andon laterones, the cat'sbehavioroftenseemed tobea"detailedrepetition" of that observed on thefirst trial. In not a few cases, one sample of release behavior of a givenanimalwasindistinguishablefromanotherinpracticallyeveryaspect.Somevariation in responsewas noted:minor differences in escape postureswerefairlycommonand,insomecases,markedlydifferentmodesofescapewereused by the same animal on successive trials. Guthrie and Horton were,however,most impressedby thestereotypyof thebehavior, rather than thevariations in it. (Thevariationswereaccreditedby themto the influenceofslight changes in the stimulus situationoccasionedbydifferent approachestothepole,accidentaldistractions,andother"interfering"factors.)

FIG. 41. The apparatus used by Guthrie and Horton in theirphotographic study of response variability in cats.The front wall of the"puzzlebox"wasmadeofglasstoallowthetakingofpictures.Thestripedpole, when pushed in any direction, would operate the door-openingmechanism and allow the animal to emerge from the box to reach a foodreinforcement lyingoutside.Theclockgave the time taken togetout aftertheanimalwasputin.(AfterGuthrieandHorton,1946.)

SomewhatdifferentfromtheGuthrie-HortonstudywasanearlieronebyMuenzinger (1928), who taught thirteen guinea pigs to press a bar in athree-chamberproblembox.Eachanimalenteredtheresponse-chamberfromastarting-boxandwasgivenaccess toa food-chamberwheneverheopenedthedoortoitbypressingalever.Assoonasthefoodwaseaten,theanimalreturnedorwasforcedback to thestarting-boxwhereheremaineduntil theexperimenterwasreadyforthenexttrial.Asarule,fifteentrialsweregiventoeachof thepigsdailyanda totalofeither600or1000solutionsof theproblem was accumulated. These solutions were not photographed, butMuenzinger was able to distinguish nine different "patterns" of successfulresponse (three right-paw and three left-paw patterns, a two-paw pattern, ahead-movementandabitingorgnawingresponse).Hefoundthatonlyoneanimal,apigthatnevermadethebitingresponse,failedtodisplayallninepatternsatonetimeoranotherduringtheexperiment;andhenotedthattheanimalschangedfromonepatterntoanotherasfrequentlyattheendoftheirtraining as they did in the earlier trials. Some degree of stereotypy didappear; there was a decrease in the averagenumber of different patternsdisplayedfromthebeginningtotheendoftheexperiment,andmostoftheresponses at the end fell within one or two of the nine categories. ButMuenzinger emphasized the "plasticity" rather than the "mechanization" of

the behavior. Even when thesame pattern occurred on many successiveoccasions, he tells us that it was seldom duplicated exactly from trial totrial, and might be suddenly supplanted by an entirely different pattern.Onlythreeofthethirteenanimalsshowedapreferenceforasingleresponsepatternfromstarttofinish.

Close examination of these studies suggests that the observed behaviorof Muenzinger's guinea pigs closely paralleled that of the Guthrie-Hortoncats. The principal difference involved seems to be one of descriptiveemphasis. In one case, the stereotypy or mechanization of response isstressed; intheother, thevariationandplasticity.Muenzinger, inaddition,seemstohavebeenmorealert to thechanges invariability thatoccurredasaresultofprolongedtraining.Bothexperiments,however,raisedquestionsof real importance to our understanding of behavior. Some of thesequestionswillbeconsideredhere;othersarestilltobeinvestigated.

ResponseInductionWe have seen, under the heading of generalization in the preceding

chapter,thatthereinforcementofaresponseinthepresenceofonestimuluswill strengthen it in the presence of other stimuli when these haveproperties in common with the first.There is, apparently, something likethis on the sideof response.The strengtheningof one responsemaybringabout a strengthening of certain others, where there is no change in theexternal stimulus situation. Thus, the reinforcement of one act of bar-pressingwillaffect,toalesserdegree,thestrengthofotheractswhichdifferfromthereinforcedoneinseveralidentifiableaspects.

Oneway inwhicha responsemaydiffer fromanother is intopography,by which we mean, in this context, theform orkind of a response.Themovement of one leg, for example, is topographically different from themovement of another, or from themovement of an arm or a finger.Also,responses, even when topographically similar, may differ in theirforce ort hei rduration, properties which may be subjected to quantitativemeasurement.Thus,thebar-pressingresponse,evenwhenquitestereotyped,will vary appreciably in its intensity or force, aswell as in the amount oftime that the bar is held down on each successive response.The questionwe now ask is this: will the strengthening of a response having a certaintopography, force, or duration lead to or 'induce' a strengthening ofresponseswhichdifferfromitinoneormoreoftheseproperties?

Letusbeginwithtopography.Kellogg(1939)conditionedtheflexionofadog'srighthindlegtothesoundofabuzzer.Theresponsewasoriginallyelicitedbyshockanditsstrengthwasmaintainedthroughshockavoidance.Along with the establishment of this foot-withdrawal, Kellogg noted thatoccasional movements of the other legs appeared. The number of suchflexionswas greatest for the left hind leg, next greatest for the right frontleg, and smallest for the left front leg.That is, a sort of 'topographical

gradient'wasobserved.Hil-gardandMarquis(1940)havetreatedthisasanexampleofresponsegeneralizationandhavesuppliedotherobservationsofa similar sort. For ease of reference,we shall speak of such phenomena ascasesofresponseinductionor,simply,induction.

Aninterestingmaze-learningexperiment,suggestingresponseinduction,was carried out byMacFarlane (1930).After training one group of rats toswim throughamaze,heinsertedafloorbeneaththesurfaceofthewaterata depth sufficiently shallow to force them torun the maze. For anothergroup, theprocedurewasreversed.When theratsofeachgroupwere testedafter the shift, no appreciable increase in errors resulted. Transfer wasapparently complete, in spite of the fact that the responses utilized by theanimals were observably different under the conditions of swimming andrunning.

Other cases suggesting topographical induction appear in variousexperiments commonly treated under the heading oftransfer of training.Starch (1910) discovered that human subjects who were given practice intracingtheoutlineofastar-shapedfigure,whenthefigurewasseenonlyinamirrorandapartial reversalofone's customarymovementswas thereforerequired, showed a gradual improvement in performance which was notlimited to the hand used during practice. In right-handed subjects, the lefthandprofitedconsiderablyevenwhenithadbeengivennotrainingatallinthetracingtask.Otherinvestigatorshaveconfirmedtheseresultsandaddedinformationon theproblem.Thus,Bray (1928), employinga technique inwhich subjects struck at a mirrored target, showed that the transfer effectwas not merely bilateral (from the practiced to the unpracticed hand), butalso extended to the homolateral foot (i.e., right hand to right foot). Suchresults remind us of Kellogg's findings with dogs, but they are probablynot to be interpreted as simple cases of induction. Bray was able todemonstrate that other factors, such as a "transfer of methods," wereinvolved. Human beings tend to verbalize what they are doing, and anyinductionthatexistsbetweenhandandfootinthiskindofexperimentmaybemediatedorsecondary,ratherthandirectorprimary.

FIG.42.Amirror-drawingapparatus.Thesubject'shandishiddenfromhisdirectview,buthecansee itand thepattern tobe traced in themirror.The reflection, of course, reverses the actual movements to be made intracing.(CourtesyofC.H.StoeltingCo.)

When we come to consider quantitative induction, the case becomessomewhat clearer.Take the property of response intensity or force. It hasbeen demonstrated that a rat, even after long practice in bar-pressing, willcontinuetoshowvariationsintheamountofpressureexertedonsuccessiveoccasions. In an experiment by Skinner (1938), ratswere first conditionedin the usual manner and regularly reinforced during several trainingsessions,afterwhich theresponsewasextinguished.Byconnecting thebarwithapendulumandsuitablerecordingapparatus,itwaspossibletoobtaincumulativecurvesofresponseintensitiesduringboththereinforcementandnon-reinforcement sessions. The slopes of these curves provided anindicationoftheaverageforceofresponseaswellasthedegreetowhichtherats deviated from this average under the two experimental conditions. Itturnedout that, under regular reinforcement, the average force amounted to35-40grams,althoughmanyof the responseswereweakerorstronger thanthis. (A ten-gram force was required to depress the bar, so practically allpressingswerestrongenoughtobereinforced.)Underextinction, therewasat first a tendency for the animals to respond with more than the averageforce during conditioning, but this was followed by a marked decrease astheprocessnearedcompletion.

Dataonthedistributionofforcesinsuchasituationhavebeencollectedby Hays andWoodbury (reported by Hull, 1943) in theYale laboratory.They reinforced bar-pressing with food whenever the response intensityamounted to 21 grams ormore, until 100 reinforcements had been given.The frequencywithwhich different intensities appeared, in the case of oneoftheirrats,isshowninthesecondcolumnofTableVII.Thethirdcolumnof this table shows the values obtained with the same animal when aminimalforceof38gramswasneeded.

It is obvious from these data that complete stereotypy in the force orintensity of the bar-pressing response is not achieved under theseconditions. The reinforcement of a response having a certain intensityapparently suffices to strengthen topographically similar responses havingwidelydifferentintensities.

TableVIITHEDISTRIBUTIONOFRESPONSEINTENSITIESINTHE

HAYS-WOODBURYEXPERIMENT,UNDERTWOCONDITIONSOFREINFORCEMENT

(DatafromHull,1943,p.305)

*Noresponsesreinforced,†Nineresponsesreinforced.‡Sevenresponsesreinforced.

Thisamountofvariationshouldnotbesurprising.Itisobviousinmostof our own actions and has often beenmeasured.TableVIII gives sampleresults from an experiment ofThorndike's (1931) in which subjects, witheyes closed, were asked to "draw a four-inch line with one quickmovement."Duringsuccessivesittings,atotalof3,000lineswasdrawnbyeach subject, under the same experimental conditions and withoutknowledge of results, the only reinforcement being the approval of theexperimenter fordoing thework.On the firstday, inwhich192 responsesweremade, therangeof line lengths foronesubjectwas4.1 to5.7 inches;on the twelfth day (175 responses), the range was 4.5 to 6.2 inches.The

distributionoflengthsonboththesedaysisshowninthetable.Asidefromaslightshiftupwardintheaveragelengthoflinedrawn,whichisnomorethanadailyvariation, thetwodistributionsareessentiallythesame.Ifthisresponse underwent any decrease of inductive effect, itmust have done sobefore theseperiodsofprolongedpracticewere instituted.While thespreadof responses in session twelve appears less than that in session one, thisdoesnotindicateaday-to-dayreductionwithpractice,sincetheinterveningdaysshowconsiderablefluctuation.

TableVIIITHEDISTRIBUTIONOFLINELENGTHSDRAWNBYASUBJECTDURINGTwoPRACTICESESSIONSINONEOFTHORNDIKE'S

EXPERIMENTS.(DatafromThorndike,1931,pp.8-9)

TheWhyofVariabilityOne fact emerges plainly from the discussion in the two preceding

sections. Under all conditions of reinforcement or non-reinforcement thus

far described, some degree ofresponse variability survives—completestereotypy is never achieved.This appears in theGuthrie-Horton study nolessthaninthestudiesofMuenzinger.It ispresent,too,inall thecasesofinductionthatwehavecited.Butwearestillwithoutanexplanationofthisvariability. Even assuming that the strengthening of one response willstrengthenother, slightlydifferentones,howdoes ithappen thataspecificresponse with one specific set of properties does not, by virtue of morefrequent reinforcement, come to be theonly response emitted in a givenexperimentalsituation?

The answer to this question demands that we postpone for a while theconsideration of the second major topic of this chapter, and turn ourattention to a factor that we have hitherto encountered only in its grosseraspects. Our side trip will not, however, be wasted, because we shalldiscoversomemattersofconsiderableinterestandsignificance.

VariabilityandNegativeReinforcementLetusgobackforamomenttotheGuthrie-Hortonexperiment.Suppose

that,inthisstudy,oneoftheircatsfirstoperatedtherelease-mechanismbyfallingbackwarduponthepolewhile tryingtoclimbthenear-bywall.Thefalling response would presumably be strengthened through its positivereinforcement, but the fall might also provide stimulation of anegativelyreinforcing sort. The consequences of the fall would thus exercise adepressive effect upon the immediately preceding response, just as amildelectricshockmightdo.Inlinewiththiswayofthinking,whenthecatre-enteredthesituationonthenexttrial,hewouldnotbeaslikelytoreinstatethesuccessfulbehaviorashewouldifsomeotheractionhadbroughtabouthis escape from the box and getting of food. Some alternative mode ofresponse, whether previously rewarded or not, would be expected, andanothersolutionmightsoonbeforthcoming.

Or,supposethatthecat'sfirstpositivereinforcementcamefromstandingonhishindlegsandstretchinghisbodytoreachthetopofthepole.Thesestraining movements would also provide a negatively reinforcing state ofaffairs, albeit of a mild degree.With repeated trials, we would look for amore or less gradual reduction in the amount of energy expended in hisresponse to the pole. In everyday language, we would say that the catshould learn to take the easiest, rather than themost awkward or tiresomewayofreachinghisobjective.(Itmightbe, too, that theawkwardresponsewould be a somewhat slower one and be at the additional disadvantage ofhavingitsreinforcementdelayed.)

There is not much evidence in the Guthrie-Horton account, or inMuenzinger's report of the behavior of his guinea pigs, to suggest theoperationof this factor in theseexperimental situations,butThorndikehasgivenus somedramatic instances.He taughtbothdogs andcats to lickorscratchthemselvesinordertoescapefromtheproblembox,andfoundthat

theseoperantsdecreasedinmagnitudeunderregularreinforcementuntiltheywere mere vestiges of the original.Thus, licking might be reduced to amere jerk of the head downward. Similarly, Lorge (1936) trained rats tomake either a face-washing, standing-up, 'begging,' or scratchingmovement, reinforcement being escape from a problem box to food. Henoted a "short-circuiting" of these responses with successive trials. "Theresponses became more perfunctory and stereotyped. The 'face-wash'changedfromavigorouswashtoarapidmovementofbothforelegs to theface; the 'scratch' changed to a rapid flexion of the hind-leg to the flank,only remotely reminiscent of the first response to irritation." Whenindividualratsweretrainedinallthreeresponsesofface-washing,standing-up,and'begging,'theytendedgraduallytoeliminatethefirsttworesponsesentirely, securing reinforcement by making a "perfunctory" beg. Relatedobservationsweremadewhenother ratswereconditioned to touchanyoneoffourdifferentiallyaccessibleprojectionsintheirproblemboxwitheitherforepaw. The animals ended by responding most frequently to theprojection that required the least movement and was closest to the doorthrough which reinforcement could be reached. The last-mentionedexperimentisremindfulofanearlieronebyGengerelli(1933),whotrainedring-tailmonkeys todepress two (or four) levers in sequence for escape-to-food.Heobserved thathis subjectspassed throughsuccessivestagesof (1)''over-exertion," in which they often climbed up on the levers; (2) two-handed clasping and tugging at the levers; (3) a one-handed grasp-slapcombination of movement; (4) a mere downward slap; (5) an ineffectual'pass'at the lever;and (6)evencases inwhich theanimals"would rushupto the lever, then precipitately turn to run to the next one," without anylevercontact.

Stillanotherexample,at thehumanlevel,comesfromanexperiment inwhichThorndike (1931) requiredstudents,duringa fourteen-dayperiod, tomake3,360wordcompletionsofalonglistofsuchword-beginningsasab,af, bo, and the like. In some cases, the same beginning was repeated asmany as twenty-eight times in the course of the experiment, and it waspossible to note any change in the nature of the completion that occurred.Thorndikediscoveredastrongtendencyforthesubjectstoreducethelengthof the complete words as the amount of repetition of the word-beginningincreased. Thus, one student, upon the first eight presentations of el,responded withelephant five times,elevate twice, andelf once; upon thelasteightpresentations,herespondedwithelfexclusively.Presumably,theeffectofverymildnegativereinforcements,inconjunctionwiththepositivereinforcement supplied by any adequate and rapid completion of the task,was sufficient tooffset the influenceofpositive reinforcement for the first-made "long" solutions.This appears tobe related to the fact that themostfrequentlyoccurringwords in theEnglish language tend tobe shorter thanthoseappearinglessoften.It isalsoreflectedinthevarioustruncationsand

substitutions observed when initially long words come into common usewithinoneoranother"verbalcommunity."Wespeakofour car, ourauto,orourbusinsteadofourautomobile;welookafterastrepthroat;wesendawire or use thephone; and we present anSD rather than adiscriminativestimulus. In addition, it may account for some of the slurring tendenciesoften observed in colloquial speech, as whenerrors becomeairs, flowersbecomeflars,andborrowbecomesbar.

NegativeReinforcementandthe"LawofLeastEffort"Another line of evidence for the depressive effect of response-produced

negative reinforcement comes from studies ofalternation behavior inanimals, especially the rat. Psychologists have for years been interested inthefactthatrats,whengivenachoiceoftwoequallylongroutestofood,asinasingle-unitT-maze, tend toavoid repetitionof the last-made response.Thus, a run to the left or to the right is commonly followed on the nexttrial by a run in the opposite direction. Some researchers (e.g., Hunter,1928)haveevenspokenofan'innate'or'natural'tendencyoftheanimalstobehave in this manner. Solomon (1948) has recently reviewed theobservations in this area and, revising an earlier formulation by Hull(1943), has proposed an explanation that jibes well with the analysis thatwe have presented in the preceding section. In a choice situation, withconditioned responses of presumably equal strength, the emission of oneresponseprovidesforitsowntemporarydepressionandthuspavesthewayfortheemissionofanalternative.Solomon(1946)foundthatgreater"effortrequirement inT-maze running is accompanied by a greater frequency ofalternationbehavior."

One would not expect such a minute amount of response-producednegative reinforcement to be long-lasting in its depressive effect; and theshortnessof effect is confirmed in an experiment byHeathers (1940).Thisinvestigator found that the amount of alternation in a single-unitT-mazedecreasedasthetimebetweensuccessiverunsinthemazeincreased.

Various researchers and theorists have treated observations of the sortdescribedhereasillustrationsofalawof"leastaction,""minimaleffort,"or"less work." Such a formulation is non-analytical and should properly beconsideredasacorollaryofsomethingmorebasic (Hull,1943);butweareallfamiliarwiththekindofbehaviortowhichitrefers.Althoughthesearethe more obvious instances, something of the same sort probablyaccompanieseveryresponse,becauseeveryresponserequireseffort.

Undercertaincircumstances,ofcourse,a responsewillbemaintained inconsiderable strength even when it seems to be inefficient and awkward.This situation occurswhen reinforcement has regularly been denied to anyeasier variations of the response or, as we shall see presently, whenmotivation ishigher.Gilhousen (1931) showed that rats, afterhavingbeen

trainedtojumpfromonesmallplatformtoanotherofastraight-awayseriesleading to food,persisted in their jumping responsemuch longer thanwasnecessary after an unobstructed pathway was set up alongside of theirplatformroute."Doingitthehardway"isapparentlyascharacteristicofratsas of human beings, and has the same explanation: the hard way has toooftenbeentheonlyreinforcedwayavailable.Unlessnegativereinforcementisextremeandisclearlycontingentuponaspecificmodeofresponse,allofus waste our energies needlessly. Industrial psychologists and 'efficiency'experts'arewellawareofthis,andmuchoftheirresearchisdirectedtowardidentifying less effortful ways of reacting which, in the course of dailyoccupational routine, would never be appreciated or adopted by theindividualworkersunderobservation.Howmanyofuswoulddiscover,byourselves, that one brand of typewriter may involve only three-fourths asmuch work in its operation as another?And how many "hunt-and-peck"typistswould be ready to change their style, in spite of the fact that theymay know of a better method?The degree of negative reinforcement thataccumulates as a result of our hunt-and-peckprocedure is not great enoughtoact as a specific depressant of the responses that brought it about.Wemay,byourselves,makenotableprogress:wereducesomewhattheamountof unnecessary bodily movements; we no longer punch the keys with aforce that makes our fingers sore; wemay even become two-finger touch-typists, thus eliminating certain movements of the head and eyes. Butseldom, ifever,doweattain theefficiencyofawell-trainedperformerwhohas had the advantage of an experienced teacher.LikeGilhousen's rats,wepersist in our energy-wasting ways, and our resistance to change is thestrongerbecausethebetterwayrequires,at theoutset,anextinctionofold,andoftenverystrong,responsesaswellasaconditioningofnewones.Addtothis thefact thatreinforcementfor thenewwayisprobablynotasgreat,atthebeginning,asthatprovidedbytheold.It isnowonderthatweclingtoourown"jumping"responses.

Resistance to the adoption of improved working methods, so oftenencountered by the applied psychologist, is, one must admit, not alwaysdue to the fact that less satisfactory methods have been more oftenreinforcedor that newonesmustbe conditioned.Sometimes theorganismhas no choice in the matter. The standard keyboard of a typewriter isundoubtedlyinefficientintermsofenergyexpenditure,yetit isunlikelytobereplacedbyonethatwouldprovideamoreequabledistributionofeffortfor the fingers of each hand—for commercial reasons that have nothing todo with the unwillingness of a typist to give up the old and take on thenewwaysofbehaving.

MotivationandVariabilityThe exclusive reinforcement of a single response variant apparently

reduces the probability that other variants will occur, but there is still

anotherfactorthatdeservesattention.Understrongmotivation,anincreasedstereotypingofresponsemayalsobeobserved.InanexperimentconductedbyElliott(1934),ratswerepermittedtoreachfoodattheendofanyoneoffive short alleys that diverged from a common starting-point.The alleyswere of equal length and, when hunger motivation was weak, the ratsshowed no marked preference for one alley over another. However, whenhunger was increased, each animal tended to take one pathway to theexclusionoftherest.Moreover,thisdecreaseinvariabilitywasirreversible:a return to conditions of weak motivation did not reduce the amount offixationuponthealleychosenwhenthemotivationwasstrong.

Quantitativedataare lacking for theeffectofmotivationupon the rangeof bar-pressing movements, but frequent observations suggest that anarrowing-down effect occurs. For example, when the response removes anoxiousstimulussuchasastronglight,awell-conditionedratmaybeseentomaintainasingle,crouchingpostureclosebesidethebarduringmostofthe experimental session, even during the faint light of the reinforcementperiods, andwith all hisbar-pressingmovementsgreatly restricted in theirvariety.Occasionally,hemaydepart fromthis routine,possiblybecauseofthenegativereinforcementproducedbyprolongedmusculartension,buthisbehavior is, in general, strikinglymachine-like in quality, and it seems toretain this character for some time after the motivating stimulus has beengreatlyreducedinitsintensity.

Everywhere about us we see human actions that appear to be at leastroughly analogous to such laboratory phenomena as these. Heightenedmotivation seems tobeone sourceof the routinizingand stereotyping thatmarksourdailyhabits;andweadheretosuchpatternsofconductwhenthepressureno longerexists.Yet, there isprobablynosingleexample thatwecould give which would not also suggest the operation of more than onefactor; and it would certainly be foolish at this time to ascribe tomotivation alone all of the tenacity and case-hardening of human behaviorthatwemayobserve.What is needednow,more than anything else, is anexpansionofresearchinthisveryimportantsphere.

Extinction,Reconditioning,andVariabilityA recent study byAntonitis (1950) sheds some light upon the way in

whichvariabilityisalteredduringtheregularreinforcement,extinction,andreconditioning of operant behavior. Rats were permitted to run from arelease box across an open area to a horizontal slot (50 cm. long) in thewall, twelve inches from,and facing, the release-boxdoor. Insertionof therat's nose into the slot at any point interrupted abeam to a photo-electriccell, causing the animal to be photographed in position, whereupon hisreturn to the release box was reinforced with a pellet of food.Experimentationwasbegun after one operant-level session inwhichbeam-breaking responses were recorded but not reinforced. Five daily

conditioning sessions, in which a total of 225 reinforcements was given,were followed by two one-hour extinction sessions, one session ofreconditioning (50 reinforcements), twomore sessions of extinction, and afinal day of reconditioning.Figure 44 shows the way in which thedistribution of slot-response positions varied for one animal under thesedifferentprocedures.During the first fivedaysofconditioning, therewasadecrease in the variability of response positions; in the next two days,during extinction, the variability increased; and this was followed by amarked decrease on the first day of reconditioning. This decrease afterextinctionwascharacteristicofalltheratsandamountedtoagreaterdegreeof stereotypy thanhadbeen achievedduring the fivedaysof conditioning.It remained at approximately the same level in the final reconditioningsession, after the third and fourth hours of extinction.A control group ofanimals, not subjected to the extinction periods but treated like theexperimentalanimalsineveryotherrespect,showednocomparableincreaseofstereotypybeyondthefirstfiveperiodsofconditioning.

FIG. 43.Plan ofAntonitis's apparatus for studying response variabilityin the white rat. Photographs of each nose-insertion response to the slotallowed the investigator to take twomeasuresof the response: the locusofthe response along the slot, and the angle of approach to the slot. Bothmeasuresyieldedinformationaboutthedegreeofresponsevariability(oritsconverse, stereotypy); and both measures, as it turned out, told the samestory.Thelabeledpartsoftheapparatusareasfollows: A—electric counter;B—mirror tilted at a 45° angle to the base of the experimental cage; C—reflected image of the 50-cm. response slot;D—clear plastic rear wallcontainingresponseslot;E—whitelinespaintedonblackfloorofcage;F—starting-feedingcompartment;G—food tray;S—spotlight;P—photoelectriccell.(AfterAntonitis,1950.)

FIG.44

LEGENDFORFIG.44Daily frequenciesof responsepositionsalong the50-cm.slot foroneof

the experimental rats inAntonitis's study.The experimental sequencemaybefollowedbytakingthegraphsfromlefttorightalongeachrow,startingwith the top row. Notice: (a) the wide variability of the operant levelresponses;(b) the trend toward lessenedvariability during the five originalconditioning days, when each nose-insertion response was reinforced by afoodpelletprovidedinthetrayofthestarting-feedingcompartment;(c) thereappearanceofgreatervariabilityduringextinctiondays1and2(aswellasthe smaller number of responses on extinction day 2 as against extinctionday1);(d)theheightenedstereotypyonreconditioningday1;(e) the returnof wide variability on extinction days 3 and 4; and (f) the extremestereotypyonreconditioningday2.(AfterAntonitis,1950.)

SUCCESSIVEEXPERIMENTALPHASES

FIG. 45. The course of response position variability throughoutAntonitis's experiment.The statisticalmeasure of variability plotted is theaverage deviation of the individual animals' positions around the groupmedian.The O.L. (operant level) and original conditioning (5 days) datapoints are based on the combined group of 12 animals, since all receivedthe same training in these stagesof the experiment.Variabilitydiminishes(stereotypy increases) as conditioning progresses, with the groupapproaching a final asymptote.Thereafter, the group was divided into anexperimental and control group of six animals each.On extinction days 1and 2, only the experimental group was extinguished, while the controlgroupremainedinitslivingcages.Asseen,extinctionraisedthevariabilityof the experimental group. On reconditioning day 1, the experimentalgroup'svariability falls significantlybelowwhat itwasbothon extinctiondays 1 and 2, and on original conditioning day 5;while the variability ofthe control group, which "sat out" the two extinction days, is notsignificantly different from that of original conditioning day 5. Just aboutthe same effects were obtained on the second two-day extinction session(extinction 3 and 4), which the control group again "sat out," and thesecondone-dayreconditioningsession(reconditioning2).(FromAntonitis,1950.)

TheDifferentiationofResponseDespite the degree of stereotypy that may be achieved through the

influencesjustdiscussed,acompletemechanizationofresponseisprobablynever reached.To thepersonwhosetsahighpremiumuponefficiencyandprecision ofmovement, this failuremay represent a fundamentalweaknessofourbiologicalfunction.Fromanother,andwiserviewpoint,itisaboon,sinceitpermits thedevelopmentofadaptivebehavior thatotherwisemightneverhavebeenincludedinanorganism'srepertory.

The procedure by which we producenovel responses in an animal or ahuman being is simple to describe, although not at all easy to execute. Inessence, it amounts to this: we select one (or more) of the "natural"variations of a well-conditioned response and give itexclusivereinforcement; the remaining variations are subjected toextinction. If wepickout, inadvance,avariationthathasbeenoffairlyfrequentoccurrence,and if we apply this selective reinforcement rigorously, we can soonproduce an increase in the frequency of the response that possesses theproperty or properties (for example, the force) thatwe have chosen.At thesame time,we decrease the frequency of those responses that do notmeetourspecifications.

Skinner (1938) has called this shift in frequency through selectivereinforcement thedifferentiation of response, to distinguish it from the

discrimination of stimuli, to which it is analogous; and he conductedseveralexperimentson thedifferentiationofbothforceanddurationofbar-pressing. Itwillbeprofitable toconsider someof thisworkbriefly, to seehow the principle operates. In doing so, we shall limit ourselves to hisobservations on changes inforce of response.We take, as our point ofdeparture, the experiment mentioned earlier in this chapter under theheadingofinduction.

When the normal force of the bar-pressing response, and the range offorces, had been determined by Skinner for a given animal, selectivereinforcementwas supplied toabove-average forces that had occurred onlyoccasionally beforeThe result was an almost immediate shift of responseintensi. ties to a higher level.The successful force, which had previouslyoccurredaboutonceineverytenpressings,nowappearedasoftenasonceintwo pressings. In addition, the entire range of force values was shiftedupward appreciably. Something like this may be seen in the Hays-WoodburyfiguresofTableVII (here).The second column of this table, asmentioned above, shows the distribution of forces when there was a 21-gramminimumrequirement toobtainreinforcement; the thirdcolumntellsus what happened in the same animal when a new and greater force (38gms.)wasdemanded.

The shift in distribution of forces that takes place under suchexperimental conditions shows us the way in which we can set up aresponse that has a forcenot previously observed in the animal's bar-pressing. If, now, we selectively reinforce some of the more intenseresponseswithinthenewrange,wemayshiftthedistributionagain;andwemay continue in this fashion until our animal is exercising his greatestpossible force upon the bar. Skinner was able, through this procedure, toreach a force-level of one hundred grams, nearly one-half of the animal'sbodily weight. This could only be achieved, of course, when the forcerequired at a new level had been of sufficiently frequent occurrence at thelevelpreceding;andeven then thenewlevelwasreachedverygraduallybythe animal. The entire procedure bears a marked resemblance to thatdescribed inChapter 4 for the establishment of high response rates underperiodicreinforcementatafixedratio.Indeed,wewouldhavebeenjustifiedintreatingfixed-ratioP-Rinthepresentcontextasaratedifferentiation.

Since force differentiation, in our example, had to be established in theface of the negatively reinforcing effect of theextreme effort involved,setbacksinprogresswerequitecommon.Thiswasespeciallythecaseatthebeginningofaday'sexperimental session,whenconsiderable 'warmingup'was required before the rat reached a level where practically all of hisresponses produced food.Aswith the licking response ofThorndike's cat,the animal tended to revert to an easier way of behaving—fromwhich hehadtobe'lifted'oneachsuccessivedayofwork.

The warming-up effect observed in differentiated bar-pressing has its

parallel in various human activities. Coaches of athletics have longrecognized the value of a limbering-up process in such sports as baseball,track, and crew-racing, and this apparently holds for any performance thatrequires above-normal exertion. Even such small-muscle activities astappingwith the fingers (Wells, 1908) or repeating the alphabet backward(RobinsonandHeron,1924)showsimilarchangesduringpracticesessionsinwhichahighrateofrespondingisrequired.

Astudyofforce-differentiationinasituationcomparabletotheoneusedwith rats has been made by Murphy (1943) with human subjects. Thisinvestigatoremployedamodificationofthewell-known"pin-ball"machineinwhich a small plungerwas pulled back against a spring and released tosendatwenty-grammarbleupaslightinclineintoahole.Thehole,aswellasthecoursetakenbythemarble,washiddenfromthesubject'sview,butaflash of light at the front of the machine indicated the correctness of theresponse.This light-flash was under the control of the experimenter, whocould make it contingent upon any degree of plunger-pull or who couldeliminate it entirely if hewished.The distance of each plungermovementwas recorded vertically and cumulatively on the paper-covered surface of akymographdrumthatjumpedaheadhorizontallyaboutone-eighteenthofaninchwith every completed response.Rate of respondingwas recordedonaseparatedruminamannersimilartothatusedwithrats.

Murphy explored (1) the 'normal force' and 'normal rate' of plunger-pullingwhen all responseswere followedby a light-flash; (2) the effect ofperiodic reinforcement upon response rate andupon the subject's resistanceto "shifting" fromone required response-magnitude to another; and (3) thedifficulty of setting up differentiations when different "margins of error"were permitted.His resultswere in general agreementwith those obtainedfrom rats in the bar-pressing studies, but were different in certain details.Forexample, thedistributionofforceswhenallresponsesproducedflashesof light bears little resemblance to that obtained by Hays andWoodbury(seeTableVII).This ispresumablydue to thefact, recognizedbyMurphy,that there were certain differences in the procedure employed in the twostudies. In terms of the organism's capacity, the range of response forceswaslessinthisthanintheexperimentusingrats, thepracticeperiodswereprobably not of comparable duration, the form of reinforcement wasdifferent,andsoon.

Murphy'sstudydidnotinvolvethereinforcementofallresponsesabovea certain force-level, but specified an upper, as well as a lower, limit ofmagnitudes.This requirement brings his procedure somewhat closer to theeveryday type of human differentiation where the range of responseproperties must not exceed a fairly narrow band of tolerance if ouradjustment is to be effective. The results indicate that (1) as theexperimenter restricts the tolerated range the number of incorrect responsesrequired before a correct one ismade grows larger; (2) there is an apparent

improvement in the ability of subjects to shift back and forth from oneresponse range to another as the number of such shifts increased; and (3)any form of periodic reinforcement is more effective than regularreinforcementinbuildinguparesistancetoshiftingfromoneforce-leveltoanother. Such findings are clearly consistent with those obtained fromstudies with rats, and point to the fruitfulness of extending this area ofresearch.

The differentiation of force is, of course, but one form ofchange thatmay result from the selective reinforcement of response. Topographicaldifferentiation is just as commonand is considerablymoredramatic, sinceit may lead to actions that are more clearly 'new.' Although fewexperimental studies have been aimed specifically at this problem, it isobviously present, along with stimulus discrimination, in many forms oftraining,anditisreadilydemonstrableatanon-quantitativelevel.Arat,forexample, may be led, by gradual stages, to walk on his hind legs, jumpthrougha loop,carryamarble inhismouth,orperformsomeotheractionthatwouldneverhaveoccurred inagivenexperimental settingwithout theapplicationofadifferentiativeprocedure.Thetechniqueisalwaysthesame:someonevariantofanalreadyconditionedresponseisselectivelyreinforceduntil it becomes more frequent than the others; when this is achieved, avariantof thenewresponse is treated in thesameway.Throughaseriesofsuccessive approximations to the desired reaction, the behavior is altereduntilitcomestobearlittleornoresemblancetothefirst-conditionedform.

Whenwethinkofdifferentiation,weperhapsthinkfirstofanimalstuntsor athletic skills, but this ismuch too narrow an application.Teachers ofspeech or of foreign languages, experts in time-and-motion study,instructors in the various arts and crafts—all these, and many others, areregularlyengagedin theformationofdifferentiatedresponses. Indeed, thereis hardly a human pursuit which, in one aspect or another, does notillustratetheprocessinsomedegree.

Theaccuracyorprecisionofadifferentiatedoperantisordinarily,perhapsalways,afunctionofthediscriminativecapacityofthepersonwhosuppliesthereinforcement.Theteacherofwhateverartorskillmustbeonewhocandistinguishminutechangesinthebehaviorofhispupil.Hemustbeabletosingle out that variation of response which is in the direction of theimprovement he desires, and hemust do this rapidly if he is tomake hisreinforcement optimally effective. Regardless of his own differentiativeability, he must be able to detect those shades and nuances, those slightdeviations and minor refinements in the actions of others which must bestrengthenedorextinguishedwhentheyappear.

ResumeAwordofrecapitulationat thispointmaynotbeoutoforder.Wehave

seen,inthepresentchapter,thatorganismscharacteristicallyshowresponse

variabilityevenintheabsenceofdetectableenvironmentalchange.Adegreeof stereotypywill result from regular reinforcement, but this stereotypy isnever complete.We have also seen that the reinforcement of one specificresponse probably strengthens others (response induction). It has beensuggested that the sourceof variability lies in the fact that evenpositivelyreinforced responses are subject to momentary depressions in strengththroughthenegativelyreinforcingconsequencesoftheiremission—howeverfaint and fleeting these consequences may be. Such depressions pave theway for the emission of alternative responses, some of which may alsoobtainpositivereinforcementandmaybeevenlessfreightedwithnegative,so that they takepriority in termsof later occurrence.This happensunlessthere are strong counter-influences—that is. unless the first response hasbeen the only reinforced one or, perhaps, has been established understronger motivation. This explanation of variability may be related to"alternation"studiesand tocommonlycitedexamplesof "least effort,"butwhat we need most is further research in this field. With respect todifferentiation, there is less room for uncertainty than in thediscussionofvariability. Given a degree of variability, however accounted for, we havenotroubleinmodifyingbehaviorindependentlyofchangesinexteroceptivestimulation.Amarkednoveltyofresponsemayevenbeachievedsimplybyreinforcing successive approximations to a desired mode of reaction andwithholding reinforcement for responses that do not change in thisdirection.Inthiseffectwehavethebasisfordevelopingvariousskills.

DifferentiationandDiscriminationWe began our discussion of operant behavior, inChapter 3, without

considerationofthestimulitowhichsuchbehaviormightberelated.Later,i nChapter 5, we showed that operants may come to be emitted in thepresenceofsomestimuliandnot in thepresenceofothers. In thischapter,we have argued that responses are never completely stereotyped, and thatthey may be changed in their character independently of any change instimulus conditions.Our examples of response differentiationwere chosenwithaneyetopointingupthis independence.Yet itmusthaveoccurredtoyou that many, if not all, differentiated operants are alsodiscriminativeoperants. Very early in an organism's behavioral development a specificmode of response becomes related to a more or less identifiable form ofstimulation. Obviously, a rat is unlikely to engage in bar-pressingmovements in the absence of a bar that can be seen or touched, just as ahuman subject in a reaction-time experiment is unlikely to make a key-tapping movement when no reaction-key is present. In many cases, ofcourse, the relation of response to stimulus may not be so apparent, aswhenarat is reinforcedforstandingonhishind legsorastudenthasbeenreinforced for raising his hand in class, but even in such instances acorrelation may often be discovered; a rat does not emit his standing

responsewhenhe isno longer in the experimental situation, anda studentdoesnotordinarilyraisehishandintheabsenceofhisinstructor.

A simple example of the tie-up between discrimination anddifferentiationisthebehaviorofananimalinaT-box(here).InthepresenceofoneSD,he turns to theright; in thepresenceofanother,he turns to theleft. Similarly, in a reaction-time study, a person may respond with onehand to a green light andwith the other to a red one.At a slightlymorecomplex level, we have the subject's behavior in a situation where thespecific stimuli and responses are greater in number, as in Merkel'sexperiment (here) which employedas many as ten alternatives. Thissituation, in turn, resembles theone inwhichstudentsmaster InternationalMorseCode,wherenofewerthanthirty-sixbasicstimuluspatternsareeachconnectedwithadifferentiatedresponsepattern.Whenone learns to receivecode, his problem is mainlydiscriminative, since the written or spokenresponses have already been well differentiated; insending code, however,the problem is one of differentiation, since the discriminative work wasdonewhenthestudentlearnedtoreadhisABC's.

The early development of a child's vocabulary providesmany excellentillustrations of the combined processes of discrimination anddifferentiation.The vocalization of the child contains, at the beginning, asmallassortmentofbasicsoundelements(phonemes).Withthematurationof the vocal mechanism, the number of these elements increases and, astheygroupthemselvesintopatternsbyvirtueoftheprincipleof"chaining"(seeChapter 7), they are also subjected to differentiation through theselective reinforcement supplied by parents and others.At the same time,the emission of many of these patterns comes to be connected withrelatively specific stimulus situations.Mama comes to be emitted mainlyin the presence of the mother;dada may be adopted by the father ashisname, and so on.At first, the reinforcement for this naming behavior isgiven lavishly, no great discriminative acuity or differentiative precisionbeing required of the child. Later, as the educational process gets underway, reinforcement is more sparingly doled out, being given only whenhighlyspecificresponsesareemittedinequallyspecificstimulussituations.In fact,onemightsay that theeducationalprocess itself is largelyamatterof establishing connections between well-discriminated stimuli and well-differentiatedresponses.

In the differentiation of verbal responses, a frequently utilized trainingprocedure is that in which reinforcement is made contingent upon theresemblance of a response to an SD that is itself the verbal response ofanotherperson.We see thiswhenever a child utters aword in imitation ofoneemittedbyaparentorteacher.TheimportanceofthisSDcanhardlybeappreciatedbyonewhohasneverobservedthedifficultywithwhichtotallydeaf children learn to speak intelligibly. The very nature of their defect

prevents such children from matching their own vocal productions withheard models. A substitute, for the deaf, is provided by vibratory ormovement cues given to the child when his hand ispressed to the jawregion of a speaking person.This stimulation may be matched to someextentby thevibrationsarousedin thechild'sownspeechapparatusduringvocalization; but thematch is far from perfect and other cuesmay also beemployed—for example, a visual matching of lip-movements through thechild'suseofamirror.Even then,however, the reinforcementprovidedbythe teacher's approval of the approximated sound-pattern remains, for theseunfortunates,thebasicfactorintheeducativeprocess.

FIG. 46. Actual records of the formation of a differentiated responsechainby twostudents learning to sendaMorseCodesignal.Oneach trialthe student tried to imitate with his own telegraph key the signal assounded for him by a professional. The figure shows only the first 12attempts by these students, who were without any prior experience inreceiving or sending the code, but who rapidly achieved an acceptablesignal.(CourtesyofM.P.Wilson.)

Training in speech differentiation has a close parallel in the procedurethroughwhich radio operators learn to transmit code signals. Here, too, awell-accepted teaching device is that in which the student attempts toreproduce a model signal transmitted by the instructor. In the absence ofthisauditorySD, progress is seriously retarded, and even a highly trainedoperator may find his task upsetting when no tone-patterns or associatedclickings result from his movements.Although experimental evidence islacking, there iseveryreasontobelieve that,undersuchcircumstances, theintelligibility of his transmission suffers appreciably.As compared withspeech differentiation, of course, this skill requires very little precision ofmovement,butthefundamentalprocessappearstobethesame.Ifadequateresponses are to be established or maintained in the absence of heardmodels, the presentation or withdrawal of reinforcement becomes all-important.

NOTESMuenzinger'sobservationsofvariabilityinthelever-pressingresponseof

guinea pigs were confirmed in a later study byMuenzinger, Koerner, andIrey(1929).Usinga lever thecross-barofwhichwasremoved toeliminatetwo-paw responses, they noted the persistence of variability in a singlemode of response during 600 solutions of the problem.Three pigs weregiven"guided"training,withonlyright-pawresponsesreinforced,andthreeweregiven"unguided"training,any effective responsebeing reinforced.Asyoumightexpect,otherthanright-pawmovementsdecreasedduringguidedtraining, but the amount and change of right-paw patterns was about thesame for the two groups.One importantdifference appeared: therewas, inthe guided group, a greater development of "accessory" movements—useless responses that accompanied or preceded the effective ones.Thus, aheadmovementmight accompany lever pressingwith the paw, or a seriesof light taps might precede a strong downward push. This unnecessarybehavior is attributed to thegreaterprecisionof response required from theguided group, and is likened to the grimaces and tongue-twistings of achild when learning to write. Similar observations have, indeed, beenreported of human behavior in learning situations where the tasks were ofconsiderable difficulty, and it has been noted that the amount of suchaccessory respondingdecreasesas the tasknearscompletion (Stroud,1931;Ghiselli,1936).

Entertaining items, suggestive of response induction, often turn up inunexpected places.There is, for example, the tale of the British surgeonwho trained his right-handed pupils to perform operations with the lefthand, and who is reported to have said: "Train the left and the right willlook after itself." And there is also the report that an increase in themusculardevelopmentofonearm, as a resultof special exercise,will lead

toanincreaseddevelopmentoftheother.Beforeinterpretingthistomeanadouble return for work done, we should remember that, according to theinductionprinciple,theexerciseitselfwasnotrestrictedtoonearmalone!

Hull's(1943)conceptionofthewayinwhicharesponsemayproduceitsownmomentary depression in strength is attributed by him to two of hisformerpupils.MowrerandMiller.Mowrer'sversionoftheideaisfoundina collaborative study with Jones (1943); Miller's is presented in a book,Sociallearningandimitation (1941),ofwhichDollardwasco-author.TheMiller-Dollard book is a good companion-piece to the present text, beingone of the early approaches to the viewpoint here outlined.The Mowrer-Jones study dealt with resistance to extinction as a function of the forcerequired of rats in depressing a bar. Their findings point to an inverserelationbetween thevariables—thegreater theforce thefewer theresponsesduring extinction.Solomon's (1948) reviewof the entire problemcontainsmuch more material than we have cited. For example, he connects theprinciple with studies of work and fatigue, extinction and spontaneousrecovery, discrimination (by rats) of pathway distances and inclinations,psychophysical judgments,and the tendencyofhumanbeings toavoid therepetitionofidenticalguesses.

Researchers and theorists whose names have often been connected withthe 'lawof least effort' areWheeler (1929),Gengerelli (1930),Tsai (1932),Lorge (1936),Waters (1937), Crutchfield (1939), and Zipf (1948). Hull'sinteresting treatment of the concept is to be found in hisPrinciples ofbehavior(1943).Sometellingcriticismsof 'leasteffort'asabasiclawmayalsobefoundinGuthrie'sPsychologyoflearning,1935.

7

CHAINING

IN LEARNING [the Lord's Prayer] we repeat it; that is we pronounce thewords in successiveorder, from thebeginning to theend.Theorderof thesensations is successive.Whenwe proceed to repeat the passage, the ideasof thewords also arise in succession, the preceding always suggesting thesucceeding, and no other.Our suggestsFather, Father suggestswhich,whichsuggestsart;andsoon,totheend.Howremarkablythisisthecase,any one may convince himself by trying to repeat backwards, even apassagewithwhichheisasfamiliarastheLord'sPraver.

JamesMill,AnalysisofthePhenomenaoftheHumanMind,1829

ThePrincipleofChainingThe quotation above, from a famous British thinker, gives us a ready-

madeintroductiontothecentralthemeofthischapter.Whenstrippedofitssubjectivity, it constitutes a clear recognition of the fact that responsescommonlyoccurinseriesratherthanasisolatedbehavioralunits.Statedintermswithwhich you aremore familiar, it amounts to this:one responsecommonlyproducesthestimulusforanother.

Allalongintheprecedingpagesofthisbook,wehavetreatedthesingleresponseorthesinglestimulus-responserelationastheprincipaltopicwithwhich psychology is concerned. Except for a brief consideration ofcompound stimuli inChapter 5, we have been careful to postpone thediscussion of more complicated matters until the simple ones had beenmade as clear aswe couldmake them.This has been a necessary approachandtheoneemployedinallscientificexposition,but itmayhaveoccurredto you that our treatmentwas incomplete—thatwe failed to recognize thefact thatonestimulus-responserelation isseldomisolatedcompletelyfromthosewhichprecedeor follow it.Sucha criticism is justifiedalthough, incarrying out our scheme of presentation, it would have been premature todiscussthematterbeforewereachedthischapter.

Ifyouweretodescribeindetailthebehaviorofawell-conditionedwhiterat when placed in the bar-pressing apparatus for a period of regularreinforcement,youmightcomeoutwithsomethinglikethis:

Heranimmediatelytothefrontoftheboxwherehecametoastopinapositionfacingthebar; thenheraisedhimselfonhishindlegsandputhisforepawson thebar;withpaws inplace,hegave aquickdownward thrustwhich depressed the bar sufficiently to activate the food-magazine anddischarge a pellet of food into the tray below; then he lowered himself to

thetrayandseizedthepelletwithhisteethandpaws,afterwhichhesettledbackonhishaunchesandbeganeatingthepellet.

Inthisfairlyaccuratedescription,thereareatleastsixdistinctreflexesuponwhich good observers might be expected to agree.Without going into anunwieldydegreeofspecification,wemaylistthemasfollows:

Reflex Stimulus Response5 Bar-location Approach4 Visualbar Rising3 Tactualbar Pressing2 Apparatusnoise Lowering1 Visualpellet Seizing0 Pellet-in-mouth Chewing

Thediscriminativestimulifortheseresponsesarenotaseasilyidentifiedas theresponses themselves.This isespecially thecasefor theendreflexesintheseries(bar-locationandpellet-in-mouthareratherill-definedstimuli),but itholdsalsofor reflex1(thepelletmightbe touchedorsmelledratherthan seen) and, to a lesser degree, for all the rest. It must be rememberedthatwearedealingwith a seriesofoperants the stimulus control ofwhichis never as strictly or clearly defined as when we deal with respondents.Control may, of course, be established—we have shown this inChapter5—andwearesafelyconservativeinassumingtheoperationofSD's in thepresentsituation,butahighlyspecificdesignationofstimulicanhardlybemadeintheabsenceofsuitabletests.

The responses of our list are not exactly equal in status.The approachresponse of reflex 5 will not often occur as a member of the series underconditionsofregularreinforcement.Afterhisfirstbar-pressingresponse,therat is in position for the next response andwill usually remain so duringmost of an experimental session.When the food has been swallowed, hewillusuallyraisehimselftothebarimmediatelyandpressagain.Wemightevenexpect that the ingestionof food itselfwouldcometoserveasanSDfor the reinstatement of the rising response. The other responses of theseriesare,ofnecessity,alwayspresentandmayeasilybeobserved—withasingleexception:thechewingresponsemayoccasionallybeabsent(aswhentherat loseshispellet)ordifficult todetect (aswhensoft food isusedandeatingsoundscannotbeheard).

Inadditiontothesesixreflexes,wecouldhavesuggestedmore.Reflexes

of ingestion are known to follow the chewing response; and "approach" ismadeupofseveralreflexesratherthanone.Ifweattemptedtospecifythemall, it would be a difficult technical task and no purposewould be servedhere. In fact, for our present purposes, it will be simpler and equallylegitimate tomove in theopposite direction and to consider the followinglistasrepresentative.

Reflex Stimulus Response4 Visualbar Rising3 Tactualbar Pressing2 Apparatusnoise Lowering1 Visualpellet Seizing

This reduction leaves us with a clearly observable and regularly recurringsequence or chain of responses, with stimuli that are effective underordinary experimental conditions.Yet we know that a visual bar is not anecessity,sinceadarkenedboxdoesnoteliminatebar-pressing;similarly,avisual pellet is not essential to eating behavior, because a blind ratwouldalso eat.We know, too, that, in addition to noisemade by the apparatus,such as the click of the food-magazine, a discriminative cue may beprovided by the sound of the pellet as it drops into the food-tray; andweknow that all of these auditory stimuli may be eliminated or reduced inintensitywithoutdisrupting thebehavior sequenceappreciably.CompoundSD's are present at every stage of response; and these SD's may beexteroceptive, originating outside of the organism, or proprioceptive,arising directly from the muscular movements themselves, that is, insidethe organism. The important fact is that each response is undoubtedlydependentuponsomeformofdiscriminativestimulation.

Theparadigmshownbelowmayhelpyoutoclarifytheoperationofourreducedchain.Thisparadigmmaybereadasfollows:

SD4(thevisualbar) leads toR4 (the rising response).This is followedby

SD3(the touchof thebaron thepaws),which leads toR3 (pressing).The

pressing is followed by SD2 (apparatus noise), which leads to R2 (the

lowering response); and this response is followed by SD1 (sight of the

pellet),whichleadstotheseizingresponse,Rl,whichinitiateseating.

ChainsandNewUnitsHere, then,wehaveachainof reflexes inwhicheach responseproduces

the stimulus for the next.A striking aspect of such a chain is the over-allsmoothness of transition from one link to another. Indeed, a well-conditionedratgivestheappearanceofmaking,notfourresponses,butone.There are no pauses, no hitches, no jerks in the sequence. One responseseems to floweasilyandrapidly into thenext. It isnothard tounderstandwhyanobserverof theanimal should speakofa singleact of bar-pressingin afield of stimulation. It is easytooverlook thegenesisof thebehaviorfrom a series of discrete units linked into a continuous and efficientperformance.

The multi-membered nature of the act can be demonstrated, however,even after strong conditioning. Suppose, after considerable training, wepermit an animal to press the bar and produce the apparatus noise, butwewithhold the food.This is the usual extinction procedure, and you knowwhatkindofextinctioncurvetoexpect.Intermsofourchainingparadigm,wehavebrokenthesequenceatapointbetweenR2andS

D1.Thefirstthreereflexesoccurwithdecreasingfrequencyuntilaverylowlevelofstrengthisreached. But what happens to the fourth reflex, the pellet-seizing? Has itbeen extinguished too?You can guess the answer: it has not.The ratwillstill seize any pellet that is placed before him.And this suggests that atleastthefinallinkinourchainisfunctionallyseparablefromanythingthatprecededit.

Butwhatabout thebehavior thatpreceded thepellet-seizing;may itnotbeanaturaltotalityratherthanachain?Againwecanputthemattertotest.Supposewe begin extinction by permitting the bar-pressing response (R3)tooccur,butnotpermittingeithertheapparatusnoiseorthefoodtofollow.Now, in terms of our paradigm, we have broken the chain at a pointbetweenR3 and S

D2, one link in advance of the former break. Suppose,further, thatwe carry out this extinction until very few pressing responsesarebeingemittedbyour rat.When this stage is reached, letus reintroducethe apparatus noise but continue to withhold food reinforcement. Now,whentheanimalraiseshimself tothebaranddepressesit,heisstimulatedbythenoisewhichwasformerlyapartoftheSDcompoundthatledhimtodropdownto thetray.Whatwillbe theeffectof thisstimulationuponhissubsequent behavior? The question has been answered—experimentally.Following the reintroduction of the apparatus noise, the almost-extinguishedbar-pressingwillimmediatelyrecoverstrength.Thatistosay,thebar-pressingisreconditionedforatimeandthenyieldsanewextinctioncurve.This is shown in Figure 47,where the arrow indicates the point at

whichtheapparatusnoisewasreinstated.

FIG. 47.The separate extinction of chained reflexes. Extinction of bar-pressingwas begun in the absence of the SD (click) for tray-approach.Atthe arrow, each bar-presswas again permitted to produce the click,with aresultantspurtinresponding.(AfterSkinner,1938.)

Twosignificantfactsemergefromthisresult.First, themereproductionofthepreviouslyabsentSDstrengthenstheresponsesthatcamebeforeitinthebar-pressingsequence.Whyithas thiseffect isaquestion towhich thefollowingchapteraddressesitself,soitneednotbeansweredhere,butthereis no denying that it does act in such a way.The apparatus noise, underthese experimental conditions, clearly serves to reinforce the behavior that,earlier,hadbeenpracticallyextinguished;anditdoessoinspiteof thefactthat it isnot,duringextinction, followedby food reinforcement.Secondly,thenear-extinctionofthefirsttwomembersofthechainapparentlydidnotaffectthethird.Thisanswersourquestionaboutthetotalityofthebehaviorthat preceded the pellet-seizing reflex. Whatever the oneness that thisbehaviorexhibited, itschainedorigin is inescapable.Thebehavior ismadeupofelements thatarebynomeansarbitraryandundemonstrable.Wecansingleouttheseelementsifwetry.

It would, of course, be wrong to say that a closely-bound chain ofresponses does not come to function as a unit.Whenwe break a chain atsomepoint,alloftheresponsesemitteduptothatpointdeclineinstrengthtogether in an orderly fashion. It is this very orderliness that justifies ourspeakingofa responsewhen the entire chain is emitted—althoughperhapsitwouldbe lessconfusing ifwespokeof thechainasanact composed ofseveral responses.Moreover, when viewed by an observer, the elementarycomponents of the act ripple off in rapid succession andwith considerablestereotypyonalmosteveryoccasionoftheiremission.

Ordinarily, in such a chain as that involved in bar-pressing, we recordthe occurrence of only one movement—the actual depression of the bar—

but there is no reasonwhywe could notmeasure the frequency of others.Wemight,forexample,recordtheapproachestothefood-traywhichfollowthepressings,ortherisingstothebarbeforethepressings.Bysuchmeanswe could study in more detail the building-up and breaking-down of thebar-pressing chain. Further investigation will undoubtedly employ suchprocedures; indeed, a number of exploratory attempts have already beenmade(e.g.,Arnold,1947).

An interesting variation in the technique of chaining is shown in anexperiment by Gellermann (1931).This investigator found that monkeys,after many practice trials, could solve adouble-alternation problem inwhichtheyhadtoliftthelidsoftwoadjacentboxes,oneattheleft(L)andone at the right (R), in a sequence LLRRLLRR, for a food reward thatappearedaftereachresponseof theseries.When thischainwasformed, themonkeys displayed an ability to extend the alternation of paired responseswell beyond the point at which the series was ordinarily terminated. In asimilar experiment, Schlosberg and Katz (1943) demonstrated that whiteratscanlearntoalternatebetweentwoupwardandtwodownwardpushesofabarwhen food is given at the endof a four-membered series.These rats,like the monkeys in Gellermann's experiment, were able to make manysuccessivedoublealternations.

Relevantobservations, althoughnotdescribed indetail,haveapparentlybeenmadebythepsychiatrist,Masserman(1946).Hetaughtcats toobtainfoodbypressingapedalthreetimesinsuccessionbeforerunningtoafood-box.When these"threes"werewellestablished, theprocedurewaschangedto permit asingle pressing to secure reinforcement. Later, when this newmanner of responding had replaced the first-conditioned threes, it wassubjected to extinction. As the "singles" fell off in frequency, the catsregressed (here) to the threes. Such observations are interesting andprovocative,andtheyshouldbecheckedwithfollow-upstudies.

In the field of verbal behavior, it would seem that nothing is moreobviousthantheunitarycharacterofchains.Nearlyeverywordonthispageis an apparent example of a unit which was originally no more than avocalized series of unrelated and disjointed elements.What is the spokenwordexample itself but a temporal pattern of sounds that were at firstdistinct and isolated but are now so well chained as to be practicallyirreducibleexceptunder the specialconditionsof training inanalysis?Andthis is not all; larger groupings than individual wordsmay act in unitaryfashion.Forexample, inasmuchas, in fact,and so forth—all possess thischaracter.Still largergroupingsmaybetray thesamecohesion,although toa lesser degree. Complete the following:bread and; salt and

timeand;hueandOr these:asquickasa;as sly as a;asstrongasan;aswiseasan;asquietasaFormostofus,themissinglinkintheseoft-recurringchainsisquicklyadded,althoughthecompletionwillnot be the same for everyperson.Amoment's thoughtwill suggest other,

andperhapsbetter,examples.The problem of the functional unit of behavior,whether verbal or non-

verbal, is one that has not yet received the experimental attention itdeserves.Itwouldappearthattherearemanychainsofbehaviortheunitsofwhicharethemselveschainsofstillmorebasicunitsthathavebeenweldedtogether by the influenceof reinforcing agents.Table IXgives data fromastudy by McCarthy (1930) on the relation between the age of pre-schoolchildrenandtheaveragenumberofwordsusedinfiftysentencesemittedina simple observational situation.The figures in this table, which show aregular increase in sentence lengthwith increasedage, are typicalof resultsobtained by other investigators for these age-levels. In addition, Davis(1937)hasshown that the lengthofspokensentencescontinues to increasewithageuntil,atnineandahalfyears, theaveragesentencecontainsaboutsevenwords.Whenwrittencompositionsaremeasured(HeiderandHeider,1940),thereisevidenceforanincreasefromabouttenwordspersentenceatthe eight-year-old level to about fourteen words at the fourteen-year-oldlevel.All this, of course, is exactlywhatwewould expect if newunits ofverbal behavior were developed on the basis of the chaining of smallerunits.Infact,itispossiblethatthenumberoffunctionalunitsemployedatdifferentagesdoesnotchangeverymuch.Aword,atanearlyage,maybeasingleunit,whereasatalaterage,itmaybemerelyapartofasingleunit.

TableIXTHEAVERAGELENGTHOFSENTENCESINSPOKEN

LANGUAGEASAFUNCTIONOFAGE.(FromMcCarthy,1930)

HomogeneousandHeterogeneousChainsSome writers have distinguished between two kinds of chaining (e.g.,

Arnold,1947). Inone,agivenstimulus is followedbya specific responsewhich produces another stimulus which is, in turn, followed by anotherresponse,andsoon.This type, inwhich the successive linksaredifferent,is calledheterogeneous chaining.The bar-pressing chain is of such a sort,andso is thechain involved insaying"Allmenarecreatedequal."On theother hand, we can imagine a chain composed of stimuli and responseswhicharepractically identical ineachsuccessive link.Thishasbeencalledhomogeneous chaining. The triple pedal-pressing of Masserman's catsapproximatessuchasituation,asdoesthe"rah-rah-rah"ofacollegecheer.

Cases of pure homogeneity in chaining are rare; strictly speaking, theyprobably do not exist. Even the "rahs" of a cheer may be discriminably

different to the listener in some aspect, say their force; and the successiveresponsesintriplepedal-pressingmaydifferslightlyinoneormoreoftheirproperties.Moreover,inmanyinstances,thehomogeneouslinksareclearlybounded by heterogeneous ones in the complete chain.Thus, prior to thefirst of three pedal-pressings is an approach response, and after the thirdthere is an advance to the food-tray. At the verbal level, the rarity ofhomogeneouslinksisreflectedbythescarcityorcompleteabsenceofwordsin any language where the same syllable is repeated more than twice insuccession.InChinese,whichisamonosyllabiclanguage,repetitionofthesameword three times is so infrequentas tobenoteworthy.Anexception,kankankan (Mandarin forlookandsee) is uttered inalmost amonotone.TheShanghaiform,kukuku,exhibitsagreatervariation inbothstressandpitch.

Related to the above is the difficulty encountered when we attempt toteach rats or other animals to solve the double-alternation problemmentioned in the preceding section. Some experiments make use of thetemporalmaze(seeFigure48).Inthismaze,thesubjectmayberequiredtomakealeft-left-right-rightsequenceofturnswhenconfrontedwiththesamechoice point on four successive occasions during his running. Mastery oftheproblemmaybe impossible forsomeanimalsbecauseof thesimilarityof the repeated elements (the SD's as well as the responses) in the chain.The situation resembles that in which a verbal chain is composed offormally similar responses each of which produces the SD for the next."Tongue-twisters" are constructed on this basis.Pluto, the tutored poodle,practised triple pedal-pressing. The difficulty of emitting such a verbalsequence rapidly may partially explain the infrequency of polysyllabicEnglishwords inwhich the same syllable precedes twodifferent syllables.Witness the trouble we have with such a common word asstatistics.Wecan sympathizewith the childwho struggled to say to his father:You putyourtoothbrushinmycupandyouputmytoothbrushinyourcup!

FIG.48.Floorplanofa temporalmaze.E is thepointofentrancefromwhich the animal moves down the central pathway to the choice point,where he may turn right or left. The hinged doors (1, 2, 3, 4) aremanipulated by the experimenter, who prescribes the path by which theanimalcanreturntoE.Reinforcement,suchas food,maybeprovidedatEafter each run, or only after a series of runs, such as left-left-right-right.(AfterHunter,1928.)

It has been suggested by Skinner (1938) andMowrer and Jones (1945)that the procedure ofperiodic reinforcement is conducive to the formationofnewunits,aswhenanumberofbar-pressingsarefollowedbyapelletoffood. Mowrer and Jones carried out an experiment in which resistance toextinction of this response was related to schedule of reinforcement. Fivegroups of rats were used. One group was regularly reinforced duringtraining; one group was aperiodically reinforced; and three groups werereinforced at fixed ratios of 1:1, 2:1, and 3:1 respectively.Their findingsled them to conclude that new "patterns" or units of response could besetup; and that the number offunctional responses during extinction wasnearly thesamefor thedifferentgroupsofanimals, inspiteofan increasednumber ofpressings as the ratios became larger. How far we can gowiththis typeof reasoning isquestionable. It seemsunlikely that the ratwouldbeabletomakemorethantwoorthreebar-pressingresponsesintoasingle

unitthroughchaining,butfurtherexperimentationoughtsoontogiveusananswer. Close observation should readily disclose whether two or morepressingscometoactlikeone,eitherduringfixed-ratiotrainingorduringasubsequentextinctionperiod.

The degree of similarity among the successive links of a chain wouldseemtoseta limit to thesizeofaunit thatcanbedeveloped.Thenumberof different responses available to an organism is by no means infinite.Soonerorlater,asattemptsaremadetoaddlinkstoachain,generalizationand induction will take their toll. More and more discrimination anddifferentiationwill be required, until a saturationpoint is reached andnewlinksareaddedonlyattheexpenseoftheold.

ChainingandDelayedReinforcementFacts sometimes lose their simplicity when analyzed but, at the same

time, get to be more understandable.We now deal with such a case.Atseveralpoints in this text, ithasbeenstated,or implied, thatanoperant isconditioned more rapidly when reinforcement followsimmediately uponemissionof the response.This is supportedby anumberof investigationsand isa truismforanimal trainersandeducators invarious fields.The factthat conditioning is slower as the reinforcement is more delayed has ledsometheorists,likeHull(1943)tospeakofareinforcementgradient'andtospecify the limits of delay beyond which reinforcement is no longereffectiveinstrengtheningaresponse.

Delay of reinforcement can now be seen to involve the formation ofreflex chains.Whenwe say that a bar-pressing response, for example, hasbeen reinforced after a five-, ten-, or fifteen-second delay, what we reallymean is that we haveimmediately reinforced some other response whichoccurredfive,ten,orfifteensecondsafter thebarpressing.Duringthetimeof 'delay,' the animal does not stop behaving; he merely does somethingelse—overwhichwemayhavelittlecontrol.Wesetupa chainof reflexes,of which bar-pressing is an early link. The bar-pressing link cannot bestrengtheneduntilwehavestrengthenedthe laterones,andthis takes time.Eveninwhatwetreatasacaseofimmediatelyreinforcedbar-pressing,thereisat leastone later link in thechain—the response to the food-tray,whichfollows the actual pressing and is conditioned first. The immediatereinforcementforthepressingisofa'secondary'sort(tobedescribedinthenextchapter),butitisnonethelessimmediate(Spence,1947;Grice,1948).

The presence of the last (and first-reinforced) link in such a chain issometimes unsuspected and may go unobserved. For example, when oneconditions awhite rat to turn off a light, in themanner describedonhere,thesituationispresumablyoneof'immediate'reinforcement:thelightgoesoffwithin a fractionof a secondafter thebar-pressing response ismade. Itmaybeshown,however, that thereinforcingeffectof light-removalcomeslater, when the animal is doing something else—holding down the bar,

turning away from it, rising above it, and so on. Probably because of thevisual lag involved, the effect of light-removal is not felt at the instant ofbar-depression. We do not reinforce first the response that we aremeasuring; rather, we make way for a 'superstitious' response to develop,andwearecompelledtoestablishalongerchainthanweintended.

A comparable situation exists in certain sports and games, where thefirst-conditionedresponseactuallyfollowstheeffectiveone.Inbowlingandin billiards, the last link in the chain is the one often referred to as "bodyEnglish"—the posture or the movements which are the trueaccompaniments of a successful throw of the ball or shot with the cue.Similar embellishments may be observed in the behavior of amateurbaseballpitchers,basketballplayers,andgolfers.

FIG.49.The effect of delaying theprimary reinforcement upon the rateofmasteringadiscrimination.Sixgroupsofhungryanimalsweretrainedtomake a white-black discrimination in a discrimination box where the"correct"choice(responsetoSD)ledtofood.TheresponsetoSD,however,wasreinforcedafterdifferenttimeintervals,thedelaysvaryingfrom0to10seconds.Thecurvesshow,forsuccessiveblocksof20trials,thepercentofcorrectchoicesmadebytheanimalsineachgroup.Notethatthelongerthe

delaytheslowerthelearning;until,witha10-seconddelay,theanimalsdonotgetpastachance(50percent)performance.Thisexperimentwascarriedout under conditions especially designed to eliminate any 'secondaryreinforcement' intervening between the choice-point response and theultimate primary reinforcement. This elimination reduced the size of themaximum delay previously thought possible, andwas the effect predictedby the experimenterwho believed that long delays involved the formationof a response chain composedof secondarily reinforcedunits. (AfterGrice,1948.)

ExteroceptiveandProprioceptiveSD'sinChainingThe preceding sections may have suggested to you another distinction

that is sometimesmade between reflex chains. Inmost cases, the obviousdiscriminative stimuli for the successive responses of a chain are in theexternal environment—they belong within theexteroceptive class. Oneresponseputs theorganism in aposition such that anewoutside stimulusis presented.That is, one responseproduces the external stimulus for thenext, as in the behavior of bar-pressing and maze-running.Yet you mayhave surmised that chains are sometimes composed of responses theprincipal stimuli for which seem to beother responses.This seems to beespecially true of verbal behavior.When we repeat a well-worn phrase, itoften seems that oneword leads to anotherdirectly, rather than byway ofthe exteroceptive SD's (sounds) produced by our speech. We emit theaccustomedsequenceinamoreorlessautomaticfashion,andwewouldbehard put to show that each componentword depended upon some outsidestimuluswhichtheprecedingwordhadproduced.Wearemorelikelytosaythat one response led to another without any reference to changes in theexternalenvironment.

Wehaveassumedallalongthatresponsesareresponses-to-stimuli.Now,when we talk about a response thatleads to another, we imply that aresponsecanbeastimulusforanother.Buthowcanwespeakofaresponseas a stimuluswithout dealing in theworst sort of confusion?Fortunately,there is a well-accepted answer to this question. Early in the nineteenthcentury it was discovered that we possess amuscle sense—that there areactual receptor organs within our muscles, tendons, and joints which areexcitedbythemovementsoftheseeffectors.Thiscametobeknownlaterasthekinesthetic (movement-perceiving) sense and to it was ascribed oursubjectiveawarenessof thepositionor changes inpositionof themovableparts of our bodies.As a more objective psychology developed, the termproprioceptors was adopted as a name for these sense-organs andproprioceptivewasappliedtothestimulithatexcitedthem(here-here).

Evidence for the importanceof this typeofmovement-produced stimuliaccumulated throughout the years.Today we have every reason to believe

that responses may producedirectly the discriminative stimuli for furtherresponses, even in the absence of exteroceptive changes. Thus, we arewilling to ascribe to the influenceof proprioceptiveSD's the fact that eye-movementsmayprovidecuesforourjudgmentsthatobjectsareatdifferentdistances; that arm, wrist, and hand movements give the cue for judgingone lifted weight to be heavier than another; that widespread musculartensions may set off the responses that help us to maintain posture andright us when we are off balance; and so on. It is difficult to isolate andobserve the action of such stimuli, because of the locus of the receptorsinvolved and because they commonly accompany the stimulation of othersenseorgans,butwecan feel relatively safe in asserting thatwheneveroneresponse leads to another, and no exteroceptive cue can be discovered, aproprioceptivecuecanbeinferred.

Probably thebest examplesof chaining inwhichproprioceptive stimulimay predominate are to be found in the sphere of verbal behavior.Verbalchains,onceestablished,mayoccurintheabsenceofexteroceptivestimuli.Spokenwords,of course,produce soundswhichbecomeexteroceptivecuesfor thespeakerhimselfand thusplayan importantpart in the formationofchains; but when these chains are once set up they may be emitted sub-vocallyorsub-audibly, inwhichcasetheprincipalSD'sareproprioceptive.Whenwesilentlyrehearseaspeechorutteraprayer,eachresponseseemstodepend upon the preceding one for its emission, which is another way ofsaying that each response is to the proprioceptive stimulus aroused by theprecedingresponse.

CovertResponseTomostpersonsthereisnothingobjectionable in theviewthatovertor

observable behavior may be caused by factors that are hidden from theobservation of others.They will tell you without hesitation that many oftheir actions are internally instigated. They will often report a train of"associated ideas"which led to the performance of some overt action.Butthey will not as readily identify these "ideas" with muscular responses toproprioceptive stimuli. Common sense tells them that there are ideasandactions,eitherofwhichmaylead to theother;commonsensedoesnot tellthem that associations of ideas are equivalent to chains of reflexes whichdiffer from observable ones only in the magnitude of the responsesinvolved.

Yet, throughout the ages, the notion has often been advanced thatthought and action are not always distinct—that some thinking, at least,was no more thaninner speech. Plato, among the ancients, held such anopinion, and so did others who came later. In the nineteenth century,Alexander Bain, one of the founders of modern psychology, treatedthinking as "restrained speaking or acting." Later still, John Watson

describedthinkingexclusively in termsofcovert response,eithersub-vocaltalking or "sub-gestural" gesturing.He argued that children, at first, thinkoutloudandcometospeaksilentlyonlythroughparentaladmonition;thatadults who are withdrawn from social contacts tend to think, as well asread, aloud; and that deaf-mutes think and dream with their fingers. Hesuggested that sensitive measuring instruments would ultimately disclosetinymuscularmovementsineveryinstancewherethinkingoccurred.

Support for such beliefs has come, in recent years, from severalexperimental studies inwhichminute electrical effects or "action currents"of muscular contractions have been magnified and recorded. Jacobson(1932) gave human subjects extensive training in relaxation (to reduce thegeneral level of electrical discharges from the muscle groups to be tested)andthen,withelectrodesplacedintheregionofcertainmuscles,askedeachsubject, while in the relaxed state, to engage in imaginal and thinkingactivities. In one experiment, when the electrodes were fastened near theflexormuscles of the right arm, the subjectwas told to imagine raising acigaret to the mouth.When such imagining was carried out, there was acorrespondingelectricaleffect—aneffectthatwasnotregisteredintheright-arm electrodeswhen amovement of theleft armwas imagined.Similarly,and more strikingly, action currents were recorded from eye-movingmuscles when the subject was asked to visualize such an object as theStatue of Liberty or the EiffelTower. Finally, with electrodes attached totongue- and lip-moving muscles, currents appeared when the subjectsengaged in mental multiplication or when they recalled, in thought, thewordsofasongor themeaningofsucha termasincongruous oreternity.These and related observations led Jacobson to conclude that covertmuscularresponsewasanessentialconditionofthoughtandimagination.

The experiments of Max (1935, 1937) give dramatic support toJacobson's conclusion. Max used as his principal subjects nineteen deaf-mutes who were adept in sign-language. By attaching electrodes to bothforearms of a subject simultaneously, he was able to get action-currentrecords of finger and handmovements under conditions inwhich no overtresponsecouldbedetected.Recordsweretakenwhilethesubjectwasinthewaking state preparatory to sleep, and while he was actually asleep.Withrespect to the sleep records, Max discovered that when strong action-currents appeared and his subjects were awakened immediately thereafter,they almost invariably reported havingbeen aroused froma dream—whichtheyoftendescribedingreatdetail.When,however, theywereawakenedintheabsenceof thesestrongdischarges, theyalmostneverreportedthat theyhad been dreaming. (Contrary to the popular belief that fairly lengthydreams may take place in only a few seconds of time, Max found that"dream reactions" in his experiment usually involved two and one halfminutesormoreofresponding.)Whenmildtactualstimuliwereappliedtothe subjects during sleep, covertmuscular responses followed in about 65

per cent of the cases. In some instances, these stimuli were sufficient toinitiate "dream reactions." Moreover, when deaf subjects. in the wakingstate, were given various "abstractthinking" problems to solve, theyshowed a far greater degree of covert finger and arm movement than didnormalsubjectsunderthesamecircumstances.

FIG. 50.Action currents from arm muscles during real and imaginalweight lifting.The upper curve is for the amount of electrical dischargewhen the weights indicated on the abscissa were actually lifted.The twolower curves are for imaginal lifting of theseweights, one for timeswhen"clear"imageswerereported,andonefor"fair."(AfterShaw,1940.)

AdditionalsupportforaresponseinterpretationofmentalactivitycomesfromastudybyShaw(1940).Thisinvestigatortrainedanumberofnormaladults to reportupon the relativeheavinessof several smallweightswhichthey lifted one at a time in random order during several experimentalsessions.These reports presumably depended upon proprioceptive stimuliarousedby the liftingmovements themselves—that is, the lifting responseprovided theSD for theweight judgment.But this is not themain point.When his subjectswerewell trained in lifting the differentweights, Shawused the electrical recording technique and found consistent differences inthe amount of action-current produced when the subjects were asked toimagine lifting each of the weights. The amount of electrical dischargevaried togetherwith the amount ofweight that the subjects imagined theywerelifting.

TheseresultsareinaccordwiththoseofEwert(1933)whomeasuredtheeye-movements of college students during the reading andrecall (oral orsilent) of stanzas fromJohnGil-pin'sRide. Ewertwas led to his study byobserving, in one of his students during an examination, eye-movementswhich seemed to resemble those made in reading a textbook assignment!His experimental results showed that, in oral or silent recall of previouslyreadmaterial, the frequencyofmovementswasapproximatelyequal to thatobserved in the original reading.While this studywas not concernedwithtruly covert responses, since the eye-movements (ordinarily unnoticed)couldbesubjected todirectobservation, itdoespoint to the importanceofresponseinwhatiscommonlythoughtofasan'imaginal'sortofactivity.

CovertResponseinCodeLearningWe have, then, considerable evidence pointing to the existence, in

human beings, of stimulus-response relationships which are ordinarilyhidden from objective observation. In view of this evidence, we do nothesitate toaffirm theexistenceofchainsof reflexes inwhichat least someof theelements are covert.Even ifwe refused to trust introspective reportsofsuchchains,wewouldprobablyfeeljustifiedinassumingtheirreality.

Theassumptionofcovertchainingservestoclarifyandbringwithinthesame conceptual framework a numberof apparently unrelatedobservations.Let us consider, in the present section, the process of learning to receiveMorse Code. If, under the code-voice method of training described inChapter 5 (here), one observes carefully the behavior of a beginning code-

student, two things soon become obvious. (1) The time required forresponding to a signal by printing a character (letter or digit) is graduallyreducedaspracticecontinues.Earlyintraining,thestudentmakesuseofallthe time available between the presentation of the signal and theannouncement, by the instructor, of its name. Later on, the appropriateresponse comes more and more quickly, well in advance of theidentification.The latencyof the responsebecomes shorter and shorter. (2)Alongwith the decrease in latency goes another change.A studentmay atfirst exhibit amarked degree of apparently superfluous activity during thelatent period.After the signal sounds, and before he prints his character,various interesting reactionsmayoccur.Hemay tapwithhis pencil or hisfoot,makenoddingmovementswithhishead,whistlesoftlytohimself,ormurmursomeequivalentofthesignal,suchas"di-dah"or"di-di-dit."Onlyaftersuchactivitydoeshemaketheprinting-response.Gradually,however,there is an abbreviation of this activity, until overt intervening responsesmaynolongerbeobserved.Yet,evenat thispoint, thestudentmayreportthat he thinksabout the signal before he makes his final reaction to it.Thus,hemayechoit,visualizeit,orrespondinsomeothercovertfashion.Ultimately, this covert activity also disappears, and hemay report that heprintsor"copies"hischaractersautomatically,withoutanythoughtofwhatheisdoing.

A tentative analysis of such changeswould run as follows. In the firststage of code-receiving, the signal occasions various responses, sometimesovert,whichserveinturnasstimulifor thecopyingresponsethatendsthesequence. Later, these intervening responses become covert, although stillpresent asmembersof the chain.Finally, they are eliminated entirely, andthe observed decrease in latency is thereby made possible. The latencydecreasemay probably be explained in terms of two principal factors: thefailure of long-latency responses to receive reinforcement, and theaccumulation of negatively reinforcing effects produced by the continuedrepetitionof(unnecessary)interveningactivity.

CovertResponseandReactionTimeInourtreatmentofreactiontimesinChapter5,wehavenotedtheclassic

distinction, by Donders, of theA-, B-, and C-reactions.We might alsohave considered another type, called byWundt the D-reaction, which wassaid to be distinguished by the fact that it involved arecognition oridentification of the stimulus prior to the overt response. What wasessentiallythisD-reactioncamelatertobecalledthesensorial reactionandwas contrasted with themuscular reaction (the successor of Donder'sA-reaction). In the sensorial reaction procedure, a stimulus, say a sound, ispresented and the subject responds only after it has beenapprehended;whereas, a muscular reaction requires the subject topay attention to theresponseonly.ThesensorialreactiontimewasfoundbyLange(1888)tobe

aboutonehundredmillisecondslongerthanthemuscularreactiontime.If you have followed our discussion and are ready to consider the

recognition, identification, orapprehension of astimulus as basically thesamesortofphenomenonasthinkingaboutastimulusbeforerespondingtoitovertly,youcanseewhywedeferredthetreatmentoftheD-reactionuntilthis time.The reason for the lengthened latency of the sensorial reactionbecomes thesameas thatwhichholdsback theprogressof thebeginner incopyingMorseCode.An interveningcovert response—anextra link in thechain—isinvolvedinbothcases.

An additional factor adds weight to this analysis. In the history ofreaction-timestudy,thedistinctionbetweensensorialandmuscularreactiontimeprovedverydifficulttomaintainunderexperimentalconditions.Whensubjects were repeatedly asked to make the sensorial reaction, their timestended to approach closely the values obtained for the muscular reaction.Practiceseemedtodecreasethedifferencebetweenthetwo.Justasthecodestudent comes to respond automatically to the presented signal with aminimal latency, so does the laboratory subject find it all too easy toeliminate thecognitive activity which characterizes the sensorial reaction.Our explanation of the two changes would be much the same. It isunrewardingworktomaintainthecovertmemberormembersofthechain.

TheContextTheoryofMeaningOurmention, in theprecedingsection,of suchmattersascognition and

apprehension, leadsquitenaturally into thediscussionofanotherproblem,that ofmeaning. Common sense tells us that objects, ideas, and actionsordinarilymean something; and the psychologist is not infrequently askedtogivehisexplanation,tosaywhy.

One answer to this question is that offered byTitchener (1915), in theearly years of the present century, and it may be of more than passinginterest to consider his formulation and the bearing of our own positionuponit.Titchenerwasanintrospectionistinpsychology,whobelievedthata man's conscious processes were basically analyzable into elements,suchassensations andimages; that these elements were usually compoundedwithinperceptionsandideas;andthatanobviouscharacteristicofthelatterwas themeaning they possessed. (It was, of course, assumed thatsensations,images,perceptions,andideasweresubjecttotheintrospector'spersonalobservation.)

"Meaning,"saysTitchener,"... isalwayscontext,onementalprocess isthemeaningofanothermentalprocessifitisthatother'scontext."Contextitselfisnothingmorethan"thefringeofrelatedprocessesthatgathersaboutthecentralgroup (or "core")of sensationsor images." Inperceptionand inidea there iscore plus context, and the latter "carries" the meaning of theformer.

Titchener offers a number of illustrations to show the wisdom of thiscore-contextdistinction.Contextmay, in somecases,be stripped from thecore—aswhenwerepeataloudsomeworduntil thecontextdisappearsandthewordbecomesmeaningless;contextmaybeaddedtocore—aswhenwelearnthemeaningofsomestrangedesignorforeignword;contextandcoremaybedisjoinedintime—aswhenweknowwhatwewanttosaybutneedtime to find expressive words, or when the point of a joke is delayed inappearance; the same core may have several contexts—as implied in ourworry about the true meaning of a chance remark...; and so on. (Keller,1937)

Titchenerhad stillmore to sayabout thematter.Meaningashe saw it,is originally derived from an organism's movement: context is initiallykinesthetic or muscular sensation, such as that aroused by our bodilyorientation toward a stimulus. Later, the contextmay be visual, auditory,and so forth. Ultimately, however, the meaning becomes almost entirelyverbal—when the context is added by something we say (to ourselves)about a stimulus. Or it may even reduce to a "brain habit," without anyconsciouscontext;werespondautomaticallyandappropriately,butwithoutanyfringeofconsciousprocesswhichsupplementsthecore.

However strange this theorymay sound to you at first, a little thoughtwillmakeitlessso.WhatTitchenercallscontextisnomorethanachainofcovert responseswhich intervenes between the initial, observable stimulusand the final, overt reaction. Recall, for a moment, our code-learningexample.Acodesignalhashardlyanymeaningfor thestudentat thestart;itevokesnomorethanakindof"what is it?"response.Later, itsmeaninggrows as the intervening responses (the "di-dahs," the visualizing, etc.)come in. Finally, as these responses drop out, the 'meaning' of the codedecreases. Clearly, intervening covert response is the counterpart ofTitchener's"consciouscontext."

Youcangoonwiththetranslation,pointbypoint—itisnotdifficultifyouhaveunderstood themain theme.The translationwillnotprovideyouwith a complete theory of meaning but it is encouraging to find that anobjectivescienceofbehaviorisablesoeasilytomakeuseofthefindingsofpre-behavioralintrospectivepsychology.

TheAssociationExperimentRelated to the principle of chaining, as well as to the reaction-time

study, is the well-knownassociation experiment—a technique which hasbeenusedextensively inpsychological laboratories andclinics since1879,whenSir FrancisGal-ton introduced the basic procedure.Galton presentedhimselfvisuallywith75stimuluswords,oneata time, fromapreviouslyprepared list, andmeasuredwith a chronograph the time elapsing betweenhisfirstviewofeachwordandthearousalofthefirsttwoideaswhichwere

suggested by it. After each stimulus-response sequence of this sort, heundertook to examine thenature of the ideas (whether they werevisualizations, verbalizations, or some other acts) and theirorigin in hisown life history (whether from boyhood and youth, from subsequentmanhood, or from more recent experiences). By repeating his list fourtimes, at one-month intervals, he was led to the following conclusionsamongothers: (1)45percentofhis responseswerepurelyverbal,32.5percentwere in "sense imagery" (e.g., visualizations), and 22.5 per centwere"histrionic" (postural); (2) the most frequently recurring ideas in the fourtests dated fromhis boyhood; (3) the average time estimated for the recallof a single idea was1.3 seconds; and (4) some of the ideas recalled wereunfitforpublication,sincetheylaidbarehis"mentalanatomy"!

In the modern form of the association experiment, words are stillemployedas stimuli,but theyarealwayspresentedbyanotherperson, andthetimemeasuredisusuallythatwhichelapsesbetweenthestimuluswordand the subject's overt responsewith the "firstword that comes tomind."As a rule, the stimuli are drawn from standard lists and presented toindividualmembersofvariousgroupsofsubjectspreviouslyselectedonthebasisof theirage,sex,occupation,educational level,andsoforth.Insomecases,theexperimenter'sprincipalinterestliesinthenatureoftheresponsewordsemitted;inothercases,thereactiontimesalsoreceiveattention.

One of the better-known studies emphasizing the kind, rather than thespeed,ofresponseisthatmadebyKentandRosanoff(1910),inwhichonehundred familiar nouns and adjectives were presented vocally to 1,000subjects,mostly adults, whowere selected,more or less at random, fromthe general population.When all the response words were tabulated forthesesubjects,afrequencytablewasconstructedtoshow,foreachstimulusword, the number of subjects who responded in identical fashion.Thus,lightwasgiveninresponsetodarkby427ofthe1,000subjects,night wasgiven by 221,black by 76, and so on. As this example indicates,considerable agreement in respondingwas shown.The individual stimuluswords, however, varied greatly in the number of different responses theyoccasioned.Anger, at one extreme, led to 276 kinds of verbal response,whereasneedle at the opposite extreme, led to only 72.Anger, we mightsay, has many more meanings thanneedle. Moreover, individual subjectsdifferedmarkedly in the frequencywithwhich they tended tomakeuniqueresponses. Some tended to "follow the crowd" while others leaned in thedirection of the bizarre and eccentric, indicating very different histories ofreinforcement.

Many studies of associative reactions have been made withthe Kent-Rosanoffandotherwordlists,usingagreatvarietyofsubjects,anumberofdifferent experimental situations, and several systems of classifying theresponses.Thesestudiescannotbereviewedhere,butafewof thefindingsmaybementionedasworthyof interest.Forexample,children, incontrast

with adults, respondwith (1)more completions of, or enlargements uponthe stimulus words (dark—night, red —wagon, soft—pillow); (2) moredefining responses, often involving severalwords (table—made of wood);(3) fewer opposites (dark—light, soft—hard); and (4) more sentenceresponses (soft—snow is soft). Very young children may often show atendency,notentirelyabsentinadults,torepeatthestimulusword(lamp—lamp).

A thoroughgoing analysis of the association experiment is yet to bemade.The basic principle involved is quite clear, however, and has oftenbeen recognized,openlyorby implication. Itmaybe stated simply. In thehistory of an organism, one verbal response, through chaining, becomesthe SD for another.The stimulus word is part of a chain which has theresponseword as anothermember.A single stimuluswordmay, however,belongwithinmore thanonechain,and the likelihoodofagiven responseto such a word will then depend upon a number of factors, operatingindividually or together. These include: (1) the amount of reinforcementwhich a specific chain has been accorded; (2) the amount of collateralstrengthening due to the presence of certain SD's in the experimentalsituation; (3) the presence of strong motives or emotions; (4) the recencywithwhich the responsehasbeen reinforced; and (5) the amountof energyexpenditureinvolved.

An example may illustrate the possible operation of these factors.Thestimulus word isdark. (1) Adults have long used this word injuxtaposition withlight; children have more frequently followed it withnight; photographers have often combined it withroom; and so on.Different responses have different reinforcement histories. (2)Any one oftheaboveresponsesmaybegivenadditionalstrengthbysomeaspectoftheimmediate stimulus situation.Light may be helped by some conspicuouslightingeffectintheexperimentalchamber;nightmaybepartially initiatedby the gathering darkness observed outside the laboratory window; androommaybe aidedby the dimness of the chamber(This room isdark) oreven by some confining aspect of the situation(Notmuch room here). (3)Forahungrysubject,growingeagerforhismeal,darkmayoccasionmeat;and for a subject in whom dark has become a conditioned stimulus for aviolentfearreaction,allresponsesmaymomentarilybeinhibited.(4)Forasubject who came to the experimental session directly from a politicaldiscussion or talk of a coming election,dark may well evokehorse. (5)Finally, of several possibilities, some responsesmaybe less effortful thanothers.Considerthefollowinghypotheticalresponsestodark:

(1)theforestprimeval(2)complexion(3)light(4)barkClassically these responses might be said to fall within a scale of'meaningfulness-superficiality,'with themostmeaningful at the top of the

list and the most superficial at the bottom. If 'superficial' is taken as theequivalent of 'economical' (as it sometimes is), and ifwe see the reactionsas running from least tomostenergy-conserving, theyalso fitwell intoaneffortful-effortless scale. It is then quite understandable that increasedsuperficiality of response has been claimed by researchers to accompanystatesoffatigueandrepetitionofstimuluslists.

Our list of determining factors may not be exhaustive, but it shouldsuggest toyou that the emissionof averbal response, like anyother,maydependuponavarietyofcircumstances.Moreoften thannotour responsesdepend upon the joint operation of several factors, rather than the isolatedaction of one. Clever experimental design and close weighing ofprobabilitiesmayberequiredtoilluminatetheinfluenceofasinglevariableupon the behavior of an organismwithsuch a complicated past as that oftheaverageadulthumanbeing.

When we turn to the matter ofreaction times in the associationexperiment, we are immediately confronted with a large body of factualinformationmostofwhichaddslittletothepresentdiscussion.Perhapsthetwo outstanding facts, noted by many investigators, are (1) that thereaction-time values in this type of experiment are usually much greaterthan in the caseofword-reactions tonon-word stimuli; and (2) thatother-word reactions are considerably slower thansame-word (repeating)reactions. Such findings are in accord with the fact that covert responsewordsarecommonlyreported to intervenebetween theverbalstimulusandthe (overt) verbal response.Moreover, since any stimulus word may havebeen linkedwithseveralwords inpreviouslyestablishedchains,wewouldexpect occasionalinterference of responses—an expectationwhich is oftensupported by subjects who claim that response words "get in each other'sway." There is also the fact that some stimuli, by virtue of earlier,respondentconditioning,mayservetodepressforashortperiodalloperantbehavior, including the verbal; and this would obviously lengthen thereactiontimesofsomesubjectstosomeofthewordsofastimuluslist.

'Complexes'and'Guilt'It is partially due to the last-mentioned fact that the association

experiment has been used as acomplex-indicator and as aguilt-detector.Carl Jung (1919), the famous Swiss psychotherapist, was the first to seethat 'complexes' of repressed behavior might be brought into the open byprobing with stimulus-words; and he made an exhaustive analysis of theassociativereactionswhichpointedtotheexistenceofthesecomplexes.Hiswork amounted to a recognition of the fact that certain, ordinarily neutral,wordsmight have become emotionally 'charged' for a patient by virtue oftheir connectionwith some earlier experience, generally involvingnegativereinforcement, and that the responses normally found linked with suchwordswere thereby less likely tooccur.Whenpresentedwith the stimulus

wordsofa'free-association'test,thepatientmightrespondtosomeofthem(1) with very unusual, far-fetched, or highly personal words; (2) byrepeating the stimulusword beforemaking an other-word response; (3) by'misunderstanding' the stimulus word; (4) by giving exceptionally longreaction times or no word-response at all; and (5) by obvious signs ofemotional upset, such as blushing, stammering, whispering the responseword, and so on. Jung found that testing and re-testing with carefullychosenstimuluswordsoftenaidedhimgreatlyingettingattherootsofhispatients' troubles, thus paving the road to therapy; and his lead has beenfollowedbymanyclinicianssince.

Theclinicaluseof the associationexperiment is like its use as a 'guilt-detector,'butwithadifference."Thedetectiveknowsthecrime,butnottheculprit; the psychotherapist knows the culprit but not the crime."(Woodworth,1938).Wemayaddthatthe'crime'inquestionislikelytobecrime by courtesy only, since it is more often an affair of the classroomthan of the courts. In the simplest demonstration of guilt detection, twostudentsmaybedrawnfromaclassandgivensecretinstructionsastotheirconductduring,say,aten-minuteperiodoutsidetheroom.Onestudent(the'culprit') is sent into a course of action which is likely to produceembarrassment.Forexample,hemaybeasked toenteraprofessor'soffice,during his absence, and engage in some trivial activity such as arrangingpapers on a desk. In the course of this activity a 'mishap' of some sort isarranged to occur and the 'unsympathetic' professor unexpectedly returns.Theotherstudent issubjected tonosuchexperience,but,at theendof theten-minute period, bothmen are brought back to the classroom and giventhefree-associationtestindividually.Theclassisacquaintedwiththenatureof the 'crime,' but does not know the culprit; and the latter has beeninstructedto hide his guilt if possible.The same set of ordinarily neutralstimuluswordsispresentedtoeachstudent,butseveralofthesewordshavebeenchosenbecauseof their possible significance to the 'criminal' throughhis recent experience. Response words are recorded and reaction times aretaken with a stop-watch. In this 'experiment' the class, as a rule, has nogreat difficulty in detecting the culprit, either by his unusual responses tosuch common words asink, desk, office, Professor, and the like, or byunusually long reaction times. Unfortunately, the validity of this type ofdetectionisnotgreatenoughtojustifyitsexclusiveuseinthedetectionofreal criminals.Whencarefully administeredandaccompaniedbyotherdata(e.g., changes in blood pressure or the galvanic skin response) it hassometimes been successful in evoking confessions, but it has seriouslimitations,sinceitisnotsure-fireandmayevenpointthefingerofguiltatpersonswhoknowaboutthecrimebutdidnotcommitit.

MazesandNonsenseSyllablesA discussion of chaining leads us naturally into two other well-known

spheres of psychological research.One of these has been chiefly concernedwiththeabilityofanimals(whiterats inparticular) torunmazes; theotherhas dealt with the rote memorizing of verbal materials (for example,nonsensesyllables) by human beings. Both have often been cited as casesof "serial learning," which gives you a clue as to the reason for theirinclusioninthepresentchapter.

Hundreds of maze-learning experiments have been carried out since theturn of the century, when the technique was first introduced, with resultsthat often impress observers as small, considering the amount of laborexpended.Whatwasconsidered,at theoutset, tobeavery simple formoflearning,andthepossibleprototypeofhabit-formationinhumanbeings,isin realityverycomplex.Moreover,wenowsee theproblemmerelyasonecallingfortheapplicationofexplanatoryprinciples.

Wearenowable(aswewerenotinChapter3) toenumeratesomeof themore important factors involved in themasteryof anymazehabit.Wecansay,withassurance, that thesolutioninvolvestheestablishmentofa chainof reflexes, initiatedwhen an animal is placed in themaze and terminatedwhen he arrives at his reward. From our analysis of bar-pressing behaviorwe can assert that successivediscriminations anddifferentiations areinvolvedatsuccessivepointsinthemazepathway;andwecansee,asbasictothese,theprinciplesofoperantconditioningandextinction.Also, andofextreme importance, we can assume the operation ofsecondaryreinforcement,afactorwithwhichweshalldealinthecomingchapter.

FIG.51.Curverelating the lengthofnonsense-syllable lists to the timetakentomemorizethelists.Thesubjectwasgivenonetrialperdayuntilagiven list was mastered.The values on the abscissa and ordinate are theactual lengths-of-list and corresponding memorizing-times. Note thepositive acceleration of the curve arising from the ever increasinggeneralizationasthelists(chains)growlonger.(AfterLyon,1917.)

With respect to rote-learning studies, the situation is muchthe same.Since1885,whenEbbinghaus reportedhis famousexperiments, thousandsof students have memorized thousands of lists of words or nonsensesyllables under various conditions and to various degrees ofmastery.Rateof learning, asmeasured by number of list repetitions or time required toreach a given standard of performance, has been related to length of list,meaningfulnessoflistmaterial,spacingofpracticeperiods,andmanyothervariables.Similarly,theretentionofsuchmemorizedlistshasbeenstudiedasafunctionoftime-since-mastery(andthenatureoftheactivitythatfilledthistime),rateofinitiallearning,amountofover-practice,andsoon.

Theproblemofrotememorizationissimilar to thatofmazelearning.It

haspracticalsignificancebecauseit is likesomeofthetasksthatfaceusineveryday affairs. Like maze learning, it presents complex matters foranalysis(Hull,etal.,1940)andinvolvestheoperationofmorefundamentalmechanisms,butwearenotyetatastageofdevelopmentwherewecandomore than suggest the manner of their interaction.The over-all picture isone of verbal chaining, in which the emission of each response comes todependupontheSDfunctionof the response,or responses,preceding it intheseries.

NOTESSome exploratory observations by Charles Ferster, in the Columbia

laboratory, bear upon the problem of 'homogeneous' chaining and theformationofnewresponseunits.Fersterwasabletotrainseveralrats,withsome success, to make paired responses for food reinforcement. Heobserved,amongother things, that,whenpairingwaswellestablished, thetwomembers of the pair were seldom if ever of the same form.The firstresponsemightbewithoneforepawandthesecondwiththeother;thefirstmightbeafairlyprotractedpushofthebarandthesecondmightbeasharpslap; and so on.Heterogeneity of the two links, rather than homogeneity,was the rule. This gives support to our statements in this chapterconcerningtheimprobabilityofpurelyhomogeneouschaining.

More than thirty years ago, Swindle (1917, 1919) published severalreportsonthebehaviorofanimals(e.g.,thepeckingbehaviorofcockatoos)in theBerlin zoo. Such responses, in spite of a high degree ofstereotypy,showedsomevariation in force,evenwhenemittedataveryconstant rate.Swindle also pointed out that what we call areaction time is dependentupon the number of responses that intervene between the stimulus and theresponsethatwehaveselectedforobservation.Weareremindednotonlyoftheassociativereaction, but also of such 'latencies' as aremeasured in therun-wayandthebar-pressingexperiments.

An excellent account of theassociation experiment is given byWood-worth in hisExperimental psychology (1938).You should now be able toappreciatesomeof thefactors involvedin thiskindofexperimentationandfind, in Woodworth's examples, plenty of food for thought and somesuggestionsforfutureresearch.

The "dream reactions" reported by Max (here) to result from mildstimulation of his sleeping subjects, makes more credible many non-experimental reports of induced dreaming. Thus, it has been said thatexposure of a sleeper's feet to cold may evoke dreams of, say, Arcticexploration; that therepeated tippingofasleeper'sbedmay lead todreamsof ocean voyages; that the entanglement of legs or arms in bed-clothesarousesdreamsoffrustratedaction;andsoon.

Early investigations (e.g.,Thorsen,1925)of the relationbetweencovert

responseand'innerspeech'or'thinking'apparentlysufferedfrominadequateexperimentaltechnique.Modernmethodsofrecordingtheelectricalchangesthat arise from covert muscular movement (as in the Jacobson and Maxstudies) open up a large field of research. Studies by Davis (1948) andothers suggest that a frontal attack is possible upon such 'intervening'responsesasthoseinvolvedinassociativereactions,meaning,andthelike.

Oneof thefirstandmostelaborate treatmentsofcovert responseand itspsychological significance is that of Kantor (1924) who found in suchresponse the basis of many allegedly 'mental' activities—perceiving,imagining, thinking, feeling, remembering, and so forth.Kantor's account,althoughnot obviously guidedby experimental findings, hasmanypointsof contact with the one of this chapter.The advanced student will not letthe difficulty of this book deter him from making the effort required toappreciateitscontents.

8

SECONDARYREINFORCEMENT

THECAWINGofarook...initself,iscertainlynotagreeable.Thissound,inthecaseofthosewhohavelivedinthecountryinearlylife,andenjoyedits scenes and its adventures, is well known to become a particularlyagreeableone....The explanation is that thisparticular sound,havingbeenheard again and again among surroundings... which have a markedaccompaniment of pleasure... produces a faint re-excitation of the manycurrentsofenjoymentwhichaccompaniedthese.

JamesSully,TheHumanMind,1892

APauseforReviewThe last chapter's discussion of response chaining has no doubt

sharpenedinyourmindaquestionthatwasformingearlier.Inachain,eachresponseproduces thestimulus for thenext response,butwhyshouldeachproducing response bemade—what keeps the chain together?The roots ofthis question go back to our first descriptions of operant reactions, whereforthesakeofsimplicitythequestionwaslaidasidesothatouranalysisofbehavior could get underway.We have now come to the pointwhere theproblem must be aired and dealt with. In so doing, we encounter a newprinciplewhichfillsakeyplaceinthejig-sawproblemofbehaviorthatwehavebeencarefullyassemblingintheprecedingpages.Thisprinciple,ifwemay name it shortly before explaining it, is that ofsecondaryreinforcement.With its aid, we can make new and material progress inroundingoutourpictureofthatextraordinaryobjectofstudy,thebehavingorganism.

Perhapsthebestwaytocomeattheprincipleofsecondaryreinforcementis by a review of the ways in which a stimuluscan affect behavior. Ifwelookbackoverwhat has been said so far,we can summarize the functionsof stimuli quite briefly. Stimuli may be (1) eliciting, as in respondentreactions;(2)discriminative,inthatthey"settheoccasion"fortheemissionof an operant response; (3)reinforcing, as when a newType S reflex isbroughtintobeingbytheunconditionedstimulus,oranemittedoperantisstrengthened by the presentation of a food pellet. Reinforcing stimuli inType S reactions are correlated with others which then become capable ofeliciting the response. InType R, the correlation is with a response, thecontingency between response and reinforcing stimulus serving to increasetheemissionfrequencyof that response.Wehavenotspokenasyetof twootherfunctionsofstimuli,thedrive-arousingandemotionalizingfunctions,but thesewillbeconsideredin thenext twochapters.Tothe threewehaveenumerated, however, we wish to add as a fourth that of secondary

reinforcement,attimescalledconditionedreinforcement.

ANewFunctionoftheStimulusWe are led to the realization of a new kind of stimulus control over

behavior,byinquiringwhattheeffectuponanon-reinforcingstimulusmaybe when it accompanies a reinforcing one. When, after many suchassociations, we apply the erstwhile non-reinforcing stimulus to theorganism,we look to seewhether its action has undergone a change.Theappropriate experiment can be easily arranged. But before any laboratorytrialismadeatall,areconsiderationofalreadywell-knownexperimentsandobservationsleadsustothehunchthatastimuluswhichisnotoriginallyareinforcing one (or which is not, as we often say, a "primaryreinforcement") canbecome reinforcing through repeatedassociationwithone that is. That is, reinforcing power may be acquired by a stimulusthroughbeingpresentwhenanoriginalreinforcementisgiven.

Workingfromthissamehunch,scientistshavenowestablisheditstruth.In thediscussionof theprinciple that follows,weshall report someof theevidence they marshalled for it; we shall consider some of its recentexperimental extensions; and we shall indicate, at the same time, itsmultifoldimplicationsforhumanandanimalbehavior.

SecondaryReinforcementofTypeSReactionsYoumay recall that in Chapter 2wementioned reports, from Pavlov's

laboratory, ofhigher-order conditioning. In those experiments, conditionedstimuliwereusedasreinforcementsforothersintheabsenceoftheoriginalunconditioned stimulus. For example, a buzzer conditioned to elicitsalivationwas then pairedwith a visual stimulus like a black square, andpresumably sufficed to establish the square as a conditioned elicitingstimulus of thesecondorder although the latterwas never once reinforcedby pairing with the food.The crucial point about these experiments wasthat a stimulus seemed able to acquire reinforcing power throughconditioning,andthereaftertoactindependentlytoconditionanotherreflex.Despite the admitted weakness and instability of this higher-orderconditioning, the phenomenon aroused much discussion amongpsychologists. Behavior theorists appealed to the principle in explainingthosecomplexactsofeverydaylifewhichareneitherinitiatednorsustainedby unconditioned stimuli like food,water, and sex. Littlemorewas doneexperimentallywithType S conditioned reinforcement after Pavlov, but itiswell that theprinciplewasnot lostsightof,becauseaparallelfor itwasfoundinoperantbehavior—onethatassumedevengreaterimportance.

SecondaryReinforcementofTypeRReactionsA stimulus is said to be reinforcing if it possesses the capacity to

FIG. 52. Conditioning of the bar-pressing response by rats withsecondary reinforcement alone. Priorto this, the sound of the foodmagazine was associated with thedelivery of pellets, no bar beingpresent.When thebarwas introducedandeachpressproducedthemagazinesound, the cumulative responsecurves shown above were obtainedfrom four animals. (From Skinner,1938.)

increaseormaintain reflex strengthabove operant level. In operantconditioning, it must be able toraise the strength of the responseproducing it, or be able to sustainthe strength to some degree in theface of extinction. (In respondentconditioning, it must increase theeliciting strength of a stimuluspaired with it.) Here, then, is thecriterion to be met by asecondaryrei nforcer:it must, throughconditioning, have acquired thepower to condition. Moreover, wemayrecognizeatoncethat,aswithany conditioned stimulus, asecondaryreinforcerusedrepeatedlywithout further association withprimary reinforcement willextinguish, that is, lose its powertoreinforce.

We may summon a fewintroductory examples todemonstrate the operation ofsecondary reinforcement.Remembering that the signs of aneffective operant reinforcer are anincrease in emission rate, or theretardation of extinction, thefollowing experiments illustrateboth of them. Thus, Grindley(1929) found that chicks increasedtheir speed of traversing a runwayduring the first trials of anexperiment in which the onlyreinforcementwas the sight of ricegrains from which they wereblocked off by a sheet of glass.These chicks were accustomed toriceas food, and in the experimentthevisualstimulusaloneactedasasecondary reinforcement for

strengthening the running response.Williams (1929) found that rats, dailytrained in using thewhite compartment of adiscrimination box as a route

to food,wereable tomaster a fourteen-unitT-mazewithnoother terminalreinforcement than thewhitecompartment itself.Asimilar resulthasmorerecently been reported by Saltzman (1949); and other investigators haveshownthatanimalswill learnmazesorotherresponsesif thereinforcementprovided is simply a return to their living-cage homes. Bruce (1932)showed that theextinctionof a simplemazeperformancewas slowerwhenfoodwas present though inaccessible (behindwiremesh) at the end of therun than when no food at all was present. Bugelski (1938) got slowerextinction of bar-pressing in ratswhen each response produced an auditoryclick which had regularly accompanied food-reinforced bar-pressing duringconditioning,thanwhentheextinctionresponsesproducedneitherclicknorfood.

Still other examples of secondary reinforcement come to light if wesuddenly bring to bear upon an extinguishing response a stimulus whichhad previously been correlatedwith reinforcement.Thus, if bar-pressing isbeingextinguishedinarat,andwhentheresponseratehasfallenquitelowwe reconnect the empty magazine so that a response now produces theaccustomedclicks(butstillnofood), theextinctioncurveshowsaburstofresponses.This effect wasmentioned in the discussion of chaining in thel as tchapter (here), and it is a neat case of secondary reinforcement.Notterman(1950),usingratsassubjectsinarunwayexperiment,duringthetrainingcorrelateda lightwith foodas the terminal reinforcement for runs.Later, duringextinctionof the running response, he re-introduced the lightattheendofruns,andfoundthatitactedasasecondaryreinforcersincetherunning speed quickened significantly and durably. Notterman ran controlanimals to prove that it was not just any stimulus change at the end of arun which could be expected to produce this effect, but that the stimulusused as a secondary reinforcer needed to have a historyof associationwithprimaryreinforcement.

SDandSr

The reader who has trained himself by this time to raise his ownquestions about details of behavior, will not have failed to ask what wemean by "a stimuluscorrelatedwith reinforcement" in describing an Sr.(We shall from now on use Sr to denote a secondary or conditionedreinforcement; the symbol SD, for a discriminative stimulus, is alreadyfamiliar.)Whatkindofa"correlation"mustitbe:whattemporalrelationtothe reinforcement, and what sort of association with it? That merecontiguity with reinforcement is not sufficient to make an Sr out of aneutral stimulus is shownby an experiment ofSchoenfeld,Antonitis, andBersh(1950a).Thesementrainedanimalstocometothecage'sfoodtrayatthe sound of a pellet falling into the tray.After an animal had seized his

pellet, and while he was eating it (which took about ten seconds on theaverage), a light was turned on for one second. One hundred such light-pellet associationswere given each animal, and then the possible Sr valueofthelightwastestedbyseeingwhetherbar-pressingcouldbeconditionedby giving only the light as reinforcement. The net result of theseassociations was zero, and since this type of correlation did not work, itbecame clear that one had to discover justwhat kinds of associationcouldgive rise to a secondary reinforcer.Though not so plentiful as we wouldwish,alltheevidencewehavesofarpointsinthesamedirection.In ordertoactasanSrforanyresponse,astimulusmusthavestatusasanSD forsomeresponse.

The foregoing sentence, standing alone, is not easy to grasp right off,andthreepointsofexplanationprobablyoughttobemade.

1. From the viewpoint of procedure, the training conditions forobtaininganSr are the same as those for obtaining an SD.Wemust firstmakeanSDofthestimuluswewishtouselaterasanSr.

2.Theresponseforwhichthestimulusisspecificallyused asSD is nottheonlyone forwhich itcanbeusedasanSr.For example, in a chainofheterogeneous reflexes, a response which produces the SD for the nextresponseistherebystrengthened,thatis,theSDforthenextresponseinthechainactsasanSron the responseproducing it even though that responseis different from the next. It is worth recalling here that in Skinner'soriginal conditioning experiments the same feature appeared.His ratswerefirst trained to come to a tray for a pellet at the sound (SD) of thedischargingmagazine.Afterbeingwellaccustomedtothisreaction, thebarwasintroduced,andpressingwasconditionedbythemagazinesoundwhichthe rat himself produced by his pressing. In these instances, we see thestimulusactingasa reinforcingandadiscriminativeoneat the same time.We shall shortly return to this matter in discussing how chains arecementedtogetherandthe"generalityofSr."

3. The equality and interchangeability of the SD and Sr values of astimulusaresuggestedbyanexperimentofDins-moor(1950).Thisworkergave several groups of rats the same amount of light-dark discriminationtraining,usingthebar-pressingresponseandfoodreinforcement.Aseriesofextinction sessions followed in which the SD established during trainingwas used discriminatively (i.e., preceded the response, and "set theoccasion" for it) with one group of subjects, and reinforcingly (i.e., wasproduced by the bar-pressing responses)with another group.No differencein extinction responding occurred between these two groups.Moreover, in

continuedextinctionsessions, theprocedureswere interchanged,so that forone group the stimulus hitherto used discriminatively was now usedreinforcingly, and vice versa for the other group. This interchange ofstimulus functions did not produce any differences in the performances ofthe groups.Apparently, establishing a stimulus as an S D establishes itequallyas anSr, and extinguishing its power to function in oneway alsosubtracts, in like degree, from its power to function the other way.Thisconclusion, tentatively drawnby the investigator, leads us towonder howfar the exact equality between the SD and Sr functions of a stimulusmayextend. Dinsmoor's study points up the question and provides a basis forfurtherwork.

NotonlyaretheSDandSrvaluesofastimulusclosely interrelated,butthe same stimulus can, as in chaining, exercise both functions at the sametime.What, then, is the reason for continuing the distinction between SD

and Sr as "two functions"? The reason lies, of course, in the way theexperimenterisusingthestimulus,thatis,whetherhewishesit,inrelationto some selected response, to act discriminatively or reinforcingly.Once astimulus has been given a measure of control over behavior throughcorrelationwithreinforcement, thatcontrolmaybeused ineitherSDorSrfashion.

That the distinction between stimulus functions on the basis of use isimportanttoretain,canbegatheredfromthefollowingexampleofhowthetwo uses may lead to different results. Suppose we have established astimulus as an SD for approaching and nosing the food tray in the cage.Afterwards,desiringtotest itscontrollingpower,weintroducethebarintothecageforthefirsttimeintheanimal'sexperience.Ifwedecidetotestthestimulus as Sr, we present it after each pressing and observe that theresponse rate goes up, subsequently giving an extinction curve (sinceprimary reinforcement is not used to bolster the Sr's power). But, if wedecidetousethestimulusasanSDforbar-pressing,andsomehowmanageto have it come on before operant-level pressings, we find no increase inrateofpressing,but ratheran increase in therateof theresponsewhich thestimulus followed. Inshort,wherewedealwithanewresponsehavingnoprevioushistoryofreinforcement, theSr functioncanhavea specific effectupon the response,whereas theSD function cannot.This difference in theoutcome of the two uses shows that we are still justified in separatelydiscussing the twostimulus functions.Afterconditioning,a stimulusmayperhaps never be divested of its dual endowment, but it is necessary inanalysisto isolate the way it is acting in a given situation—just as a

materialbodymaypossessbothmassandextentat the same time,butwedonothesitatetoseparatethetwoaspectsforstudyanduse.

MoreaboutChainingWe are now in a position to say a bit more about the topic of the

preceding chapter.A chain of reflexes is built up by having each responseproduce the SD for the next response; it isheld together by the fact thateachSDactsasanSr for the responseproducing it.To indicate thedoubleaction of each stimulus, we can give it a double superscript, as Sr.D,placing the r before theD because its reference is to the prior response,while theD has reference to the following response.The general notationforasegmentofachainwouldthenbelikethis:

If we number the terms in the formula to correspond to the sequence inwhichtheyarebuiltintothechain,weget:

in which Sr0, R0, and SDo are, respectively, the final reinforcement,

finalresponse,andfinaldiscriminativestimulusinthechain; inwhichSr1is the first secondary reinforcement to be "hooked in"; and inwhich (s) isthe operant-level source of the remotest antecedent response in the chain,Rn,totheremotestdiscriminativestimulusSDn.

The preceding chapter also emphasized that the Sr.D terms in a chainmay be either exteroceptive or proprioceptive stimuli. Psychologists havefora long timebeen interested in thesmooth flowandcontinuityof reflexchains, and have been especially impressed by chains based onproprioceptive cues which are hidden from the experimenter's view. Indescribing how chains are maintained and carried through, many writershaveusedsucha termasset.Organismswere thoughtofasbecoming"setfor"ataskbywayofa"preparatoryset,"andasbeingguidedordirectedinexecuting a task by a "continuing set." The term itself was never welldefined but the observations upon which it was based are to be analyzedfrom thestandpointofchaining.The initiationofchainsbydiscriminativestimuli, and the sustainingof a sequence by response-produced reinforcingstimuli, are at bottom the real meaning to be attached to "sets," bothpreparatoryandcontinuing.

Inreflectingonthewaysinwhichresponsechainsmaybedisturbed,wenotethattherearetwo.Oneisthatofcuttingorbreakingthechainatsomepoint by withholding the next Sr.D—in short, by extinction of allsegments up to that point.A second is that of a response producing the"wrong"SD,thatis,nottheonenecessaryforgoingonwiththerestofthechain.This is an occasion for "confusion" or "interruption of the set," sothat the completion of the chain is diverted or slowed down, or istemporarily halted while the organism "readjusts," "takes stock of thesituation," or "finds another way" to resume the interrupted activity. Ahoundintentlyfollowinghisquarry throughthebrushwill takesometimetoregainhiscomposureandreturntothetrail ifheunexpectedlypokeshisnose toward a porcupine or a snake.Aman's sleepy-eyedmorning routineof washing is quickly alerted when, after drowsily squeezing a tube, hetastes shaving cream on his toothbrush; and parents are familiar with theoutrageofthechildwhodetectsamistakeinanoft-repeatedsong—aswhenthewearyparenterrs:"SimpleSimonmetapie-mangoinghometobed...."Inthese,asinmanyothercasesofinterruptedchains,thebehaviordoesnotstopcompletely,butisapttotake,atleastforawhile,anotherdirection.

TheGeneralityofSrInour treatmentofsecondaryreinforcementsofar,wehavemadeuseof

a factwhichmaynowbegiven theexplicitemphasis itdeserves. It is thisfact that makes the principle ofsecondary reinforcement so significant forcomplexactivitieslikethesocialbehaviorofhumanbeings.Hereitis:onceestablished, a secondary reinforcement can strengthen other responses thanthat used during its original establishment,and can do so under othermotives thanthatprevailingduringtheoriginal training.Thismeans thatan Sr is not limited in its range of applicability, and that it provides uswith a "purely general" reinforcement for use under any conditions wedesire.Wehavealreadyseen thatanewresponsecanbeconditionedbyanSr,andwearenowaddingtheimportantobservationthatthismaybedoneeven if the motive is shifted (e.g., Estes, 1949a, 1949b). On the animallevel, for example,wecan correlate a stimuluswithwater-reinforcementofbar-pressingbythirstyandnot-hungryrats,andthenusethestimulusasanSr toconditionchain-pullingwhen theanimalsarehungrybutnot thirsty.On the human level, the words "Well done!", or the approval of otherpeople, or money, may act as secondary reinforcers for a variety ofbehaviorsunderanyprevailingmotive.

ThefactofSrindependenceshouldnotmakeusforget,however,that(a)anSr extinguishesor loses its powerwith repeated application if separatedfrom primary reinforcement and(b) aswith any reinforcement, some drive

must be present for the Sr to be effective.We need not always pause toidentify either the primary reinforcements that continue to support thepowerofSr's,ortheprevailingdriveunderwhichtheSrisacting.Always,however, there is implied the assumption that both are present. In a broadway,youmaythinkofthesupportingprimaryreinforcementsasbeingsuchthings as our daily food and drink, or relief of pain or anxiety; and theoperative drives as occasional hunger and thirst, sex drive, and aversivetensions or fears. In a hypothetical environment shorn of primaryreinforcements, or in which organisms are devoid of motives, Sr's wouldeitherextinguishcompletelyorneveract.

Students sometimes offer examples of a social stimulus apparentlyseparated from its primary reinforcement which nevertheless indefinitelyretainsitscontroloverbehavior.Motherisnolonger,seemingly,relatedtofood, comfort, and so on, yet she remains a strong controlling stimulus.Butwemaynotethatifmotherisnolongerpresent,shecannotextinguishas an Sr; if she is present, she often remains associated with primaryreinforcement (like "Thanksgiving dinner with the folks"), or shegeneralizeswithotherpersonswhoaresoassociated.Seeming"exceptions"aredangerouswhen theycondenseawholehuman life inonesentence,notoffering sufficient data for diagnosis or discussion.We ought to prefer toremain on firmer ground, using verified principles with proper safeguardsagainst over-extrapolation. The principle of secondary reinforcement, andthefactofitsindependence,willenableustotakefurtherstepsinanalyzingcomplex behavior, and we shall find ourselves employing it in theremainingpagesofthisbook.

One final word on generality. Since the S r function of a stimulus isclosely related to its SD function, we might hypothesize that the lattershould show, to some degree at least, the same generality as the former.That is to say,wemight askwhether an SD once established for a givenresponse under a given drive, is capable of acting as an SD for otherresponsesunderotherdrives.AgaintheanswerseemstobeYes,butsubjectto the qualification that the response to which the SD is transferred mustalreadyhavebeenconditioned.Thus,Walker(1942)tookanS D previouslyassociatedwitharunningresponseandpresenteditduringtheextinctionofbar-pressing.The stimulus had never been related to bar-pressing, but bar-pressing had been conditioned beforehand with the same reinforcement(food) as had the running. The result was an increased rate of pressingduring the application of the SD.Apparently, an S D which denotes an"occasion"forreinforcementofoneresponsecanalsobean"occasion"forasecond response which has been similarly reinforced but never before

connectedwiththatSD.Data corroborating thisfindinghavebeen reportedby Estes (1943, 1948). Not so definitely answered is the experimentalquestionwhetheranSDmaybe transferred toanother responsebeingmadeunder another drive.The provisowould oncemore hold that the responseinvolved under the new drive must possess some conditioned strength,whetherthatstrengthwasacquiredunderthenewdriveortheoneprevailingwhentheSDwasestablished.Clearly, thegeneralityofSDcombinedwiththat of Sr adds up to the possibility of greatly extended behavior varietyandcontrol.

SomeParametersofSr

What variables determine the strength of a secondary reinforcement?Ofthemanywecanguessat,experimentalinformationisthusfaravailableononly a few.This area of research is growing, however, andwe are sure toincreaseourknowledgeofitgreatlyinthenearfuture.

1.Thestrengthofasecondaryreinforcermaybeexpectedtodepend,forone thing, on thenumber of times it was correlated with a primaryreinforcement.Bersh (1950)has found this tobe soexperimentally.Usinggroupsofrats,hepaireda lightwith thedroppingofapellet into thefoodtray, varying the number of these pairings from 0 to 120 with differentgroups of subjects. When he later tested the light as an S r forreconditioning the response of bar-pressing, which had been previouslyconditioned and extinguished, he found that its strengthwas greaterwhenitsnumberofcorrelationswithreinforcementhadbeengreater,andthat thestrength approached an asymptote beyond which an increased number ofassociationswithreinforcementdidnotcarryit.

2. Bersh (1950) has also shown that thetemporal relation between theSrand theprimary reinforcementwithwhich it ispairedduring training isimportant in determining the reinforcing value acquired by the Sr. In oneexperiment, a light was arranged to begin 0, ½, 1, 2, 4, and 10 secondsbefore the dropping of a pellet into the tray of the rat's livingcage.WhentheSrwas laterused toconditionbar-pressing, thegraphof theamountofconditioning obtained rose to a maximum for the Sr which had had aninterval of about one second between its onset and primary reinforcementduring training. One second is the optimal interval, and intervalseitherlonger or shorter are not so effective in establishing a stimulus as an Sr.AnotherexperimentalongtheselineshasbeenreportedbyJenkins(1950).

FIG.53.Theeffectivenessof a secondary reinforcer as a functionof thenumber of times itwas pairedwith a primary reinforcer. SN is the neutralstimulus being conditioned as a secondary reinforcer; SR is the primaryreinforcerwithwhichitispaired.(AfterBersh,1950.)

FIG. 54.The effect of the time interval separatingSN andSR (definedas inFIG. 53) upon the power acquired by SN to act as a secondaryreinforcer.(AfterBersh,1950.)

3.PursuingtheideathatanSr isestablishedthroughtrainingasanSD,Notterman (1950) did an experiment in which thediscrimination traininggivenastimuluswasvaried.Heusedgroups of rats as subjects, a runwayasapparatus,andalightastheSDcorrelatedwithfoodreinforcementwhileabsence-of-lightwasSΔonnon-reinforcedruns.Eachgroupofratsreceivedthe same number of SD trials (fifty in all over a six-day period), but foreach group the number of interspersed SΔ trials varied.One group got noSΔ trials at all; another, 10 SΔ trials; another, 25; another, 50; another,100.After trainingwas over, each groupwas extinguished on the runwayfor94trials,eachtrialbeinggivenintheabsenceoflight(SΔ)andwithoutreinforcement.ThelightwasthenintroducedasS rfromthe95thtrialontothe 105th which was the last, and the amount of facilitation of runningproduced by the light was compared for the several groups.There was asystematic increase in reinforcing power of the Sr which paralleled theincreasingnumberofSΔ trials it had been given during its discriminativetraining.

FIG. 55.The effect of increasing discrimination training upon the Srpower of an SD in Notterman's experiment. The ordinate refers to theamount of decrease in running timewhen the previousSDwas introducedastheSrafter94unreinforcedrunsunderSΔ.Thedecreaseforeachgroupisexpressed as a per cent of that group's running time just prior to theintroductionoftheSr.(AfterNotterman,1950.)

4.We may guess, in the absence to date of experimental information,that theamount of primary reinforcement associated with a secondaryreinforcer may also be of significance in determining its strength. Otherthingsequal, if theprimary reinforcement is small, then theaccompanyingstimulusmaybecomeonlymildly reinforcing; if it is large, the secondaryreinforcement may gain more strength. This relationship is suggestedthroughanalogywithbasicconditioning(here).

5.Averytemptingviewisthatthe scheduleofassociationwithprimaryreinforcement has its effect on secondary-reinforcement strength. Anyscheduleofprimaryreinforcement,whetherregularorperiodicoraperiodic,

may be associated with any schedule of presenting the potential Sr. Itwould be surprising if all the possible combinations of schedule had thesame effect on the strength of the secondary reinforcer, but whatever theeffectweshouldbegladtoknowit.

6.Again,wemayguessthattheeffectivenessofanS rdependsuponthedelaywithwhichitisappliedtotheresponse.Thiswouldbeanalogoustowhat has been said about the "delay-of-reinforcement gradient" for primaryreinforcers(here).ExactmeasurementoftheresultofdelayinganSrhasnotyet,however,beenmadeexperimentally.

7. There is also to be considered thenature of the organism as adeterminant of the length and complexity of chains in which secondaryreinforcement figures.The organism dealtwith in an experiment is one ofthe conditions of that experiment.Hewill determinewhat stimulimay beused atall forhimbyreasonofhissensorycapacities.Hewill require thisor that number of reinforcements to develop a reaction strength whichanother species might achieve with half that number. His potentiality foracquiringa longchainofreflexesmaybequite limitedincomparisonwithan organism far above him in the evolutionary scale. The resistance toextinction of a secondary reinforcement—its longevity—may beenormouslydifferent fromone leveloforganism toanother: itmayendureforlifeinaman,butexpireafterseveralresponsesinachicken.

"TokenRewards"and"Sub-Goals"In analyzing response chains, we noted, among other things, that each

SDin thechainactsasanSr for thepreceding response.We spokeof oneresponse as producing the stimulus in the presence of which the nextresponseiscalledfor.Nowsupposethatwealterthesituationandtrytogeta chain in whichthe next-called-for response is to manipulate or dosomethingwith theSDproduced by the preceding response.What successwouldwehave,andwhatcouldwelearnfromsuchanexperiment?

The experiment,we find, has been done successfully several times, andwithavarietyofanimalsrangingfromtherattothechimpanzee.Bywayofillustration, however, we shall limit ourselves here to studies of apebehavior, since these have been clear, dramatic, and highly suggestive ofhuman conduct.Wolfe (1936) trained chimpanzees to insert small discs(poker chips and brass slugs) into the slot of a vending machine whichautomatically rewarded each response by delivering a grape. Subsequently,theanimalsquickly learned tomove the leverofa "workapparatus"whichprovideddiscs that couldbeexchanged forgrapes in thevendingmachine.Inanumberofvariationsoftheexperiment,theydemonstratedtheirability(1) to discriminate food tokens (white discs) from non-food tokens (brassdiscs); (2) toselectblue tokens inpreference towhite tokens,andwhite in

preferenceto brass,when the blue procured twograpes, thewhite procuredone, and the brass none at all; and (3) to select tokens in accordancewiththeir prevailing needs—for example, black (food-getting) tokens werechosen when the animals were hungry, and yellow (water-getting) tokenswerechosenwhentheywerethirsty.Evenwhenthedisc-for-grapeexchangewasdelayed, the subjects stayedatworkobtainingdiscs for a considerabletime and saved themuntil they could be used.Discs that could be 'tradedin' immediately had somewhat greater incentive value than those whichcould not be exchanged until some time later; but the animals wouldreadily lay in a supply of twenty or thirty discs for future use. Cowles(1937) confirmedmanyofWolfe's observations andaddednewdata.Moststrikingly, he showed that apes were able to form a number ofdiscriminations, some of them quite complicated, solely on the basis oftoken reward. Thus, they were able to master color, pattern and sizediscriminations in a standard discriminative situation, after which theycarriedtheirtokenstoanotherroomandexchangedthemforraisins.

There is no reason why you should not have been able to predict inadvance the outcomeof these studies, since you are already equippedwithample background. The first stage of Wolfe's experiment involved astraightforward operant, disc insertion into a machine for food-reinforcement, that isanalogous to therat'sbar-pressing. (Thisappliesalsotoahostofstudiesinwhicha toolora"manipulandum"isusedtoobtainfood—for example, string-pulling, use of rakes, box-stacking, and so on.)In thepreferential selectionofblue (two-grape) tokens, the reflexSD (bluedisc)RhastheadvantageofagreateramountofreinforcementthantheSD(white disc) R reflex, just as if there were two bars in a cage, with a ratlearning to press one predominantly because it gavemore or larger pelletsthan the other. Related to this choice is also the action of the negativelyreinforcing character of the "hard way" (here), with the blue disc againbenefitingbecauseitgivesmore-return-for-less-work.Inthesame situation,the brass disc response is extinguished as the usual result of non-reinforcement, although occasional selection of these discs would beexpected(andwasobserved)onthebasisofstimulusgeneralization.InbothWolfe's and Cowles' studies, the discs were employed in the chain assecondary reinforcements for conditioning other responses. The fact that,after procurement, the next response was a manipulation of the Sr.D (thedisc)is,onsecondthought,actuallynodifferentfromtheSD(visualpellet)R (seizing) in the chain that follows bar-pressing.The "hoarding" of discsis itself learned through the ultimate, though delayed, reinforcementprovidedwhenthediscscanonceagainbeexchangedforfood.Andsoon.There is apparently no aspect of this research which taxes unduly theprincipleswithwhichwearefamiliar.

Experiments of this sort have also been called "sub-goal" studies.Both

"sub-goal"and"token-reward"arewaysofdenotingtheactionofsecondaryreinforcements, and we should not permit ourselves to be disturbed bysynonyms."Token-rewards"weresocalledmerelytodistinguishthemfrom"primary" reinforcement. The chimpanzee experiments arouse our interestbecause they come so close to our own social and economic behavior. Inhuman society, money and other secondary reinforcements such as"prestige" and "community approval" assume, through training, the statusof generalized rewards that have potency for practically everything we do.Chimpanzees,likemen,willapparentlydomanythingsformoney.

SecondaryNegativeReinforcementWhile we have spoken thus far of conditioned reinforcers which act

positively,wemustnotoverlookthepossibilityofconditioningastimulusasasecondarynegative reinforcement. In theexperimental studyofanxiety(seeChapter9),anSDisarrangedtoprecedebyafixedtimetheonsetofaninescapablenoxious(negativelyreinforcing)stimuluslikeanelectric shock.Theresult,brieflystated, is thatbar-pressingintheinterval-before-shockisdepressedorhalteduntil theshock ispast.Weaskwhether theSD in thissituationbecomesalsoasecondarynegativeSr,andwecanconfirmthefactthatitdoesbyshowingthatananimalcanbereinforcedbyitsremoval.

FIG. 56.An apparatus for studying escape from a secondary negativereinforcer.The leftcompartment ispaintedwhite,andcontainsa floorgridfordeliveringelectric shocks.The rightcompartment ispaintedblack.Theanimal can escape from the left to the right by operating a door-openingmechanisminthewallbetween.(FromMiller,1948.)

Evidence for such an effect has been obtained by several investigators.Hefferline (1950) presented a series of faint clicking sounds regularly inadvance of a strong light from which rats could escape by bar-pressing.Whenever an animal responded during the SD period, the clicking was

stopped and, subsequently, no light was presented.After long-con tinuedcombination of clicking-followed-by-light, it became clear that the bar-pressing response could be kept going for a time solely because it cut outthe clicks.This is related to an earlier study by Mowrer and Lamoreaux(1942) in which a buzzer was sounded for six seconds prior to an electricshockwhichcouldbeeliminatedwhenthesubjects(rats)ranfromonehalfof an experimental chamber to the other. If the rat made the runningresponseduring the six-secondperiod, the soundwasdiscontinuedand theshockwas not given; if he did not respond during this period, the buzzerwas continued, with the shock added, until the escape response occurred.Tendaysofbuzzer-shockcombination,withtencombinationsperday,weresufficient to establish the negatively reinforcing effect of the buzzer—thatis,tostrengthentherunningresponsethroughtheterminationofthesound.Inalaterexperiment,MowrerandLamoreaux(1946)demonstratedthesameeffect by conditioning a buzzer-terminating response (jumping into the air)thatwasdistinctfromtherunningresponsethatendedtheshock.Stillmorerecently,Miller(1948)showedthatbothawheel-rotatingandabar-pressingresponse could be strengthened when the only reinforcement was escapefromawhitebox(terminationofanegativeSr) inwhichtheyhadformerlybeenshocked.

FIG.57.Afterconsiderable shocking in the left (white)compartmentofthe box inFigure 56, here without any opportunity to escape, the wheelwasmadefunctional,and theanimalswerepermitted toescape to theright(black) compartment by rotating the wheel to open the door. The curveshowstheaveragelatencychangesofthewheel-turningresponseonthefirst16trials.(AfterMiller,1948.)

FIG. 58. After the wheel turning response (see Figure 57, here) waslearned, the wheel was made non-functional and the new response of bar-pressing to open the door was substituted. The white compartment stillretained its negative character, since the animal stopped rotating thewheelandacquiredthebar-pressingresponseasshowninthiscurve.(AfterMiller,1948.)

In thisconnection,wemaycite twoobservationsmadebyWolfe in thecourseofhis token-reward studies.Oneofhis apes, accustomed to theuseof light-blue tokens to obtain 'activity privileges' (return to living-quartersor play with the experimenter), twice employed the discs in what wasapparentlyanattempt toescape from the test situation: oncewhen awhiterat was introduced into the experimental room, andonce when aphotographer appeared on the scene to take pictures of the animal'sbehavior!

One of the commonest appearances of secondary negative reinforcementis in the reflexchains associatedwith themoving-aboutof anorganism in

its environment. Locomotor responses which bring an animal (or humanbeing)intothepresenceofSD'swhichhavebeencorrelatedwithpainfulornoxious stimuli (primary negative reinforcements) are depressed orweakened,sothatheseemstoavoidthosesituations,stayingin'safe'areas.Moreover, an experiment byBugelski andMiller (1938) showed that rats,placed in a runway at different distances from the point atwhich they hadreceivedelectricshocks,ranawaywithaspeedthatwasproportional to thenearnessofthepunishmentarea—thefartherfromthenegativeSrtheslowertheyran.

FIG. 59. The tendency to move away from a place where negativereinforcement was applied. The strength of the tendency is measured interms of the time taken by the animal to get started when placed at threedifferent distances from the point at which he had been shocked. Theexperimenters refer to this as a "spatial gradient." (After Bugelski andMiller,1938.)

Wecanobservethisphenomenoninmanyotherformsofbehavior.It isseenespeciallyinthemasteryofthoseskillswhereaslipmeansdisaster.Achoking mouthful of water keeps the beginning swimmer's head up andmouth closedfor a long time afterward; and the 'no-hands' cyclist quicklyseizesthehandlebarswhenalossofequilibriumisthreatened.'Safe'placesare those in which positive, or at least no negative, secondaryreinforcementsarefound;andtheyare'preferred'(theorganismspendsmostof his time in them) to areas in which the SD's have become negative

reinforcersthroughassociationwithnoxiousstimuli.'Home'issuchaplaceto many persons; whereas strange places may be crowded with unknownand untrustworthy SD's, foretelling unpredictable, even fearsome, events.What we call 'homesickness' may be in good part the absence from one'senvironment of accustomed, and positive, Sr's.A child raised by much-visiting andmuch-traveling parentsmay be 'at home' anywhere, especiallyifhisparentsarewithhim.Notunlikethis, is the 'hospitalitis' reportedbynurses and physicians in patients of long residence in a hospital who aredejectedby theprospectofdischarge, andwhomaydevelopa relapse, realorfeigned,thatprolongstheirstayinafriendly,protectiveenvironment.

SecondaryReinforcementand"Feelings"Among the topics which psychologists, ancient and modern, have

traditionallydiscussedisthatoffeelings.Thistopiccantakeoneintomanyblindalleys,especiallyifoneweretofollowtheeverydayusesoftheword.In conversation, the word has no rigorous meaning.We use it in suchstatements asI feel reluctant, My feeling is that one should compromise,Thatfeelscold,Ifeelsad,IfeelthatImust,andGoya'setchingshavesuchfeeling!Thereare thousandsofphrases inwhichweemployeither theverborthenoun,andthoughtfulstudentshavelongagoagreedthattheseusagescannotpossiblydelineateasingleproblemorfieldofinvestigation.

Progress in finding a psychological meaning for the term began in thelast century and has continued into the present. In one famous attempt todelimit the field, Wilhelm Wundt (1896) proposed a 'tri-dimensional'theory.All true feelings were tobedescribed in termsof threecoordinates:degree of subjective pleasantness-unpleasantness; tension-relaxation; andexcitement-calm. Thus, any feeling could be adequately described byreference to its place along each of these continua. One feeling might bepleasant, relaxing, and calming; anothermight be unpleasant, tensing, andexciting; and others would portray still different combinations.This viewcarried enough prestige to initiate many experimental attempts to findbehavioral counterparts (changes in blood-pressure, pulse-rate, respiration,and so on) of each of the alleged dimensions.A great deal of conflictingdata resulted, and the whole enterprise is now generally conceded to havebeen relatively unproductive. Titchener (1908) offered a pared-downintrospective account of feelings which eliminated two of Wundt'sdimensions, retaining only pleasantness-unpleasantness, but the search forbodily correlates of this dimension was still dogged by failure. Beebe-Center (1932), a present-day. authority in this field, finds that no singlerelation between pleasantness-unpleasantness and either respondent oroperant behavior has been acceptedwithout question.The best that he cantell us is that all theories of the relation between feeling and 'locomotor'(his equivalent ofoperant) response "depend on the correlation of

pleasantness with seeking, approach, acquisition; and the correlation ofunpleasantnesswithavoidance,withdrawal,rejection."

The bearing of reinforcement upon such verbalizings as "pleasant,""unpleasant,"andtheirsynonymshasnotbeenfullyexplored,butis likelyto prove the most profitable approach of all to this problem.Reinforcements appropriate to our "drives" (Chapter 9) commonly evoke"pleasant" or its equivalent—as food when one is hungry; while non-reinforcement, negative reinforcement, and the removal of reinforcementapparently occasion "unpleasant." Evenmore to the point is the action ofsecondaryreinforcers.Manyfoodsarenourishingbutnotequally"pleasing"(Mother's cookingis the best!). It would seem that the discriminativestimuliwhich foretell positive reinforcement raisewithin us "pleasant andjoyful anticipation"; those foretelling negative reinforcement cause"unpleasant and fearful anticipation" or "anxiety." But still we are facedwith the task of isolating the SD's that lead us to emit such words.Theoristshavelongarguedthat"feelings"arerelated,nottoobjectsorotherexternalSD's,buttotheeffecttheyhaveuponus("Ifachildispleasedbyanew red toy, thenboth red andpleasantness are features of his experience,butitisthetoythatisredandthechildwhoispleased"—Beebe-Center).Itmay be that whenever we say an object is pleasant or unpleasant ourstatements are based upon intervening responses (of approach,withdrawal,or some other sort) which follow directly upon exteroceptive stimulationand themselves determine the overt spoken outcome. In such a case,"feelings" and "meanings"wouldhavemuch in common, and theproblemof reported "feelings" would be similar to the problem of reported"emotions" (seeChapter 10). This parallel is supported by the fact thatreaction times of the verbal responses or judgments 'pleasant' and'unpleasant' are generally slower than ordinary reaction times—a findingthatwouldbeexpectedon thebasisof the intervening response that servesastheSDforthejudgment.

Thesuggestionisstrongthatpositivesecondaryreinforcementfiguresinthe explanation of 'joys'—like the joy of revisiting 'the scenes of ourchildhood.'Apparently thesource lies in thereinforcementsconnectedwithfamiliar places, people, things, and actions. The loss or absence of oldsecondary reinforcers, on the other hand, seems often to be at the heart ofour 'sorrows.'Homesickness;dejectionatthedeathofafriend;regretatthepassing of old landmarks; yearning for 'the snows of yester-year'—do notthese suggest the loss of Sr.D's? Indeed, the tie-up of secondaryreinforcement and "feeling" seems to pervade our whole lives as socialorganisms. The process of socialization has to do with acquiring thosemodes and standardsof behaviorwhich are typical of the society inwhichthe individual grows. Because all of our reinforcements occur in a givensocial environment, and because society often makes reinforcement

contingentuponourdoingandsaying 'theproper things,'wemaycometolikeandpreferas'naturallypleasing'ourownparticularwayoflife,andthisorthatstyleofmusicorart.Evenourgroupprejudices,oursmalltalk,ourambitions, our religion, our special forms of virtue (and vice), appear tohave their roots in this basic principle of human behavior. The verylanguagewe speak is an edifice reared upon secondary reinforcement. In aforeignland,wethrilltothesoundofournativetongue;athomeweenjoythe writer or speaker in whose verbal skill dwell many of our strongestsecondaryreinforcements.Wefind thatpoetsatisfyingwhoartfullysetsupinustheverbalchainsthatheisusing,sothathisnextword,andhisnext,comealmostasthecompletionsofwhatwewereabouttosay!Intheverbalinterplay betweenwriter and reader, between speaker and listener, it is thewriter's or speaker's art to start up inus those 'thoughts'whichagree withhis, bringing us together in the same responses, making us say the samethings,orsupplythesamerhyme.Suchanoutcomeistruly'pleasant,'evenoverpowering,becausethepoetandwehavebeenspeakinginthesamewaytogether, exchanging our 'desires' and 'passions' and our tears in perfectharmony—a relationship as near to the purely social as we can conceive,and asmatchless an example of secondary reinforcement asman can createorfind.

SecondaryReinforcementinSocialBehaviorAsimpliedabove,astimuluscontrollingbehaviorasanSD,Sr,orSr.D

neednotstemfromthe inanimateenvironmentalone.Otherorganisms(or,the stimuli emanating from them) can act in these ways, too, and thereinliesa factofutmost significance tohumanaswell asotherbiological life.Socialbehaviormaybedescribedasbehavior forwhich thereinforcing ordiscriminative stimuli are, or have been, mediated by the behavior ofanotherorganism.By"mediated"wemean"arisingfrom,or inconnectionwith," and there is no intention of straining the word's connotation.Wewould neither (1) include as social the delivery of a pellet by theexperimenter to a bar-pressing rat—it may be a social situation for theexperimenter, but it is not for the rat; nor (2) exclude as non-social thebehaviorofthemaroonedsailorwhospeakstohimselformakesclothesoutofskins,sincebothactivitieshavebeensociallyacquired.

From birth on, social stimuli play a large part in the life of humanbeings. Many scientists, indeed, have thought that society itself has itsoriginsintheprotractedandutterdependenceofthehumaninfant.Howeverthat may be, parents, and especially the mother, are among the firstsecondary reinforcers of a social sort to enter the infant's ken. Theirdiscriminative and reinforcing potency are quickly established by theircontinual associationwith food,warmth, relief frompain, and the like. If,however, the child is reared by a nurse, then she becomes the ever-presentsecondary reinforcement, and it is commonly seen that attachment to the

nurse replaces that to the mother. Psychiatrists have pointed out that, inadolescence and before, the first sexually interesting objects may be theparents, brothers, or sisters. Within the relatively restricted socialenvironment of the child, the few organisms who serve as theaccompanyingstimuliwhenreinforcement isgivenorwithheldcanacquirean overspreading and life-long grip on his behavior.With increasing age,the child's widening excursions beyond the home provide an increasingrange of secondary reinforcers to control his reactions. School, friends,clubs, and related activities of all sorts—these push upon him the stimuliwhich are the ever-present signals and accompaniments of ultimatereinforcement,theSD'sandtheSr'swhicharethewarpandwoofofhislifeinsociety.

Although the theme ofanxiety will be developed later (seeChapter 9),weshouldrecognizeinpassingthatsocialstimulicanalsoactassecondarynegative reinforcers. As we shall see, the behavior which occurs in theinterval between an SD and a negative reinforcement is characterized by adepression of operant activity and the onset of respondent changes. Thelaboratory study of anxiety (induced experimentally by using the SD-negative reinforcement sequence) has numerous implications for everydayaffairs. A child raised in a rigid and over-disciplined home, will suffermany punishments through the prescriptions and prohibitions imposedupon him.The adolescent, having to copewith his newly acquired socialstatus, encounters many pitfalls and rebuffs before he learns acceptablemodesofbehavior. In these and similar instances,persons are the appliersof punishment or emotion-arousing stimuli, and, by this association,themselves become secondary negative reinforcements. Through stimulusgeneralization, other persons may be included within this category. Theresultmaybeadepressionofactivityintheirpresence,avoidanceofpeople,seclusiveness; in short, there may develop anti-social and maladjustedbehavior so alarming to the clinical psychologist or mental hygienist.Where external positive reinforcements of the secondary sort are radicallyreduced,andwherenegativereinforcementscrowdtoostronglyuponaman,thefinaloutcomemaybecomplete'withdrawalfromtheworld'—asseeninthepsychosisofschizophrenia.Eveninthenormaldevelopmentofchildrenit has long been remarked that age is accompanied by an increase in thesheer number of things feared.The six-month-old infant has few fears; thesix-year-old has many. This is the upshot of increasing experience withnegative reinforcement in an ever enlarging world, with the consequentmultiplicationofSD'sforfear.Loweranimalsmaylearnmostoftheirfearsat the hands of their natural environment; man gets most of his at theheedlessorunmercifulhandsofhisfellowcreatures.

TakingStockandLookingAhead

Farbackin thisbookwesaid thathumanbehavior is thefinalobjectofinterest tomostpsychologists,asit is tothelayman.Inreachingourgoal,the principle of secondary reinforcement will be of great analyticalassistance.When added to the other functions of stimuli, it gives us apowerful and indispensable tool for the solution of many vexing andabsorbing problems of human action. It will not escape the thoughtfulstudentthatthefollowingpointsgofartowardexplainingtheelaborateandramified behavior of organisms high in the evolutionary scale, and of oldorganisms as against young. Given an undeveloped creature to whom fewprimary reinforcements may be relevant, the following facts allow us toaugment our control over hismaturing operant repertory in ever increasingfashion.

1.A stimulus that occasions or accompanies a reinforcement acquiresthereby reinforcing value of its own, and may be called a conditioned,secondary, or derived reinforcement. A secondary reinforcement may beextinguished when repeatedly applied to a response for which there is noultimateprimaryreinforcement.

2.A secondary reinforcement is positive when the reinforcement withwhichitiscorrelatedispositive,andnegativewhenthelatterisnegative.

3.Once established, a secondary reinforcement is independent and non-specific; itwill not only strengthen the same responsewhichproduced theoriginal reinforcement, but it will also condition a new and unrelatedresponse.Moreover, itwill do so even in the presence of a different basicmotive.

4.Throughgeneralization,manystimulibesides theonecorrelatedwithreinforcementacquirereinforcingvalue–positiveornegative.Thispointwasnotstressedinthepresentchapterbutitshouldrequirenoelaborationhere.

Finally, it should be remembered that, in everyday life, stimuli are notsingle, that responses are not all alike, and that learning is not all doneunder the same drive.Many stimuli are present when a response isconditioned,allbecomingdiscriminativeandsecondarily reinforcing;manyresponses are capable of obtaining the same reinforcement; andmore thanonedrive(hunger,thirst,etc.)maybesatisfiedatvarioustimesbythesameresponses and in the presence of the same stimuli. Truly, the study ofbehaviorisaloftychallengetoscientificimaginationandmethod!

NOTESThere isnobookonsecondary reinforcement towhichwecansendyou

for additional information. Hull, however, in hisPrinciples of behavior(1943),hasdevotedachapter to this importantprinciple. (Almostany textin social or abnormal psychology will, of course, provide you withnumerousexamplesof itsoperation.)Yououghtnow,however, tobeabletofollowwithlittletroublethegrowingexperimentalliteratureinthisarea.Access to these reports may be gained by way of thePsychological

Abstracts,a journal thatcontainsshortsummariesofallarticlesandbooksassumedtohaveanyinterestwhateverforpsychologists.

In termsof the relationof behavior to the environment of anorganism,thepresent chapter takesus almost as far aswecango.Nomore than twofunctionsofstimuliremaintobeconsidered,andnothingnewwillbesaidabout thebasicprocessesof conditioning, extinction, and the like. In fact,ifyouhavefollowedourdiscussionuptothispointwithmoderatesuccess,youpossessmostof theavailable tools forbehavioral analysis.Fromnowon,weshallbeinterestedinthekindofenvironmentalcontrolthatinduceschangesinthestateofanorganism—aswhenwedepriveananimaloffoodorwater. In our talk ofmotivation and emotion in the next two chapters,youwill see thatwe do littlemore than shift the focus of our attention—say, from reinforcement to somepreviouslyunmentioned conditionsunderwhichstimulibecomereinforcing. InChapter11,we shall try topointoutthe direction one might profitably take in carrying the principles into therealmofhumaninteraction.

9

MOTIVATION

IT IS by the pleasure of exertion, and the pain of inexertion, that we areroused from that indolence, intowhich... we otherwisemight sink: asweareroused,inlikemanner,bythepleasureoffood,andthepainofhunger,we take the aliment that is necessary for our individual sustenance; andthough themerealiment is, indeed,more important for life, it isnotmoreimportant forhappiness than thepleasureofactivitywhichcallsand forcesusfromslothfulrepose.

ThomasBrown,LecturesonthePhilosophyoftheHumanMind,1822

ANewLineofInquiryWe have been occupied in exploring the principle of reinforcement and

the manner in which the environment controls organisms by way ofstimuli. From this single starting point, we have been able to take largestridesinunderstandingwhymenandloweranimalsbehaveastheydo.Yetstudentsofpsychology,inpasttimesandpresent,havefeltorknownthatadescription of behavior would be incomplete without taking into accountanotherkindofcontrollingfactorwhichtodaywecallmotivation.

Common experience reveals the existence of this factor so vividly thatmen everywhere have evolved a vocabulary and set of ideas for explainingand speaking of it. Growing up in a social community as we do, we aretaught the prevailingwords and concepts.These seem consequently, to beright, natural, and but common sense. Unhappily, there are few areas inpsychologywherepopularnotions contain amore alluringblendof correctandincorrectobservations,ofvalidandbiasedthinking,ofwiseandfoolishconclusions.Ourfirst task, ifwe are tomake progress in this new line ofinquiry,istobeginaright.

Ordinarily, our questions about motives occur in connection withcomplex types of human interaction.We feel that unless we know theunderlyingmotiveswe shall not be able to deal effectivelywith ourselvesand others in the many important affairs of everyday life.Why does eachpersonseekpopularity?Whydomenmarryastheydo,andwhomtheydo?Why domen fight and take pleasure in killing? Our analysis cannotstartwith such samples of behavior, but rather with fundamentals andexperimentallyverifiabledata.The initialaimis togetaproperfoundationfor furtherbuilding, not ablueprint for a superstructurewhichmay requireendless revision, or may be totally useless, as the basic facts come to beknown.Only in thiswaycanwemake the scientificprogresswewantandneed.

TheNeedfortheConceptofMotivationOddly enough, it is the very first step toward comprehending motives

(drives is a synonym) which often proves the hardest for the beginningstudenttotake.Whatisdemandedisthathelayasidelong-favoredpersonalopinions, and examinede novo the reasons for believing that a science ofbehavior cannot get alongwithout motivation! Only if there wereobservationsofbehaviornotencompassedby theprinciplesset forth in theearlier chapters, would we be required to formulate new concepts to dealwith the data. The present chapter is, in fact, devoted to just suchobservations.Letustakethe'hungerdrive'asanexample.

Neitherratsnormeneatcontinuously,butatfairlydefiniteintervalsandin fairlydefiniteamounts.This fact isbanalenough tomostpersons,whowould probably say that an organism eats when 'hungry' and stops when'full.' But it may do us good to question the obvious, wherein at timeshave lain concealed some astonishing natural phenomena. So, with asuspicion that this is a phaseof behaviorworthinspecting,we take to thelaboratory;andsoonourcriticalsenseisproddedbyfurtherdiscoveries.Wefind that a bar-trained rat, left in his box orworking-cage to obtain all ofhismealsbypressing,willeatinafairlyregularcycleandonlysomuchatany one time. In a twenty-four-hour period, he may eat on ten or twelveoccasionsonly, each timeat a slowbut constant rate.Whengivenbut onemeal a day, always at the same hour, a curve like that in Figure 7 (here)willbeproduced.Whenplacedinthebox,rightafterameal,hewillnoteatat all.Moreover, an animal cannot be conditioned to salivate (Type S) orpressabarforfood(TypeR)unlesshungry.

These factsmay leave you unmoved.Didwe not assume all along thathunger or some other drive is required in experiments dealing withstimulus-controlledbehavior?Yes,wedid;butnowwerealizethatwehaveon our hands a phenomenon that deserves special consideration. For,trimmeddown,ourobservationsarethat(1)deprivingananimaloffoodisa way of increasing the strength of a conditioned reflex like bar-pressing;thatconcurrently,(2)manyotherreflexesriseinstrength,suchasreaching,seizing, and chewing; that (3) with sufficient intake of food(satiation),thesereflexesdropinstrengthtozero;andthat(4)food-deprivationisitselfa prerequisite for using food as reinforcement—that a reinforcer is such byvirtueofsomeoperationthatmakesitactso.

It is in this way that the need for the concept of motivation as a newvariableinbehaviorarises.Itisbecauseresponsescanbecontrolledinotherways thanby reinforcement, thatanewdescriptive term iscalled forandanewbehavioral concept emerges.What shallwe say about occurrences likefeeding and fastingwhich affect reflex strength?Howmany types of suchoccurrencesare there?Howcanweexercisecontrolover themand, throughthem, over the organism?Howmany degrees of each can be set up?Howcan they be measured?What new experiments do they suggest, and what

new knowledge will they generate? How are they related toconditioning,extinction,discrimination,and the like?Thesequestionsandmanysimilaronessoonlosetheir'obvious'character.

TheNatureofDrivesTo the questionWhat is drive?, wemust now answer that drive is the

name for a fact—the fact that certain operations can be performed on anorganism (for example, depriving it of food) that have an effect uponbehavior which is different from that of other operations. Drive is not athing, but simply a word we use to show our recognition that behavioralfunctions which may depend on reinforcement are also modifiable byanother influence, one exerted by occurrences which do not involvereinforcement.

FIG.60. Illustratinghowvaryingdrive levelsduringextinctionactasaparameter of the extinction curve. The higher the drive, the greater thenumberofresponsesmadeandthemorerapidlytheyareemitted.

The observation that 'neither rats normen eat continuously' requires ananalysis that, after all, is rather distant from everyday levels ofconversation. Words likemotive anddrive, drawn from the common

vocabulary, do not explicitly denote the relation with which we are nowconcerned.Perhaps the ideawill becomemore acceptable aswe examine itfrom different angles and express it in different ways, with a fewexperimentalexamples.

1 .Drive in extinction. It has been quite well established (e.g., Perin,1942) that the process of extinction is influenced by the drive factor.Theexperimental design, urged by the initial observation that a bar-trained ratwill not presswhen satiated, is simply this: Expose groups of animals toequalamountsof trainingunderequaldeprivation,and thenextinguish theresponseunderdifferentlengths-of-deprivationtoseewhetherthenumberofpressings they make co-varies with deprivation time at extinction.A fairnumber of experiments have been made on this question, with goodagreement that deprivation time does act as an important variable. Theidealized finding is shown inFigure60 indicating that, in addition to theheight of the cumulative response curvereached, the early rate of emissionis also affected. Such experiments give substance to the concept ofmotivation as a determinant of behavior supplementing reinforcement.Technically,wemay say that drive, here specified in terms of deprivationtime,isa"parameter"ofextinction.

FIG.61.Experimentaldataon the relationofdrive level to thenumberof responses made in extinction. Four groups of rats were conditioned tobar-pressingunder the sameconditionsofhungerandgivenequalnumbersof reinforcements. Subsequent extinction was carried out with each group

underadifferentamountofdeprivation:1,3,16,and23hours.Extinctionwas carried to a criterion of five minutes. The plot shows the averagenumber of extinction responses for each group. (From Hull, 1943, asadaptedfromPerin,1942.)

2 .Drive in periodicreconditioning. Our secondexampleofhowthemotivationconcept derives from the effectof an operation like food-deprivation, is taken fromperiodic reconditioning. Ananimal placed on a schedule ofP-R at a fixed interval of, say,three minutes, will givecumulative response curves ofvaryingslopes.Thatis,therateof responding, within widelimits, increaseswith increaseddeprivation (seeFigure 62).Similar results are obtained if,instead of varying deprivationtime,we deprive allanimals atthesametimebutprovidethemwith different amounts of foodjustbeforeusingtheminaP-R

session(seeFigure63).This introduction to the

nature of drive needs roundingout,andweshallattempttodothis in succeeding sections.Before going on, however, weshould mention twomisconceptions of drive whichour analysis enables us tosidestep. These are the twinerrors of identifying the driveexclusively with either thereflexes or the operationsinvolved.Taking hunger againas our example, it would bedangerous to say that, merelybecause an animal eats, he ishungry. There are many

variables which may influenceeating behavior. Emotionalexcitement canmake an animalstop eating, or the introductionof another animal into thesituation may cause him toresume eating after he hasstopped; and human beingsmay eat for social reasons orstarve for political ones. Mereeating at any given time is aninadequate criterion of hungerunless we know what precededthe eating. On the other hand,identifying a drive with itsestablishing operation, such asdeprivation, is also wrong.There are many deprivationswhichareofnoconsequence indetermining behavior (forexample, depriving a rat oforgan music)—which do notestablish a drive.The truth ofthe matter is that when wenotice changes in reflexstrength we look forcorresponding operations, thetwo thingstogether leading usto infer the drive. It is becauseeveryday observation gives usprior knowledge that fooddeprivation has an effect onresponsethatweaccept italoneas 'hunger'. We would gainlittle from argument as towhether a starving mystic, inthe throes of his elation, is'hungry' or not! And phraseslike 'a twenty-four-hour hungryrat' or 'a drive level of sixteenhours,' commonly used as aconcession to convenience,shouldalwaysbeunderstoodtomean 'twenty-four or sixteen

FIG. 62. Cumulative response curvesfor one rat under P-R on six successivedaysinwhichnofoodwasgivenbeyondthe pellet reinforcements during thesessions.Theincreasingdriveisreflectedin the rising slopes of the curves. (FromSkinner,1938.)

hoursofdeprivation.'

FIG. 63. Cumulative response curves for the same rat under P-R withvarying amounts of pre-feeding.The curves were taken on different days.Theratwas24-hourshungryeachtime,butwaspre-fedwiththeamountingramsindicatedoneachcurve.(FromSkinner,1938.)

Driveasan"InternalState"It would be cumbersome in talking and thinking about drive to have

always to do so in long and meticulous statements about its nature as 'achange in reflex strength attributable to some operation performed eitherexperimentally or naturally. . .' and so on.The word drive is a shorthanddevice,aconvenienceofexpression,forboththeconceptofmotivationandthe facts subsumed thereunder.We do no harm in using it as long as we

know that it is an ultimately dispensable word, and that we can return toouranalysisofitwhennecessary.Itislikethewordforceinphysicswhere,intheend,itmustberefinedandanalyzed,butwhereitisahandywordinmanyconnections.Thedangerassociatedwith thepsychologist'sdriveandthephysicist'sforceisthatoflettingoneselfbemisledintothefalseissuesthatareraisedbyvagueeverydayreferences.

Similar considerations apply to the designation of drive as an 'inferredinternalstate.'Thisphrasesimplyrecognizesthattheeffectsofanoperationlike food deprivation are ramified throughout both the physiology andbehavior of the organism. Physiologically, there are numberlessreverberationsandcorrelatesoftheoperation.These,inascientificdivisionof labor, are within the province of the physiologist, the chemist, thebiophysicist.The concept of drive, however, is a behavioral one; and thephysiological effects are not the drive. Behaviorally, a drive ismarked byconcurrent changes in the strength of many reflexes. A food-deprivedanimalnotonlypressesabarmorefrequentlyandeatsmorevigorously,butmanyother reflexes change at the same time (e.g., climbing, sniffing, andrunning about) if the bar is not present, giving us the impression ofincreased general activity and restlessness.This makes it appear that theoperation has changed the creature in some over-all way, and makes thenotionofa'stateoftheorganism'morepalatable.

Very fewpersonswould question that the state is 'infer' rable' from theobservations, or that it must be 'internal' in nature; and they wouldprobablyapproveaformulasuchasthis:

Yet, despite the validity of the observations, the phrase 'inferred internalstate' adds nothing to our knowledge of drive, because it denotes nothingbeyond that which is contained within the observations themselves. It is,once again, a convenience of expression, and wemight dispense with theterm altogether if it were not for the effort involved in straining fortechnicalpurity.

We have given much space to the clarification of the nature of drives,

and thedifficultieswhichcomefrompopularusages. Ifourpointsarekeptinmind, itwill beunnecessary forus to forego theadvantagesofeaseandsuccinctness which inhere in our ready-made and ingrained vocabulary.Instead of devising new symbols, we can adopt familiar speech if at thesame timewe remember thatwe have an agreed-upon analysis to fall backon whenever it is needed to reason out newmotivational problems or re-examine old ones. We shall, then, henceforth use expressions like'establishing a drive,' 'reducing a drive,' and others, because they are neat,but ineveryinstanceourabbreviatedphrasecanbetranslatedinto themorecorrectandexpandedformthatourdiscussionhassupplied.

DiscoveringDrivesOneofthecommonestquestionsaskedofstudentsofmotivationisHow

manydrivesarethere?To this there is no answerbut a tentativeone.Thenumberofdriveswhichexistforanyspeciesoforganismisnotdeterminedbyanyman'ssay-so,andnolistcanbegivenexceptinsofarasthedrivesare known today. One who wishes to draw up a list must be prepared todefend itbycitinghisevidenceforeachone thathe includes,sinceadriveis not a matter of opinion but of proof. He must be able to demonstratethat, as a consequence of some operation, reinforcement and conditioningare made possible; that the operation leads to changes in strength of apreviouslyconditionedreflexand,concurrently,ofothers;andsoon.And,later, if new evidence is forthcoming to show that a drive has beenunwittinglyomitted,hemustbereadytoacceptitandaddittohislist.

The fact that evidencemust be found for a drivemeans, of course, thatdrivesarediscovered.Thereisnowayoftellinginadvancehowmanywillbe finally found, and all attempts at enumeration are provisional. Thesituation isnotunlike thatsurrounding thediscoveryofchemicalelementsbeforeMendeleev'speriodic table,or sincemodernatomic research showedhow to create new substances. At neither time could the number ofobtainable elements be predicted with certainty. The explorer into drivesfaces two problems. On the one hand, he may set out to discover drivefactorsinbehaviorwhichwerenothithertosuspected.Anexampleisthatofthe so-called "sub-hungers" which are established, not by gross fooddeprivation, but by particular deficiencies in diet, such as lack of calcium.Sub-hungersactonreflexstrengthtodirectananimal'schoiceoffoods(hisselectiveeating)fromamongmanythataresimultaneouslyproffered.Thesesub-types of hunger were unearthed in recent years after chemistry anddietetics gave inspiration to the idea and made possible the appropriatedeprivingoperations.Ontheotherhand,theexplorermayguessthatadriveinfluence is at work but may not have the establishing operation at hiscommand.Theoperationsforhungerandthirsthavebeenknowntoallmenthroughout theages,but this isnot trueof suchawell-recognizeddriveassex.Although, in the females of many species, this drive has long been

knowntoundergocyclicchangesasafunctionoftime,ithasonlyrecentlybeen possible to exercise regulation through glandular extracts, surgicalincursion into sex tissues, and so on. The discovery, classification,measurement, and study of any drive are inextricably related to theidentificationof(and,hopefully,masteryover)itsestablishingoperations.

Inadditiontotheirincompleteness,listsofdriveswouldbedifferentfordifferenttypesoflivingorganisms.Anoperationwhichstronglyaffectsthebehavior of one speciesmay have little or no import for another. Broadlyviewed,theamountofsimilarityinmotivationfromonespeciestoanotherdependsupon their evolutionarycloseness,but thisdoesnot alter theneedforseparatelistingsandproofs.Fortunately,wearenothereconcernedwithcataloguing, but with an introduction to the nature of drives, to someimportant ones that human beings share with other organisms, and to theproblems of experimental investigation. It is to these ends that our laterdiscussionofafewrepresentativedrivesismainlydirected.

TheMeasurementofDrivesIfyouhavefollowedtheprecedingpoints,youshouldhavenodifficulty

with the next. The matter of measuring drives (or drive level, or drivestrength)isalwaysapproachedincorrectlybyonewhothinksofamotiveasa substantive thing rather than a set of relations between an establishingoperationandbehavioralchanges.Youcanreadilyseethatwhatwemeasureisbehavior;that,inpractice,anymethodpurportingtomeasuredriveisonethat measures reflex strength as it changes with different degrees of aselected operation.The establishing operation is our independent variable,thebehaviorourdependentvariable;theformerisspecifiableastokindanddegree, the latter is measured for extent of change. The concomitantvariationofthetwogivesriseto,anddefines, theconceptandtheproblemofmotivation.

It follows, therefore, that thereareasmanywaysofgettingatanydriveas there are behavioral effects that can be measured. There is no singleconsequence of an operation like, say, food deprivation, that must bedepended on; any concomitant change in response may be taken. Instudying or discovering drives, it is important to find an appropriateresponse that undergoes clear enough, and large enough, changes to yieldsignificantco-variationwiththeoperation.Apoorchoicecanmakeitseemthat the operation has had no effect on behavior (is not really drive-establishing),buttrialwithanotherreflexmaygiveimmediatesuccess.

It would be superfluous, even if it were possible, to catalog all thereflexes that have been measured in studies of motivation.The particularoneselectedbyaninvestigatorisonlyincidentaltohispurposeoflearningsomething about drive. Fortunately, the number of reflexproperties andcharacteristics that canbemeasured isnot sogreat as tomakeburdensomeour task of selection. Here we find, as you would expect, such familiar

aspects of responding as rate, latency, force, and the like, all ofwhich areoldfriendsintheanalysisofbehaviorandrequirenofurthertreatmenthere.In our later discussion of several representative drives, we shall see withgreaterclarityandinmoredetailhowthesemeasuresareemployed.

TwoClassesofDrivesAswe have pointed out before,many historical observations of human

and animal behavior have contained elements of great accuracy andfaithfulness to nature.The age-old division of motives intoappetites andaversions is based upon such observations, and remains useful tomodernscientists.

The twofold classification of drives arises from certain naturaldifferences.(1)Theactualoperationswhichestablishdrivesmaybethoseofdeprivation (for example, of food or water) orstimulation (as by electricshock or painfully strong lights or noises). (2)The types of reinforcementwhichareeffectiveafter,orappropriate to, these twooperationsalsodiffer.(3)Appetites can be reduced or satiated,whereas aver sions cannot.Thus,given sufficient food, hunger is erased; but an animal that is aversive toelectricshockorstrong lightcannotbesatiatedwithno-shockordarkness.(4) Most deprivations must extend over some length of time before theirdrive effects become evident. The build-up may require hours, days, orweeks (think of thirst, sex, and the sub-hungers). By contrast, aversivestimuli act, for all practical purposes, immediately. No sooner are theyapplied than they result in widespread reflex changes and set up theconditionsforthestrengtheningofresponsesthatremovethem.

With respect to aversions, it is likely thatany stimulus, sufficientlyintense,may be drive-inducing.The intensity of stimuli forms a physicalcontinuum going from zero to extreme magnitudes. At low but supra-liminalvaluestheymayserveanelicitativeordiscriminativefunctionwhileremainingmotivationally neutral, in that the animalwill neither 'work forthem' nor 'seek to get rid of them.'Atmoderate to high values, theymaytake on an aversive quality, and responses may be strengthened by theirremoval. Extreme intensities of stimuli may, through their 'emotional'effect, be no longer suitable for discriminative or aversive purposes. (Forfurther discussion, seeChapter 10.) At any rate, an attempt to list theknowndrivesshouldprobably,therefore,includeunderaversionsareferenceto all the stimuli to which an organism is responsive, with theunderstandingthatthisholdsforagiven,butasyetindeterminate,rangeofintensityvalues.

TableXRELATIONOFDRIVE-OPERATIONSTOREINFORCEMENTAND

SUBSEQUENTBEHAVIOR

SomeMisconceptionsaboutDriveThemodernconceptionofmotiveshas takenfirmshapeonlyin the last

fewdecades.Beforethat,motivationwasdiscussedfrommanystandpoints,and, since the available facts were few, it is not surprising that manydifferent conclusions were drawn. Our heritage from those years, as is sofrequently the case in the history of science, is a mixture of valuableobservations and now-discarded theories.We willingly accept the former,but hasten to correct the latterwhen they creep intomodern thinking. Forthisreasonwemustpauseheretoindicatesomeoutmodednotionsofdrivewhichseemespeciallyproductiveofconfusioninapresent-daycontext.

1 .A drive is not a stimulus. Historically, a stimulus was treated as agoad(stimulus is the Latin word for goad) and was often confused withmotivation(motive means "movement-initiating"). These are, of course,only manners of speaking and are of little factual or theoretical value.Adrivehasneitherthestatus,northefunctions,northeplaceinareflex,thatastimulushas.Itisnotapart,orachangeinapart,oftheenvironment;itisnot,initself,eithereliciting,reinforcing,ordiscriminative;andit isnotcorrelated with a single response (as is the stimulus) to give us ourbehavioralunit, thereflex.True, thedistinctionbetweenstateandstimulusis hard to maintain when we consider the aversions, which arestimulus-producedstates,butwemustnotconfusetheeffectofthestimuluswiththestimulus itself.We would say that the aversive stimulus sets up a drive,one effect of which is to change the momentary strength of a group ofreflexes and another to make reinforcement possible. But to say that astimulusexcitesoneormorestimulus-responserelationsdirectly is todealinabsurdity,leadingustosuchaparadigmasthis:S——(S——R).

Anotherdifficulty inkeepingdrives and stimuli distinct lies in the factthat there are internal stimuli whichaccompany drives and may serve asSD's for a response. Thus, an operation like food-deprivation whichestablishes the hunger drive also gives rise to internal stimuliwhichmay,in turn, occasion a response of one sort or another—perhaps theverbalizationI'm hungry. There is no cause to equate two things merelybecause theyco-exist,whenactually theyhavedifferentpropertiesandplaydifferent rô1es in the control of behavior.A drive is not identifiable withthestimuliitmayitselfevoke.

2.A drive is not a response.Although a motive involves behavioralchanges, it is not in itself a response. An organism does not 'respond'(except in a figurative way) to an aversive stimulus or an appetitivedeprivation by developing a drive state. Drive is established by theseoperations, but is not a response to them in anyproper senseof theword.Toputit inanotherway,amotiveisneitherasmooth-muscleorglandularrespondentnoraskeletal-muscleoperant.

3 .A drive is not a physiological concept. Everyone acknowledges thatbehavior is the behavior of an organism, and is consequently accompanied

by physiological, chemical, mechanical, electrical, atomic, and otherprocesses in that organism. One must also acknowledge that drive-establishingoperationshavesimilarbroadconsequences.Butthesefactsdonot compromise the status of drive as a behavioral concept.The conceptarises from behavioral data, and it is in a science of behavior that it isrequired.The physiologist has no need for drive in his account of, say,blood changes during hunger. He correlates these changes with food-deprivation, just as we do reflex changes; and his correlations arephysiological principles, while ours are behavioral ones. The laws ofbehaviorarenot the lawsofphysiology,neurology,chemistry,orphysics.Yet,onedisadvantageof treatingdrive,evenfictitiously,asa 'state,' is thetendency it fosters in the beginning student to think of it as somethingphysiological which intervenes between the establishing operation and thereflex changes. This habit of thought is readily acquired by one whobelieves that behavior 'cannot really be understood' without reference to'underlying bodily processes.' In the end, however, the'understanding' ofbehavior depends upon finding lawfulnessin behavior and, once found, itshouldnotbeignoredorrenouncedinfavoroflawfulnessinphysiologyoranyothersubjectmatter.

4 .A drive is not pleasure-directed. Motives involve neither thepurposive nor the pleasure principle commonly discussed by thephilosopher.Whenone says that anorganism 'wants toobtainpleasure' or'works to get the satisfying reward,' he uses a manner of speech whichdimlyexpressestheactionofreinforcementinstrengtheningaresponse.Wehave noted (here) thatpleasant is a human verbal response that may beattached to certainSD's and is related to the appropriatenessof a reinforcertotheprevailingdrive.Hedonicphilosophiesdonotstickclosetothefactsofbehaviorinascribinganobjectiveexistencetopleasure,andstressingtheprocurement of pleasureper se as amotive or the purpose of allmotives.Wedonotdepriveanorganismofpleasure,butof food;wedonot reducehunger with pleasure, but with food, and the purpose of the organism isirrelevanttoeitherdeprivationorsatiation.

SOMEREPRESENTATIVEDRIVES:ACTIVITYDRIVE

Everyresponse,whetheroperantorrespondent,isanactivityand,inthissense, every psychological experiment deals with activity. But the termsactivitydrive andgeneral activity have a different and definitemeaning asthey are applied to the motivational and behavioral characteristics of anorganism.Activity drive is classed among the appetites, and is the sourceof action when other reasons for behaving are ruled out or controlled. Innature, periods of activity are preceded by periods of inactivity duringwhich the need increases, just as hunger drive appears in an alternation of

eatingandnot-eating.The fundamentaloperation, then, isdeprivation,andactivitydrivecanbecontrolled in the laboratorybyexperimentally varyingthedeprivationtime(i.e.,byimposingvariableperiodsofinactivity).

Thedrive ismanifested in general activity, bywhich ismeant the totalover-allmovement,orgetting-around,of theanimal.Asweshall see, it isnotnecessary inmeasuringgeneralactivity todissect it into itscomponentspecific responses like running,climbing, scratching,andsoon. Itmayberecorded as though itwere a single thing, although actually it is the grosssumof all the reflexeswhichhavebeen increased in emission rate becauseof the drive operation. It is possible, however, and desirable in manyconnections, to use one component of activity as ameasure of thewhole,because the rise in strength of the component is a reflection of the generalincrease.Generalactivity is influencedbyothermotives thanactivitydrive(alldrivesproduceconcurrentchangesinthestrengthofmanyreflexes),anditmay then serveas an indicatorof thepresenceof theseothers. It isonlywhen these others are not present that one speaks of a genuine activitydrive.

AnIllustrativeSeriesofExperimentsIn 1922, Richter introduced a new method for recording the general

activity of the rat.This made use of a triangular cagemounted on rubbertambours whichwere connected by air tubes to a kymograph recorder.Asthe animal moved about the cage, the shifting weight on the tamboursaffectedtheairpressureinthesystem,producinggreaterorlessagitationofthekymographpendependingupon theamountofmovement, thusgivingameasureoftotalactivity.

In one experiment, with other drives and external stimuli controlled,evidence for the activity drive was sought, and found, in continuoustwenty-four-hour recordsof theactivityoccurring"spontaneously"with themere passage of time.Active periods were seen to appear (ten to fifteentimesperday)withgoodregularityofspacingandduration.Thisdependedsomewhat upon the age of the animal, both young and old animals beingless active than those in their prime.Although there were individualdifferencesamonganimalsat everyage, eachone revealed typicalburstsofactivity.The demonstration of pure activity drive has been paralleledwithotherorganismsthantherat,andthedrivehasturnedouttobeafactorthathastobereckonedwithinagreatvarietyofpsychologicalexperiments.

FIG.64.Richter'stambour-mountedactivityapparatus.(AfterRichter,1922.)

With the same apparatus, Richter went on to study the relation ofvarious drive-establishing operations to general activity.He found that hismethod, although lumping together all changes in reflex strength into thesinglemeasure of activity,was sensitive enough to showup the effects ofthedifferentoperations.Thus,inanexperimentwithhunger,agroupofratswasput on a twenty-four-hour eating rhythm.Each animalwas allowed toeattosatiationfromanunlimitedsupplyonceadayatthesamehour,freshwater being always present. Half of the group was then deprived of bothfood andwater, while the other half was deprived of food alone, and thedeprivationswere continuedwithoutbreakuntil all the animalsweredead.Each rat lived in his own tambourmounted cage for the duration of the

experiment, and continuous twenty-four-hour activity records were taken.The results showed several things. In the first stageof the experiment, thetwenty-four-hour rhythmic hunger sent activity above the usual level forsatiatedanimals.Moreover,thetemporalpatternoftheactivityoverthedaytook on a characteristic form, there being a rise about twelve hours after ameal and another as the next mealtime approached.When the continuousdeprivation was started, the rats without food and water showed regulardecreases in activity until the fifth day when all activity ceased (shortlybefore death), whereas those with food-deprivation onlyincreased theiractivityforthefirstfewdaysandthendroppedofftozero(eighthday).Theaddition of thirst to hunger, then, gave a different result from a mereincreaseintheeffectofhungeralone.

Richter also tested the effect of temperature on activity. Threetemperature levelswereused:10-15°,23°,and29-30°Centigrade.The ratswere on a twenty-four-hour eating rhythm, with all other conditionscontrolled.He found lessactivityat either extreme thanat23°Centigrade.Experiments by othermen, andwith other organisms, support the generalfindingthatatemperaturedrive,certainlyamongwarm-bloodedanimals,isas valid a drive as any.Mass migrations, and the perennial flight of citydwellers to the seaside during a heatwave, bearwitness to the power of afewmillimetersofmercuryinathermometercolumn.

Instillanotherexperiment,Richterstudiedthe'nocturnality'ofrats,thatis,theeffectofilluminationonactivity.Theanimalsspentalternatetwelve-hourperiods in the lightandin thedark, theseperiodsbeingcontrolledbythe experimenter without reference to the natural day-night cycle. Beforerecordingwas started, theywere put on a twenty-four-hour eating rhythm.To insure no contamination of the records by activity related to feedingtime,halftheratswerefedjustbeforethetwelve-hourlightperiodandhalfbefore the dark.Activity was found to be higher in the dark than in thelight,withthedifferenceincreasingforolderanimals.Thisresult;hasbeenconfirmedbyotherinvestigators(HuntandSchlosberg,1939)andisrelatedto thedepressive effect of light uponbar-pressing (Skinner, 1938), aswellasthestudyofthelight-aversiondrivetowhichweshallreturnpresently.

FIG.65.Anactivitywheel.

Inadditiontothetambour-mountedcage,Richteremployedalaboratorydevice which had been in use for some time before his work, theactivitywheel.Thisapparatusisanenclosedwheel,suspendedbyitsaxlesothatitrotates easily, the arrangement being that of a circular tread-mill. From alivingcagesetby itsside, theratcanenter thewheelwhenever thedoor isopened, and run asmuch and as fast as hewishes,with thewheel turningfaster as he runs faster.Eachrotation is automatically tallied on a counter,andreadingscanbetakenforanyperiodwithoutdisturbingtheanimal.Thewheeldoesnotprovidean indexofgeneralactivity in thesamewayas thetambour-mounted cage, since only straight-away running is counted andneither lateral movements within the wheel nor partial revolutions are

recorded.Nevertheless, data obtainedwith this device are highly correlatedwith those from recorders like Richter's, and can therefore be taken as anequivalent measure. Using the wheel, Richter and other workers haveconfirmedand supplemented the earlier findingsonpure activitydrive andthe effects upon activity of hunger, age, and other factors. Of especialinterest is the finding that activity varies as a function of deprivation. If arat is confined to a small cage and denied access to thewheel except for ashort daily period, he will do more running in the time allowed than hewould do otherwise; the drive apparently works like any other appetite inthisrespect.

OtherIllustrationsofActivityDriveThere are many indications of activity drive in non-experimental

situationswhichpointtoitsreality.Cagedanimalsinzooshaveperiodsofactivitywhich visitors try to catch. Lions and tigers pace, elephants swayand 'dance,' seals swim and dive, monkeys climb and chase—althoughother drivesmaybe taken care of and the activitymay seem 'unnecessary.'City dogs, ordinarily in leash, jump and run and roll on the groundwhenfreed in a park or meadow, not unlike children let loose from school orspending the first hours of a country vacation. Some observers have evenreported thathuman infants,overly restricted inmovement, 'workoff' theiractivity in rhythmic swayings, rockings, and jouncings within their play-pens or cribs. Such instances seem reasonably clear in terms of activitydriveandactivitydeprivation.

ActivityandtheOperantLevelIt was noted earlier (here) that the operant level of any representative

reflexisrelatedtotwothings: thegeneralactivityof theorganism,andtheeaseof conditioning that response.The introductionof a drive likehungeris relevant tobothof these.Notonlydoes itput intoourhands thepowerof reinforcement, hence of conditioning, but it also expedites theconditioning by raising the general activity. This rise will, if we haveproperlychosenourreflex,carryalongwithit theoperant level,sothat theresponse becomes more available for reinforcement and our chances ofreinforcingandconditioningitquicklyareenhanced.

Otherfactors,ofcourse,mayenterinsettingtheoperantlevel.Therearespecies differences (think of the relative activity of a turtle and a mouse);thereareagedifferenceswithinanyonespecies;andyoucanprobablythinkof half a dozen additional determiners. Species differences are especiallyimportant when it comes to the form of the response to be conditioned.Pigeonscanbetaughttopeckattargets,ratstopressbars,monkeystoturnkeys or lift latches, and so forth.The response, naturally,must bewithinthe organism's capacity.Usuallywe select one that is easy, or natural, forthe species in question. If one decides to use a response which, though

possible for the organism but having an almost-zero operant level,conditioningwillbeslowandmayevenrequireagradualdifferentiation.

HUNGERDRIVEThebeginningofscientificresearchonhungerisofinterest,amongother

reasons, because it shows how, in the gradual refinement of a concept,different pathsmay be taken by different people; how, by experiment andhypothesis, exploration and reasoning, proof and counter-proof, there isfinally left a residue of agreed-upon facts and conclusions which areincorporated within the body of scientific theory. In using hunger as ourmodel in theprecedingpages,wehavealready said agooddeal about thisdrive.Wemay,however,tellabitoftheearlyhistoryofhungerresearchasasettingforsomefurtherfacts.

Richter's work on activity and hunger drive, together with informationalready at hand, containedwithin it the essence of a proper formulation ofhunger and, indeed, of all drives.He had performed a deprivation; he hadobserved concurrent changes in the strength of many reflexes ('heightenedactivity'); and he had two techniques for measuring these changes (thetambour-mounted cage and the running wheel). Moreover, his techniqueswere able tomeasure thedegree of the drive, that is, they showed the co-variation of amount of response changewith amount of deprivation.Withthis ground-work laid, the concept of drivewas ready to be developed.Atthis point, however, he and other workers were diverted by a strongcontemporary interest in the physiological correlates of hunger, and theappropriatebehavioralconceptofmotiveswasside-trackedforseveralyears.The physiological studies were centered upon events in the stomachresulting from food deprivation, and it was to these that attempts weremadetorelatetheactivitychangesinhunger.

HungerandtheStomachIn 1912, a decade or more before Richter began his work, W. B.

Cannon, in collaborationwithA.L.Washburn, published a paper entitledAnExplanationofHunger.Tracingbacksomesixty-sixyears the idea that"hungeristheresultof(stomach)contractions,"theyfeltthattheywereabletoofferdirectproofofthecorrectnessofthisview.Theirdatawereobtainedbymeansofasimpleapparatusconsistingofasmallrubberballoonwhichwasswallowedbythesubject(Washburn,principally)afterhisstomachwasemptiedoffood.Tothisballoonwasattachedathinrubbertubethatleduptheoesophagusandout the subject'smouth.Theballoonwas then inflateduntil itassumedthecontoursof thestomach.With the tubeconnected toawater manometer, any pressure upon the balloon could be observed andrecordedbyafloatingmarkeronthewater.Aftersomepractice,thesubjectswere able to retain the balloon and tube without nausea, andexperimentationwasbegun.

FIG. 66. Portion of a record ofhuman stomach contractions duringhunger. The white signals on thebottom line are the subject's reportsof "felt" hunger.The reports coincidewith peak contractions. The secondline is a time line, marked off inminutes.(FromCannon,1929.)

The findings may be brieflysummarized. A few hours aftereating,whenthestomachisempty,powerful stomach contractionsbegin. These contractions are ofabout thirty seconds duration, andthey occur at thirty- to ninetysecond intervals.Theymaygo on,in this rhythmical fashion, forthirty minutes or more, to befollowed by quiescent periods ofthirty toninetyminutes.When thesubject, unaware of the record hisstomach was writing, obeyed theexperimenter's instructions andsignalled(bydepressingatelegraphkey) whenever he 'felt hungerpangs,' his signals were found tocome when "contractions wereinvariably being registered" andnear the timewhen their forcewasmaximal.

From their data, Cannon andWashburn came to the conclusionthat hunger, which they describedas "a dull ache or gnawingsensationreferredtothelowermid-

chest region and epigastrium,"wasdue to the stimulationprovidedby thestrong contractions of the empty stomach. So great was the impressionmade by their research, that this stimulus theoryof hunger prevailed formany years.We know, today, that there is more to hunger than stomachcontractions. It has been shown, for example, that food-getting activitycontinues with undiminished vigor when the stomachs of rats have beensurgically isolated to the degree that stomach contractions could notpossiblyserveasstimuli foranyresponses(Bash,1939); thatanimalswilleat prior to the appearance of the contractions and long after thesecontractions have disappeared; and that changes in response rate as afunction of deprivation are inconsistent with any known properties ofstimuli, in the stomach or elsewhere.Yet the stimulus theory of hungerpersisted,andsostrongwasitsinfluencethattheoristswereledtopostulatea stimulus nature for all other drives, even those which, like sex and thesub-hungers, shouldhave encouraged a reconsideration of the evidence andledtoatruerconception.

Before Richter's time, there were two major obstacles in the way of a

correct interpretation of drive.Onewas thePavlovian notion of stimuli aseliciting agents. In looking for the cause of behavioral changes associatedwithdrive,theonlyknownpossibilitywastheelicitativeactionofstimuli.Therecognitionofoperantbehaviorstilllayahead.Thesecondobstaclewasthat the subject, as in the Cannon-Washburn study, could report the'experience' of hunger pangs. They, and later workers, took the verbalresponses and signals, not as behavioraldata to be explained, but asobjective indicatorsof thedrive'spresenceandaction.Todayweknowthata drive-establishing operation has among its consequences the creation ofinternal stimuli which can act as SD's for operant responding. Stomachcontractions are stimuli of this sort.They are one bodily result of fasting,and they are certainly discriminable by the organism, as Cannon andWashburn discoveredwhen their signals 'I am feeling a hunger pang'wereseen to be occasioned by the contractions. But, while the contractions areSD's for the verbalization or the report of a 'sensation' or 'experience' ofhunger, they are not the drive itself.The 'feeling' of hunger is not hunger,or,tosayitdifferently,hungerisnotinthefeelingofit.

TheMeasurementofHungerAswithanydrive, themeasurementofhunger ismadepossiblebyany

response change which co-varies with the degree of the drive-establishingoperation—amount of deprivation. Thegeneral activity method gives theneededrelationbetweendeprivationandresponsechange.Inacrudeway,itprovides ameasure of response rate, sincewith it one obtains the total ofresponseemissionsoveraratherbroadperiodoftime.Boththesethingsaretrueof thesecondtechnique, theactivitywheel.Theobstructionmethod isone in which the response must be made in the face of a deterring oropposing'resistance'likeelectricshock.Thatis,theanimalmustsubmittoa negative reinforcement before he can proceed to the positive.The animalisacclimatized inabox,with twochambers,andallowedfirst to run fromone to theotheracrossanunchargedgrid toa reinforcement, suchas food.Then the shock is turned on, and a test period of, say, twentyminutes isgiventoseehowmanytimeshewilltaketheshocktoreachthefood.Aftereach crossing, he is allowed a nibble and then quickly transported back tothe starting chamber for another trial. This method has features whichcomplicate the data got from it. Principally, one must remember that itdepends upon aconflictful situation in which positive and negativereinforcement are combined and in which two drives, an aversive and anappetitive,arealwaysoperatingtogether.

FIG.67.FloorplanoftheColumbiaObstructionBox.Thelabeledpartsare:A—starting chamber;B—passage-way with grid floor;C andD—reinforcementchambers.(AfterWarden,1931.)

A more recent, and very satisfactory, method employs response rateunderP-R.Co-variationbetweendeprivation and response is shown in theemissionrate of the response,which is a fundamental indicator of operantstrength.Figures 62 and 63 (pages 267 and 268) have already illustratedthis method as it was first used by its originator with the bar-pressingresponse. If you will look back at these graphs, you will see that thechangingslopeofaP-Rcurve isadmirablysuited tomeasuringdegreesofhunger.Themeasure isuncomplicatedbya simultaneousaversivedrive; itissensitivetosmallchangesindrivelevel;itisapplicabletoasinglereflexrather than a mass of reflexes; the emission pace is not restricted byarbitrary trial imposed on the organism by the experimenter; it is usefulwith a single animal, anddoesnot require the averagingofdata formany;and, finally, it yields ameasure of reflex strength at any desiredmoment,instead of a gross one for a long test periodwhichmay run to a full day.Such a procedure is superior to the use ofeating curves (whichmay havesuggested itself to you) because, in the latter, the rate of respondingdepends somuch upon the rate atwhich the animal can eat (seeFigure 7,here).Aboveacertainpointofdeprivation,theanimaleatsasfastashecananyhow, with the result that the slope of an eating curve is a relativelyinsensitivemeasure of drive strength. P-R curves depend less upon eatingtime,andtheirslopesarefreertovary.

HungerDriveinManAsis trueofsomanybehavioralproblems,directexperimentationupon

human subjects is very difficult to arrange in the case of hunger. Socialpressures militate against the use of infants, and adults are poor materialbecause of their long and unknown personal histories of conditioning andmotivation. Yet, there is no reason to believe that hunger is any less

important for man than for lower organisms. History records hugemigrations in search of food, and wars over bountiful lands; there arereportsofhumancannibalismunderextremestarvationdespiteallprevioustraining; and there is, of course, Napoleon's classic and highly crediblestatement that "an army travels on its stomach."The psychologist,Wada(1922), who studied the relation of hunger contractions to activity ininfants, felt that hunger was initially the strongest of all drives, since thechildwhennothungryisusuallyasleep.

Several studies have shown that, among the human responses which adrive like hunger can affect, we must include the verbal. Discriminativestimuli normally insufficient to evoke food-word responses may becomeincreasingly effective under stronger degrees of hunger. Thus, Sanford(1936) sought to find the relation of hunger to the number of foodwords(namesof foods,meals,etc.)emittedbyagroupofcollegestudents.Testsofwordassociation,wordcompletion,andthelike,weregiven(withoutthesubjects' knowledge of their true purpose) at various between-meal periodsand after a twenty-four-hour fast. Results indicated an increase in foodwords as a function of increased time since eating. In a similar type ofexperiment(Levine,Chein,andMurphy,1942),volunteeradultsunderwentfooddeprivationsof one, three, six, andninehours, afterwhich theywereshown various ambiguous or 'nonsense' figures briefly behind a ground-glassscreenandasked toname the 'objectspictured.'Thepercentageof thetimesthatfoodswere'seen'wasgreaterforthehungriersubjects.

Suchresultsremindoneofthegreatamountofsexualconversationheardamong the members of isolated and sex-deprived groups—soldiers,convicts, andothers.Also in linewith this, it iswell-known that extremedegrees of drive may precipitate 'illusions' wherein very weak or usuallynon-generalizing stimuli may be responded to in a manner that seemspathological to theunmotivatedonlooker.Apointmay be reached, in fact,wheredrive is so strong thatnoexternalSDat allmaybe required for theresponsetoappear:astarvingmanmay'see'hisfavoritedishesbeforehim.

THESUB-HUNGERSThe possibility of demonstrating sub-hungers depends on the fact that

the withholding of food need not be an all-or-none matter, but can belimitedtoparticularconstituents.Inplaceoftotaldeprivation,ananimalisoffered all he can eat of a diet which is unbalanced or deficient in somerespect.The question iswhether such operations (selective deprivations orselectivesatiations)haveeffectsonsubsequentresponseswhichwarrantourspeakingofvarioussub-hungerdrives.

TestingforSub-HungersThereareseveralwaysinwhichsub-hungershavebeendemonstrated.In

t hesatiation method, as used by Bousfield (1938), the animal is firstallowed to eat until satedwith one food, afterwhich that food is removedand one ormore others are offered one at a time.When catswere used assubjects and four kinds of food were rotated in order of presentation onsuccessive days, he found that satiation with one food might lead to nofurther eating when the others were presented, whereas satiation withanother might be followed by successive satiations on all those thatfollowed.

A second procedure is that called thefree choice orselective eatingmethod.This gives the animal anopportunity to select, in any amount hewishes, from among two ormore foodsmade available at the same time.Themethodcanbeusedonanimalsthathavebeenheldforsometimetoadietdeficientinsomeelements,orwithanimalsthat,althoughrearedonanadequatediet,havecreatedtheirownselectivedeprivationbyvirtueofone-sided choices from among foods offered. Numerous experiments with anumber of animals (pigs, cows,chickens, mice, etc.) agree quite well inleadingtotheconclusionthat,givenachoice,organismswilltendtoselectthe one food needed for an optimal diet or will draw in time frommanysourcestomaintainnormalhealthandgrowth.Moreover,thechoiceamongrations is governed by the changing requirements of the organism duringmaturation,pregnancy,illness,andsoon.

TableXIDATAFROMONECATINANEXPERIMENTUSINGTHE

SATIATIONMETHODINTHESTUDYOFSUB-HUNGERS

(DatafromBousfield,1938)

Anexperiment,performedbyDavis(1928)onthreeinfants,supportsthesurmise that the phenomenon of sub-hunger is not limited to loweranimals. Just afterweaning, the childrenwere started on a selective eatingprocedurewhich lasted sixmonths for two of them and an entire year forthe third. At each mealtime, a number of natural and cooked foods,including liquids, were placed on a tray before them.They ate with theirfingers, or spoons, ormore directly, as theywished, choosing freely fromamong theofferings.Over the long courseof the experiment, they showed

themselves to be omnivorous, and their choices were nutritionally broadenoughtopromotehealth,energy,goodsleeping,andsomewhatbetterthannormalweightincreases.Occasionally,oneofthechildrenwoulddevelopastrong preference for one food (e.g., cereal or meat) which would last fordays, but would then ease off in favor of a wider selection. It was evenrecordedthatoneinfant,whohadricketsatthebeginningofthestudy,tookcod-liver oil until the condition disappeared, and then stopped taking anymorel

FIG. 68. Model of an apparatus used by Young for studying foodpreferencesandsub-hungers,A—entrancechamber;B—choicechamber;C1and C2—proffered foods. (From Harlow, inT. G.Andrews, Methods ofpsychology[JohnWiley&Sons,Inc.,1948].Bypermission.)

SomeFinalObservationsThesefindingslendcredencetonon-experimentalreportsofthewaymen

andanimalsadjusttheireatingtotheirdietaryneeds.Deertravelfartoreachlickswhen they require salt; cattle in phosphorus-poor areaswill gnaw onbones to supply the lack; dogs may eat grass when ill; malnourishedchildren are said to eat earth or schoolroom chalk because of themineralstheycontain.Withinthelaboratory,ifratsinacolony arenotproperlyfed,cannibalismmayappearasameansofdietarycompensation; and the samereasonmayleadparturientfemaleratstodevourtheirnew-bornlitters.

Although these things seem dramatic at first reading, it should berecognizedthatanimalsinthewildstate,grantedavariedandfullsupplyof

food,arealwaysfacedwiththeproblemofwhat theyshouldeat.That theyhave survived, matured, and multiplied points to the biological value ofsub-hungersinthelifeoftheindividualandthespecies.Contrariwise,it iscurious to note how human adults, under the direction of social learningandthepressureofcustomorfashion,maysufferfrommalnutritiondespitean abundance of food. He whose menu is dictated by the socialacceptability, rather than the adequacy, of foods, needs to have his naturalsub-hungersgivenahelpinghandbythescienceofdietetics.

SEXDRIVEThere is a large amount of scientific information and general

misinformation concerning sex—more thanwe can outline (or counteract!)here.Wemustlimitourselvestosomeintroductoryfacts.

Thesexdrive isclassedwith theappetitesbecause it involvesresponseswhich change in strength as a function of deprivation and satiation.Withthefemalesofmanyspecies,thedriveismarkedbytemporalcyclessolongas satiation (or insemination) is not provided. Seeking, approaching,courting, and copulating are broad categories of responses affected by sexdrive. Changes in the strength of such responses are fairly clear in lowerorganisms, but obscure in man owing to long and meticulous socialtraining. As a basic motive in animals, sex has been investigated withmethodssimilartothoseemployedwithhungerandotherdrives.

Inour short treatmentof sexdrive,weshall referoften tophysiologicalcorrelates of the drive. Much work has been done by other sciences onsexual mechanisms. But sexdrive is still abehavioral matter, and astimulusorglandulartheoryofsexbehaviorwouldbeasextraneoushereasastomach-contractiontheoryofhunger.

SomeClassicalStudiesofSexBehaviorAfewpioneerstudiesmaybecited,bothfortheinformationtheysupply

and as illustrations of the experimental approach to the problem. Thesestudiesmaylogically,andforhistoricalcontinuity,beorganizedaroundtheapparatusandmethodsemployed.Twolinesofdevelopmentwillsufficeforourpurposes.

1.Studieswith theactivitywheel. In1923,Wang,working in thesamelaboratory with Richter, reported his findings in an experiment on theactivity of adult female white rats. In the wheel, each female showed ahighlyconsistentcyclicalriseandfallinheractivity,withpeakseveryfourto five days and declines on intervening days. Wang followed up thisobservation,andprovedthatthecyclewasintimatelycorrelatedwitheventsin the reproductive organs of the female, events called the oestrus rhythm.Activity is greatest at theheight of oestrum, atwhich time there is a flowfrom the vagina and the female is said to be 'in heat.'At this time, too,ovulationoccurssothatthefemaleismaximallyreadyforinsemination, is

mostreceptivetosexualadvancesbythemale(whennotinoestrum,sheislikely to resist copulationand fightoff themale), and ismost stimulatingto themale byway of olfactory cues given off by the vaginal discharges.The rise to, and recession from, the peak of oestrum coincides withincreasing vaginal discharge (and the cellular content of the flow is acontinuallychanging indexofoestrusstages),whilebetweenpeaks there isno discharge at all. (An interesting comparison may be noted hereparenthetically.Therat'soestrusperioddoesnotcorrespondfunctionally tothemenses of human females. In contrastwith the rat, the latter are fertileapproximatelymidway betweenmenstrual periods, and infertile at menseswhen the ovum is being discarded in the vaginal flow.Also, for mosthuman females, the susceptibility to fertilization is low justbeforemenseswhen the uterus is preparing for menses, and again just afterward whenovulationmaynotyethaveoccurred. It is these featuresofhuman fertilitythatarethebasisoftheso-called'rhythmmethod'ofbirthcontrol.Someofthe higher apes are thought to resemble the human female in this respect,but it is not yet clear just where in the evolutionary scale, and why, thiscuriousreversaloccurred.)

FIG. 69.The effect of the oestrus cycle upon the activity of the femalewhite rat. Notice the regularity with which the peaks of activity occurcomingattheheightofoestrum.(AfterWang,1923.)

Wangandlaterworkerswentontoshowanumberofothercorrelations.The characteristic activity cycle of the adult female rat is missing beforepuberty,andcommencesat pubertywhenoestrumdoes; it isabsentduring

pregnancy, and after parturition (while lactation is going on), two periodswhen the oestrus rhythm is in abeyance; it is restored afterweaningof thelitter,whenoestrus rhythmreappears; it canbepermanentlyerased, just asoestrumis,bysurgicalexcisionoftheovaries,aneventthatalsolowersallactivityandrendersitmoreuniformfromdaytoday.

Themale rat does not exhibit the sex-activity cycle seen in the female.Male rats, like human males and others, differ in this regard from manyspecies in which males as well as females have special mating seasons.Strength changes in all sex reflexes are the result of deprivation andsatiation.Themale'ssexdriveislowestrightafteraperiodoffreeaccesstoa female in heat, but recovers much of its strength within a few hoursfollowing.

2 .Studies with the obstruction method. The first to devise anobstruction apparatususing electric shock for the studyof drivewasMoss(1924).HisapparatuswasaforerunneroflaterimprovedmodelssuchastheColumbia apparatus constructed by Warner and described earlier (here).WithMoss's ideas of drive (namely, that drive is an "impelling force" thestrength of which could be measured by the amount of "opposingresistance" or "repelling force" it would overcome) we are no longerconcerned. But his device and his findings were the instigators of muchworthwhile research and are still, therefore, of interest. Having undertakentomeasure the sexdrive of the rat by its "willingness" to undergo electricshockinordertoreachasex-object,hecomparedsexandseventy-two-hourhungerbypermittingratstochoosebetweenacompartmentcontainingfoodand one containing a mate. His observations (on too small a number ofanimals to be significant) were that females in heat cross the grid moreoftentoreachthemalethanthemaledoestoreachthefemale;andalsothathungercouldbea strongermotive than sex, sinceabout80percentof themales chose food rather than a female. Tsai (1925), using a method ofsimple choicewithout any shock, came to a similar conclusion.Hismalerats, only twenty-four-hours hungry, chose food rather than a female inabout 75 per cent of the chances given.While sex has been condemned orextolledas themostpowerfulofdrives, this is apparentlynot trueevenofinfra-humananimals.

The strength of sex drive as related to length of sex deprivation wasinvestigated byWarner (1927) with the Columbia obstruction apparatus.Male rats under six degrees of sexual deprivation were tested for gridcrossings to a female in oestrum within a standard observation period oftwentyminutes.Thesixdeprivationperiods,timedfromthelastsessionofunrestricted copulations, were 0, 6, 12, 24 hours, 4 and 28 days. Heconcluded that maximal drive was reached after about twenty-four-hours'deprivation, with subsequent slight decline to the twenty-eighth day. Abreakdown ofWarner's data byLeuba (1931), however, raised some doubtabout thedeclineaftermaximum.If thegridcrossingsare tabulatedfor the

four 5-minute parts of thewhole 20-minute session, the crossings after 24hoursand28daysdistributethemselvesasfollows:

AVERAGENUMBEROFCROSSINGSFORGROUPSOFANIMALS

Theaccelerationofcrossingsbythe28-daygroupwithinthe20-minutetestperiod, despite its smaller total indicates two things: (1) after 28 days,adaptation to the shock builds up more slowly; and (2) that a 20-minutetestperiod is too short tohavea full indicationofdrive strength.There isreason to believe that sex drive in male rats does not decrease after amaximum, but instead that after the first rapid increase therearediminishing increments with prolonged deprivation. It may be mentionedincidentally thatWarner, too, showed that sex can be outweighed by bothhungerandthirstasmotives.

Theobstructionmethodhasalsobeenusedtostudy 'maternalbehavior.'Inthefemalewhiterat,thisisacomplexofmanyactivities.Objectively,itappears as nest-building, nursing, retrieving the young or returning to thelitter when separated from them, and more. Nissen (1930) found that thenumberof grid crossings a parturient femalewouldmake in order to reachher litterwasmaximal just after birth, and decreased until the youngwereof weaning age. Many observers have recorded their belief thatreinforcement is the basis of this reaction: regular suckling by the youngrelievesbreastcongestioninthemother,andallowsnormalfunction.ingofthemilkglands;thedeclineinlactationcoincideswithweaningtime(abouttwenty-one days after birth), and return-to-litter behavior fades out.Although physiological correlates are not known as yet for the othercomponentsofmaternalbehavior, it is a fact that they all lose strengthupto weaning time, after which the female's oestrus rhythm re-emerges.Wemay note in passing that the principles of stimulus generalization anddiscrimination show themselves in the operant reactions that compose thematernalpattern.Femaleswillretrieve,andcarrybacktothenest,notonlytheirownwanderingoffspring,butwilltakeyoungfromthelittersofothermothers if the age-size difference from their own young is not too great.

When the difference is large, say ten days or more, they show somediscrimination and will reject the strange young. Moreover, females willcross a grid to litters not their own, will suckle other young, and, morestrikingly, will retrieve even inanimate objects like a piece of wood or asandbagifitapproximatesthesizeoftheiroffspring.

Becauseofoursocialtraining,wetendtothinkof 'maternalbehavior'astender, loving, protective and care-providing. We call women who actotherwise 'unnatural mothers.' But the female rat, though she will takemanyshocks to return toher litter,will alsoonoccasion, if severelyupsetemotionally or subjected to dietary deficiencies, kill or devour her young.And if we get outside the confines of our society, we find practices likeinfanticideacceptableundercertaincircumstances.It isalwaysdangeroustomistake our individual ethical ideals for either the facts or the potentialityofhumanandanimalbehavior.Apracticalprogramofethicaltrainingmusttakeaccountoffact,butfactpaysnoheedtoourethicalgoals.

HumanSexualityAlthoughthesexdrivewasalwaysacceptedasimportantforinfra-human

organisms, the emphasis upon its significance in human living has comeabout in relatively recent times.At times its role among men has beengreatly, even grotesquely, exaggerated. Controlled studies of sex amonghuman beings are naturally difficult. Society's laws, personal reluctance,and the researcher's own sympathies stand in the way of experimentation;and field studies are handicapped by socially induced attitudes that blocktheinformation-seeker.Theresultisthatrealevidenceonthecharacteristicsandmodesofhumansexbehavioranddevelopmentisveryscanty.Yetitisjust here that interest is greatest, so that speculations and theories are putforth in bewildering and distorted profusion. Where facts are few andinterest ishigh, the temptation tomakeaguessand tovoiceanopinion ishardtoresist.Onlyinthelastfewyearshastherebeeninprogress(Kinsey,Pomeroy, and Martin, 1948) the first large-scale attempt to get someelementary statistics on the sex activities of typicalAmericans of all ages,social levels,educationalbackgrounds,andofbothsexes.Thisstudy, it isestimated, will take some ten to fifteen years, will cost somemillions ofdollars, and will end up with a sample of 100,000 persons in the totalpopulation of 140,000,000. While this return may seem small, it farsurpasses any effort to date, and represents the lengths towhich onemustbepreparedtogoforstatisticsonhumansexuality.

Beyondthestatisticaltabulations,however,remainstheproblemofhowsexactivitiesreflectbehavioralprinciples.Apremonitionofthecomplexityandmultiple causalityof copulationbyhumanmales, for example, canbegot from castration studies of the male rat (Stone, 1927). The effect ofcastrationontheratdependsprincipallyontwofactors,theageandamountofprevious experiencewith sex at the timeof castration.Castrationbefore

pubertyeliminatescopulationafterpuberty;andcastrationofanadultmalewhohadbeenraisedinisolation,willkeephimfromcopulatingwhenheislater given access to a female. If, however, an adult male with ampleprevious experience is castrated, his sex behavior may continue for sometime albeit at diminished frequency. Medical literature contains very fewpost-castration studies of men injured accidentally or operated on formedical necessity. Those available indicate that frequency of intercourse,and the pleasure reported, may be little, if any, reduced by the operation.The action of secondary reinforcements in these instances shows that thereareotherreasonsthanglandularforsexualintercourse,justasthereareotherreasons than stomach contractions for eating.These sexual S r's probablyincludebothexteroceptiveandproprioceptivestimulation.

SexualDiscriminationsandDifferentiationsAmong infra-human animals, it is frequently seen that the stimuli and

responseswhich are correlatedwith sexual reinforcementmay be operativewithaminimumofconditioning.It isoftensodifficult totellwhetherthestimuli are eliciting or discriminative, that instances have sometimes beencalled"unlearnedbehavior."SomeEuropeanzoologistshavesimplyspokenof"innate releasingmechanisms" in theirattempt toexplain the interactionofcertainanimals(e.g.,themaleandfemalesticklebackfish)intheirsexualbehavior. Animals at the mammalian level may show a similar quickestablishment of sexual chains.The fact is that one cannot yet tell withassurancewhenheisdealingwithaccountsofspeedilyconditionedoperantsorofrespondentsrequiringnoconditioningatall.

Male rats raised in isolation from females will mount and copulate,withoutovermuchdelay,wheneventuallypairedwithafemaleinheat.TheSD's for the male rat are largely olfactory, but anosmic males (with thesenseofsmelldestroyed)willstillcopulate,sincethereareotherSD's—forexample, the short, jerkymovements and leaps of the female in oestrum.Before puberty,male rats are not responsive to the smell cues of oestrumnoranyotherSD's from the female.At thehuman level,males respond tosecondarysexualcharacteristicsofthefemale,suchasshape,contour,voiceand others, as SD's, and in addition are affected by social SD's likefashionablenessofdress,skillwithcosmetics,flirtatiousness,andsoon.Incontrastwiththerat,too,thehumanmalemayneedsometimetolearntheresponsesinvolvedincopulation.

Under certain conditions, sexual discriminations and differentiations ofhigheranimals,andespeciallyman,arenotonly 'incorrect,'butmaytakeaturn conventionally termed 'abnormal.'These deviant responses are learnedin accordancewith the same lawswhichgovern the acquisitionof 'normal'behavior, but are deemed queer because they are out of step with more

common patterns or conflict with the codes of the society in which theyoccur.Thus,inoneofMoss'sstudies,maleratswererearedtogetherfor50-150dayswithout females,with the result that theyseemed to lose interestinfemalesandmountedeachotherinpreference.Asimilarfindinghasbeenreported for birds, with males reared together preferring males to females.Another example of early reinforcement interfering with 'normal' sexualityis that in which birds raised with members of another but related speciesmay later prefer them as mates rather than their own species. Humanhomosexuality involvesmuchmore intricate setsofpersonal relationships,but there is littledoubt thatdiscriminative trainingplaysan important roleinit.

Aswithothermotives,highsexdrivemayproduceresponses tostimuliwhich are not ordinarily sexual. Deprived persons may see a sexualsymbolism in things likemountainsandvalleys,gunsand targets, laddersand windows, and many other objects. Under high drive, stimuli of allsorts may generalize as SD's and become metaphors, as it were, of truesexualSD's for evoking sex talkor action that isbizarre to anobserver.Aphrase withdouble entendre, an unconscious pun, a mispronounced ormisspelled word, oftimes embarrassingly reveals the motive underlying aspeaker's or writer's avowed intention. Such 'slips' of the tongue or penwere cited by Freud as the "psychopathology of everyday life," and heascribed them to the workings of a "subconscious mind." There is,however, nothing esoteric about motivational strengthening of reflexes,evenwhenthepersoncannottellus,oris'unawareof,'hispresentmotives.

THEAVERSIVEDRIVESAversions, like appetites, form a major class of drives.There are two

criteriafordeterminingwhetheragivenstimulusisaversive:(1)aresponsemade in itspresence should be strengthened by itsremoval; and (2) aresponse made in theabsence of the stimulus should be depressed if it isfollowed by theadministration of the stimulus.Youwill note, of course,that these are also the criteria for a negative reinforcer. In this chapter,however, our interest will be in the motivating function of such stimuli,which may, therefore, be named aversive stimuli. The three aversionschosen for comment here—light aversion, sound aversion, and shockaversion—bynomeans exhaust the list.Others are known, still others aresuspected, and the total number of them, as in the case of appetites, is amatterofempiricaldiscoveryforeachspeciesoforganism.

LightAversionThisdrivecanbedemonstrated rathereasily in thewhite rat.Richter, it

will be recalled, found that the rat, a 'nocturnal' animal, is less active inlightthanindark,suggestingthedepressiveactionwhichwenowexpectof

an aversive stimulus; and several experiments in recent years have shownthat light-removalmayconstituteapositive reinforcement forconditioninga response (Keller, 1942).The second criterion of an aversion to light canbemet by administering the stimuluswhenever a given response ismade.Inatrialexperiment,Schoenfeld(1946)usedahungryratdailyinone-hoursessions in a bar-pressing apparatus with no positive reinforcement givenforpressings.Wheneach response to thebarwas immediately followedbyfive flashes of a bright light the rate of operant-level pressing droppedswiftly and remained low for some days after the punishing stimuluswasdiscontinued.

Acombinationof theabove-mentionedprocedurescanbeverystrong initseffect.Hefferline(here) riggedhis apparatus so that a rat'sdepressionofthebarwouldturnoffalight,butreleaseofthebarwouldbringitbackonagain.He reasoned that the bar-pressing response has two components, an'up' and a 'down' response.The 'down' received a positive reinforcement(light-removal) and the 'up' a negative (light-onset); the positivereinforcement could be obtained, and the negative reinforcement avoided,only by pressing andholding the bar down. The rats in this situationconditioned quickly to prolonged holding, so that theywere immobilizedonthebarforasmuchas95percentofeachexperimentalhour.Thisfailureto 'unmake' a response is related to those preference situations in whichanimalsmoveinto,andstayin,darkratherthanlightedareas,aswellastoother instances of sticking to 'safe' environments. It also helps us to seewhy, in experiments using aversive stimuli, some animals cannot beconditioned at all. If a partial reinforcement canbeobtained in someotherway, as by hiding one's head or covering one's eyes in a light-aversionsituation, theanimalmaybe immobilized through thecombinationof thisreinforcement with the punishment consequent upon moving from theposition—with the result that a more effective reinforcement is neverachieved.

An amusing aspect of the effect of combining positive with negativereinforcement in the manner just described is that provided by a simpleexperiment routinely carried out by first-year students of psychology inColumbia College. Some object like a metal coin or piece of wood isplacedintherat'scageandabrightlightisturnedonoverhead.Assoonasthe animal, in moving about, makes contact with the coin, the light isdiminished; and, as soonashe leaves the coin, the light returns.Within afewminutes, theanimal showsaprototypical 'fetishism':he staysclose tohis coin, taking it in his teeth, chewing on it, and carrying it about withhim wherever he goes.Although one would hesitate to say that the ratworshipped the object for itsmagical powers, one cannot but be remindedoftherabbit'sfeetandluckycoinsatahigherevolutionarylevel.

SoundAversion

An aversion to certain sounds is found inmany animals, including thestudent who is annoyed by such things as the squeak of chalk on ablackboard. Among rats, 'audiogenic seizures' are known to result fromstimulation with escaping air, jangling keys, electric bells, and the like.These seizures are marked by periods of wild, erratic behavior alternatingirregularly with trembling, prostration, and bodily rigidity; an unnaturallack of sensitivity to other stimulus influences (as when the animal ispokedwith a stick); and a condition somewhat like the "waxy flexibility"of certain patients in mental hospitals whose limbs and body are passiveand may be molded into various positions. Recovery may seemingly becompleteonthedayfollowingtheseizure,butanewepisodeismoreeasilyinducedonlateroccasions.

Inanexploratoryexperiment,tworatsweretrainedinaworking-boxthatcontainedaloudbuzzer.Indailyone-hoursessions,thebuzzerwassoundedand bar-pressings were regularly reinforced by sound-removal for one-minute periods.The responsewas quickly strengthened and soon came tobe very stereotyped in appearance, with the animal remaining close to thebarinthesilentperiods.Thelatenciesoftheresponsetosound,whichwerecharacteristically short, showed an interesting cyclical effect thatmay alsobeobservedoccasionallyunder lightaversion. Inashort seriesofstimuluspresentations, they became gradually of shorter and shorter duration; thenthey would suddenly lengthen and begin another gradual decrease; withsuch lengthenings and shortenings continuing throughout most of anexperimentalhour.Apparently,thelatencydecreasedtoapointatwhichtheescaperesponsecamesocloselyon theheelsof thestimulusonset that theresponse was virtually punished, thus bringing about a temporarydepression of the sound-removing behavior that showed itself in alengthenedlatency.

As in other aversion experiments, those with sound require rapidconditioning if the animal is not to be 'beaten down' by the stimulus andfail torespondinthedesiredmanner.Itwouldprobablybeofadvantageinsuch experiments to use as the to-be-conditioned response one which iseasier still than bar-pressing and has a higher operant level.The faster theconditioning, the less should be the likelihood of producing audiogenicseizures.

ShockAversionElectric shock is perhaps the most generally aversive of all stimuli for

most organisms. In the rat, it easily satisfies both criteria for an aversivestimulus:a responseremoving it isstrengthened,andone thatbrings itonis depressed.Wemay takeMowrer's (1940) study as an example sincehisresults resembled those obtained under light aversion. He used a box inwhich electric shock to the animal's feet built up graduallyfrom zerointensitytoonethatwashighlyupsetting,withthe 'painthreshold'reached

at approximately the end of the first minute of application.As the shockincreased above this point, the rat became very active, even frantic. If,duringthisactivity,hebumpedorpushedapanelatoneendofthebox,theshock returned to the zero level and gradually built up again in intensity.Conditioning was rapid; the successful response became very stereotyped;'freezings' to the panel were often observed following shock-removal.Allthese phenomena are equally typical of light-aversive behavior. It isimportantwhenusingshockforconditioning,evenmorethaninthecaseoflight or sound, to avoid over-stimulation, else the organism may beimmediatelythrownintoaflurryofconvulsiveactivitywhichprecludesanyformofadaptivebehavior.

HumanAversionsThat human beings possess aversive drives seems beyond question.

Infants, by their squirmings and waitings, and adults, by their verbalreports, indicate that they find some stimuli noxious, painful, orunpleasant.They easily learn to get rid of strong shocks, glaring lights,gratingnoises,putridodors,prickingpins,andthelike.Yet,whenbroughtintothelaboratory,theiraversionisnotalwaysobvious.Whenweattempt,for example, to condition a finger-withdrawal under shock-aversion, oursubjectmaythwartusbytakingtheshockwithoutcomplaintunlessitisofexceptionalseverity.Ourexperimenthasperhapsnottakenintoaccountthesubject's reinforcement history, inwhich theremay have been givenmuchpositivereinforcementforstoicalbehavior.

It is a commonplace of human experience that stimuliwhichwere onceunpleasantcanlosethischaracterandbecomeacceptable,evensoughtafter,as a result of training.Adultsmay like bitter and spicy foods that infantswillreject;theycometoenjoymusicthatiscacophonytochildrenorotheradults; they welcome such otherwise annoying stimulias the vibrationsapplied by the barber to a balding head.We have alreadymentioned suchmatters inourchapteronsecondaryreinforcement.It isnotsimplythat theaversion is outweighed by conflicting drive, but rather that the aversivestimulus becomes a secondary reinforcement.This is related, not only toour subject's failure to withdraw his finger from shock, but to the agingboxer'srefusaltoquitthering,andtheneurotic'smasochism.

A simple experiment, easily carried out in an undergraduate laboratory,provides the prototype ofmasochism. Hungry rats are conditioned to bar-pressing for food pellets, some receiving regular and others periodicreinforcement.Eachreinforcement isaccompaniedbyamildelectricshock.After an hour of such training, extinction of bar-pressing is begun, withneither food nor shock given for the response.When extinction is welladvanced, the shock alone is reintroduced as an accompaniment of bar-pressing. The result is commonly a rapid acceleration in response rate,which appears as a sizable bump in the cumulative response curve.To an

uninformed bystander watching the animal's behavior, it appears that theratsenjoytheshocksandaredeliberatelypunishingthemselves

ANXIETYANDAVOIDANCEPavlov once wrote: "... it is evident that under natural conditions the

animal must respond... to stimuli which signal the approach of harm orgood...." We now ask what happens when a discriminative stimulusprecedes an aversive one.We approached this problem in the precedingchapter, in our discussion of secondary negative reinforcement. Twoexperimentscanbedesigned. Inone, theSD is followed inevitably by theaversive stimulus; in the other, some response to the SD suffices to keepthe aversive stimulus from coming. These two arrangements provideoperationaldefinitionsof"anxiety"and"avoidance"respectively.

AnxietyIn an investigation of anxiety, Estes and Skinner (1941) dealtwith the

problem in the followingway.At intervals during periodic reconditioningof thebar-pressing response, a soft tone cameonand sounded for aperiodof fiveminutes, at the end of which an electric shock was delivered.Thetone,attheoutset,hadnoappreciableeffectuponbar-pressingrate,butverysoon it took on amarkedly depressant character at each presentation, fromwhich therewas immediate recovery as soon as the shock had been given.Bar-pressing during the tone ceased almost entirely, with the undoubtedaccompaniment of other reflex changes which might easily be labeled'anxious.' Extinction of this effect was accomplished by presenting tonecontinuously,withoutshock,duringa thirty-three-minuteperiodofP-R inwhich the rats gradually returned to their normal rate of responding.Spontaneous recovery of the effect, on the following day, was, however,practically complete, suggesting that the tone had by no means lost itspowertosetuptheanxiety.

Wemayspeakofthisanxietyasastateinducedbyastimuluswhich,inthepast,hasprecedednegativereinforcement.Itarisesfromexperienceand,ifitistobeclassedasadrive,itshouldberegardedassecondary.ItbeginswithanSD that has become a secondary negative reinforcer and endswiththe primary aversive stimulus. When the warning is over and thepunishment borne, normal behavior is restored—reminiscent of the waymendescribetheirreliefwhenaperiodofanxiouswaitingisended:"Well,that'sover!"

Anxietyis,unfortunately,morecommoninmenthaninrats.Daily,andinmanyways,we are subjected topunishment and threats of punishment.Law-abidanceandeducationareboth fields inwhich it isoftendifficult toachieve results without recourse to such motivation. Children are all too

often made literate by the same aversive motivation that is used to keepmen out of jail—this is the philosophy uponwhich a whole society maystand. Is it odd that such a society is plagued by a high incidence ofanxiety-ridden members who may, at one time or another in their lives,need therapeutic assistance? If threats fromone's personal environment andsocietyingeneralbecomeexcessive,anindividual'severydayandeveryactmay be dominated by anxiety.And if, as a child, he does not knowwhatresponsesmay be punished, but only that it will be strong, frequent, andinevitable, all the ordinary stimuli about him become 'dangerous' andanxiety-producing.Thefilesofpsychologicalclinicsarefilledwithcasesofmorbid and obsessive anxieties which are clearly the outcome ofdisciplinary and social training overburdened with threats of punishment.Ascomparedwithothers,includingso-called inferiororprimitivecultures,oursocietyissorelybesetinthisrespect.

FIG. 70. The effect of an anxiety-inducing stimulus on rate of bar-

pressing under P-R.T denotes the onset of a 5-minute tone; S, theoccurrence of shock.The four curves (A, B, C, D) are for four successivedays.Theresponseratedeclinesduringsuccessiveperiodsofanxietyuntil,bythefourthday,itisalmostzeroduringthepresentationofthetone.Eachcumulative response curve is an average for the group of six rats. (AfterEstesandSkinner,1941.)

AvoidanceInourselves, inothers, and inmanyanimals,we seem to see responses

thataremadetopreventtheimpactofnoxiousstimuli.Itwouldappearthatsuch avoidance responses depend on some SD or other that signals theimminent arrival of a negative reinforcer—or, more concisely, that theseresponses are discriminative operants. But what is the reinforcement forthem?Anyanswer that proposes the response's "actual success in avoidingthe stimulus" is inadequate because it glosses over the difficulty itself, towit, how the absence or non-occurrence of a stimulus can be reinforcing.Wereitnotforverystrongevidencethatavoidancecanbemanifestedbyanorganism,onewouldbeinclinedtodenyitsverypossibility.

Apartialanswertothisquestionhas been suggested inChapter 8,wherewesawthattheremovalofaconditioned negative reinforcementcould act as a positive reinforcer.Therearemanysituationsinwhichwe commonly speak of avoidancewhen an animal escapes from suchan SD—for example, in theexperiments ofMowrer (1939) andMiller (1941). In these cases, thebest avoidance results areapparentlyobtainedwhentheSDiscontinuously present until broughtto an end by some response. Butthere are other cases in which theSD is of short duration and theprimary negative reinforcementdoes not follow until some timehas elapsed.Warner (1932) taughtrats to escape shock by jumpingover a barrier in an experimentalchamber. When the response waswell established, he introduced a

FIG. 71. Four-chambered octagonalbox used in studying escape andavoidance behavior by rats. Theanimal, when shocked by the floorgrid in any compartment or warnedby some signal, can move into anadjoining chamber. (After Hunter,1935.)

one-second warning buzzer thatpreceded the shock by intervals of1, 10, 20, and 30 seconds. Usingdifferent groups of rats with eachdelay interval, he gave fifty trialsperdayundertheseconditionsforaperiodas longas twentydays.Theratsinthe1-secondgrouprequired,

on the average, three days to meet a criterion of six successive avoidingresponses—thatis,theyjumpedthebarrierwithinthe1-secondintervalanddid not receive the shock. In the 10-second group, the median animalreached this criterion in the fourth day and one animal failed to solve theproblem.The 20-second group required seven and a half days, with fouranimalsunsuccessful;andallof theanimals in the30-secondgroupfailed.In addition to demonstrating avoidance in the absence of a continuousexternal SD which could be terminated by the response, Warner'sexperimentshowsus thatwhateverdiscriminativeornoxiousvalue theSDacquired was dependent on the temporal remoteness of the negativelyreinforcingstimuluswithwhichitwascorrelated.

Hunter (1935), employing a similar experimental procedure with whiterats, sounded a buzzer for one-fifth of a second; two seconds later a shockwas presented unless the animal jumped into an adjacent compartmentduring the period between the two stimuli. Trials were spaced by 58secondsandamaximumof150trialswasallowedperanimalformeetingacriterion of ten successive avoidances. Of 91 rats, only six failed to reachthecriterion.Theaveragenumberoftrialswasalittleoversixty.

Hefferline(1950)showedthatrats,withlongtraining,couldbetaughttoavoidanothertypeofnoxiousstimulus,stronglight,whenitwasprecededfor fifteen seconds by a series of clicks at half-second intervals. This ishardly comparable to the studies just described, since the SD wasessentially continuous, but Hefferline was also able to show avoidanceconclusivelywhenno change at all occurred in the externalsituation.Thiswasinhis"holding"experiment,whereinratswerenotedtokeepabarina"down"positionforstretchesoffortyminutesormore,when'lettinggo'ofthebarwasalways followedby theonsetof anegatively reinforcing light.Heffer-line ascribed this holding behavior to the fact thatproprioceptivestimuli, arousedwhen the rat began to let up on the bar, had come to benoxious and "anxiety-inducing" in themselves.Maintenance of the "down-response"wasthustreatedessentiallyasanescapefromthestimuliarousedwheneverthe"up-response"began.

Hefferline's explanation hasmuch in commonwith the view, expressedbyMowrerandLamoreaux(1947), thatavoidancebehavior isreinforcedbyanxiety-reduction;but itgivesusaclueas to thepossible stimulus factors

involved. Further research will undoubtedly clarify this situation, but atpresent we seem to be justified in saying that the most common acts ofavoidance are positively reinforced by the termination of anexternal SDwhich is also a secondary negative reinforcer, as in the experiments byMiller, Mowrer, and others (pages 250-253). Other avoidances, like thatdescribedbyWarner, inwhich there isa lapseof timebetween theSD andtheprimarynegativereinforcer,areconceivablyreinforcedbytheremovalofproprioceptive stimuli or other parts of the external stimulus compound(Schoenfeld, 1950). For convenience of expression, wemay also speak ofavoidance as anxiety-reducing, just as we sometimes speak of a food-reinforcedresponseashunger-reducing.Thisaccordswellwiththeteachingsof some psychotherapists who insist that a wide diversity of behavioralsymptoms are to be understood as 'escape mechanisms'—ways ofalleviating anxiety or avoiding 'unpleasant thoughts or ideas.' You willcorrectly infer, fromthis lineof reasoning, that theextinctionofavoidancerespondingdependson theextinctionofanxiety-inducingstimuli.So longas the SD is capable of exciting anxiety, just so long will an avoidanceresponsewhichcutsoffthesestimulibepositivelyreinforced.

Theextinctionofavoidancerespondingmaybecarriedoutintwoways.First,wemayhavetheSDalwaysfollowedbythenoxiousstimulus,withthe responseno longer effective in removing theSD.This is, of course, areturntotheoriginalanxietysituation,andwillresultinaweakeningoftheavoidance response as well as a resumption of the customary anxietysymptoms—depressed activity, cowering, and so on. Secondly, we couldhave the SD presented on schedule, but never followed by the noxiousstimulus.Thiscaseduplicatesthesituationthatprevailswhentheorganismis successfully avoiding; and youmight be led towonder how the animalwould ever 'discover' that the noxious stimulus would not come. Youwouldhave torecall that thereinforcementforavoidancebehaviordoesnotlieinthenon-appearanceofanything;rather,itliesinstimulusremoval.

Actually, extinction occurs in this second case, because the SDmaintains its negatively reinforcing status only when it is at leastoccasionallyfollowedbytheprimaryreinforcer.Withouttheshockorothernoxious stimulus, its power dissipates (along with its anxiety-producingfunction). This is related to the experimental observation (Schlosberg,1936, 1937; Sheffield, 1948) that the latency of the avoidance response totheSDgoesthroughacyclicallengtheningandshortening.Inthecourseofastringof successfulavoidances, the latency increases from trial to trialastheSD'sanxiety-producingpowerextinguishes.When,finally,theresponsecomestoolate, thenoxiousstimulusstrikes theanimal,andacts torestore

the SD's power.The next latency 'snaps back' to a low value, only to gothrough another lengthening phase with each successful avoidance.However, in extinction, it has been observed (Culler, Finch, Girden, andBrogden, 1935) that if an avoidance response passes by the time whenshock is customarilygiven, itwill continue toweaken.Thisweakening issometimesstrikinglyrapidanditmaybethatitdependsupontheviolationofatemporaldiscriminationbuiltupbythetrainingschedule.

Wehaveagreatdeal to learnaboutavoidancebehaviorandanxiety.Ourpresent knowledge, gleaned mainly from animal study, indicates that theextinction of an avoidance response is often extremely difficult, even inwell-controlled experimen tal situations. It suggests that, at the level ofhumanbehavior,whereallmannerofstimulimay,bydesignorcaprice,accompanypunishment,theproblemofeliminatinganxietyisoneofthefirstmagnitude.This is especially trueofourownsociety,which sowantonly,deliberately, or systematically relies on punishment in the control ofbehavior. Ideas on the subject abound, among laymen as well asprofessionalworkersinpsychotherapy,butthescientificvalidationofmostofthemstillliesahead.Isitwisetosendtheaviatoraloftimmediatelyafterthe crash? Should we force the person with a phobia to face the fearedobject? Can a child's dread of darkness be removed by keeping him in adark room?These and related questions are of deep concern to thementalhygienist, who has evolved his own set of rules for the prevention orremoval of behavioral 'maladjustments'—rules that have a pragmatic basisin the benefits sometimes derived from their use, but which as yet haveonlybeenpartiallysubjectedtoanalysis.

TheConflictofDrivesThe laboratory, which permits the analysis of natural processes into

components rarely or never seen in the massive events of daily life, alsopermits us to re-create themore complex events step-by-step for deliberatestudy.Althoughwehavesofardealtmainlywithmotivesoperatingsingly,itmustbeevident toall that in theusualaffairsofanimalsandmenmorethan one drive may be present at any time. Before this already lengthychapteriscloseditmaybeworthwhiletolookbrieflyatsomeoftheknowneffects of co-existingmotives upon behavior. In this connection, attentionhasoftenbeenfocussedupontheconflictofdrives.

Aword of explanation is needed about "conflict of drives" lestwe losecontactwithourearlierdiscussions.Amoment'sconsiderationwilldisclosethat it is not drives that conflict, but rather responses or responsetendencies.AsShaffer (1936)hasput it: "The term conflictmeans that theindividual confronts a complex situation to different aspects of which hehaslearnedtomakeantagonisticresponses...[which]cannotbemadeatthesame time...." We speak of drive conflict when opposition between

incompatible responses arises from drive operations like deprivation orstimulation.This is in contrastwith conflicts originating in otherways—for example, from the direct stimulation of antagonistic musclessimultaneously, or the simultaneous use of SD's for incompatibleresponses. Direct stimulation has been investigated mainly byphysiologists; the "conflict of cues" (or SD's) has been of interest in thestudy of special discrimination problems (e.g., Nissen and Jenkins, 1943)likesensory"illusions."Always,however,'conflict'referstotheinstigationof incompatible movements, movements which cannot occur together andmay be thought of as competing. In the discussion which follows,therefore,ouruseof 'driveconflict'willmean"responseconflictcreatedbydrive operations"; and when we speak here of 'response conflict' we shallintendthistomeancompetitionthatisdrive-induced.

The co-existence of several drives canmean harmonious performance ifthe responses involved are simultaneously possible; or it canmean, if theresponsesare incompatible, thekindofbehavioralupset thatallofushaveexperiencedatonetimeoranother.Thesignsofconflict,whicharealsotheresponse characteristics measured in experiments, include increasedhesitation or latency of response, vacillation or alternation betweenresponses, frequency of 'blocking' or failing to respond at all. Thesemeasures have been used by several researchers who, within recent years,havebegunapromisinglineofexperimentalattackupontheproblem.

Conflictsituationsdiffer incomplexitydependinguponthenumberandvariety of competing responses.Two experimental arrangements, however,because of their simplicity, provide a good starting point. In each case letusassumethattheresponseisfullyconditioned,andthatothervariablesarekept constant, so that only the experimenter's drive operation ismanipulating the strengths of the responses involved. (In this way, weavoid thecomplications thatresponseswithunequal trainingaredifferentlyaffected by the same drive operation (Sears and Hovland, 1941), and soforth. While these are doubtless important variables, we may forsimplicity'ssakeignorethemhere.)

1.Ananimal isplaced intoanalleyapparatus inwhichhe is trained tomove toward or away from either end by the positive or negativereinforcements placed in the ends. Locomotion is a kind of response, andthissituationisoftendescribedintermsof"approachandavoidance."Thisarrangement is valuable because it affords a good picture of conflictingresponses and allows relatively easy measurement-The student shouldremember, however, that the spatial results of the animal's responding arenot the prime consideration, and that analysis could as well be made ofresponses thatdonotrequire travelby theorganism(Miller,1948).Brown(1948), usingwhite rats and such an alley, attached a small harness to hisanimals and was able to measure the pull exerted by the animals as they

tried tomove towardapositive reinforcer (food) andaway fromanegative(electric shock).The graphs in Figure 72 show how the force of the pullvariedwithdistanceof the rat in thealley fromthepointof reinforcement.The two gradients of approach and avoidance, each obtained from animalswho had had only one type of reinforcement, led to the followingobservations(Miller,1944):(a) the tendency toapproachagoal isstrongerthe nearer the subject is to it ("approach gradient");(b) the tendency to goawayfromaplaceorobjectavoidedisstrongerthenearerthesubjectistoit("avoidancegradient"); (c) the strength of avoidance increasesmore rapidlywithnearnessthandoesthestrengthofapproach.Withthisknown,wemaygo on to establish three conflictful situations by simultaneously placingreinforcementinthealleyendsaccordingtothreecombinations.a.Bothalternativespositivelyreinforced.Thus,we could use food and

waterwithanimalsbothhungryandthirsty;orevenfoodatbothends,withonlyhungerdrivepresent.This isan"approach-approach"conflict like thatofthemythicalassbetweentwobalesofhay.Theass,itistold,starvedtodeath, but you will readily agree that this is not a likely outcome!Thestory's humor arises from the fact that its ingredients are conceptuallypossible,but theirattainmentactually impossible: theass isnotapoint inspaceequidistant from thebales; thebalesarenotequalSD's, nor do theyoccupymerepointsinspace.Inactuality, theslightestdeparturefromidealconditionsmeansanincreasingpullinonedirectionwithacorrespondinglydecreasingpullintheotherasthedistancefromitincreases(seeFigure 72).Conflicts like these are easily resolved for the organism.Whenmen speakof being torn between equally attractive courses of action, there is usuallythe unspoken element that one coursewill berewarding and theother not,so that there is an undercurrent of anxiety lest thewrong choice bemade.But in a world where all choices are rewarded, the approach-approachsituationshouldproducenonebutfleetingconflicts.

FIG.72.The intersectingapproachandavoidancegradientsobtainedbyBrown (After Miller, in J. McV. Hunt, 1944.) In connection with thesegraphs,seealsoFigure59,here.

b. Both alternatives negatively reinforced. Thus, either end of thepathwaymight be charged to give a painful shock.The organism tends toretreatfromeitherend,butcannotgetoutofthedilemma.Movementawayfrom the center, in either direction, increases the repelling power of theapproachingaversivespot,sothatheiscaughtandsuspendedbetweenthemin an agitated and conflictful immobility. If he were not confined to thealley, the animal would flee from the situation altogether, probably in acompromisedirectionatrightanglestothelineconnectingthetwoaversivespots.A phenomenon similar to this was demonstrated by Hovland andSears (1938) with human subjects. The instructions were to draw on asquare sheetofpaperadiagonal line from themidpointof thenear side tooneof thefarcorners.Eachfarcornerhada lightwhichcouldbeflashedatthe will of the experimenter, and the subject was to respond to the lightwhichdidflashbydrawingthelineasquicklyashecould.Aftertrainingtoboth corners, a conflict was set up by flashing both lights at once.Theresult inmany caseswas a blocking that caused lines to be drawn straightaheadbetween the two far corners rather than diagonally. In contrast with

this,wecaneasily suppose that if anorganismwere ringedby an aversivecirclewith escape impossible, hewouldbogdownat the center, just as inthe alley where the walls prevent leaving the situation. The effect ofhemminganorganisminonallsidesbynegativereinforcementsisofgreatinterest to students of abnormal behavior.The "escape from reality" and"retreat into phantasy" which they observe in many patients seem to beconsequences of life situations in which punishment awaits every move.Wherethereisnoescape,somethingisboundtogiveway.c.Onealternativepositively,onenegatively, reinforced. Thus, a choice

ofonedirection is favoredbyboth itsownpositivevalueand thenegativevalue of the other. This is clearly a combination in which no conflictresides. One has only to answer such questions as how the two responsetendenciessummate,whetherarithmeticallyorinsomeotherfashion.

2.Thealleyapparatuscanbeusedforaconflictarrangementthatisevensimpler than theabove.This involves locating the reinforcements,whetherpositive or negative, at only one end so that we can consider locomotiontoward thatendasasingleresponse thatmayberewardedorpunished.Noproblem arises when: (a) only positive reinforcements have been given atthat end; or(b) when only negative reinforcements have been given. Ineither case, there is no conflict about approaching, or retreating from, theplaceof reinforcement.The instance thatmostconcernsus is that inwhichtheanimalshaveonsomeoccasionsbeenrewardedforapproaching,andonother occasions punished for the same response.The result is that he hastendenciesboth toapproachandavoid,andweshouldbeable todeduce inpartfromFigure72whathisbehaviorwillbe like.Since theapproachandavoidance gradients differ in steepness and the point of intersection, wemight expect that therewill be a place in the alley at some distance fromthe end, where the animal will tend to come to rest.Thus, if we put theanimal down in the alley beyond that point, he would approach until theavoidance forceexceeded thatofapproach.Hisapproachmomentumwouldprobablycarryhimpastthatpoint,whereuponhewouldslowupandbegintoretreat;theretreatmomentumwouldagaintakehimtoofaraway,andhewould thenbegin toapproachoncemore;andsoon. Inshort, thereshouldappear not only a point of equilibriumbut also a vacillationofmovementdirection around that point, with the animal unable to advance or retreatdecisively.Justthistypeofthinghasbeenobtainedexperimentally(Miller,Brown, and Lipof-sky, 1943; Rasmussen, 1940). As Miller has written(1944), thesituation is reminiscent of human dilemmas like that of thehard-hit but bashful lover who vacillates helplessly at a distance from hisloved one; or like that of a man eager for a higher salary but afraid toapproacha'tough'employer.

Of the situations outlined above, there are two that seem to beoutstanding generators of conflict and probably contain the germ of theadmittedly more complex conflicts of daily life. They are, if we may

recapitulate, the double avoidance case where either of two alternateresponses is punished, and the case where a single response has had bothpositive and negative reinforcement applied to it. Before taking up someexamples of conflict that are on the next level of complexity, it may benoted that the amount of conflict engendered between two responsetendencies is inversely related to the difference in strength of the twotendencies; or, the amount of conflict increases as the strengths of theopposing responses approach equality.This has been tested and confirmedin several studies (Sears and Hovland, 1941; Godbeer, 1940; see also areviewbyBitterman, 1944).Barker (1942), for example, gave ten-year-oldboys the task of indicating which of two proffered liquids they wouldchoose to drink. In one part of the experiment, both alternatives wereunpleasant (as determined in advance of the experiment), but differentiallyso.Barkerfoundthatthemorethealternativesdifferedinattractiveness,thefasterthechoicewasmadeandthefewerthevacillationsinmakingthefinaldecision. This example is one of the double avoidance type, but theavailable evidence shows that thehypothesis is probablyvalid for a singleresponsewithacombinedhistoryofrewardandpunishment(Miller,1944).

Our final case of conflict involves two incompatible reactions, either ofwhich is reinforced but at the same time results in the loss of the otherreinforcement or is punished. In other words, one positive reinforcementmust be foregone (or punishment endured) when the subject chooses theother.Thesituationcomprisestwointeractinginstancesofasingleresponsethat is both rewarded and punished. It has been recognized (e.g., Hovlandand Sears, 1938) that the case resembles more closely the conflictfulcontingencies of real life, and it has been studied by several investigators.Inaseriesofexperimentswithconflicting light-aversionandhungerdrivesin the white rat, Tolcott (1949) employed an apparatus in which twoincompatible responses were available for separate reinforcement withdarknessorfood.Locatedatoneendofaboxwasabarthat,whenpressed,deliveredfoodpellets,whileattheotherendwasaplatformorpedalwhichcontrolleda flashingbright lightplaceddirectlyover thebox.Theanimalswere first trained, in separatepreliminary sessions, to procure foodbybar-pressing and to keep the light off by stepping or sitting on the platform,whengettingoff theplatformmeant the immediate returnof the light.Thebarandplatformwereseparatedbyadistancegreater than the lengthof theanimal'sbody,so that in theconflictsessionswhen theywerehungry theycouldnotmaintainplatformcontactwhilestretchingovertopressthebaratthesametime(asmanytriedtodo).

Inoneexperiment,Tolcott compared thebehaviorofhis subjects in theconflict situation with that under hunger alone and under light aversionalone.Twoofhisinterestingfindingsmaybequoted.

1.When the twodrivesarearousedsimultaneously(conflict) the typicalbehavior pattern is an alternation between the responses—short periods of

pedal-holdingalternatewithbriefflurriesofbar-pressing.2. During these conflict sessions, more than half the experimental

interval is spent on the pedal; nevertheless the same number of bar-responsesaremadeasoccurredduringthesessionsofappetitealone.Thisisaccomplished by an increase in the net rate, or rate at which the bar ispressedduringthespurtsofbar-pressing.The latter is an interesting point and is in line with the finding thatpunishmentisnotnecessarilysufficienttoreducethestrengthofaresponse.Indeed,itlooksasifitisthisveryinefficacyofpunishmentthatallowstheconflicttocontinueunabated.

In another experiment, Tolcott lessened the degree of conflictsystematically by lowering the hunger drive, and comparedbehavior underthevarious strengths of hunger as it competedwith the light aversion thatwas held constant.The use of light aversionwas somewhat different here,sinceit involvedtheperiodicandsignalledpresentationoflightratherthancontinuous light. He was able to demonstrate that with increased pre-feeding, thewarningsignal foronsetof lightcame tohaveagreatereffect.Lowered hungermeant less conflict,more frequent returns to the platformin the warning period, less vacillation between the two incompatibleresponses,andsoon.Thedata,therefore,oncemoresupportthehypothesisthat theamountof conflict is inversely related to thedifference in strengthofcompetingincompatibleresponses.

A later study byWinnick (1950) produced data that are significant forourunderstandingofconflictfulbehavior,andthatarerelevanttoourearlierdiscussion of experiments like Hefferline's on avoidance behavior. Shereplaced the platform ofTolcott's apparatus with a small vertical hingedpanelagainstwhichtheratpushedtokeepoff theaversivelight.Thepanelcould be pushed some distance beyond the point at which a switch brokethelightcircuit,andapenwasattachedtothepanelthatgaveacontinuousrecordofitsmovementsonakymographtape.Asbefore,abarat theotherendofthebox,wherebyfoodcouldbeobtained,createdaconflictsituationwhentheanimalswereplacedin theboxhungry.AsTolcotthadobserved,animals vacillated between panel-pushing and bar-pressing in the conflictsessions.But therecordofpanel-pushingrevealed thatconflictwaspresentevenwhen no oscillationwas apparent to the eye.The panel-pushingwasnot steady in force or extent, but was marked by large variations eventhoughtheanimalmightnotfora longtimereleaseitsufficientlytoallowthe light to come on. Incipient movements toward the bar that stoppedshort of the light-switching-on point, alternated with retreats fromtheswitching-onpointandpushingwithrenewedvigor.Figure73 is a samplebitofWinnick'srecordsofthisconflict.

Thisingeniousexperimentdeservestobefollowedupbecauseofwhatitreveals about conflict and the way it bridges the topics of proprioceptiveSD's, covert response, anxiety and avoidance, and conflict behavior.The

beginnings of one response, as we have said elsewhere, can generateproprioceptive SD's which serve as anxiety arousers because they arecorrelatedwith ensuing punishment if the response goes to completion. Inthis way, anxiety SD's of muscular origin give rise to 'inhibition' or'repression'ofbehavior.Asimilaremphasisontherô1eofproprioceptionisto be found in the writings and actual therapeutic practices of somepsychoanalysts(e.g.,WilhelmReich,1949).Weshallnotdigress intothisarea, but there can be little doubt that thework of the clinician awaits thefruitfultouchofthelaboratoryscientist.StudieslikethoseofLuria(1932),Godbeer (1940), and others, show thatlaboratory studies with humansubjects in conflict situations are feasible. Some of the findings withanimals have already been corroborated with human beings. In the end,however, our knowledge of man, though it may start with humblerorganisms,mustberoundedoutbystudyingmanhimself.

FIG.73.Sample recordsofpanel-pushingbehaviorby rats inaconflictsituation. When the recording pen was at the base line, the aversivestimulus (light) was present. The animals were hungry and the jaggedrecords show that, though they often kept the light off, there wasconsiderable vacillation of pushing behavior because the food-getting barwaspresentat theoppositeendof theresponsechamber.Therecordsaretobe read from right to left.Three 4-minute samples of behavior are shown.(FromWinnick,1950.)

NOTESOurchapterhasbeenlong,butitwouldhavebeenlongerifwehadpaid

attention tonon-experimental studiesofmotivation,especially to theworkof Sigmund Freud (1856-1939) who, more than any other, is identifiedwith the modern emphasis upon the sexual drive in human beings. ForFreud, this drive was the foundation for his general theory ofpsychoanalysis, thebasic tenetsofwhicharenowavailable inmanybooksand articles written for laymen as well as students and professionalpsychotherapists.Wecannotgo into the intricaciesof this theoryhere,butwemay note that many of its concepts (e.g., regression, symbolism, andrepression) have found their parallels in experimental research, even at thesub-human level. The future may show that this development has evenmorepointsofcontactwithbehavioralsciencethanarenowapparent.

Thebest single textonmotivation towhichwecandirect you isP.T.Young'sMotivationofbehavior(1936)whichcontainsanexcellentaccountof the research in this field prior to 1936. Young has himself madesignificant experimental contributions to our knowledge, especially instudyingthesub-hungers.

The distinction between appetites and aversions is an old one, goingback at least to the time ofThomas Hobbes (1651).Young, in the bookmentioned above, treats these as two classes of drives, and so does E. C.Tolman, in a thought-provoking little book calledDrives toward war(1942).Tolman's book, which leans heavily upon observations of animalbehavior, also includes some interesting speculations concerning therelation of basicmotives to several psychoanalytic concepts and to certainsociologicalandeconomicteachings.

10

EMOTION

THE TYPICAL varieties of emotion are each connected with certaincharacteristic...trendsofactivity.Angerinvolvesatendencytodestroyandforcibly to break down opposition.... Joy involves what we may callexpansive activity.... In grief there is a general depression and disturbanceof thevital functions.... Fear... arises in a situationwhichdemands actionfor averting, evading, or escaping a loss or misfortune which has not yettakenplace.

G.F.Stout,TheGroundworkofPsychology,1903

APreliminaryQuestionIn this chapter, as in the preceding, we are concerned, not with a

principle, but a problem.The problem, in large part, is that of decidingwhattodoabout thetermemotion. Is thiswordtobekept inserviceasanaid to our understanding of behavior, or should we retire it from activescientificduty?Thequestionmaystrikeyouasafoolishone.Emotion,onehears, is something that colors human life frombirth until death. It is theessenceofpleasureandthecompanionofpain;itisthespiritofecstasyandthesoulofdespair;itisthefriendofcreativeeffort;itpromoteswell-beingand ruins digestion.Dowithout it?Wouldwenot be leftwith a colorlessandcoldexistence?

This attitude misses the point of our problem.We did not propose toshirk our obligation to deal with those aspects of behavior which, incommon speech throughout the ages, have been called 'emotional.' Weaskedonly ifaclearandusefulmeaningcouldbegiven to theword.Whatwe sought to emphasizewas thedanger that lies in adopting, for scientificpurposes, a term that has such a variegated history of usageas this one.Othertermsofancientlineage,suchasinstinct,havesufferedfromthesamedefect, and have been dropped from the vocabulary of many present-daypsychologists. Still others seem destined for a similar fate. If we are toretain thisone (aswe really intend todo),we shallhave tohedge it aboutcarefully.Wemustrestrictandqualifyitsuseinawaythatleavesnoroomfor misunderstanding.At the same time, we must not strip it so bare ofmeaningthatitnolongerpertainstothecountlessobservationsfromwhichitoriginallystemmed.

SomeObservationsIthasbeenprofitableforus,inprecedingchapters,totake,asastarting-

pointfordiscussion,thebar-pressingbehaviorofthewhiterat.Supposewereturn to the laboratory again, this time with an eye to any details of his

responding which we may have slighted in our earlier descriptions.Whatwillwefindthatisrelevanttothetopicnowathand?

A number of items immediately present themselves. We note, forexample, that an initial period of'acclimatization' to the experimentalsituation isgenerallyprovidedbeforeanyattempt ismade tocondition thebar-pressing response.Without suchaperiod,orperiods, the time requiredto set up the habit is often quite prolonged. Familiarizing procedures areused in almost every type of animal experimentation. 'Gentling' subjectsbefore training;giving themunreinforced 'pre-trials' in runwaysandmazes;acquaintingthemwithfood-boxes;accustomingthemtoclicks,buzzes,andotherapparatusnoises—all thesedevicesareconducivetorapidlearning.Asimilar effect has also resulted from the use of sound-proof experimentalrooms,constantconditionsof illumination,andso forth. In theabsenceofsuch precautions, rats are not only slowed down in their rate of learning,but may also be seen to react in unusual ways.When introduced to theexperimental situation, they may wash or groom excessively, crouch incorners, and void urine or feces. Movementsof exploration may be slowand cautious or absent entirely; and the slightest alteration of stimulusconditionsmaystartlethemorcausethemto'freeze'inposition.

A second item of interest turns up in connectionwith theextinction ofType R responses after a period of regular reinforcement. Here, as weremarkedearlier (here), three things are obvious: (1) an initial high rate ofbar-pressing;(2)recurrentperiods inwhichbar-pressingmaybecompletelyabsentormuchreducedinfrequency;and(3)correspondingperiodsofrapidresponding which seem to compensate, in decreasing measure, for theperiods of no response (seeFigure 16). If we watch the animal closely,again we note bar-gnawing, urination, defecation, grooming, and otherresponses, together with attempts to get out of the situation. Theseresponses,andthecyclicaldeviationsinbar-pressingrate,arenot so typicalofextinctionsthatfollowperiodic(oraperiodic)reconditioning(here).

Easily observed changes of a similar sort appear when somenegativelyreinforcingstimulus is applied to our animals duringperiods of reinforcedor non-reinforced bar-pressing. Besides the more or less completesuppressionof theoperant itselfwhentheanimal isslappedorshocked,heshows crouching, climbing, jumping, approaching and retreating, rapidbreathing. These and other movements are characteristic of his behaviorwhenthenegativelyreinforcingstimuliarefrequentorintense.

Weak but clearly discernible effects upon bar-pressing rate may appearunderotherexperimentalconditions.TheintroductionofadimlightwithintheresponsechamberwillregularlyproducealoweredfrequencyofresponseduringP-Rsessions;andevenamildtoneorbuzz,tobeusedasanSD oran SΔ in a discrimination experiment, may halt responses completelyduring their first few presentations. Extreme intensities of these stimuli

may render the animal incapable of anymovement whatever.They are nolonger useful as SD's, nor can they serve as aversive stimuli, thetermination ofwhich is reinforcing—the animal cannot make thecoordinatedmovementswhichwouldremovethem.

These behavioral changes are not peculiar to any one experimentalsituation. Many studies have been made of the effects of lights, tones,shocks,withdrawalofreinforcement,andsoforth,uponotheranimals thanthe rat, and in relation to other activities than bar-pressing. Our ownobservations merely confirm those which have been described by variousinvestigators.

Itrequiresnogreatimaginationforonetodrawupcomparableexamplesof his own or others' behavior. Who has not been upset in strangesurroundings or in the face of unexpected change? Who has not beenfrustrated by his inability tomeet a familiar situationwith an accustomedmode of action?Andwho has not experienced, at least for amoment, thebehavioral disruption which can follow an injury, an explosion, a loss ofequilibrium,orsomeotherformofsudden,intensestimulation?Wewouldhardlysaythatourrat'sbehaviorisunique.

Such phenomena as those we have described above lead us to a fewgeneral statements. Certain experimental operations, such as the takingaway of positive reinforcement or the giving of negative, result in certainoperantandrespondentchanges.Theseeffectssometimesfadeoutgraduallywith continued exposure to the situation or with removal of the noxiousstimuli.When one observes all or some of them, he feels fairly safe insaying that emotion exists.This is not the entire story, but it gives us apointofdepartureforwhatfollows.

EmotionandDriveIfwewere todefineemotion tentatively,andbroadly,asastate inferred

from the change in strength of a number of reflexes as a result of certainoperations,weshouldnotonlyepitomize theforegoingdiscussion,butweshouldalsobemakinga statement thatappliesequallywell to theconceptofdrive aswe treated it in thepreceding chapter.Moreover,wewouldnotbethefirsttorecognizethissimilarity.Manystudents,ancientandmodern,havenotedasmuch.Descartes, in theseventeenthcentury,arguedthat"theprincipal effect of the passions [emotions] in man is that they incite anddispose [motivate] the mind to will the things to which they prepare thebody." William James (1890) asserted that "instinctive reactions [readmotives]...shadeimperceptiblyintoeachother.Everyobjectthatexcitesaninstinct excites an emotion as well...."AndWilliam McDougall (1923)postulatedformostofthebasic 'instincts'(again,motives)acloselyrelatedemotionalcounterpart—flight,an 'instinctive' impulse,wasassociatedwithfear, an "affective state"; pugnacity was associated with the emotion of

anger;andsoon.Inourowndiscussionofaversions,itmusthaveoccurredtoyouthatthe

effects of such stimuli as electric shocks were certainly 'emotional' in theeverydaysenseof theword;and to thisweare forced toagree. Inaddition,we would have to admit that the scientific, no less than the popular,distinctionbetweenemotionandaversionisperhapsinsupportable.Thelineof demarcation is at best very faintly drawn.The operation by which wedefineaversion is thatofpresentingstimuli—shocks, slaps, lights, tones,andsoon;but this isalsooneof theoperationsweuse todefineemotion.Onedifferencebetweenaversionandemotionseemstolieintheintensityofstimulithatwepresent.

Thisdifferenceleadsustolookatthematterinanotherway.On herewestatedthatstimulihavedifferentfunctions.Beyondtheirreinforcing status,they may beelicitative (the respondent case),discriminative (setting theoccasion for reinforcement of an operant),drive-inducing (aversive), andemotionalizing.We now suggest that the intensity continuum cannot beignored.Mildormoderately intensestimulireadilybecomeSD's; stronger,negatively reinforcing stimuli, the removal of which is positivelyreinforcing, may be calledaversive; and extreme intensities are those wecommonly treatasemotional.TakeaweaklightasanS D(ashockor tonewould serve as well).A rat can be easily trained to press a bar wheneversuch a stimulus is presented. If, however,we increase the intensity of thisstimulus, we have trouble in separating the purely discriminative oreliciting status from the aversive—that is, we find that the animal can betaught to turnoff this light.Now, if the intensity is still further increased,the motivating function of the stimulus is obscured—the animal mayrespond for a while in an adaptive fashion, but gradually his behaviorbecomesmore andmoredisorganized.He runs thisway and that; he leapsand climbs and falls about in his cage; and he may even end by lyingprostrate on the floor, unable tomake anymovement thatwould free himfromhisplight.

Webegin to see, then,why it is that emotionsanddriveshave so longand so persistently been confused with each other. Ultimately, we maydecide that the fence between the two concepts does not justify the laborspent in its upkeep. For the present, however, we shall assume that thesuggesteddifference inoperationsandbehavioraregreatenough towarrantadistinction.

EmotionasanInferredStateInourtentativedefinition,wespokeofemotionasan'inferredstate.'We

should now add, as in our discussion of drive, that such inferring isconvenient,butnotstrictlynecessary,andsometimesdangerous.Ifweslipinto the practice of assigning to this hypothetical state any properties not

given in our observations, we jeopardize the clarity of our thinking. It isespecially necessary to avoid treating the state as something 'mental' thatcausesourbehavior.Insodoing,wemaybemerelytryingtoliftourselvesby our own bootstraps—we infer somethingfrom behaviorwhichwe thenusetoexplaintheverybehaviorfromwhichtheinferencecame.

Whenwe thinkof thestateasphysiological orneurological, rather thanmental, we would seem to be on better ground.There is indisputableevidence of widespread organic changes which accompany responsesregularly called emotional.Yet, even here we must be wary. Our lack ofdetailedknowledge,bothphysiological andbehavioral, doesnotpermitusto go very far in relating the two areas.We are onmuch better footing atpresent ifweconsideranemotionalstateasabehavioralconstruct,keepinguppermostinmindtheoperationsthatprovideforchangesinresponse.

ThreeTheoriesToillustratethecomplexityofourproblem,withreferencetothepoints

justmade, letusconsiderbriefly threeways inwhich theemotionof 'fear'has been treated by psychologists and physiologists.A very old, and stillpopular, view runs something like this. Fear is an episodicmental state—an unpleasant, all-pervasive, and 'stirred-up' consciousness, known at firsthandonlytoitspossessor.Itisarousedbycertainobjects,events,orideas;and itis expressed by various 'involuntary' or 'voluntary' changes inbehavior. It arisesnaturally, aswhenone quakes at a loud sound heard inthe night, orthrough training, as when one trembles at the prospect ofmaking a speech before a large audience. In its extreme form, terror, itsexpression is very dramatic.The eyes and themouth openwide; the skinpales and cools and sweats; breathlessness occurs, and the heart beatsagainst the ribs; saliva stops flowing and the mouth becomes dry andsticky; thevoicebecomeshuskyandindistinct,orfailsentirely; thepupilsof the eyes enlarge and the hairs stand erect on the head and body surface.Thenormaldigestivefunctionsarehaltedandbowelorbladdercontrolmaybelost.Thehandsmayberaisedhigh,flightmayensue,oronemaystand,trembling,rootedtothespot.

In this classical picture, the sequence of events is (1) stimulation, (2)subjectively 'felt' emotion, and (3) bodily 'expres-sion'—behavior. Thesame sequence,with differences in content,would be characteristic of joy,sorrow, anger, andother time-honored categories. (We ignore, for thepresent, the problem of how an 'idea' may be a stimulus, how a 'mental'state may be known except through behavior, or why 'introspective'descriptionsoftheemotionsseemalwaystopointtoSD'sarisingfromthebodilyresponseitself.)

William James, in 1884 and later, proposed a drastic revision of theearlier theory. Supporting his view by an appeal to a similar position

arrived at by the Danish physiologist, Lange, he argued that the feltemotionfollows, rather thanprecedes, the bodily expression.For him, thestimulating object or event led first to theexpressive movements whichthemselvesproduced the 'awareness'of theemotion."Commonsensesays,we loseour fortune,aresorryandweep;wemeetabear,are frightenedandrun; we are insulted by a rival, are angry and strike....The more rationalstatement is that we feel sorry because we cry, angry because we strike,afraid becausewe tremble." Statedmore baldly than Jameswould perhapshaveliked,hispositionwasthis:astimulus(e.g.,abear)sometimeselicitswidespread changes in a person's behavior, and these changes in turnbecomediscriminativestimuliwhichpromptus,or the individualhimself,toassertthatfearexists.

The James-Langedoctrineno longerupsetsus.Theobjectivelyorientedperson,trainedtodistinguishbetweenwhatisobservedandwhatismerelyinferred, even finds it easier to understand than its predecessor. But avigorousattempthasbeenmadeto turnour thoughts ina thirddirection—toward brain physiology. In 1927,Walter B. Cannon and two associatesdemonstrated that the surgical removal of the sympathetic division of theautonomicnervoussystemdidnoteliminatetheovertresponses'expressive'of rage, fear, or joy in cats. The same investigator adduced evidence toshow that there are no response changes (in heart-rate, digestive activity,sweating, hair-erection, and the like) which are distinctive for the variousemotions.Allof themarecharacteristicof strongexcitement, andmanyofthem are present in fever, strongexercise, and other presumably non-emotional states. From such considerations, Cannon was led to reject theJames-Lange view that emotion, in general, is the 'awareness' of response,and that different response patterns give rise to the specific emotions ofrage, fear, and so on. Instead, he offered a theory which relates theemotional 'consciousness' to the activity of a special brain center (thethalamus)whichmaybeexcitedbystimuliindependentlyofanyexpressivereactionsofthemusclesorglands.

We need not enter into the intricacies of these theories or the debateswhich have revolved about them. Rather, let us remind ourselves that allthree suffer from the defect that we described in the preceding section. Inascribing either a 'mental' or a physiological character to emotion, they donot clarify our thought.Thebasic facts are these: something is done to anorganism—a stimulus is presented orwithdrawn, a reinforcement is givenortakenaway;and,asaresult,widespreadchangesinreflexstrengthoccur.Someof these changes, like thoseof digestive activity, aredetectible onlywith the tools of the physiologist; others, like the depressions oraccelerations of on-going operants, are visible to anyone's naked eye. Inorder to circumvent the round-aboutness of this account, we say that theorganism isemotional. To assert that the emotion is mental, does notadvanceourscientificunderstanding,since there isnoknownprocedurefor

investigating themental as distinct from the behavioral.To say that it isphysiological, has more justification, since we have some independentevidenceofphysiologicalchangeswhichaccompanythebehavioral,butwestill have a long road to travel before all the physiological 'correlates' ofbehavior are determined. Certainly, theprediction andcontrol of behaviorisnotfaradvancedbytheassumptionthateithera'felt'emotion(mental)orits'expression'(behavioral)iscausedbybrainprocesses(physiological).

ResponsePatternsandEmotionFor many generations, men have treated certain responses or groups of

responses as indicative of emotion; and, down through the years, attemptshaveregularlybeenmadetodistinguishoneemotionfromanotheronsuchabasis.Theclassical theorywasoriented in thisdirection, and sowas theJames-Langeviewthatfollowedit.Eachmajoremotionwas tobedefined,at least in part, in terms of a pattern of bodily changes, alike in allindividuals, with components that were presumably apparent to any naiveobserver. It behooves us, therefore, to ask whether this practice is sound,whetheremotionscanbedescribedasspecificresponsepatterns.

1 .Facial expression.The answer to this question is not as simple asyoumightthink.Itseemsobviousthatthefaceofanenragedmanisunlikethat of a frightenedone; and, seemingly, noone couldmistake the face ofjoy for that of sorrow.Yet, when we look at the matter closely, we findourselves in trouble. Ismy expression of fear or joy like yours?Does theanger of theAsiatic resemble that of the European?Are these expressions'natural' or are they 'acquired'? How can we tell whether they are 'real' or'feigned? Are the responses "universal and immutable," as argued byDuchenne (1862), orwasWilliam James right in saying that "everyone ofus,almost,hassomepersonalidiosyncrasyofexpression?"

The history of facial-expression studies gives us a rather confusedpicture, but one or two conclusions may today be drawn with someassurance. First, it would appear that a certain limited number of facialresponse patterns are recognizable in human children shortly after birth.Aclearly distinguishablesmile, which precedes and is closely related to thelaughing response, may be evoked after two to fifteen weeks of infancy(Jones, 1926). It has beenpointedout byYoung (1943) that both smilingand laughing tend tooccurwith(a) the satisfaction of hunger or thirst,(b)relieffromdiscomfort,(c)presentationofcertaincolors,sounds,tastes,andcontacts, and(d) the free movements of play or dancing. In a word, theycome as a consequence of presenting positive reinforcers or removingnegative reinforcers.Frowning, which is related tocrying as the smile isrelatedtothelaugh,alsooccursininfancyasarecognizablepattern.Youngsays that this response accompanies(a) hunger or thirst,(b) painfulstimulation,(c) dazzling lights, loud sounds, or other intense stimuli, and(d) frustration or blockage of response. That is to say, negative

reinforcementorthewithdrawalofpositivereinforcement.Secondly,wenote that such expressions can apparently be strengthened

andmodifiedoperant-wise.Thechild,wesay,comesto 'use'anexpression'forhisownpurposes.'Hedevelopsan 'ingratiating'ora 'disarming'smile;he 'getshisownway'bycrying; and soon.His facial expression, likehisvocal,isshapedinonewayoranotherbytheselectivereinforcementwhichhisfamily,hisfriends,andothersprovide.The 'natural'smileofbabyhoodmaybedisplacedalmost entirelybya smile that is 'artificial.' 'studied,' or'forced.' So powerful are these strengthening influences that the naiveexpression isoftenobscured. In someadults, a fairlycomplete immobilityof features—for example, the 'dead pan' of the professional card-player—maybeachievedwhenitisofadvantagetodoso.Or,attheotherextreme,as in the expressions of the skilled actor, a variety of subtle variations ofresponse may be developed through their effectiveness in molding thereactions of others. As with many common forms of behavior, facialresponses come under the control of the organism's social environment. Itshould, therefore, occasion little surprise to find that the facial habits ofadult Chinese are as different from our own as forks are different fromchopsticks(Klineberg,1940).

Thirdly,wemaysay that,as faras theadulthumanbeing isconcerned,facialexpressionsarepoorindicesoftheprevailingemotionalstate.Awell-known study by Landis (1924) points this up rather dramatically.Maturehuman subjectswere presented with a variety of presumably emotion-arousing stimuli (electric shocks, explosions, etc.). Resultant changes inpulse, respiration, and heart-beat were recorded and, in addition,photographs were made of the facial expression of each subject in eachemotional situation.Analysis of these photographs showed that there wasno facial expression typical of any stimulus situation or any verballyreportedemotion.Norwasthereanyrelationbetweenreportedemotionandthe recorded internal changes. Landis was led to conclude that thebehavioral distinction between emotions lay, not in the facial responsepattern or the organic changes, but inthe nature of the stimulus situationand the degree of general disturbance evoked.Verbal reports of such statesas surprise, anger, exasperation, and disgust were paralleled mainly by adecrease in theamount of facial movement, in the order here named; andeach individual characteristically used no more than two or three distinctpatternsof expression in all the situations towhichhewas exposed. Ifweconsider these reports as discriminative operants based upon similar statesin all the subjects, we can only conclude, with Landis, that facialexpression is a very untrustworthy guide to the type of emotion involved,atleastforadults.

FIG. 74.Two representations of "startle," an unconditioned pattern ofmuscular responses to such stimuli as pistol shots.This is one of the fewrelatively stable 'emotional' patterns in human adults. This characteristicspasm is over in less than a second. (From Landis andHunt,The startlepattern[Farrar&Rinehart,1939].)

2 .Patterns in infancy. The story of infant expressions of emotion ismuchthesame.Exceptforsuchpatternsofresponseaslaughingandcryingwelookinvainforspecificfacialorbodilymovementsthatwillpermitusto distinguish between emotional and non-emotional behavior or between

one emotion and another. Watson (1924), from his study of infantresponses to loud sounds, bodily restraint, fondling, and other forms ofstimulation, was led to argue that three basic patterns could bedistinguished in newborn babies, and he named these patternsfear, rage,an dlove. Today, however, in the light of similar studies by otherinvestigators, we questionWatson's view. Sherman (1928), for example,has shown us that the observation of response patternsalone (aswhen thestimulus conditions are deleted from moving-picture records of theexperiments)provideforagreatdealofdisagreementamongobserversastotheemotionsactuallydisplayed.Unanimityofjudgmentbynurses,medicalstudents, and others, depended largely upon their knowledge of thecircumstances inwhich the behavior tookplace. In this field, as in that ofadult facial expression or of 'physiological' reactions (blood-pressure,respiratory, and other changes), the attempt to find uniform emotionalpatternsofresponsehasbeensingularlyunsuccessful.

Youmayhaveanticipatedsuchfindingsbecauseofthewayinwhichweapproachedourproblem.Wesaid thatemotion is to bedistinguished frommotivation on the basis of the operations involved and the widespreadchanges in reflex strengthobserved, butwedidnot attempt to specify anyparticularresponsesorpatternsofresponsethatwereuniquelyemotional intheirnature.Perhaps thereare such,observableat thesub-human levelandin the early years of human growth.A pattern of 'rage,' for example, hasbeen observed in cats, dogs, and other animals, and we have alreadymentionedthesmilingandfrowningresponsesofbabies.Butthereis littlereason to believe that we shall ever be able to classify man's emotionsexclusivelyintermsofhisbehavior.

TableXIITHEIDENTIFICATIONBYADULTSOFEMOTIONS"EXPRESSED"

BYINFANTS(DatafromSherman,1928)

Some of the names given by graduate students in psychology to infant"emotional" behavior (as seen in moving pictures) when (a) the initiatingconditions wereshown (or named, as for "hunger") and(b) when theinitiatingconditionswerenotshown.Thetablemaybereadasfollows:13studentscalledtheinfant'sbehavior"anger"whenthebehaviorwasarousedby"hunger"buttheinitiatingconditionswerenotshown,etc.

SpecificEmotionsWhat, then, canwe say about the traditional categories of emotion that

we commonly respect in our everyday speech and which have been thesubject matter of so many psychological treatises? Are we justified intalkingof fear, rage, joy, sorrow, and the like,when theyhave never beendefined to the satisfaction of the objectively oriented scientist? It will notdo, in answering these questions, to argue that, since we have used suchwordsor their equivalentsdown through theages and inmany lands, theymust represent fundamental realities; nor will it do to appeal to personal'experience' or 'consciousness' of these different emotions. Error, no less

than truth, may persist for long periods of time; and one's report of an'experienced' emotion does not ordinarily tell uswhat SD's determined it.Wemight rather take the same path that we followed in our treatment ofmotivation. Let us ask about the effect of specific operations upon thechanges in strength of groups of reflexes.Perhaps, in the end,we shall beable to justify the popular distinctions, but we ought not to begin byacceptingthemtooreadily.

1 .Fear. In our discussion of anxiety inChapter 9, itmay have struckyou thatwe could as easily have used thewordfear for this inferred stateand called itemotional.This is, indeed, the case. It is difficult to draw asharp line between themotivational and emotional aspects of anxiety, andthe difference betweenfear andanxiety is hardlyworth arguing about.Theimportant thing is not the name that we use but the fact that someoriginallyneutral stimuluscomes tobea secondarynegative reinforcerandalters the probability that certain responses will occur. A distinctionbetween fear as emotion and as drive is possibly useful in some cases, aswhen the organism seems unable to make the responses which wouldremove the activating stimuli. Fear then qualifies as distinctly emotionaland is sometimes calledterror. However, we could study the matter indetail without ever mentioning any of these words;the experimentaloperationsandthebehavioralchangesare the trulysignificant things. Ifwechoosetoapplytheseclassical terms, it isonlybecauseseriousstudentsofmotivationandemotionhaveoftenusedtheminconnectionwithsituationsand behavior that are indistinguishable from those with which we areconcerned.

Fromtheearliestdaysoftheorizingaboutfear,emphasishasbeenplacedupon responsesmediated by the autonomic nervous system—respondents,likepaling,sweating,shivering,andsoon.Buttheseresponses,asCannonhas made clear, are not sufficient to define the emotion adequately.Theymay occur in other states, some ofwhich, like fever,would not be calledemotional. The changes in strength which help us most are thosedepressions of on-going operant behavior which are obvious to anyobserver.Theorganismshowsamarkeddecrease in the frequencyofmanyofhisnormal,everydayresponses.Therat,forexample,atthesoundofthetonewhichwas followedby shock in theEstes-Skinner experiment (here),no longer pressed the bar at his accustomed rate; and we have reason tobelieve that all of his other operant behavior in the experimental chamberwas similarly affected.The human being, in a comparable situation, talksless,makesfewermovements,ishyper-sensitivetosmallstimuluschanges,approaches new objects gingerly, or brings his routine activities to acompletestandstill.Thesearesomeof thechangeswhichserveasSD's forusinascribingfeartoanother.

Theoperationwehave selected in defining fear is thepresentation of a

secondary negative reinforcer. It is unlikely that this is the only effectivecondition. Primary negative reinforcers (shocks, loud sounds, etc.) mayhave the same behavior outcome; and there may be still other inducingfactors. It is best, however, to leave it for future research to broaden thegroupofoperationsandtomakefurtherdistinctions.

In line with the present analysis, any stimulus situation may becomefear-instigating if associated with aversive stimuli.This has long beenrecognizedbyspecialistsinthestudyofmorbidfearsor 'phobias.'Scarcelyan object or event can be named which has not been for some person atsome time a source of overpowering fear. Sharp objects, open spaces,running water, horses, dirt, automobiles—these, and countless othercomponents of our everyday outside world may come, throughconditioning, to alter severely our response tendencies. Even internal,response-produced stimulimay be effective, as whenwe are said to fear a'thought,'an'idea,'ortheperformanceofsomeact.Whenthecircumstancesoflifehaveconspiredtogiveanegativelyreinforcingstatustotoomanyofthe aspects of the environment, outer or inner, itmay evenmake sense tospeakof'generalizedfear.'

2 .Anger.When positive reinforcement is suddenly withdrawn, a rat'sbar-pressing, as well as his general activity in the experimental situation,will showaquickacceleration, andhemayevenbe seen to 'attack' thebarvigorouslywith his teeth. Pigeons,when crowded into narrow quarters orconfronted by trespassers upon their well-staked-out preserves, will fighteach other viciously (Craig, 1921). Other animals respond similarly insimilar situations. Hebb (1946) has recently given us an especiallyinteresting account of chimpanzee behavior, fromwhich it appears that thebreakingof anoperant chain is a fundamental factor in producing suddenand violent attack behavior, self-injury (pounding the head on the floor),screaming, and, on some occasions, a response depression or 'negativity'whichmayreachthepointwheretheapewillturnawayentirelyfromsomeformerlypositivereinforcer.Hebbappliesthewordsrage, temper tantrums,andsulking to suchbehavior, and speaksof the initiating circumstances as"the removal of a desired object," the "failure to get something," and the"frustration of an expectation." The breaking of a chain of on-goingresponses may, however, apparently be effected by the withdrawal ofpositive reinforcers and the presentation of negative. Thus, Hebb speaksalso ofteasing, aswhen one ape repeatedly interruptsanother's activity-in-progressby screaming, spitting at him,ormaking threateninggestures. (Itseemsobvious that 'teasing'might alsobeapplied to thosecases inwhichan'expectation'issetupandthen'frustrated.')

In his treatment of infant emotions,Watson (1924) cited restraint ofmovement' as the principal cause of rage.We can see in physical restraintanother way of interrupting a chain, and some credence may therefore begiven toWatson's position. But it should be noted that the restraining

action must be imposed upon movement-in-progress—it must be truly'thwarting.'Dennis (1940)haspointedout thatanew-born infant's randommovements may be restricted greatly and for long periods of time, as byswaddling clothes or cradle-boards, without noticeable emotional upsetunlessthe'restraint'isforcefulenoughtoconstituteanegativelyreinforcingstimulation.

The operation of chain-breakingmay be detected or suggested inmanyeveryday situations allegedly productive of anger. In a study by Gates(1926),women studentswere asked to report all instances of this emotionduringaone-weekperiodofself-observation,givingineachcaseanaccountof the precipitating circumstances. Among the situations mentioned bythese subjects were scoldings, criticisms, refusals of requests, tardiness offriendsinkeepingappointments,gettingawrongnumberonthetelephone,failure in operation of such objects as watches, fountain-pens, andtypewriters, delays in bus or elevator operation, clumsiness in dressing orsewing, disobedience of dogs or children, interrupted sleep, and loss ofmoney.These occasionswere reported to engender "impulses" to (a)makeverbal retorts,(b) do physical injury to the offender,(c) injure inanimateobjects,(d)getoutofthesituation,and(e)cry,scream,orswear.Theorderof frequencywithwhich these tendencieswere reported is indicated by theorder of mention. The verbal retort was specified five times as often ascrying, screaming, or swearing. Also,persons, rather thanthings, werereportedasthemainsourceofanger.Obviouslynoonepatternofresponsesfor all subjects was peculiar to any one situation, since the responsesinvolvedweredependentupon the individual'shistoryofconditioning,yetsome similarity of outcome is indicated by the fact that the responses fallintothesefivegroupings.

Inspiteofthefactthatnotwopersonsreactinexactlythesamewayasaresult of the anger-producing operation, there appear to be certain fairlycommon elements of such behavior as we ordinarily observe it. Increasedmuscular tension (sometimes seen in the clenched fist, thegrimly set jaw,and the rigid posture of an angry man), movements of advance or attack,stamping of the feet, elevation of the voice, and so on, are often enoughseen to suggest the state even when we are unaware of the initiatingconditions.

3.Sorrow and joy.The changes in rat behavior resulting from stimulithat have been regularly associatedwith electric shock often lead observersto attributefear to these animals. The sudden burst of bar-pressing andbitingresponsesatthebeginningofanextinctionsessionisfrequentlysaidto showanger.Wenow see that the use of these two terms is justified insomedegree.Sorrowandjoy,ontheotherhand,areseldomimputedtotherat in any situation, even by personswho regularly employ such terms inconnectionwiththebehaviorofdogs,chimpanzees,andhumanbeings.Arethereoperations,differentfromthoseweusetodefineangerandfear,which

arerelatedtospecialchangesinreflexstrength?Weareloathtoreplytothisquestion,forseveralreasons.Psychological

literature does not abound with treatment of these concepts. One of themost authoritative discussions in recent years (Young, 1943) gives thembarely a mention; and most textbooks do little more than assume theirexistence. Unlike anger, and especially fear, they have seldom been thesubject of scientific investigation. Moreover, those who attempt to givemeaning to these terms are in no great agreement, possibly because theyhave so often been concerned with the fruitless task of identifying theseemotionsbyfacialexpression.Yet, it is improbable that thepresent readerwillhavetheleastdoubtabouttheexistenceofsorrowandjoy.

Itwouldperhapsbewisetosaynothingmoreaboutthematterhere.Yetwe cannot resist mention of two possibilities that present themselves inconnection with familiar descriptions of sorrow and joy.To us it seemsclear that thecomplete removalof secondarypositive reinforcers occasionsa widespread behavioral depression which involves changes commonlyreferred to as those of grief, sadness, or sorrow.The irretrievable loss ofmoney, home, possessions, family, or friends seems to bring an obviousaltering of muscular tonus (one is 'bowed down' under the 'weight ofsorrow'),aslownessorabsenceofreactiontocustomarysocialstimuli,anda decrease in the frequency and force of speaking, not to mention theappearanceofsobbingandmoaning.

The withdrawal of strong secondary reinforcers implies the breaking ofresponse chains, and this is an anger-producing operation.How, then, canwecall it a conditionof sorrow?Theanswerheremaybe that anger is thefirst stage in the production of sorrow. If the loss of secondaryreinforcement is impermanent or restricted to but a few chains of action,angermay be the result, but if the loss is prolonged and related tomanychainssorrowwillensue.Accountsofsorrowfulbehaviorseemtobeinpartaccounts of anger. Thus, Foley (1935) has described the grief ofchimpanzeesasdue tosuchfactorsasa refusalordelay incomplyingwiththeir expressed wants or desires, operations which Hebb (here) treats asproductive of anger.And have we not all observed that great losses arecommonlymetattheoutsetbycriesofrageandresentment,tobefollowedsoonby thedeepest of depressions?Perhaps, even, the 'sulkiness' sooftentakentobeaformofangeris,ingoodpart,thebeginningofsorrow.

If the removal of secondary positively reinforcing stimuli is importantforsorrow,certainlythepresentationofsuchstimulishouldfacilitatethoseresponseswe calljoyful.Again,wehave little evidence topresent, but thenon-experimentalobservation of human beings and certain animals pointsconvincingly in this direction.Almost everyone has noted the changes inthebehaviorofchildrenattheapproachofadotingrelativeorfamilyfriend,the announcement of an unexpected holiday, the promise of a trip to thebeachorthezoo,orthesoundofthebellthatsignalsarecessfromschool.

Thereactionpictureonsuchoccasionsisoneofexcitement,justasmarkedin its own way as the depression of sorrow. Lively, free, and extensivemovementsofarmsandlegs(thechild 'dancesforjoy'),smiling,laughing,excessive vocalizing—these and other activities are commonly observedwhenachild ispresentedwithverbalorotherstimuli thathave in thepastbeen followed by strong positive reinforcement. Adults are usually lessexpressive, and react in a less diffuse,more differentiated fashion than doyoungsters,butsomechangesintendencytorespondareusuallyapparent.

Thesesuggestionsareputdownwithhesitation,because theproblemofidentifying specific emotions is a complex one.We have not made anexhaustive attack upon this problem, but an important aimwill have beenaccomplished if we have shown the way in which an objective,experimental approach must proceed. One thing is certain: there areoperations, involving stimulus presentation and removal, which do morethan change the strength of a single response (as in conditioning andextinction); they change the strength ofmany responses.We have arguedthattheselatterchanges,takeninconjunctionwiththeoperationsinvolved,seemtoberelatedtotraditionaldescriptionsoffear,anger,sorrow,andjoy.But,evenif thiswerenotentirelytrue,wewouldstillhavetheproblemofaccountingforthechangesresultingfromtheoperations.

EmotionalConditioningIt is unnecessary to say much about the conditioning of emotions, in

view of the instances of itwhich have already accumulated in this text. Ithas long been recognized thatemotion involves a great deal ofautonomically governed behavior; and, sinceChapter 2, you have knownthat this kind of behavior is readily conditioned inType S fashion. Ourtreatmentof the effectsofpunishment, inChapter4,made explicit note ofemotional conditioning; and statements of similar import have occurred inourdiscussionofseveralothertopics.Emotionalconditioningisafact,andso is emotional extinction, although our knowledge of these processes isstillatafairlyprimitivelevel.

Aclassicinstanceofconditionedemotionmaybecitedhere.WatsonandRayner(1920)presentedanine-months-oldboywithawhiteratatthesametime thata loudsoundwasproducedbyhammeringonasteelbar.Therathad not previously disturbed the child, whereas the noise had elicited aviolent 'fear reaction.' Only a few combinations of the two stimuli wererequired to give the rat a strong conditioned-stimulus function.Moreover,the effect was shown to generalize to other furry animals and severalinanimateobjectsofa'furry'nature—eventocottonwool.

Criticism was later launched atWatson'sdefinition of fear (here), butseveral studies have confirmed the fundamental observations of aconditioned emotional change, and a few investigators have gone further.Jones (1924a, 1924b),workingwith a three-year-old boywho came to her

with an already well-established emotional reaction to rabbits, rats, andother animals, undertook to eliminate this effect. The procedure firstemployedwasthatofintroducingtherabbittothethechildgraduallywhilethe latter was playing with three other, unafraid, children.This techniquewasworkingwellwhenanunfortunateexperiencewithalargedogservedtorecondition the emotion.Themethod then adoptedwas that of presentingtherabbitattimeswhenthechildwaseatingahighlydesiredfood.Inordertoavoidaviolentupsetonanytestoccasion,theanimalwasneverbroughtso near to the child as to depress his eating behavior.This procedure ofcombiningtheextinctionofsomeresponseswiththepositivereinforcementof others was successful in eradicating the fear. In addition, Jones foundthat the extinction generalized to such a degree that the other, formerlyfrightening,animalsandobjectslosttheirpowertoevoketheemotion.

In view of such findings as these, you may be surprised to learn thatvery few follow-up studies of the conditioning and extinction of humanemotions have been made. Obviously, much work remains to be done inthis area. It is understandable, however, that experimentalists (andparents)hesitate to subject children to emotional conditioning unless there is acontrol of factors, in the home and elsewhere, such that the often lengthyprocess of extinction can be carried to completion. The usefulness ofpreliminaryresearchwithanimalsisclearlyindicated.Wedowell,perhaps,in following the lead of medical scientists and others who test out theirproceduresataphylogenetic levelwhere itwillnotendangerhumanhealthanddevelopment.

ObservingOurOwnEmotionsItwas stated earlier (here) that our 'consciousness' of our own fear, joy,

and so forthwasnot a satisfactory reason for assuming the realityof thesestates of reflex strength. Before leaving the topic of emotion, somethingfurthershouldbesaidaboutthisstatement.

Whenwe say thatweobserve anything,wemeannomore than thatwemake a specific response to a discriminative stimulus or stimuluscompound. Such stimulationmay, in psychology, arise from the behavioro fother organisms, in which case we ordinarily speak of 'objective'observation.Itmayalsoarisefromourownresponses,affectingusthroughour eyes, ears, or the sense-organs in ourmuscles (propriocep-tors).Whenyou say, of another person, "He is angry," you are behaving inmuch thesame way as when you respond to your own behavior by saying "I amangry."Thedifference liesmainly in thediscriminativestimuli.Moreover,your responses to either of these two stimulus situations aremadebecausethey have in the past beenreinforced in their presence—you have actuallybeentaught to discriminate by someone in the course of your behavioraldevelopment.

If thisisclear, let'sgoabitfurther.Whenyouweretaughttosay"Iam

angry,"your teacher couldnot himself observe (respond to) all the stimulipresent, because some of them were produced by your covert responding.Hecouldobservetheexternalstimulussituationandsomeofthechangesinyour behavior that followed, but not all.Which stimuli became importantforyouasSD'shecouldnottell.Yourteachermayhavesaidonly"You'reangry, aren't you," when he saw your overt behavior.When you come tosay"Iamangry,"yourSD'smaynotbe limited to thosewhich led tohisresponse.The astonishing fact is that, after such training, there is sooftengoodagreementbetweenone'sownreportofanger(orsomeotherstate)andthe report of an onlooker—not seldom is the assertion "You look angry"answeredbytheassertion"Iamangry."

Wehave already said, however, that the response "I am angry," aswellas such typicalnon-verbal accompaniments as fist-clenching,maybemadeindependentlyof anyanger-producingoperationwhatever.Theuseof thesewords and gestures may be effective in producing other kinds ofreinforcement than that provided by our 'teacher.'They may be used, forexample, because they have been reinforced by flight, or some form ofcompliance, in another. In such cases it is sometimes hard for us to saywhetheraperson is 'reallyangry,' 'justbluffing,' or 'hunting for sympathy'unless we know his history very well and the situation in which he isbehaving.

Bynowyou shouldbe able to seewhypsychologists havehesitated tolean upon verbal reports of emotion. These reports some are of thediscriminativeSD'sareprivateresponsesandoneasmuchcannotasbeanysureother,thatbuttheyalwaysgohandinhandwiththosethatarepublic.In any case, you should appreciate that one's reports of his own emotionsarederivedinitiallyfromthediscriminativetraininggivenbysomeoneelse.Hence, they can hardly be used as the test of anyobjective distinctionbetweenanger,joy,andanyotherstates.

EmotionandFeelingInChapter 8 (pages 254-256), we noted that the problem of reported

'feelings' is similar to that of reported 'emotions.'When someone tells usthat an object, a color, or a design is 'pleasant' or 'unpleasant,' he isreporting upon his own reactions to that object, color, or design. Apositive,oranegative,reinforcerhasbeenpresentedorwithdrawn;achangein behavior has taken place, including, perhaps, incipient movements ofapproach orwithdrawal; thesemovements (or others) provide the SD's forhis verbal responses—his 'affective judgments.' The situation isundoubtedlysimilartothatwhichobtainswhenweasksubjectstoidentifytheir emotions.We can easily understand why feelings and emotions arecharacteristically lumped together, and why people have often tried to

classifyemotionsas 'pleasant'or 'unpleasant.'Theoperationsofgivingandtaking away reinforcement are important in both emotion and feeling;whetherwereportoneortheotherdependsupontheparticularSD complex(aswellastherequestforareport)thatisoperativeatthemoment.

Finally, we can see why feelings and emotions are generally treated as'subjective.'Wemight suggest thatyouextend this lineof thought a littlefarther.Where, inessence,would"beauty"and"ugliness" lie?Andwhat isthe basis for our ethical judgments of "good" and "bad"? These areinteresting questions, in psychology as in other disciplines, but this ishardly the place for their discussion.We have perhaps equally interestingmattersforconsiderationinthechaptertocome.Letus,then,moveon.

NOTESAnexcellent reviewofpast treatmentsofemotionbypsychologistsand

physiologists may be found in Chapters XI and XII of Woodworth'sExperimental psychology (1938); and you may find interesting the relatedchapters(XandXIII)dealingwithfeelingandthegalvanicskinresponse. Itwillalsopayyoutoreadthediscussionsofemotional 'expression'(Landis)andthephysiologicaltheoryofemotion(Bard)intheHandbookofgeneralexperimentalpsychology, edited byCarlMurchison. For an account of theJames-Lange theoryof emotion, youought to go to thatwrittenby Jameshimself in hisPrinciples of psychology,Vol. II, Chapter 25 (1890).Themost inclusive modern text on emotion isYoung's Emotion in man andanimal (1943). The intrinsic interest and readability of these accountsshould not blind you to the fact that they do not attempt a systematicbehavioral analysis of the problem—one that is free of mentalism andphysiology.

11

SOCIALBEHAVIOR

SOCIAL BEHAVIOR comprises the stimulations and reactions arisingbetweenan individual...andhis fellows....Socialpsychologymustnotbeplaced incontradistinction to thepsychologyof the individual; it is apartofthepsychologyoftheindividual,whosebehavioritstudiesinrelationtothatsectorofhisenvironmentcomprisedbyhisfellows....

F.H.Allport,SocialPsychology,1924

IntroductionInone sense, this chapter is anti-climactic; in another, it is preparatory.

Your introduction to the problems of general psychology was essentiallycompleted with the discussion of motivation and emotion. The presentchapterrepresentsinasmallwayhowourbasicprinciplesmaybeextendedinto areas of behavior which, though complex, are compelling in interestand importance.Psychologistshavepainstakinglyamasseda largebodyofinformationabout the social conductof animalsandhumanbeings. Itwillbe our task here to show how this area may be approached with theanalytical tools now at your command. Only considerations of spaceprevent our doing the same thing for other special areas, such as abnormalpsychology,developmentalpsychology,andsoon.

We have said before (here) that social stimuli do not differ from otherstimuli in their dimensions. Rather, the difference is one oforigin. Theyarisefromotherorganisms,theirbehavior,ortheproductsoftheirbehavior.Moreover, social stimuli do not differ in theirfunction from those ofinanimate origin; they act as eliciting, reinforcing, discriminative, and soon. Social life arises becausesocial stimuli cometo exercise thesefunctions.These facts giveus a solid stepping-off point for thediscussiontocome.Itwillhelp,too,torecallthatsub-humanorganisms,nolessthanman,exhibitsocialbehavior.Indeed,fromthemwemaylearnanumberofthings thatwillenableus toseeman's socialactivities inamoreobjectivelight.

In venturing to extend our principles into social behavior, wemust becautious because the experimental or field data under examination haveoftenbeengathered under complex conditions.Complex studies, inwhichmany (and sometimes unknown) variables are atwork, allow only generalinterpretations and force any detailed analysis to be too speculative forscientific comfort. Thedetails of data may resist analysis even when wediscern in them the operation of some underlying principle that we knowwell.

SocialBehaviorinAnimalsBehavioral interactionmaybeobservedat thebiological levelofsingle-

celledcreatures.TheplantVolvox, for example, lives in a colony towhichthe offspring remain connected until they break away to form their ownaggregations; andParamecia, which are normally solitary animals, cometogether for the purposes of sexual reproduction. At this level, actionsprobably depend on physico-chemical changes that function exclusively inanelicitingfashion.

Higher in the animal scale,insects have long provided us withcomplicated examples of social interaction. Everyone has heard of thedivision of labor and specialization of functionwhich characterize ant andbee communities. Some species of ants even protect and rear otherorganisms(aphids)withintheirnests,and'milk'themlaterforthenutritiveliquid they secrete. Much of this and other insect behavior is probablyelicited respondent-wise. Nevertheless, insects can also learn in Type Rfashion.We know that they may be taught to run simple mazes for foodandtoescapefromnoxiousstimuli;anditmaybethatsocialstimuli,suchasthoseinvolvedinreturntothenest,mayalsoactasoperantreinforcers.Birdsprovidefurtherexamplesofsociallymediatedreinforcement.Many

species habitually live in flocks, but evenwhere this is not the case, theymay form groups under certain conditions, as at times of migration.Seasonalchangesintemperatureandindiurnalilluminationproducebodilyeffects(e.g.,inthesexorgans)thatraisetheactivitylevel.Thegatheringofamigratorygroupatgivenlocations,theformationoftypicalflightpatternsintheair, theflight leadershipbyolderandmoreexperiencedbirds—alloftheseprobably involveoperant-respondentoverlap,but social interaction isobvious. Most birds show a degree of care for the young, and they mayprolong this care beyond the time when the offspring are capable ofindependent flight and self-support.Even the songs andcall-notesofbirdsmay be in part a product of social control. In the oft-cited experiment byScott (1901),youngorioleswere separated from theirparents and raised inisolation, so that they would not be exposed to the normal oriole song,with the result that a new kind of song appeared. Conradi (1905), later,placed sparrows with canaries and observed that the sparrows tended toreproducethecanarycalls.Insuchcreatures,theremaybeanaturaltendencyto echo sounds, but even if there were not, it is possible that, throughsocially-given differentiation training, such behaviormight arise, with theparent bird providing food for the young when certain sounds wereapproximated.Inthisway,ayoungstranger in thenestmightgraduallybetaught the 'language' of his foster parents, even as we do with our ownchildren.

Social behavior is, of course, found in extraordinary richness amongmammals, particularly theprimates. Habitual living may be in herds,flocks, and packs; family units are often clearly defined; and sexual

attachments may be monogamous or polygamous, seasonal or perennial,and sometimes even homosexual. One has but to think of domesticatedanimals like the dog and horse to realize that social stimuli may becomepowerfully discriminative, reinforcing, and drive-arousing for sub-humanbehavior.Monkeys,baboons,andtheanthropoidapeshavebeenstudiedinsome detail by psychologists and zoologists, and much is known abouttheir actual or potential social activity. Investigators all attest to thestrength of friendships and enmities among these animals, and to variousotherwell-delineatedinter-personalrelations.

The exigencies underwhich social stimulation controls animal behaviorare numerous. Sexual needs give reinforcing value to the mate whomediates the reinforcement. Parental care, especially among animals inwhich the young go through a long period of dependency, also favors thedevelopmentofsocialSD'sandSr's.Huddlingbehavior,wherebyphysicalwarmthisprovided,mayleadtothesameend.Evenfood-gettingmaygiverise tobehavioral controlby such stimuli, aswhenpredatoryanimalshuntinpacksandworktogetherinrunningdowntheirprey.

In a famous experiment by the Kelloggs (1933), a seven-months-oldchimpanzee,Gua, was brought into their family and raised with a ten-months-old child of their own,Donald, during a period of nine months.Every attempt was made, during this period, to treat the youngsters inidentical fashion.They lived and played together like any ordinary pair ofchildren, even as twins might; they wore the same clothes, followed thesame daily schedules, ate the same foods, slept in similar beds, werespoken to in the sameway, and so on.The 'children' accepted each otherfully and affectionately; and Gua's development testified clearly to thepotentiality of chimpanzee behavior. She was hardly to be outdone byDonald in a number of activities. She acquired, for example, a good'passive'vocabulary,reactingconsistentlyandcorrectlytomanywordsusedby the Kelloggs. In other respects, she definitely surpassed Donald, as inher climbing ability.The experiment could not, of course, be continuedindefinitely, since the child had clearly less to gain from the association;butitprovidesuswithaclearindicationofalmost-humansocialcapacityatthesub-humanlevel.

SampleAnalysesofSocialInteractionOf the several typesof social relationamonganimals thathave received

the attention of psychologists, we may select as examples for our closerattentionthoseofdominance,cooperation,andimitation.

1.Dominance.Ananimalmayachievedominant statusover anotherbyvirtue of the actual or threatened negative reinforcement it applies. Incompetition for food,mates, sleeping quarters, or whatever, the dominantindividual wins out by lowering response strength in a rival through

punishment. He may thus become a strong secondary negative reinforcerandSD,totheextentthatameregesturewillleadthesubmissiveanimaltobeatahastyretreat.Schjelderup-Ebbe(1935)hasnotedthatdominancemayoften be established at the first meeting of two birds, with the dominantanimaldrivingawayorpeckingtheotheratwill.Theavoidanceandescapereactions of submissive birds may persist even when the dominant birdshave become old and weak, if only the battle posturings of the lattermaintain their SD and Sr functions. The fact that submissive animalssometimes 'rebel' isconsistentwithourknowledgeof theimpermanenceoftheeffectofpunishment;andwemaysuspectthatrebellionsaremorelikelytooccurwhenmotivation ishighand thedominantanimalblocks thewaytoreinforcementlikefoodoramate.

Dominance through physical force has also been observed amongbaboonsbyMaslow(1936a,1936b)andamongratsbySeward(1946),andhas been seen to involve stimulus generalization.A badly-thrashed animalwill transfer his submissiveness to other animals than the one that firstdefeated him. A single thrashing, if severe enough, will bring aboutsubmission to all comers, even when the loser in a fight has previouslybeenthechampion.

Dominance-submission relations, it is often said, may bebased uponother factors than general physical strength. This is true, since theapplicationofpunishmentbyoneindividualisnotrelatedsolelytoover-allphysique. The determining factors are thehistories of the interactingorganisms.Thus, an old bird, or a relatively feeble young one, may bedominantbyvirtueofthesubmissiveone'sexperiencewithsimilarwar-likestances assumed by a former victor; and a nimble or resourceful ape maydominate a strongeronebecausehe is neverthelessmore adept in applyingpunishment. Circumstances like these are not far different from those ofhumanexperience.However,thereisnoreasonwhy,inhumanaswellasinanimal life, dominance-submissionneednotbebasedon a combinationofpositive and negative reinforcement, or on positive reinforcement alone.Any definition of dominance is arbitrary. In some animal species, forexample, a normally submissive female may gain marked control, inmating season or at other times, because of sexual changes in the malewhich raise her reinforcement value for him. Or, if a person's own effortsoften result in failure and there is someone who unfailingly suppliesrewards, independent efforts may cease and excessive dependence on thewilling benefactor may develop. Furthermore, the dominance of oneindividual and the submission of another may be specific to certainsituationsandactions, inaccordancewiththehistoriesofbothindividuals.Itisnotalwaysthesamemaninacollegeclasswhodominatesinboththeclassroomandthegym.

2.Cooperation. Cooperation, like dominance, is not a single response

entity,butdenotesmanykindsofrespondinginmanytypesofsituation.Itmaybedefinedasthecaseinwhichthecombinedbehaviorof twoormoreorganismsisneededtoprocurepositive,orremovenegative,reinforcementforeither.Twoorganismsmayperformthesameact, likehauling togetheronaropetoobtainfood,ordifferentacts,aswhenonestepsonatreadletoopenadoorwhiletheotherpullsinfood.Thissortofcooperationiseasilyobtained inhumanadults,byverbally instructing themin their rôlesor bylettingthemdiscovertheefficiencyofjointaction.Withanimals,aswellasyoung children, we have to train them from the start and can, therefore,observehowcooperationdevelops.Theeffectivenessofverbalinstructioninthehuman adult depends, of course, upon a longhistoryof training, fromchildhoodon;and,evenwithus,cooperativeendeavorinsomeactivitiesisoftenachievedonlywithdifficulty—sometimesnever.Inanycase,whetherwithhumanbeingsoranimals,andwhateverthetask,cooperationinvolvestwothings:(1)eachorganism'sactionmustbediscriminativefortheother'sperformance;and (2)eachorganismmustbe reinforced for thepart itplaysinthecooperativescheme.

The investigations of Daniel (1942, 1943) are good illustrations of theway inwhich cooperation is acquired by animals difficult to train in suchbehavior.Inhisstudies,ratswerefirsttaughttoobtainfoodfromadishonthe floor of their experimental chamber, and then to turn off an electricshockfromthegridfloorbyclimbingontoasmallplatform.Theplatformwaseightinchesfromthefooddish,makingitimpossiblefortheanimaltoeat and avoid shock simultaneously.When the two responses had beenindependently well established in each rat, training in cooperation began.Twohungryanimalswereplacedinthechamberatthesametime,withthefloorelectrifiedandfoodin thedish.Thus,oneratwasable toeatwithoutbeingshockedonlyifhispartnerstayedontheplatform.Ifbothwenttothefood dish, bothwere shocked. Experimentation took place during a forty-day period.As training progressed, Daniel observed that the rats came toexchange positions with considerable regularity, and with such efficiencythatbothgotplenty toeatandwereable toavoid shockon94-99percentofall the times thateitheranimal lefthisposition.Seldomdid theywastefeeding time by sitting on the platform together; andmore andmore theirbehavior came to be directed by the stimulation they received from eachother. Keeping at least one foot on the platform, the shock-avoiding ratwouldnudge,paw,orevenbitethetailofthefeedingratuntilthelatterleftthedishandgavehimhis'turn'ateating.

A sketchy analysis of such cooperative behavior might run as follows.Theplatformrat,thoughhungry,stayedinplacebecauseshockwastherebyavoided.Ashisanxietyabated,hisfood-gettingbehaviorbecameprepotentand he attempted to produce the SD for approach to food—to bring theother rat beside him on the platform.His nudgings, pawings, and bitingsserved in turnasanxiety-arousingstimuli forhispartner,since theyhad in

the past been followed by electric shock ifhe delayed in returning to theplatformrat'sside.Theonlyescapefromthisstimulationforthefeedingratwas, of course, by leaving the food and going to the platform;whereuponthe platform rat, having its SD for 'safety,' was free to eat, and the entireprocess was repeated with the animals in reversed positions.All that wasrequired to maintain the subsequent cycles of exchange—to keep thecontrollingstimulieffective—wasanoccasionalshocktobothratsresultingfromafailureofoneortheotherto'playhispart.'

FIG. 75. Plan of the apparatus used by Daniel (1942) in studyingcooperativebehaviorinrats.(AfterDashiell,1949.)

In the above experiment, cooperationwas based primarily upon hungerand shock-aversion. In his second study (1943), Daniel tried to see whatwould happen if, after cooperative behavior of the above kind wasestablished,theshockwassystematicallyreduceduntilnonewaseverused.In this experiment, an electrically operated cap was placed over the fooddishso thatananimalhad tobeon theplatform inorder for thecap tobelifted and the food exposed for eating. The question was whether theanimals would continue to 'take turns' at platform and dish under theseconditions.They did not, and it is not difficult to see why.The anxietythatcausedonerattoreturntotheplatformonbeingnudgedbytheotherisno longer present.The onlymotive at work now is hunger, and the onlyrewardforgoingtotheplatformandawayfromthedishisthechancetoeatlaterif thepartnercooperates in thesameway.Thereinforcementforgoingto theplatform is,atbest, longdelayed,andstimuli likenudgingsmerelysignaltotheeatinganimalthatfoodmaysoonbewithdrawn.Thiskindof"higher order" cooperation, involving long delay of reinforcement, isapparentlyverydifficult, ifnot impossible, toobtain inarat,whereas it ischaracteristic of many everyday human activities such as budgeting one'sincome in order to save for desired things like a car or the family's ownhome. Nevertheless, you can see that cooperative behavior at all levels oforganismiclifefollowsourbasicprinciples,andthatthefunctionsofsocialstimuliarethesameasthoseofnon-social.

3.Imitation.Our third illustration of social relations among animals isthat of imitation, and it, too, can give us an opening into human socialbehavior.Weneednotgointothecenturies-longdiscussionofimitation.Itisenough tosay that, fora long time,manyoutstandingmen likeCharlesDarwinandWilliamJamesbelieved that imitationwassimplyanaturalor"instinctive" thing. Early social psychologists like Gabriel Tarde (1890)regarded imitation as a fundamental trait of behavior, and used it as anexplanatory principle in itself when discussing the regulation of humancustom and the development of human society. Experimental attempts tomeasure"unlearned"imitationinmanoranimalsnevergaveanunequivocalanswer, and there has been an increasing reluctance on the part ofpsychologists to appeal to imitation as the explanation of any case of oneanimal's duplicating the behavior of another. They feel that behavioralduplication can arise frommany causes, so that not all cases ought to beclassed together. Moreover, not all behavioral duplication is "imitation,"since there are many duplications no one would be inclined to class asimitation; for example, peoplewalking to themorningbus, or a defensivefullbackchasingtheballcarrier.

Whetherornot there isanysuch thingasunlearnedor innate imitation,onethingissure.Imitationcanbetaught.Throughtheuseofconditioningprocedures—by making reinforcement contingent upon repeating another'sact—anorganismcanbemade to imitate.Anactofoneorganism ismade

the SD for an act of another, so that the action of the leader must berepeatedifthefolloweristogethisreinforcement.

As an example of this leader-follower relation, we may cite anexperimentbyMillerandDollard(1941)whoworkedwithbothhumanandanimalsubjects.Oncemore,wechooseananimalillustrationbecauseofitsrelativesimplicityanddramaticclarity.Theanimal(arat)whowastoactasleader,was first and separately taught to rundown the stemof an elevatedT-maze, and at the choice point to turn down the armmarked by a blackcard. In the usual discrimination training procedure, the position of thewhiteandblackcardswasexchanged in randomorder fromtrial to trialonthearmsoftheT,butaturntowardtheblackcard(S D)alwaysbroughttheanimal to food, whereas a turn to the white card (SΔ) wentunreinforced.After the leaderhadmastered thisdiscriminationwell, apotential follower(aratwithoutanyprevioustraining,buthungry)wasplaceddirectlybehindhim on the stem of theT.The follower received reward only if he turnedinto thesamearmof theTasdid the leader; ifhe turned into theoppositearm, he got no food.The experimental results left no doubt that imitativeresponding was learned: the response of the leader became an SD for thefollowerwho came practically always tomake the same turn he did.Thisfinding was not dependent on accidental cues from the black and whitestimuli reaching the follower, as proved by the production of imitativerespondingwithoutcardsbeingpresentatall.For thischeck, some leaderswere trained always to turn right in theT, others always to turn left.Thetested animalswerewell-trained followers, but had learned their followingwith leaders who had themselves obeyed card cues.With the new leaderswho requirednocards,andwhosometimeswere right-turninganimalsandsometimes left-turning animals, the followers continued to use the leader'sresponse as SD.You will not have missed the added implication of thisfinding that leaders will generalize for followers, so that a well-trainedfollowerwill trailaftermanyleadersindiscriminately!Inthelightofrecentand tragic historical examples we would hardly care to train into ourchildren such discriminative dependence upon others or the generalizedtendencytofollowanywould-beleaderwhohappensalong.

FIG. 76.The apparatus used in theMiller-Dollard studies of imitation.

The stemof theT is 18 inches long and the running surface is 1⅛ incheswide.The gap to be jumped at the choice point is 4¾ inches in eitherdirection.The food-cup for the leader is 16½ inches from the gap; for thefollower,8½inchesandiscoveredbyahingedcapwhich is removedaftertheleaderhaspassedoverit.(AfterMillerandDollard,

HumanBehaviorandtheCulturalEnvironmentThe community inwhichmen andwomen livemakes up an important

partoftheirtotalenvironment.Itisdifficult,ifnotimpossible,tothinkofa single operant activity of the individual which does not show in somedegree the pervasive influence of his community's teaching. The fullrecognition of this we owe to the efforts of anthropologists and socialpsychologists who have devoted themselves to the study of the form andcontentofhumansocietiesaroundtheworld,andtotheinteractionbetweenthe individual and his cultural environment. From earliest times, menknew, and were interested in, the fact that neighboring or distantcommunities existed that were different from their own, and that peoplebehaved differently in those communities. During the nineteenth andtwentieth centuries, anthropology developed as a social science, animportant sector ofwhich came to beknownascomparative ethnology, orthecomparativestudyofhumanbehaviorandcustomsindifferentsocieties.From such study, we have been led to discard some old conceptions andattitudestowardshumancommunities.

The new conception may be calledcultural relativity.Where once wewereinclinedtothinkofotherpeoplesasstrange,backward,andinferiortoourselves, the conclusions of anthropologists lead us to think otherwise.They point out that even the so-calledprimitives think the same thingsaboutus. In some societies, even thewordhuman is limited to one's owngroup, while other peoples are described as "non-men" or "sub-human."Furthermore, the primitive is not, as was once thought, childlike in hissimplicityandillogical inhisreasoning; infact,he thinksthesamethingsofusandeasilypicksoutweaknesses inourcultureand inconsistencies inour social behavior. Nor does he concede any ethical superiority to us.There are instances, for example, where he has reproached us for ourwholesaleslaughtersinwarandhasofferedtosendmissionariestoteachushowconflictsamongmenmaybesettledwithoutbloodshed.Heisamusedby some of our customs, and revolted by others. Sometimes he isimpressed by our gadgets and material achievements; or he may besingularlyunmovedby tall buildings and ships, or airplanes, pointingoutthat such things, while perhaps useful, are no guarantee of a peaceful andabundant span of years.All in all, the comparative study of cultures leadsus to conclude that such differences as exist among human groups do notform a basis for makingvalue judgments concerning them. Differences incustoms,folk-ways,manners,religiouspractices, ideals,beliefs,andsoon,

are simply differences in the way men may grow up in diverse physicalenvironments,outofdiversehistoricalstreams,andoutofa totalpotentialrepertory of behavior that is far greater than anyone culture can explore tothe full. Each human society is to be approached in a spirit of objectivityandaccordedthe tolerance(whichwewouldwishforourselves) thatcomesfrom an understanding of its origins, values, andmethods ofmolding thelivesofitsmembers.

Understandably enough, this contribution of anthropology has, in itsdissemination,sometimesbeenmisconstrued.Theunderlyingconceptionofman'salikeness must not be lost in the recognition ofdifferences.Whatmen learn indifferentsocietiesoughtnot toobscure thebasic facts thatallmen learn in the same way and are subject to the same drives. Theprinciplesofreinforcement,extinction,discrimination,andthelike,operateuniversally, though the form of the response to be learned, or the specialtype of SD to be obeyed, may be selected by the community. Similarlywiththemotivesofmen:allmeneat, thoughtheydiffer inhowandwhat;all men drink, perform sexual acts, and breathe. Further, all men havesimilar eyes and ears and noses for the reception of exteroceptive stimuli,thoughtheymaybe taught toresponddifferently to thosestimuli, throughconditioning procedures that are the same for everyone.We need to holdontotheideaofhumanabsolutesasmuchastothatofculturalrelativity.

The cultural environment (or, more exactly, the members of thecommunity) starts out with a human infant formed and endowed alongspecies lines, but capable of behavioral training inmany directions. Fromthisrawmaterial,thecultureproceedstomake,insofarasitcan,aproductacceptable to itself. It does this by training: by reinforcing the behavior itdesires and extinguishing others; by making some natural and socialstimuli into SD's, and ignoring others; by differentiating out this or thatspecific response or chain of responses, such asmanners and attitudes; byconditioning emotional andanxiety reactions to some stimuli and notothers. It teaches the individualwhathemayandmaynotdo, givinghimnorms and ranges of social behavior that are permissive or prescriptive orprohibitive. It teaches him the language he is to speak; it gives him hisstandards of beauty and art, of good andbad conduct; it sets before him apictureof the idealpersonality thathe is to imitate and strive tobe. In allthis,thefundamentallawsofbehavioraretobefound.

PersonalityandHistoryWhileallhumanbeingsobeythesamelawsofbehavior,eachindividual

ends upwith a unique behavioral equipment that defines his 'personality.'The emergence of uniqueness from uniformity is possible because of thevariations in circumstanceunderwhich thebasicprocesses areworkedout.The uniqueness stems from the reinforcement and motivational history of

the individualwhich is different from that of others.Wemay thinkof theindividual as a special (and very large) set of parametric values of basicfunctions.Thus,persistenceinthefaceoffailuremayvarygreatlyfromoneperson to another, depending upon the amount and variety of periodicreinforcement one has had in the past.Again, one may be more or lessdominantorsubmissiveinthepresenceoffriendsandstrangers,dependinguponhisexperiencewithotherpeople.

The fact that the individual's present personality is related to hisbiographyleadstoaninterestinbehavioraldevelopmentfrombirththroughinfancy, childhood, adolescence, adulthood, and old age. Students ofpsychopathic behavior know the importance of the early years in forminghabits thatmay last for life. In tracing behavioral disorders of adults theyareoftenledbacktochildhoodexperiencesasthesourcesofinstabilityanddisturbance. The importance of biography has long been known bypoliticians, fanatics, and all manner of special-interest groups, each ofwhom,inseekingtomanagesociety,soughteithertocontrolthefamilyinwhich thechildgrowsup,or togainoutrightpossessionof the child.Thetenacity of early training has been made the topic of bitter wit, as whenWilliam James said, "People often think they are thinking,when they aremerelyre-arrangingprejudices";or,asastudentoncewroteinatermessay,"A college education does not eradicate prejudices, it merely makes themmore subtle." He who would control society must ultimately reach theyouth, and there is psychological soundness in the old slogan, "Give methechilduntilseven,andyoucandowithhimasyouwishthereafter."Weneednotagreewiththecompletepessimismof thisopinion,norvouchfortheageseven,buttheideaisnotwithoutbasis.

Considerthe"Oedipuscomplex"ofpsychoanalysis.Thegeneralidea,asthe readerprobablyknows, is that theson tends to rebelagainsthis father,andwishestoovercomeordestroyhim,becausethefatherisapowerfulandpre-emptingrivalfortheaffectionofthemotherwhomthesoncovets.Thiscompetition with, and fear of, the father colors the whole psychologicaldevelopment of the child, andmaybe reflected inmanyways in the adultcharacter.The Oedipus relation was thought to be an innate, motivatingreactionofthesontothefather.Thisviewofthefather-sonrelationseems,however, to arise within families in which the father is the disciplinarianwho applies negative reinforcement. Study of other kinds of familyorganization,andinsomenon-Europeancultures,showsthat thefather-sonrelation can be quite different (Malinowski 1929, Kardiner 1939, Linton1945).Insomesocieties, thefatherhaslittleornodisciplinaryjurisdictionover the child, and acts mostly as a beneficent supplier of comforts andaffection.Itmaybethemother'sbrotherwhoactsasheadofthefamilyandadministers discipline. In this case, we would have an "uncle-Oedipuscomplex."Thereislittlereasontoresist theconclusionthat thepersonalitydevelopment of the individual is a function of his own conditioning

history.After the time when the role of the family is at its peak,formative

influences continue to work upon the maturing child through his friendsand his school. The currents and counter-currents of conditioning andextinction that stem from social origins do not stop. Joined with thereinforcementhistoryofanindividualaresuchfactorsashisgeneralhealth,physique, personal appearance, social and economic position.Other thingsequal,astrongandhealthychildismorelikelythanaweakandsicklyoneto develop physical self-reliance through the success of his running andjumping and climbing andpushing. Inour society,where 'good looks' arehighlyprized,ahandsomechildcanquicklyacquiresocialsuavitvandself-assurance (or a 'spoiled brat' character) because of the partiality andindulgenceshownbyhiselders.Personsofhighsocio-economic statusareoften the victims of obeisance and flattery—and what is flattery but theindiscriminate useof positive reinforcement?Obstinacy and arrogancemaybe the result.The 'cute' childwho is encouraged again and again to recite,totell jokes,ordance,maycontinuetodosoas longashewinsapproval,untilheendsupmakinghislivingatit.Withsuchexamples,wedonot,ofcourse,intendtooversimplifythefactorsthatcontributetoanypersonality.Thephrase "other things equal" seldomholdsoutside the laboratorywherethingsarekept equalbyexperimentaldevices. Itmustbe admittedatoncethatthedetailedanalysisofanypersonalityisanextremelycomplexmatterbecauseofthemultiplicityofco-actingvariables,pastandpresent.Inbroadoutline, nevertheless, we can see how such variables exert their force bydeterminingwhen reinforcement andextinctionwill be applied,how often,andbywhatschedule,andsimilarquestions.

Of great importance to the formation of personality, is the fact thathumanbeingscandiscriminate theirownactions,appearance, feelings,andsuccessfulness. In the course of growing up, the child comes to 'know'about himself; he becomes at least partially 'aware' of his capacities andweaknesses, his likelihood of winning or losing in given situations,hisphysical and social attractiveness, his characteristic reactions. This issometimesspokenofasthedevelopmentoremergenceofthe"Self,"awordthat ismeant todesignate theability tospeakof (be"aware"of)one'sownbehavior, or the ability to use one's own behavior as the SD for furtherbehavior,verbalorotherwise.ThesociologistMeadspokeofthe"Self"asasocial product, that is, it arises out of social interaction; but morespecifically,wecansaytodaythattheindividualistaughtbyhisfellowstorespond discriminatively to himself and his behavior. He can observehimself and judge himself with words like "good" and "bad." He canestimate his own efficacy as a social agent in pleasing people and instriving for social success; and if he discriminateswhat in his behavior iscausing failure, he may switch to new responses, that is, "improve" or"snapoutof it."The"Self," inshort, is theperson,hisbodyandbehavior

and characteristic interactions with the environment, taken as thediscriminative objects of his own verbal behavior. They are madediscriminative for him by his social community, as it teaches him hislanguage, because(a) his behavior is itself important to the community,and(b) it is also important to the community that he 'know' about hisbehavior. Byimportant, here,wemean that the smooth livingofboth theindividual and the community depends on his conduct toward others aswellashisdiscriminativeself-control.

We can deduce two things from the fact that the "Self" arises out ofdiscriminationtrainingandoutofverbalbehavior.First,thechildstartsoutin life without a "Self," and must build one up through stages of dimrealizationbyacontinuous learningprocess."Self"-learningprobablyneverends, nor is it ever perfect.There are times in lifewhen new requirementsare imposed upon the individual and new behavioral possibilities open uptohimforexploration.Atsuchtimes,the"Self"maycomeinforsearchingexamination,andmaygrowrapidlyastheindividualrecognizesnewtalentsandcapacities;oritmaysufferastheindividualfailstomeasureup towhatisexpectedofhimby thecommunity.Thus, inadolescence,asweusuallyconsider it, we see a transition (often attended by social rituals andceremonies) fromchildhood to adulthood,with the consequent expectationthat thepersonwill thereafter exhibit thebearingandbehaviorof anadult,and take on the adult's responsibility for his own behavior. Yet theadolescentdoesnotknowwhatanadulthas tobeanddo,and the learningof it is not easy in a society that does not planfully allow its children toassume adulthood by graduated steps. In our society, we speak of"adolescence" as though there couldbeonlyone, and that combiningbothphysical and social maturation. In truth, however, we prolong socialadolescence long after sexual adolescence is complete, and this createsproblems not necessarily encountered in societies that recognize thedifference. Other societies may have many "adolescences," or transitionsbetween recognizedstagesof social responsibility, rather thana single leapfromchildhood toadulthood.But, regardlessof thenumberof transitions,ateachonetherearenewlylearnedincrementstothe"Self,"aswellassomesloughing off of old parts; and while the emergence of the "Self" isexpectedtocomeinlargestepsattheseofficialtransitions,thereisinfactacontinualandprogressive flux in self-discriminationgoingonat all times.When the process is arrested or seriously distorted, societymay judge theindividual's lack of self-control or self-knowledge as pathological.Psychotherapy is then brought to bear upon him in the hope that, withspecialtutoring,thegrowthofthesocialized"Self"mayberesumed.

Aseconddeduction:apersonpossessingnoverbalbehaviorof any sortwouldnothavea"Self,"orany 'consciousness.'His reactions to theworldwouldbe like thoseofanyanimal, thoughhemightbemore"intelligent,"that is,hecould learnmore thingsandfaster thancould loweranimals.He

would go after positive reinforcements, and would avoid negativereinforcements, but would do so directly, without "reflection." To askwhether he "feels" this way or that, whether he"knows" that he is beingpained,whetherhe"realizes" thatwhathehasdone isgoodorbad,andsoon, isas idleas toask thesamequestionofarat.Withoutverbalbehaviorwhich is discriminatively conditioned to come out (i.e., to "describe," or"report"or"introspect")at theoccurrenceofstimuli,orat theoccurrenceofsome of his own behavior or behavior tendencies, there is no "consciousawareness"andno"Self."Itisnotthathepossessesthesethingsbutcannotspeakof them; it issimply that thequestionofwhetherhepossesses themismeaningless.Theyaretheproductsofverbalbehavior,notthecauses.Inthe lastanalysis, the"Self"and"consciousness"are thecreationsofhumansocietyoperatingontheindividualbymeansofverbaltraining.

In passing, it is perhaps worth noting that because the "Self" is aconstellation of discriminative operants, it is under some motivationalcontrol. We pointed out in Chapter 5 that high motivation may distortdiscriminationtoapointwhereadiscriminativeresponsemaybeemittedinthe presence of remotely similar stimuli or even in the absence of anystimulus.Somewritersspeakofthisas"autism"or"wishfulthinking."Wesee 'wishfulthinking'whenapersonspeaksof(i.e.,respondsto)his"Self"as though he had traits he does not really possess.Thus, he may say ofhimself"beautiful,witty,andwise,"whenhepossesses,not thesefeatures,butonlythedesire!

ConsistencyandIntegrationofthePersonalityBecausepersonalityistheoutcomeofreinforcementhistory,andbecause

an individual is capable of behaving in many ways, there arise the twoproblems ofconsistency andintegration. Behavior is always consistentwith the laws of behavior, but not always with logic.Amanmay act intwologicallyinconsistentwayswithrespecttothesamesubjectmatter,butifhedoessoitisbecausehehasbeenseparatelyreinforcedforbothwaysofacting. Thus, in one company, a man may decry greed or gossip orintolerance;yetatanothertime,inanothersetting,hemaypractiseanyoneof them. He has been reinforced for speaking one way, and for actinganother. It is as if different response systems, each capable of separatereinforcement and extinction, exist within a single person's behavioralrepertory,withtheseveralsystemsoperatinginwaysthatmaybeconsistentor inconsistentas judgedbysomeoutside logicalcriterion.Oftimes,whenan individualhashadan inconsistency inhisbehavioror attitudespointedouttohim,hewillsummonreasonswhichwesuspectarenottherealonesbut are only "rationalizations" for his inconsistency. Rationalizations areusuallyunsuspectedbytheonewhousesthem,butserveaseffectivesmokescreensfortherealcausesofhisconduct.

Overlapping the question of consistency is the broader concept of

integration of personality, or that harmonious functioning of all theresponse systems within the individual which is necessary for goodadjustment to his social and natural environment. Persons who arepsychologically "normal" have a certain oneness about their behavior, acontinuity or wholeness, that indicates coordination and cooperation (notnecessarilyperfect)among their several responsesystems.Theyseem tobe"in touchwith"mostof theirbehavior tendenciesorattitudes, in the sensethat they can speak of them on demand, or be taught to recognize them.Integration is not a clearly defined behavioral concept, but it figuresimportantly in the way clinical psychologists and other students ofpersonality think about the organization and functioningof an individual'sreactionrepertory.Inoneformoranother,clinicalevaluationsofpersonalityare likely to include observations upon the harmony of interaction ofwhatever "segments" or response systems have been identified in thepersonality.An integrated personality does not exhibit responses that areseverely out of jointwith the circumstances inwhich the individual findshimself, or out of joint with one another; nor is the individual entirelyunaware of the consistencies and inconsistencies among his responsetendencies.There isa threadofcontinuityandself-control running throughintegrated behavior that makes one feel that he is dealing with a single,articulated, unitary person. On the other hand, personalities said to bepoorlyintegratedgivetheoppositeimpression.Inthem,itseemsasthoughthe response systems have no contact with one another; as though therewerenoover-all direction to the individual's behavior that gets his actionsto cohere, or to correspond with what is called for in a given situation.Thus, a patientwith hysterical symptoms, such as functional paralysis, issaid to suffer from impaired integration; that is, his bodily and behavioralsegments seem to be going along independently, without relation to thepersonality's unity. Again, pathological cases of multiple personality,amnesia,somnambulism,andthelike,representfailuresofintegration.Itisdifficulttoconveytheintentionoftheconceptof"integration"becauseitisnot a rigorously definable term. Yet clinicians and psychiatrists find ituseful, andyouwoulddowell tobecomeacquaintedwith it if youexpecttoreadfurtherinthefieldofabnormalpsychology.Itshouldbemadeclear,however, that "inadequate integration is a developmental defect" (Shaffer,1936,page384).Wemightadd,also,thatitmayshowuponlywhensomeprecipitatingcircumstancebefalls theindividual(suchasemotionalshocks,severenegative reinforcement,or threatofpunishment) thathecannotcopewith.At such times, personality may disintegrate, producing the neuroticand psychotic maladjustments that are the specialized subject matter ofabnormalpsychology.

SocialMotives—AreThereAny?Theheadingofthissectionisputasaquestion,becausethatisprobably

the best way to approach the matter of social motives.The answer mustdependontheelementaryconsiderationsaboutdrivethatoccupiedusattheopening ofChapter 9. There we saw that the concept of drive has threesources:(a) an establishingoperation;(b) theeffectsof thatoperation uponthe momentary strength of reflexes apart from further operations ofreinforcement; and(c) thepossibilityof reinforcement that the establishingoperationcreates.It isagainstthesecriteriathatthestudentmustweighthestatus of any proposed "social" motive. To put the question a bitdifferently,weaskwhatisintendedbytheadjectivaluseoftheword socialasappliedtomotives.

On taking thought, the studentought to come to these conclusions: (1)Amotivemaybecalled social if its establishingoperationandappropriatereinforcement involve the withdrawal or supplying of social objects orstimuli.Thus, a female rat separated fromher litter for a timemayhave aresponse conditioned if return-to-litter is used as the reinforcement. Somehave argued that the social aspect of this motive is incidental to theunderlying factorofpain in the filledmammaryglandswhich is alleviatedby the draining of milk by the unweaned litter, and therefore that thesociality of the motive is specious or entirely derived from a non-socialdrive. Counter arguments have likewise beenmade against other proposedsocial motives such as gregariousness and sex. With further thought,however, youwill not fail to see that (2) thedistinctionbetweendrives associalandnon-socialisnotacriticalone.Classifyingdrivesasofonetypeortheotherisnotnearlysoimportantasrecognizingthatmanycommonlysupposedsocialdrivesarenotdrivesatall.

Whenwespeakcasuallyofaperson"strivingfor,"or"desiring," thisorthat, it is plain thatwemean there is somemotive present.The layman'serror,however,comes inhisassumption that the thingsstrivenforidentifythemotives.Take,asexamples,thedesiresforprestigeandsocialapproval,whichthelaymanmightproposetoincludeamonghissocialmotives.Thereflective studentwill sooneror later askhimselfwhat are the establishingoperations and drive-reducing reinforcements for the "prestige motive" or"approval motive"; and, failing to see an answer, will wisely decide thatthey are doubtful candidates.The many forms that prestige and approvaltake—awards,smiles,words,deference, invitations,money,andsoforth—areenoughtomakeuswonderinthefirstplacehowsuchdiverseitemsevercame to be grouped together. It is likely that they are all secondaryreinforcementsbecauseoftheircorrelationwithpositiveprimaryreinforcers,or because they stave off anxiety about possible punishments. Stripped ofthese correlations, the tokens of prestige and approval become empty, andineffective as controllers of behavior. The tokens, furthermore, probablyserve as the SD's for the next things to do in our enormously complexchainsofsocialactivity,sincetheyindicatewhatbehavior,successfulinthepast,isworthdoingagain.Here,too,whenthebehavioroccasionedbypast

success is no longer reinforced secondarily or primarily, the tokensextinguishasdoanySD'sthatarenolongercorrelatedwithreinforcement.

Other candidates for the position of socialmotive, such as the popularoneof "mastery,"are subject to the same reservationsmentionedabove. Ineach case, the difficulty of finding the establishing and reinforcingoperations, as well as the infinite variety of forms the "motive" takes,makesusreluctant toaccept them.Concerningdominance,cooperationandimitation,enoughhasalreadybeensaidtoindicatethatnoneofthesecanbesubsumedunderthedriveconcept.

Perhapsawordoughttobesaidaboutromanticloveasapossiblesocialmotive since it figures so large in our literature, art, and folklore. It iscertainthattwopersonsinlovehavemorebetweenthemthanthesexdrivealone,since for the latteranypartnerwouldserve;andanagedcouplemaybein lovewhensex isno longerafactor in their lives.Yetwhat it is, andeven how it is recognized by lovers, cannot easily be put down eitherintrospectively or by direct examination of specimens.Nevertheless,manypeople unhesitatingly call love a motive, partly because of our literarytradition that tells of its force. We may doubt that love is a motive,although the behavior involved still interests us.The dramatic case is thatof"loveat firstsight,"andwecanmakesomeguessesaboutit.Here,eachlover at once presents to the other an array of discriminative, reinforcing,and drive-arousing stimuli that are effective for him.These stimuli have ahistory,andthereisundoubtedlyalargeelementofstimulusgeneralizationor transference involved.The more of these each lover embodies for theother, thedeeper the attraction, until the ideal of the story-teller is reachedin the all-inclusive love.Thus, the bond that unites loversmay be tied attheirfirstmeeting,butitsstrandswereaweavinglongbefore.Thematchingoflife-historiesinthiswayisboundtobearareeventinhumanexperience,but it need not happen often for the ideal to suggest itself. Lessdramatically,lovemaybetheoutcomeofafelicitouslifespenttogether.Inthecourseoftime,eachofapaircomestopossesspositivestimulusvaluesthat may not have been present at the start and that give rise to mutualdevotionanddependence.

In either event, however, the stimulus roles played for each other bylovers are basically the same.Always, of course, the continuance of thesestimulus functions in lovedependson theultimateprimary reinforcementsprovidedorsharedbylovers, includingfoodanddrink,playactivities,restandsleep,reductionofanxiety,andthereliefofthesexdrive.Butwhatevertruemotivesbecomeinvolvedinit,loveisnotitselfamotive.

VERBALBEHAVIORIntroduction

No account of human behavior can be complete that overlooks man's

verbal activity. It ishishighest andmostvaluable formofbehavior;morethananyother, itdistinguisheshimfromloweranimals; in itare treasuredup man's cultural heritage of philosophy and science and art andtechnology, and by it the transmission of this accumulated knowledge iseffected from generation to generation. Indeed, verbal behavior has madesuch knowledgepossible. The layman may take his verbal behavior forgranted, but to the scientific mindthe forms, functions, and influences oflanguageconstitutephenomenaofthegrandestmagnitude.

Theanalysisoflanguage'sformsandfunctionshasbeenapproachedfrommany angles in years past. One after another, the grammatical, linguistic,semantic, and still other approaches have sought to clarify the origins anddevelopment of language, in the belief that thereby they could clear awayobstacles to thinking, avoid needless disputes of understanding, make forbetter communication among men, and provide a better basis for dealingwith human behavior. All of these varied attempts have in common afundamental assumption, or set of assumptions, about man and hislanguage which the science of behavior finds quite unacceptable. Thisassumptionholds that amanhas "ideas"whichhe "expresses" in language"inordertocommunicate"themtoanotherperson;hislanguageissimplyamediumwherewithheclotheshis ideassoas toproject themthroughspaceto someone else who, hearing the words, "interprets" or undresses thembackintotheunderlying"ideas."Theassumptionis thatamanisanagentwho manipulates words—finds the "right words"—for purposes ofcommunication, and that the words areindicators of ideas outsidethemselves. Such views are essentially mystical, as well as logicallycircular. In explaining verbal behavior, such views assert the existence ofthingswhichcannotbeeitherprovedordisproved, and forwhich theonlyevidence is the very language behavior which the things are invented toexplain in the firstplace.Thedual classificationofman into "mental" and"behavioral" components has been a stumbling block for theories of bothlanguage and behavior. Scientific psychology has profitably discarded thisdualism.

Youwillprobablyfindtheproblemofverbalbehaviorthemostdifficultone in this text.Themind-bodydualism is deeply ingrained in our habitsof thinking, and a newviewpointwhichdoes not comportwith the old ishard to achieve and tempting to resist.Nevertheless, it is in this directionthatwemustgo,and the lastmajor topicof this textwillbe toshowhowverbal behavior can be subsumed within objective psychological science.Our phrase "verbal behavior" covers all aspects of language—spoken,written,andgestural—butweshall limitourselvestothespoken,sincetheextension of our analysis to the other types would introduce furthercomplexityindetailbutnonewprinciple.

TheNatureofaBehavioralTheoryofLanguage

Verbal behavior (spoken) is composed of responses of muscles in themouth, throat, and chest which we may call the vocal apparatus. Theseresponses produce various combinations of sounds which serve asexteroceptive stimuli for the person hearing them and the one emittingthem. In addition, the responses produce proprioceptive stimuli in thespeaker that play a part in directing chains of verbal responses. Spokensounds,or themuscular responsescausing them,donot 'naturally' indicateobjects or events in the outside or innerworld.The fact that adult speechbears relation to the environment in a more or less lawful manner issomethingtobescientificallyexplained,ratherthantakenforgranted.Howsuch a correspondence arises is a central problem for analysis, just as it iswith an animal whose operant responses become conditioned andextinguishedsothattheyfinallyprovideacommercewiththeworldthatisbasedonitsactualfeaturesandrequirements.

Abehavioraltreatmentoflanguagewill,then,takeasitsdatathesoundsemitted by the human organism, just as it takes any observable behavior,like the rat's bar-pressing.When a person says something, our concern iswith thesaying and with the conditions thatcontrol the saying.Enunciation is an act, and to relate this act to its controlling factors is tounderstandthe'meaning'ofspeech.Ourtreatmentwillstartwiththeyoungchild and show the processeswhereby its sounds become transformed intolanguage. Our interest will not be in the historical origins of verbalbehavior in the human species, but in the genesis and development ofIanguage in the individual as an outcome of present and past variablesworkinguponhim.

That verbal responses were to be regarded as raw data was one of theremarkable things proposed over thirty years ago by John B. Watson(1919).Watsonassertedthatwhenahumansubjectinanexperimentspokeof his 'awareness,' or 'consciousness,' or 'perceptions,' it is a mistake tohypothesize thereality of these things.The subject's 'introspective' wordsdo not objectively 'report' anything, but should be recorded as additionaldata in themselves, thoughWatson was not clear on how they could beanalyzed in any useful way. He argued that inner mental events have noindependentobservable existence, thatwehaveonly the subject'swords todealwithasourdata,andthatthereisnonecessaryreasontoerectahumanpsychology on the supposed presence of an inner controllingpsyche.Watson'sviewswerereceivedwithhostilitybymostofhiscontemporariesbecausetheywereradicalforthattime,andbecausetheywerenotdevelopedto a stagewhere their fruitfulness could silence criticism.Moreover, asweshall see, they were wrong in one important way. Verbal responses dobecome correlated in the course of training with interoceptive andproprioceptive stimuli, so that verbal reports of inner events acquire somedegreeofcredibility.Abehavioralaccountofhowthesecorrelationsaresetup does not, however, depend on psychic forces that supposedly employ

words as a medium of communication.Watson was thus correct in hisgeneral approach, though he erred in detail.After all these years, we canmore justly appreciate the soundness of his thinking, and the acuteness ofthoughtwhichledtohisbeliefs.

There are many aspects of verbal behavior about which we have noexperimental data at present. Even so, the principles with which we arealreadyfamiliarenableustodissectverbalprocessesinawaythatconformsneatly with what is known about behavior generally. Perhaps the mostvaluable thing you will get from our discussion is a perspective and anattitude toward verbal behavior that will help you avoid some of themysteryinthisfieldandhelpyouthinkaboutlanguageinascientificway.

TheEmergenceofVerbalBehaviorinChildhoodVocalization alone does not constitute language, and the sounds an

infant makes are only the rawmaterial from which his verbal behavior isformed.At first, his repertory of sounds is limited, but as hematures therange of vocal responses increases. No one language makes use of all thebasic sound units(phonemes) that the human vocal apparatus can produce,but each selects out a numberwhichgo into itswords and ignores others.Thus, theEnglish languagemakes no use of a phoneme like thech in theGermannicht, or of the nasalizedon of French.The infant's repertory ofsounds grows rapidly from birth and includes more than his own nativetonguewill need, so that, in his language training, part of his repertory isretained and strengthened, and other parts extinguished. You may recallyour own amusement, and your teacher's horror, when you tried topronounceaforeignwordinhighschoolorcollegelanguagecourses.

It is, of course, of interest to knowhowphonemesbecomeavailable inthechild'srepertory,sincethisavailabilityisrelatedtotherateatwhichwecan expect children to acquire proper pronunciations.Data of this kind forEnglish-speakingchildrenwererecordedbyIrwinandChen(1946),inusualhomeenvironments,andrepresent thenormalphonemicgrowth thatoccurswithout special training in differentiation. Irwin and Chen determined thenumber of native-tongue phonemes spontaneously uttered by infants, andplotted the growth curve (Figure 77) for the number of types of speechsounds appearing at various ages.A total of 95 infants frommiddle-classhomeswasstudiedduringthefirst30monthsafterbirth.Atagivenvisit,achild's sample vocalization, as uttered on 30 respirations or breaths, weretranscribed in the International PhoneticAlphabet.The data are plotted fortime units of two months, so that the developmental curves for the 30monthsarebasedupon15points,eachpointrepresentingthemeannumberof types of speech sounds for the corresponding two-month period. Thecurveforallchildrenshowsthatfromthefirsttothelasttwo-monthperiodthe child passes from7.5 to 27 of the 35 sounds present in adult Englishspeech.

FIG.77.Theincrease innumberofEnglish-languagephonemesemittedbyinfantsatdifferentages.(AfterIrwinandChen,1946.)

In addition to variety of sounds, an infant's vocalizations may beanalyzed with respect to thefrequency of appearance of each type. It hasbeen found (Irwin, 1947) that, whereas the mastery curves for phonemetypes show a decreasing rate with age, the frequency of production is apositivelyacceleratedfunction.Thus,as thechildgrowsolder,henotonlymakes regularprogress inmasteringhisnative tongue'sphonemes,buthisuse of these sounds increases at a faster rate.While the available data foranygiven agedonot permit analysis of speech sounds intodiscriminativeand non-discriminativeoperants, it is plain that opportunities for elaborateverbalconditioningarepresentintheearlymonthsoflife.

Thefirstfewyears'growthofthechild'slanguageisswift,withsomanyfacetsdevelopingfromweektoweekthataclose,deliberateanalysisof theindividual case is rendered extremely difficult. Different workers in this

field have often used different schemes for classifying early language, theaforementioned phonemic count method being one possibility. Languagegrowth,however, runs throughwhat some investigators feel tobedifferentstages of early development that are glossed over in a simple count ofphonemes. Even these stages may be differently listed, but an illustrativelineupisthatofEisenson(1938):

a.Theundifferentiatedcry—occurringatbirthandshortlythereafter;thiscrying does not differ noticeably for hunger, thirst, noxious stimulations,andthelike.b.Differentiatedcrying—occurringafter thefirstmonthof life; thecries

differ, so that the cause may be discerned by a familiar observer like themother.c. Babbling—may begin at the end of the second month; there is a

polyglotphonemiccontent,weightedonthesideofvowels.d. Lallation—repetition by the child of his own vocal production,

leadingtoperseverativevocalization;beginsabout thesixthmonthorsoonafter.e. Echolalia—repetition or imitation by the infant of sounds made by

other persons; begins at about the ninth or tenth month and lastsindefinitely,perhapsthroughoutlife.f.Verbalutterance—theuseofrecognizablewordsinresponsetostimuli

or to control the actions of people; usually begins in the first half of thesecondyear.

The flow of speech development is, of couse, divisible into as manystages as we wish, and only practical or theoretical considerations caninform us when we are to stop.At the present time, any segregation ofstages is of limited fruitful.ness, although all attempts of this kindemphasize the large changes going on in the early years.That all childrengo through roughly similar changes is itself a provocative observation,since it indicates that there are probably uniform underlying processes atwork. For example, there is the long-known fact that congenitally deafchildrendonotlearntospeak.Theyreachthebabblingstage,andprobablygettoalimitedlallation,butneverprogressfurther.Apparentlythehearingof speech is a condition for language development. Special coachingmethods have been worked out, however, for teaching deaf children tospeak,andthereareafewinstitutionsforthispurposeintheUnitedStates.Thesemethods represent theeducationalartatahigh level,buta scientificsurveyof theprinciples involved couldwell lead to a better understandingof languagebehavior aswell as possible further refinement in themethodsthemselves.This is almost unexplored territory awaiting the attention oftrainedpsychologists.

TwoFunctionsofLanguage:TheMandandtheTactPhonemicavailabilityandthelistingofdevelopmentalstagesdonot,of

course, give us information about some things. More important for ourunderstanding of verbal behavior are theprocesses by which it islearned,and theconditions under which it isused.The forms and functioning oflanguage are our central problems.We take the operant speech repertory asthe rawmaterial onwhichour principles operate, and askhow the specificcharacteristics of verbal behavior arise.Wenote, first, that the individual'sspeech results from the training given him by those who make up his"verbal community." Without a verbal community to teach him, anindividual would never develop verbal behavior. In asking how theindividual acquires speech, therefore, we shall find ourselves dealing withotherpeoplewhoarehis"hearers,"andwhoaretheinstrumentsfortraininghisvocalizations.

Given, then, the phonemic rawmaterial, the principles of conditioningandanappropriate teacher,ouranalysisof languagecangetunderway.Wemayidentifytwobasiccategoriesofverbalbehaviorwhichappearearlyandremainfundamentalthroughoutanindividual'slifetime.Thesearethemandandthetact.

1. The mand (from the Latinmando, to order or command) is anutterancewhichprocuresaspecifictypeofreinforcementfromthehearer.Achild can reach for his milk or he can make a soundmilk and the tworesponsesarenotdifferentinanyfunctionalway.Botharereinforcedinthesamemannerexcept that in the lattercase thereinforcement ismediatedbyanother organism. The mand as a response is strengthened by itsconsequences just as any operant such as bar-pressing, chain-pulling, orrunning through a maze. In the grammarian's terms, the mand is in theimperativemood, and it includes demands, entreaty, commands, requests,and so forth.Mandsareprobably the first functional elements to appear inthelanguagebehaviorofthechild.Theyaredevelopedundertheinstigationof primary drives and are the first verbal responses to be reinforced byhearerslikethemotherornurse.Whenaninfantishungry,forexample,hisgeneral activity, including the vocal, is heightened, and different sounds,interpretedby themother as approximating intelligiblewords, are fasteneduponby her for differential reinforcement. She proceeds to condition thesesounds by supplying milk when the infant says something vaguelyresemblingmilk,supplyingatoyforasoundliketuh,andsoon.Jespersen(1922,pages154-155)hasputitinthesewords:

... The baby lies and babbles... without associating the slightestmeaningwithhismouth-games,andhisgrownfriends,intheirjoyovertheprecocious child, assign to these syllables a rational sense.... It is verynaturalthatthemotherwhoisgreetedbyher...childwiththesound'mama'should take it as though the child werecalling her 'mama,' and since she

frequentlycomes to thecradlewhenshehears the sound, thechildhimselfdoeslearntousethesesyllableswhenhewantstocallher.

The conditioning of mands, therefore, is one in which theform of theresponseisspecifictotheparticularreinforcementobtained.The'meaning'ofamandisgivenbywhatconsequencesfollowitsemissioninaparticularverbal community. Mand emission is controlled by the organism's needssince thesemands have a history of appropriate reinforcement under thoseneeds. Mands are not originally discriminative operants, but they usuallybecomediscriminativebecause the likelihoodof reinforcement isgreater insome circumstances than in others.Thus, manding food or drink usuallyoccurs in thepresenceof certainpeopleand incertainplaces.Nevertheless,aswith other discriminative operants, if the drive becomes strong enough,theymaybeemitted,evenbyadults,intheabsenceofthecustomarySD's.Sometimes, by extension or generalization, manding will occur insituationswhere the localSD'shavenever themselvesbeen reinforced.The"magicalmand"isofthisvariety,anexamplebeingthatofthebowlerwhocallsafterhisswervingball"getinthere!"

We have said that mands are conditioned by the verbal communitywhich gives specific reinforcement to specific sounds or to words of aspecific form.The sounds or words selected from the child's repertory formandtrainingdependontheconventionsoftheverbalcommunity, that is,thenativelanguage.Thereisnothinginherentinthesoundsthatnecessarilyplacestheminthemandcategory.Anywordscanserveasmandsiftheyaresoconditionedandiftheyfunctionthatwayuponhearers.Mandsaremadeso by their past history of specific reinforcement, and they disclose theircharacterinthebehaviorofthehearerwhosereactionsaretherebycontrolledin specific ways. In the case of high drive or stimulus generalization,manding responses may appear that are not actually effective in procuringreinforcement, but the behavior is traceable to prior history of the sortwehaveindicated.

2. The richness and versatility of language in human life would begreatly restricted were its content limited only tomands. Of greaterimportance is thetact function of verbal responses(tact from the pastparticiple, "tactus," of the Latin, "tango," to touch).Whereas themand isonly incidentally under the control of SD's, and is a responsewhose formdetermines a specific reinforcement, the tact is of a different complexion.The tact relation isone inwhich theformof theverbal responsewhich thecommunity reinforces is related toparticular discriminative stimuli, andthe response is given, not specific reinforcement, butgeneralizedreinforcement. We may think of tacting as a 'naming' function: if thespeaker emits the required sound when a given SD is present, he isreinforced.Thetactisnotmotivatedbyaspecialneedofthespeaker,andit

does not call for a special reinforcement. The correlation between theresponse and its proper SD must, however, be reinforced somehow, andwhatwe observe is that the reinforcement provided by hearers is ageneralone. This reinforcement may be in the nature of smiles, approval, ormoney, all of which have in the past been associatedwithmany types ofprimary reinforcement and are now, on that account, secondary positivereinforcers operating effectively in a wide range of situations. It has beensaid that the notion of generalized reinforcement is probably the mostimportant single characteristic of verbal behavior, since it gives the verbalcommunityalmostunlimitedpowertotraintheindividual.Thevitalresultof this training is that the verbal responses of the individual are made tostand in a dependable relation to the environment. In the grammarian'sterms, the tact relation is that of the declarative sentence; it is anannouncementof"fact" representingrelativelydisinterestedbehavioron thepart of the speaker, behavior for which he gets nothing in particular butonlysomethingingeneral.

Ofall theverbalbehavioracquiredby the individual, the tact relation isof the deepest value both to himself and to the community. By teachinghim to correlate words with facts and states of affairs, both inside andoutside himself, the community opens up for him the opportunity toparticipate in,and contribute to, human discourse and human wisdom. Ifhuman languagewere limited tomandsalone, itwoulddiffer little (exceptfor size and clarity of the vocabulary) from the grunts and barks of loweranimals.Withmandsalone,wewouldneverriseabovethelevelofmakingour personal needs and states discriminable to other persons who mightthen(weretheyreinforcedfordoingso)acttoalleviateourhungerorthirst.It isthetactrelation,however, thatmakesmanabletospeakof, to 'know,'andbeableto'think'about,theworldandhimself.

TheSpeaker-HearerRelationInanyverbalcommunity,theroleofspeakerchangeshandscontinually,

asdoesthatofhearer,sothatagivenpersonisnowoneandnowtheother.Ananalysis of verbal behaviormust cover this double-endedprocess ifweare to glimpse the full scope of verbal functions.A key question we cantake off from is, what reinforcement does the hearer afford the speaker forspeaking, and what is afforded the hearer for listening and reinforcing thespeaker?Who profits, and how? Clearly, both speaker and hearer mustprofitiftheyaretoactaseither,butwemustseehowthisiseffected.

The speaker's speaking is reinforced by the hearer in two ways. Hismands are specifically reinforced according to his needs, while his tactingsecures generalized reinforcement from the hearer who thus "encourages"him togoon.The hearerhashisown reasons for listeningand reinforcingthespeaker. Ifhe reinforces thespeaker'smands,hegets reinforcedhimselfinoneorbothoftwoways:generalizedreinforcementfromthirdpartiesfordoing the "proper thing," and somewhat more specific, though delayed,reinforcement from avoiding trouble caused bynot meeting the speaker'sneeds. A mother, for example, responds to her child's mands as she is"expected to"by thecommunitywhichrequiresher tobeadutifulmother;but shemayalsodo sobecause she therebyprevents such later troubles ascopingwithanundernourishedorinjuredchild.

An interesting problem, however, iswhy a hearer reinforces a speaker'stacting.Hearers reinforce tacting because theymake use of the informationtheyobtainfromthe tacting.Thus,aspeakermaysay"there is food in therefrigerator,"or"fire!"withresultingreinforcement to thehearer ifhebaseshis ownbehavior on these tacts.The hearer gets an ultimate reinforcementfor giving a speaker immediate reinforcement, since good tacting providesguides or SD's for the hearer's own successful responding. Each hearerobtains frommanyspeakers inhiscommunitymuchmoreknowledge thanhecouldobtainbyhimself;eachpersoninacommunitybenefitsallothersbyactingasadditionaleyesandearsforall,andthisisastrueforopinionsandmeditationsasforstraightreportingofenvironmentalfacts.Becauseofthis mutual profit, the community plenteously reinforces tacting in theverbal behavior of its members. Note that we are considering the"informative or communicational value" of speech, without resort to thedualisticnotionthatspeechismerelyamediumforexpressing'ideas'inthe

'mind.'Tactingresponsesareacquiredaccordingtoconditioningprinciples,andthebehaviorofhearersmaybeexplainedinthesameway.

We have pointed out earlier that verbal behavior would never developwithout a verbal community whichmediates the reinforcement for it.Thefact that reinforcement isat thedisposalof thehearermeans that theheareris the one whoacts as the teacher of speech.Any speech response is acomplex pattern of muscular action in the vocal apparatus, and as we allknow it takes a long time to acquire acceptable pronunciations andinflections, in addition to grammatical forms and sentence structures.Involved here is the matter of response differentiation, for which theresponsibility lies in thehandsof thehearer-teacher.As the childmatures,hearers use the devices of selective reinforcement for, and successiveapproximationsto,thedesiredsounds.Bythissocialinsistence,thechild'sspeech is gradually molded into accepted forms. "It is notyeah, son, butyes—can't you say it properly?",we tell our youngsters.Orwemay growangry and threaten to withhold some desired thing unless the childabandons his baby talk and speaks as "a big boy should." Furthermore,children are soon conditioned to imitate the sounds of adults so that theactiveroleofthehearerasadifferentiatormaybereducedtothepointwherehe is unaware that hehasdone anything.Youhaveundoubtedlyobserved,asdidMarkTwain, thatFrenchchildren learnFrenchandChinesechildrenChinese,withoutanytrouble!

An interesting and extremely important variety of tacting is thatwhichhasinternal stimuli as the discriminative source. From childhood on, theindividual is taught by the community to report to hearers on hisinteroceptive and proprioceptive stimuli.The adult can, for example, tactthe locationof pain,whereas the childmaybe in obvious distress andyetunable to say even what part of his body hurts. Many expressionsbeginningwith"Ifeel. . ."aretactsofinternalstimuliandstateslikepain,sleepiness, fatigue, tension, hunger. Such tacts are an essential source ofinformation about goings-on inside the individual which are otherwiseinaccessible to an observer. To the community, this information isnecessary for meeting the individual's needs, assuring his survival, andpredicting what he is going to do. The individual learns such tactingbecause it is heavily reinforced: if he says his tooth aches, it can berelieved; if he says he ismelancholy, he can be cheered; if he speaks hisfear,hecanbecalmed."Introspective"reportsoffeelings,thoughts,andthelike, are what have led to the dualistic conception of an internal mind asseparatefromthelanguageexpressingit.NoonedoubtsthathiddeninternalSD'sforsuchtactingexist,andthatasaresultoftrainingthetactsdo serveto give information about them.But a scientific theory of verbal behaviortakesthetactrelationasadatum,andrefers itsorigintoknownprinciples,rather than to some mind-agency within the individual that somehowknows which word should go with which stimulus, and thereupon sorts

them into meaningful speech. The dualistic view takes the existence ofreferents for words, the correspondence of speech with fact, and theconveying of information to hearers, as matters that need only to belogically or grammatically formalized. The science of behavior askshowthesethingscomeabout.

Youmayaskatthispointwhatmakesusdecidethatsomeintrospectivetactsarevalidandothersnot.Why,forexample,dowequestion"mind"ifapersonsays,"Mymindisworking"?Tothisandmanysimilarquestions,wecanreplythatthereisnodoubtthespeakerissaying something,butheisnotnecessarily tactingwhathesupposeshe is.Thus,apersonwhosays"I am thinking" does not have ghostly thoughts as his SD, but rathermuscular responses of the vocal apparatus (and perhaps other parts of thebody) which he tacts as "thinking."A child wrestling with a problem istold by his parent "My, but you're thinking hard, aren't you?" and hegropingly replies "am Ithinking?"; thereafter, similar activitywill also becalled thinking although he learns in later life to make vague statementsaboutthe"natureofthought."Weallthink,butwedosowithourmuscleswhich provide the only SD's for the tact "thinking." If we are asked,however,todescribe thinkingwedosoincorrectlybecausewehavelearnedtotalkaboutthinkinginacertainway.Wemaysaythatwe"feelourbrainsworking,"butthisisnotavalidtactsinceneuralactivityinthebrainisnotitselfperceptibleandnodiscriminative response can be anchored to it.Wecouldexamineinthismannerall themisconceptionsabout themselvesthatmenevinceintheirverbalresponding.Forthemoment,itwillsufficeifthebasic viewpoint is grasped to some extent, for this is perhaps the mostdifficultreorientationofallforbeginningstudentsinpsychology.

DistortionsoftheTactRelationIdeally,thetactrelationwouldcontainaresponsemadediscriminatively

toasingleSD.Inactuality,thisrelationissubjecttomanydeflections,andsome of the more important kinds may be noted. In the following cases,you might provide your own running commentary on the socialacceptability of the distortions, and of the way they are learned,extinguished,orpunished.

1.GeneralizationsoftheSD.Since the tact is a discriminativeoperant,generalization of the SD is naturally to be expected. Such generalizationsmayhavepractical use.Theyget information to a hearermorebriefly thanotherwise, or perhaps more comprehensively than by direct statement,becausethehearer'sownbackgroundofexperienceiscalledintoplay.a.Metaphoricalextension:thisisusednotonlybypoetsbutinordinary

conversation."Thinasareed,"or"strongasahorse,"or"fastasadeer,"arecommonplace similes. Contractions may occur if the speaker is sure his

hearerwillunderstand,aswith"sheisaflower."b.Generic extension: here the generalization is sanctioned by the usage

ofthecommunity.Wespeakof"players"indifferentsportsandgames,yetsomeareprofessionalswhoareclearlynot'atplay.'

c.Redintegration:aformofgeneralizationinwhichapartofastimuluscomplex is used to stand for the whole (here inChapter 5). "Countingnoses," "reciting Shakespeare," are such expressions in which noses standforpeople,andonerecitesonlywhatShakespearewrote.

2 .False and non-existent SD's. These are instances where a speakerpretends to tact although there is no SD for it, or where the speaker tactsincorrectly.a. Lying, or contrary-to-fact: thismay be done to avert the punishment

thatatruthfultactwouldincur."No,Ididnotbreakthat,Mother."Afalsetactmayalsobeusedtoobtainspecificpositivereinforcement."IgotanAinarithmetictoday,"saysJohn,ifthatistheonlymethodhehastogethiscandy. Of all people, we depend on our scientists to be truthful tacterswithout specific rewards or inducements. Imagine the catastrophe if theyhad to fabricate data in order to eat.A scientist must be free of pressuresthat turn tacts intoconcealedmands, and society setsbeforehimstandardsofhonestythataremoreheavilyreinforcedthanstandardsofsuccess.b. Exaggeration or invention: this is a way of obtaining generalized

reinforcement fromanaudience thatwouldotherwisewithhold it.Wehaveall listened to tall stories, embellished autobiographies, and dubious talesofpersonalconquest.c.Distortionsthatproducespecialeffectsonthehearer:tellingjokesand

'tear-jerkers,' 'crying wolf to make a person do something we want, andtellingstorieswithamoralareinstanceswherethehearer'sbehavioristobemanipulatedinspecialways.

Alistingliketheabovedoesnotavoidsomeoverlappingbetweentypesof tacting. Its purpose is to point updifferentways that tacting appears inthe stream of speech activities, so that you may get a feeling for theprocessesinvolved.

OtherControllingFactorsinSpeechWhereas mands are of value most directly to the speaker whose needs

they serve, the tact relation is of greater value to the verbal community.Tacting is thequintessenceof socialbehavior,andofallverbalbehavior itis thepartwearemostconcerned tounderstand.Wemay, therefore, takeabitmoretimeonsomefactorsthatcontrolresponsesofthetactclass.Someofthesefactorsmaybeobvious,buteventheyareworthmakingexplicit.

1.Textualfactors.Verbal responses aremade to certain stimuli bywayofreading, that is,wetact themarksonpaper thatmakeupwriting.Here,the form of the response has a particular conventionalized relation to the

written symbols. Schools devotemuch of their curriculum to the teachingof reading. Moreover, some written matter gives directions for continuedverbal responding by the reader—for example,mathematical symbols like

andwe then get into the problem of how suchfollow-upresponsesaretaughttotheindividual.

2.Echoicresponses.Wementionedearlierthatthereseemstobeastagewhen the child begins to imitate or echo the soundsmade by others. Butevenbeforehe learns tospeakwords,hisbabblingacquires the inflectionalpattern of his native tongue, so that he babbles English, as it were, evenbefore he can speak it.These observations indicate the importance of theheardsound.Lateroninlife,itiseasytoevokeechoicbehaviorbysimplyinstructingaperson, "repeat afterme."Moreover, it is likely that all ofustend to repeat speechweare listening toevenwithoutbeing told todoso.Wemurmurtoourselvesasthespeakertalks;weusehiswordstointroduceour own sentences; we go alongwith him and complete his sentences forhim.Inanexploratorystudyofthisproblem,Ritter(1949),usingadultsassubjects,hadthemmemorizenonsensesyllables.The instructor thencalledout one of these syllables, with the subjects instructed to respond asquickly as possible with the first syllable they thought of. They couldrespondwith anyof thememorized syllables including theone calledout.Thedatashowedanotabletendencyforthemtorepeattheheardonedespitethe fact that they had a number of other well-learned choices available.Echoic control over verbal responding seems real enough, butwe need toknowmuchmoreabout itbeforeweunderstand it fullyorcanestimate itsmagnitude.

3.Audience.The speaker's audienceprovides an immediate control overhisverbalresponding.Agivenpartoftheverbalrepertorymaybeactivatedby one audience and left untouched by another.We would not address aministeraswedoourfraternitybrothers—thetoneandcontentalikeofourspeakingwoulddiffer.Recallthatahurtchildwillcrymorewhentherearepeople about than when alone, and still more when, of all people, itsmotheristhere.Whereaudiencecontrolofspeakingfailsconspicuously,wehaveasymptomofpsychopathologicalbehavior inwhichverbaloutput is,wesay,not'incontact'withtheimmediatesocialsituation.

4.Motivation.Variations in drive strength affect the strength of verbalresponses just as they do any other kind of behavior.This is quite clearwith respect tomands, but tacts are also susceptible. Slips of the tongue,unwitting puns, and so forth, in a speaker's output often indicate somemotive at workin him that has taken priority over the usual SD's. Freudhas made extensive use of this observation, but it would be unwise toconclude thateverysliporpunhasamotivationalbasis.There ismultiple

causality inspeech,andsometimeserrorsanddistortionsoccurbecause thetongue is sidetracked by similar phonetic words or by the intrusion ofrelated themes from the speaker's history.An invention (Skinner, 1936)called the "verbal summator" brings out many of the factors mentionedhere.Combinationsofvowelsareplayedfromrecordingsatalowloudness,eachcombinationbeingrepeatedanumberoftimes.Thesubjectistoldthat"thisisatestintheclarityofspeech"andisaskedtolistencarefullyandtoput down, as soon as he thinks he has detected it, "what the man issaying." Of course, the man is not actually saying anything but a vowelseries like "uh-ee'-uh-uh-uh," but this acts to select out of the subject'srepertory words or sentences that are, for him, the stimulus. He reports"hearing" the man say such and such, but what he really hears ishimselfsayingsuch and such in response to the meaningless sounds of therecording. It is as if the repetition of incoming stimuli builds up thestrengthof the subject's latent speech, and soon those items that are closerto the emission threshold "come to the surface" and are spoken out.Theverbal summator is related to testing methods which have been called"projective techniques" in the study of personality. Different subjectsrespond in different ways to the same meaningless stimuli, and theirresponsesareoftenrevealingoftheirmotivesandtheirhistory.Tabulationsofresponsefrequenciestothesummatoroftenshowthat themostcommonresponsesrefertotheoppositesex,toworries,thewishtogohome,andsoon. When normal persons are compared with mental hospital patients(ShakowandRosenzweig,1940)grossdifferencesappear in their responsesto the summator.This is not wholly unexpected, since we already knowsomethingofhowthesecontrastedgroupsperformonmanyothertests.Theinvestigation indicated, however, that the summator technique deserves tobe explored more thoroughly in the analysis of normal speech andpsychopathicdeteriorations.

FurtherCharacteristicsofVerbalBehaviorBroadly considered, verbal behavior is subject to greater delays in

reinforcementthanaresomeotherwaysofresponding.Mands,forexample,require action by hearers who mediate the reinforcements, and often thespeaker could get his reinforcement more quickly by reaching or movinghimself. Children must learn to ask for bread rather than grab for it.Moreover,inbothmandingandtacting,wemayfrequentlybedisappointedby our hearers who fail to supply any reinforcement whatever. Even ourpoliterequestsmayoftenberebuffed.Theresultisthatweoftenspeakwithlessassurancethanweact.

Vocal language has special advantages for the range and utility of ourverbal behavior. There is no necessary reason why language could notoriginally have been gestural ortactile.As we may conceive it, however,thespeechtypewasselectedoutowingtoitsgreaterversatilityandvaluein

procuringreinforcement(perhapsbyleavingthehandsfree).Constructedaswe are, olfactory, heat, or radiational language would not be an effectivebasisforaverbalcommunity.Sightorhearingwouldbemorelikelybasesforinterstimulationatadistance.Ofthesetwo,however,alanguageofseenmovementsislimitedbythenecessityoflookingandbyitsineffectivenessin the dark. Moreover, a vocabulary of movements-to-be-seen would berestricted to thenumberofdiscriminablydifferentmovementsmadeby the'talker,'andthisnumberisnotlarge.Finally,suchavocabularyisnoteasytoexecutebutrequiresconsiderableexpenditureofenergythatisnegativelyreinforcing tosomedegree.On theotherhand,speechgetsaroundmost, ifnotall,thesedifficultiesandis,inaddition,capableofanindefinitelylargevocabulary. Responses of the vocal apparatus can be differentiated veryfinely to produce an enormous number of sound combinations that arediscriminablydifferent to thehearer.Speechmovementsaresmallandcostlittleenergy,theycanberapidlymade,theirintensitycanbevariedaswellas their pitch, they are easily combined into chains, and so on.We maythink of language functions as having located themselves in the vocalapparatus by virtue of a kind of natural selection that we can describe intermsofreinforcementprinciples.

The first years of language development in children are commonlycharacterized by the fact that their verbal behavior isaudible, that is,childrendoalltheirtalkingaloud.Itneedsschoolingandsocialpressuretoquiet down their speech so they do not say aloud everything that 'comesinto their heads.'Even as seemingly simple a thing as silent reading takesupagooddealofteachingtimeintheearlygrades.And,evenasadults,wemay talk aloud to ourselves when we are alone; and, by the same token,mental patients showing 'withdrawal from society' (i.e., a loss of controlover behavior by social SD's) talk to themselves aloud without regard forthepresenceofothers.Silentspeechisanimportantpartofthinking.Whenengaged on difficult tasks like mathematical problems, we are likely toabandon silent speech and revert to 'thinking aloud.' Perhaps thesuppression of talk is not altogether good for our intellectual efficiency. Itdoes,ofcourse,concealfromourneighborsthingswedonotwantthemtoknow.Yet, fromthestandpointofmentalhygiene, itmaybebetter to talkoutourfeelingsthantoletthemrankle.

In general, speech uses only the smaller muscles of the body, and insilent speech or thinking still smaller movements of these small musclesare involved. But man's speech is a mighty lever, and onemay truly saythatneverdidsuchsmallmusclesmovesuchlargeworlds.

ConsciousnessandUnderstanding

VerbalbehaviorcanitselfactastheSDfor furtherverbalbehavior.Thatis,wecantalkaboutourtalking,orthinkaboutourthinking.Wemayuse

the termsecondary language to designate verbal behavior that has verbalbehaviorasitsSD.Aswehavesaid,"consciousness"isprobablyreducibleintheendtotheabilityofapersontoverbalizeadequatelyhisownactions,including his prior verbal responses.We say he is 'oriented' or 'self-aware'whenhecantacthisownbehaviorinaboutthesamewaythatwewould.

Wearesatisfied,also, thatapersoncomprehendsorunderstandsus,notif he echoeswhatwehave just said, butmorebyhis ability to say it in adifferent way.Merely to repeat our words verbatim tells us only that ourown verbal chainswerememorized. If he can startwith ourwords and goon to further statements that we ourselvesmightmake, we concludewithmore confidence that he has understood us. Understanding is a verbalactivity, for the person who understands us talks as we do. Like allbehavior. understanding needs reinforcement if it is to bemade firm.Thisbeing so, onemay examine various forms of instruction used in schools,for example, to see which provides the best reinforcement forcomprehension, that is, proper speaking or thinking. From such anexamination, the lecture method of teaching, so common at the collegelevel, emerges with a poor recommendation.The lecture method providesno reinforcement for the hearer's speech except the long-delayed one ofexamination and final grade. It is not to bewondered that,with the smallcomprehension achieved by lectures, students often aim on examinationsonly to reproduceverbatimthe lecturer'swords.Thenotes thathe takesareameansof stimulating talk inhimself afterhe is alone (akindof lecturer-student conversation without the lecturer present), but can only partiallysucceedinfillingthegapleftbytheabsenceofimmediatereinforcementforthe hearer's talking. On the other hand, the laboratory method ofinstruction, and the discussion-group method, hold out the possibility ofsuperiorlearningiftheyarewellused.Onedangerindiscussiongroupsthatarenotskillfullyconductedisthetendencytoreinforceeverydiscussantforsaying almost anything as his "contribution," so that no one carries awayanymorecomprehensionthanhebroughtwithhim.Butthelecturemethodinherently promises little for any but selected audiences already so welltrained in a special subject matter that the heard words fit into, andgeneralizewith, a verbal repertory prepared in advance and primed at highstrengthwhilelistening.

Problemsof individualunderstandingare joinedwith thoseof commonunderstanding in the community, that is, with verbal usages that areexpected to have a commonmeaning for all persons.The "meaning" for aspeaker of a word or sentence is, of course, defined by the sum total ofconditions under which it is emitted; while the "meaning" for a hearer isdefined by the behavioral consequences it induces in him.As backgroundand training differ or agree for different speakers, the conditions evoking agiven word from them will overlap but perhaps not coincide exactly.Similarlywiththepartialcoincidenceofreactiononthepartofhearers.The

study ofsemantics centersonquestionsofmeaning like this,and it seemsprobablethat thebehavioralandsemanticapproachestolanguagewilldrawclosertogetherinthefuture.

It is interesting, finally, to conjecture what the impact may be uponphilosophy when the behavioral analysis of language reaches a moreformidable stage than at present.Language is, in away, both themediumand substanceofphilosophy. Itwouldbe strange indeed if an inquiry intothenatureandoriginsofverbalbehaviorhadnobearingupontheimportorvalidityofthatsameverbalbehavior.For,evenifitistruethatanideamaybevalidwithoutourknowing its source, it is also true that inmanycaseswe cannot judge its validity except from its source. In like manner, itwould be strange if our knowledge of the medium in which philosophersspeak, take issue, or agree with one another, did not help us also knowwhen their talking is soundand fury,andwhen it signifies something.Noone can say how far in the future lies the first strong impact of behavioralscience upon philosophy, but that itwill come seems certain. In a sciencethat takes the whole of behavior as its province, what part of man'sactivities shall be said to lie out of bounds and be exempt from scrutiny?Whocan justlydenyher the rightofpassage throughanymeadow,andonwhatbasisdeclarethatshetrespasses?

NOTESSocial psychology is such a large and inclusive province that there is

often little similarity in the content of textbooks within the field. Aninteresting approach is that provided by Klineberg'sSocial psychology(1940), the material of which may be used to supplement some of thetopics we have treated in this chapter.A handbook of social psychology,edited byMurchison (1935), is also to be recommended, especially for itsdescription,byvariousauthorities,ofsub-humansocialbehavior.

Anenlighteningexperimentoncooperativebehavior inchimpanzeeshasbeen carried out by Crawford (1937).This investigator's exposition lendsitself readily to the type of analysis you have been using, and supplies asplendid approach to the complexity of human cooperation. Maslow's(1936a, 1936b) observations on the dominance-submission relation amongbaboons serves a similar purpose. With respect to imitation, we havealready mentioned (page 361) the studies reported by Miller and Dollard(1941).

In our treatment of verbal behavior,we have leaned primarily upon theanalysisrecentlypresentedbySkinner(1948b)inaseriesofformallecturesat Harvard University. Skinner's approach amounts to the first importantextension of reinforcement theory into the realm of human languagefunction,andisanimpressiveattempttofreetheproblemoflanguagefrommind-bodydualism.

ALASTWORD

Thepowerofscienceisnowhereinhistorymorestirringlyrevealedthanin its first victories over themost difficult and defiant subjectmatter everpresented to it—the behavior of organisms. This book has attempted toplot, in general fashion, some of the paths that have been opened up inman's study of his own behavior and that of his fellow-creatures. Startingwith the humble analysis of simple bits of behavior, modern psychologyhas extracted a few shining conclusions that dispel somewhat the mistswhich once shrouded human and animal behavior and made anunderstandingofitslawsaseeminglyimpossiblegoal.

All behavior, aswe can nowdiscern it, is composed of variations on afew basic themes. For the first time in mankind's saga, these themes areopentoallwhowishtoseetheminthesteadylightofscience,rather thanby the rare illuminations of intuitiveminds.We are on the frontier of anenormouspower: the power tomanipulate our ownbehavior scientifically,deliberately, rationally.Howthispowerwillbeused—whether forgoodorill—nooneofuscantell.Certainit is thatwhateveruseismadeofitwillbedeterminedbythecharacterofthepersonsusingit.Butcharacteritselfisopen to a science of behavior.Weneed to hasten and train a generation ofmenofgoodwill.Howthis is tobedonemaybemankind's lastdesperatequestionofall.Withoutascienceofpsychology,noanswerispossible;butpsychology,whileofferingthemethods,cannotensuretheiruse.Itistothelatterthatwefinallycommendourreaders.

REFERENCESANDAUTHORINDEX

Thenumbers in, italicsat theendofeachreference indicate thepages inthistextonwhichthereferenceiscited.

1. ALLPORT, F. H. (1924)Social psychology. Boston, HoughtonMifflin,xiv,453PP-—35

2. ANDERSON,O.D., and LIDDELL,H. S. (1935)Observations onex-perimental neurosis in sheep.Arch. Neurol. Psychiat.,Chicago,367-385.—18

3. ANREP,G.V. (1920) Pitch discrimination in the dog. J.Physiol.,53,367-385.—18

4. ANTONITIS, J. J. (1950)Variability of response in the white ratduring conditioning and succeeding extinction andreconditioning. Ph.D. dissertation, Columbia Univ.—181, 182,183,185ANTONITIS,J.J.,seealsoNos.226,227.

5. ARNOLD,W.J.(1947)Simplereactionchainsandtheirintegration.J.comp.physiol.Psychol.,40,349-363.—203,205

6. BARD,P.(1934)Emotion:I.Theneuro-humoralbasisofemotionalreactions. In Murchison, C. A. (ed.),A handbook of generalexperimental psychology.Worcester, Mass., Clark Univ. Press.—351AYLESWORTH,M.,seeNo.282.

7. BARKER,R.G. (1945)An experimental study of the resolution ofconflicts in children. In Q.McNemar andM.A.Merrill (eds.),Studiesinpersonality.NewYork,McGraw-Hill.—321

8. BASH,K.W. (1939)An investigation into apossibleorganicbasisforthehungerdrive.J.comp.Psychol.,28,109-135.—287

9. BASS,M. J., andHULL, C. L. (1934)The irradiation of a tactileconditionedreflexinman.J.comp.Psychol.,17,47-65.—127

10. BEEBE-CENTER, J. G. (1932)The psychology of pleasantnessand unpleasantness. New York, Van Nostrand, viii, 427 pp.—255

11. BECHTEREV, V. M. (1932) General principles of humanreflexology.NewYork,International,467pp.—34BERNSTONE,A.H.,seeNo.185.

12. BERSH, P. J. (1950) The influence of two variables upon theestablishment of a secondary reinforcer for operant responses.J.exp.Psychol.inpress.—243,244BERSH,P.J.,seealsoNos.226,227.

13. BITTERMAN,M.E. (1944)Behaviordisorderasa functionof therelativestrengthofantagonisticresponse-tendencies.Psychol. 51,375-378.—321

14.BORING, E.G. (1929)A history of experimental psychology. NewYork,Appleton-Century-Crofts,xvi,699pp.—163

15. BOUSFIELD,W.A.(1938),seeB.F.Skinner(1938).—291,29216. BRAY, C.W. (1928) Transfer of learning. J. exp. Psychol., 11,

443-467.—17017. BROGDEN,W.J. (1939a)Theeffectof frequencyof reinforcement

upon the level of conditioning.J. exp. Psychol., 24, 419-431.—99,113

18. BROGDEN,W. J. (1939b)Unconditioned stimulus-substitution intheconditioningprocess.Amer.J.Psychol.,52,46-55.—49BROGDEN,W.J.,seealsoNo.40.

19. BROWN, J. S. (1948) Gradients of approach and avoidanceresponses and their relation to level of motivation.J. comp.physiol.Psychol.,41,450-465.—317BROWN,J.S.,seealsoNo.170.

20. BROWN,R.H.(1936)Colorvisionintherabbit.J.gen. Psychol.,14.83-97.—133

21. BRUCE,R.H.(1932)Theeffectofremovalofrewardonthemazeperformance of rats. III.Univ. Calif. Publ. Psychol., 6, 75-82.—235

22. BRYAN, W. L., and HARTER, N. (1899) Studies on thetelegraphic language. The acquisition of a hierarchy of habits.Psychol.Rev.,6,345-375.—153

23. BUGELSKI, R. (1938) Extinction with and without sub-goalreinforcement.J.comp.Psychol.,26,121-134.—235

24. BUGELSKI,R., andMILLER,N. E. (1938)A spatial gradient inthe strength of avoidance responses.J. exp. Psychol., 23, 494-505.

25. BULLOCK—253, D. H., and FISCHER, L. (1949)The effect ofrepeated conditioning and extinction of an operant response. (Tobepublished).—83

26. BUNCH,M. E. (1928)The effect of electric shock as punishmentforerrorsinhumanmazelearning.J.comp.Psychol.,8,343-359.—111

27. BUNCH,M.E.(1935)Certaineffectsofelectricshockinlearningastylusmaze.J.comp.Psychol.,20,211-242.—111

28. BURNHAM,W. H. (1917) Mental hygiene and the conditionedreflex.J.genet.Psychol.,24,449-488.—35CAMPBELL,R.K.,seeNo.94.

29. CANNON,W. B. (1927)The James-Lange theory of emotions: acriticalexaminationandanalternative theory.Amer. J. Psychol .,39,106-124.—333

30. CANNON,W.B.(1929)Bodilychanges inpain,hunger, fearandrage.NewYork,Appleton-Century,xvi,404pp.—286

31. CANNON,W.B.,andWASHBURN,A.L.(1912)Anexplanationofhunger.Amer.J.Physiol.,29,441-454.—285

32. CASON, H. (1922) The conditioned pupillary reaction. J. exp.Psychol.,5,108-146.—25,35CHEIN,I.,seeNo.147.CHEN,H.P.,seeNo.114.CHRISTO,I.J.,seeNo.131.CLAYTON,F.L.,seeNo.120.

33. COFER,C.N., andFOLEY, J. P. (1942)Mediated generalizationand the interpretation of verbal behavior: I. Prolegomena.Psychol.Rev.,49,513-540.—160,163COMMINS,W.D.,seeNo.257.

34. CONRADI, E. (1905) Song and call notes of English sparrowswhenrearedbycanaries.Amer.J.Psychol.,16,190-198.—354

35. COWLES, J.T. (1937) Food-tokens as incentives for learning bychimpanzees.Comp.Psychol.Monogr.,14,No.71,96pp.—248

36. CRAFTS, L. W., SCHNEIRLA,T. C, ROBINSON, E. E., andGILBERT,R.W.(1938)Recentexperimentsinpsychology. NewYork,McGraw-Hill,xiv,417pp.—14,136,137

37. CRAIG,W. (1921)Why do animals fight? Int. Jour. Ethics , 31,264-278.—342

38. CRAWFORD,M. P. (1937)The coöperative solving of problemsbyyoungchimpanzees.Comp.Psychol.Monogr.,14,No.68,88pp.—399

39. CRUTCHFIELD, R. S. (1939) The determiners of energyexpenditure in string-pulling by the rat.J.Psychol., 7, 163-178.—196

40. CULLER, E., FINCH, G., GIRDEN, E., and BROGDEN,W. J.(1935) Measurements of acuity by the conditioned responsetechnique.J.gen.Psychol.,12,223-227.—314

41. DANIEL, W. J. (1942) Cooperative problem solving in rats. J.comp.Psychol.,34,361-369.—358,359

42. DANIEL,W.J.(1943)Higherordercooperativeproblemsolvinginrats.J.comp.Psychol.,35,297-305.—358,360

43. DASHIELL, J.F. (1949)Fundamentalsofgeneralpsychology. 3rded.,Boston,HoughtonMifflin,x,690pp.—17,359

44. DAVIS,E.A. (1937)Thedevelopmentof linguisticskill in twins,singletonswith siblings, and only children from age five to tenyears.(Inst. ChildWelfareMonogr. Ser . No. 14),Minneapolis,Univ.ofMinn.Press,ix,165pp.—205

45. DAVIS,C.M. (1928)Self selectiondietbynewlyweaned infants.Amer.J.Dis.Child.,36,651-679.—292

46. DAVIS,R.C. (1948)Responses to 'meaningful' and 'meaningless'sounds.J.exp.Psychol,38,744-756.—230

47. DENNIS,W. (1940) Infant reaction to restraint: an evaluation ofWatson's theory.Trans. N. Y. acad. Sci ., 2, No. 8, 202-218.—343

48. DINSMOOR, J. A. (1950) A quantitative comparison of thediscriminative and reinforcing functions of a stimulus.J. exp.Psychol.,inpress.—237

49. DODSON, J. D. (1932) The relative values of satisfying andannoying situations asmotives in the learning process.J. comp.Psychol.,14,147-164.—111DOLLARD,J.,seeNo.171.

50. DUCHENNE,G.B.A.(1862),seeDumas,G.(1923).—33551. DUMAS, G. (1923)Traité de psychologie. Vol. I. Paris, Alcan,

xiv,964PP.—33552. EBBINGHAUS,H. (1885)Memory (Trans.H.A.Ruger andC.E.

Bussenius),NewYork,TeachersCollege,1913,viii,123pp. —78,229

53. EISENSON, J. (1938)The psychology of speech. New York,Appleton-Century-Crofts,xiii,280pp.—382

54. ELDER, J.H. (1934)Auditory acuity of the chimpanzee. J. comp.Psychol.,17,157-183.—135

55. ELLIOTT, M. H. (1934) The effect of hunger on variability ofperformance.Amer.J.Psychol.,46,107-112.—179

56. ELLSON, D. G. (1938) Quantitative studies of the interaction ofsimple habits: I. Recovery from the specific and generalizedeffectsofextinction.J.exp.Psychol.,23,339-358.—77

57. ELLSON,D.G.(1939)Theconceptofreflexreserve.Psychol.Rev.,46,566-575-—77

58. ESTES, W. K. (1943) Discriminative conditioning: I. Adiscriminative property of conditioned anticipation.J. exp.Psychol.,32,150-155—243

59. ESTES, W. K. (1944) An experimental study of punishment.Psychol.Monogr.,57,No.263,iii,40pp.—107

60. ESTES,W.K. (1948)Discriminative conditioning: II.Effects of aPavlovian conditioned stimulus upon a subsequently establishedoperantresponse.J.exp.Psychol.,38,173-177.—243

61. ESTES,W. K. (1949a) Generalization of secondary reinforcementfromtheprimarydrive.J.comp.physiol.Psychol., 42, 286-295;—241

62. ESTES,W.K.(1949b)Astudyofmotivatingconditionsnecessaryforsecondaryreinforcement.J.exp.Psychol.,39,306-310.—241

63. ESTES,W. K., and SKINNER, B. F. (1941) Some quantitativepropertiesofanxiety.J.exp.Psychol.,29,390-400.—309,310

64. EWERT,P.H.(1933)Eye-movementsduringreadingandrecall. J.gen.Psychol.,8,65-84.—216

65. FIELDS, P. E. (1932) Studies in concept formation: I. Thedevelopment of the concept of triangularity by the white rat.Comp.Psychol.Monogr.,9,No.2.—158

66. FINAN,J.L.(1940)Quantitativestudiesinmotivation:I.Strengthof conditioning in rats under varyingdegrees of hunger.J. compPsychol.,29,119-134.—112FINCH,G.,seeNo.40.FISCHER,L.,seeNo.25.FLETCHER,F.M.,seeNo.186.

67. FOLEY, J. P. (1935) Judgment of facial expression of emotion inthechimpanzee.J.soc.Psychol.,6,31-67.—345FOLEY,J.P.,seealsoNo.33.

68. FRICK, F. C. (1948)An analysis of an operant discrimination. J.Psychol.,26,93-123.—127,128GAGNÉ,R.M.,seeNo.81.

69. GALTON, F. (1883) Inquiries into human faculty and itsdevelopment.London,Macmillan,xii,387pp.—221

70. GARRETT, H. E. (1941)Great experiments in psychology. Rev.ed.,NewYork,Appleton-Century-Crofts,xxi,452pp.—14

71. GATES, G. S. (1926) An observational study of anger. J. exp.Psychol.,9,325-336.—343

72. GELLERMANN,L.W.(1931)Thedoublealternationproblem:III.The behavior ofmonkeys in a double alternation box-apparatus.J.genet.Psychol.,39,359-392-—203

73. GENGERELLI, J.A. (1930)Theprinciple ofmaxima andminimainanimallearning.J.comp.Psychol.,11,193-236.—196

74. GENGERELLI,J.A.(1933)Theprincipleofminimumpathintheringtailmonkey.Publ.U.C.L.A.Educ.Philos.,Psychol.,1,No.13,165-187.—176

75. GHISELLI, E. E. (1936) Changes in neuromuscular tensionaccompanying the performance of a learning problem involvingconstantchoicetime.J.exp.Psychol.,19,91-98.—196

76. GIBSON, E. J. (1939) Sensory generalization with voluntaryreactions.J.exp.Psychol.,24,237-253.–129

77. GILBERT,R.W. (1936)Theeffectofnon-informativeshockuponmaze learning and retention with human subjects.J. exp.Psychol.,19,456-466.—111

78. GILBERT, R. W. (1937) A further study of the effect of non-informative shock upon learning.J. exp. Psychol., 20, 396-407.—111GILBERT,R.W.,seealsoNo.36.

79. GILHOUSEN, H. C. (1931)An investigation of "insight" in rats.Science,73,711-712.—178

80. GODBEER, E. (1940) Factors introducing conflict in the choice

behavior of children. Ph.D. dissertation,YaleUniversity.— 321,324GIRDEN,E.,seeNo.40.

81. GRAHAM, C. H., and GAGNÉ, R. M. (1940) The acquisition,extinction, and spontaneous recovery of a conditioned operantresponse.J.exp.Psychol.,26,251-280.—53,55

82. GRICE, G. R. (1948)The relation of secondary reinforcement todelayedrewardinvisualdiscriminationlearning.J.exp.Psychol.,38,1-16.—209,210

83. GRINDLEY, G. C. (1929) Experiments on the influence of theamount of reward on learning in young chickens.Brit. J.Psychol.,20,173-180.—73,234

84. GUTHRIE, E. R. (1935)Psychology of learning. New York,Harper,viii,258pp.—113,196GUTHRIE,E.R.,seealsoNo.247.

85.GUTHRIE, E. R., and HORTON, G. P. (1946) Cats in a puzzlebox.NewYork,Rinehart,67pp.—164,165,166

86. HARLOW, H. F. (1948) Studying animal behavior. In T. G.Andrews (ed.),Methods of Psychology. NewYork, JohnWiley,xiv,716pp.—293HARTER,N.,seeNo.22.

87. HAYS, R., andWOODBURY, C. B. Unpublished study reportedbyC.L.Hull(seeNo.104)—171,172

88. HEATHERS, G. L. (1940)The avoidance of repetition of a mazereactionintheratasafunctionofthetimeintervalbetweentrials.J.Psychol.,10,359-380.—177

89. HEBB, D. O. (1945) The forms and conditions of chimpanzeeanger.Bull.Canad.psychol.Ass.,5,32-35.—342

90. HEFFERLINE, R. F. (1950)An experimental study of avoidance.Genet.Psychol.Monogr.,inpress.—113,250,312

91. HEIDER, F. K., and HEIDER, G. M. (1940) A comparison ofsentencestructureofdeafandhearingchildren.Psychol.Monogr.,52,No.1,42-103.—205HEIDER,G.M.,seeNo.91.

92. HENMON,V.A.C.(1906)Thetimeofperceptionasameasureofdifferencesinsensations.Arch.ofPhil.,Psychol.andSci.Meth.,NewYork,TheSciencePress,75pp.—143HERON,W.T.,seeNo.215.

93. HILGARD,E.R. (1937)The relationship between the conditionedresponse and conventional learning experiments.Psychol. Bull.,34,61-102.—66

94. HILGARD,E.R.,CAMPBELL,R.K.,andSEARS,W.N.(1938)Conditioneddiscrimination: theeffectofknowledgeofstimulus-relationships.Amer.J.Psychol.,51,498-506.—121,122

95. HILGARD,E.R.,andMARQUIS,D.G.(1940)Conditioning andlearning.NewYork,Appleton-Century-Crofts, xi, 429pp.— 13,34,35,66,169

96. HOBBES,T. (1651) Leviathan. London,OxfordUniv. Press, PartI,Ch.VI.—325HORTON,G.P.,seeNo.85.

97. HOVLAND, C. I. (1937a) The generalization of conditionedresponses: I.Thesensorygeneralizationofconditioned responseswith varying frequencies of tone.J.gen. Psychol., 17, 125-148.—127

98. HOVLAND, C. I. (1937b) The generalization of conditionedresponses:II.Thesensorygeneralizationofconditionedresponseswithvaryingintensitiesoftone.J.genet.Psychol.,51,279-291.

99. HOVLAND—127, C. I. (1937c)The generalization of conditionedresponses: III. Extinction, spontaneous recovery, anddisinhibitionofconditionedandofgeneralizedresponses.J. exp.Psychol.,21,47-62.—127

100. HOVLAND, C. I. (1937d) The generalization of conditionedresponses: IV.The effects of varying amounts of reinforcementuponthedegreeofgeneralizationofconditionedresponses.J. exp.Psychol.,21,261-276.—26HOVLAND,C.I.,seealsoNo.229.

101. HOVLAND, C. I., and SEARS, R. R. (1938) Experiments onmotorconflict:I.Typesofconflictandtheirmodesofresolution.J.exp.Psychol.,23,477-493-–319,322

102. HUDGINS,C.V.(1933)Conditioningandthevoluntarycontrolofthepupillarylightreflex.J.gen.Psychol.,8,3-51.—25

103. HULL, C. L. (1920) Quantitative aspects of the evolution ofconcepts;anexperimentalstudy.Psychol.Monogr., 28,No.123.–154

104. HULL,C. L. (1943)Principlesofbehavior. NewYork,Appleton-Century-Crofts,x,422pp.—14,26,69,74,171,172,178,196,209,261HULL,C.L.,seealsoNo.9.

105. HULL, C. L.,et al. (1940)Mathematico-deductive theory of rotelearning.NewHaven,YaleUniv.Press.—229

106. HUMPHREYS, L. G. (1939)The effect of random alternation ofreinforcement on the acquisition and extinction of conditionedeyelidreactions.J.exp.Psychol.,25,141-158.—99,113

107. HUMPHREYS, L. G. (1940) Extinction of conditioned psycho-galvanic responses following two conditions of reinforcement.J.exp.Psychol.,27,71-75.—100

108. HUMPHREYS, L. G. (1943) The strength of a Thorndikianresponse as a function of the number of practice trials.J. comp.

Psychol.,35,101-110.—100109. HUNT,J.MCV.,andSCHLOSBERG,H. (1939)The influenceof

illumination upon general activity in normal, blinded, andcastratedmalewhiterats.J.comp.Psychol.,28,285-298.—282HUNT,W.A.,seeNo.142.

110. HUNTER, W. S. (1920) The temporal maze and kinaestheticsensoryprocessesinthewhiterat.Psychobiol.,2,1-17.—177

111. HUNTER,W. S. (1928) The behavior of raccoons in a double-alternation temporalmaze.J.genet. Psychol., 35, 374-388.–177,207

112. HUNTER,W. S. (1935) Conditioning and extinction in the rat.Brit.J.Psychol.,26,135-148.—312IREY,E.,seeNo.187.

113. IRWIN,O.C. (1947)Developmentof speechduring infancy:curveofphonemicfrequencies.J.exp.Psychol.,37,187-193.—381

114. IRWIN, O. C, and CHEN, H. P. (1946) Development of speechduring infancy: curve of phonemic types.J. exp. Psychol., 36,431-436.—380,381

115. JACKSON,T.A.(1939)Studiesinthetranspositionoflearningbychildren: III.Transpositional responseasa functionof transposeddimensions.J.exp.Psychol.,25,116-124.—158

116. JACOBSON, E. (1932)The electrophysiology ofmental activities.Amer.J.Psychol.,44,677-694.—213

117. JAMES,W.(1884)Whatisanemotion?Mind,9,188-205.—333118. JAMES,W. (1890)Principles of psychology. NewYork,Holt, II,

vi,704pp.—330,351JENKINS,T.N.,seeNo.283.

119. JENKINS, W. O. (1950) A temporal gradient of derivedreinforcement.Amer.J.Psychol.,63,237-243.—245JENKINS,W.O.,seealsoNo.194.

120. JENKINS,W.O.,andCLAYTON,F.L.(1949)Rateofrespondingandamountofreinforcement.J.comp.physiol.Psychol.,42,174-181.—100,112

121. JESPERSEN, J. O. (1922)Language: its nature development andorigin.NewYork,Holt,448pp.—384JONES,H.,seeNos.177,178.

122. JONES,M. C. (1924a)The elimination of children's fears. J. exp.Psychol.,7,382-390.—347

123. JONES, M. C. (1924b) A laboratory study of fear. The case ofPeter.J.genet.Psychol.,31,308-315.—347

124. JONES,M. C. (1926)The development of early behavior patternsin young children.Ped. sem.& J. genet.Psychol., 33, 537-585.—335

125. JUNG, C. G. (1919)Studies in word association. New York,

Moffat,Yard,ix,575pp.—225126. KANTOR, J. R. (1924)Principles of psychology. New York,

Knopf,xix,473pp.—230KAPLON,M.D.,seeNo.297.

127. KARDINER,A. (1939)The individualandhissociety: thepsycho-dynamicsofprimitivesocialorganization. NewYork, ColumbiaUniv.Press,xxvi,503pp.—367KATZ,A.,seeNo.222.

128. KELLER, F. S. (1937)The definition of psychology. NewYork,Appleton-Century-Crofts,vi,111pp.—220

129. KELLER, F. S. (1940)The effect of sequence of continuous andperiodic reinforcement upon the 'reflex reserve.'J. exp. Psychol.,27,559-565.—77

130. KELLER, F. S. (1941) Light-aversion in the white rat.Psychol.Rec,4,235-250.—62,304

131. KELLER, F. S., CHRISTO, I. J., and SCHOENFELD,W. N.(1946).Studies in InternationalMorseCode:V.Theeffectof the"phoneticequivalent."J.appl.Psychol.,30,265-270.—149

132. KELLER, F. S., and SCHOENFELD,W. N. (1944) Studies inInternationalMorseCode:III.Theefficiencyofthecodeasrelatedto errors made during learning.J.appl. Psychol., 28, 254-266.—148

133. KELLOGG,W.N. (1939) "Positive" and "negative" conditioning,without contraction of the essentialmuscles during the period oftraining.Psychol.Bull.,36,575.—169KELLOGG,L.A.,seeNo.134.

134. KELLOGG,W.N.,andKELLOGG,L.A. (1933)Theapeand thechild.NewYork,Whittlesey,xiv,314pp.—355

135. KENT, G. H., and ROSANOFF, A. J. (1910) A study ofassociation in insanity.Amer. J. Insan ., 67, 37-96; 317-390.—222

136. KINSEY,A.C,POMEROY,W.B., andMARTIN,C. E. (1948)Sexual behavior in the human male. Philadelphia, W. B.Saunders,xv,804pp.—300

137. KLINEBERG, O. (1940)Social psychology. NewYork, Holt, xii,570pp.—336399KOERNER,L.,seeNo.187.

138. KÖHLER,W. (1925)The mentality of apes. NewYork, HarcourtBrace,viii,342pp.—59,65

139. KONORSKI, J., and MILLER, S. (1937) On two types ofconditionedreflex.J.gen.Psychol.,16,264-272.—63LAMOREAUX,R.R.,seeNos.179,180.

140. LANDIS, C. (1924) Studies of emotional reactions: II. Generalbehavior and facial expression.J. comp. Psychol., 4, 447-501.

—336141. LANDIS, C. (1934) Emotion: II.The expressions of emotion. In

Murchison, C. (ed.),A handbook of general experimentalpsychology.Worcester,Mass.,ClarkUniv.Press.—351

142. LANDIS, C, and HUNT,W.A. (1939) The startle pattern. NewYork,FarrarandRinehart,xvi,168pp.—337

143. LASHLEY,K. S. (1916)The human salivary reflex and its use inpsychology.Psychol.Rev.,23,446-464.—35

144. LASHLEY, K. S. (1929)Brain mechanisms and intelligence.Chicago,Univ.ofChicagoPress,xiv,186pp.—137

145. LASHLEY,K.S.(1930)Themechanismofvision:I.Amethodforrapid analysisof patternvision in the rat.J.genet. Psychol., 37,453-460.—136

146. LEUBA, C. J. (1931) Some comments on the first reports of theColumbia study of animal drives.J. comp. Psychol., 11, 275-280.—298

147. LEVINE,R.,CHEIN,I.,andMURPHY,G. (1942)The relationofintensity of a need to the amount of perceptual distortion: apreliminaryreport.J.Psychol.,13,283-293.—290

148. LIDDELL,H.S.(1938)Theexperimentalneurosisandtheproblemofmentaldisorder.Amer.J.Psychiat.,94,1035-1043.—29,139LIDDELL,H.S.,seealsoNo.2.

149. LINTON, R. (1945)The cultural background of personality. NewYork,Appleton-Century-Crofts,xix,157pp.—367LIPOFSKY,H.,seeNo.170.

150. LORGE, I. (1936) Irrelevant rewards in animal learning.J. comp.Psychol.,21,105-128.—175,196

151. LURIA,A. R. (1932)The nature of human conflicts, or emotion,conflict, and will; an objective study of disorganization andcontrolofhumanbehavior. (Trans. byW.H.Gantt)NewYork,Liveright,xviii,431pp.—324

152. LYON,D.O.(1917)Memoryand the learningprocess. Baltimore,WarwickandYork,179pp.—228

153. MACDUFF, M. M. (1946) The effect on retention of varyingdegrees ofmotivation during learning in rats.J. comp. Psychol.,39,207-240.—112

154. MACFARLANE, D. A. (1930) The rôle of kinesthesis in mazelearning.Univ.Calif.Publ.Psychol.,4,277-305.—169

155. MALINOWSKI, B. (1929)The sexual life of savages innorthwestern Melanesia: an ethnographic account of courtship,marriage and family life among the natives of the TrobriandIslands in British New Guinea. NewYork, Liveright, 506 pp.—367MARQUIS,D.G.,seeNos.95,299.

MARTIN,C.E.,seeNo.136.156. MASLOW,A.H. (1936a)The rôleof dominance in the social and

sexualbehaviorof infra-humanprimates: I.ObservationsatVilasParkZoo.J.genet.Psychol.,48,261-277.—356,399

157. MASLOW,A.H. (1936b)The roleofdominance in the social andsexualbehaviorofinfra-humanprimates:IV.Thedeterminationofhierarchyinpairsandinagroup.J.genet.Psychol.,49,161-198.—356,399

158. MASSERMAN, J. H. (1946)Principles of dynamic psychiatry.Philadelphia,W.B.Saunders,xix,322pp.—82,203

159. MATEER, F. (1918)Child behavior, a critical and experimentalstudy of young children by the method of conditioned reflexes.Boston,Badger,v,239pp.—35

160. MAX,L.W. (1935)Anexperimental studyof themotor theoryofconsciousness: III.Action-current responses in deafmutes duringsleep, sensory stimulation and dreams.J. comp. Psychol., 19,469-486.—214

161. MAX, L.W. (1937) Experimental study of the motor theory ofconsciousness: IV. Action-current responses in the deaf duringawakening, kinaesthetic imagery, and abstract thinking. J. comp.Psychol.,24,301-344.—214

162. MCCARTHY, D. A. (1930)The language development of thepreschool child. Minneapolis, Univ. Minn. Press, xiii, 174 pp.—204,205

163. MCDOUGALL (1923)Outlineofpsychology.NewYork,Scribner,xvi,456pp.—330

164. MENZIES, R. (1937) Conditioned vasomotor responses in humansubjects.J.Psychol.,4,75-120.—26

165. MERKEL, J. (1885) Die zeitlichen verhältnisse derwillensthätigkeit.Philos.Studien,2,73-127.—143

166. MILES,W. R. (1931) Studies in physical exertion: II. Individualandgroupreactiontimeinfootballcharging.Res.Quar.,2,6-13.—146

167.MILLER, N. E. (1941)An experimental investigation of acquireddrives.Psychol.Bull.,38,534-535.—311.SeeMiller,(1948)

168. MILLER,N.E.(1944)Experimentalstudiesofconflict.InJ.McV.Hunt (ed.),Personality and the behavior disorders.Vol. I. NewYork,RonaldPress.—317,318,320,321

169. MILLER,N.E.(1948)Studiesoffearasanacquirabledrive:I.Fearasmotivation and fear reduction as reinforcement in the learningof new responses.J.exp. Psychol., 38, 89-101.—250,251, 252,317MILLER,N.E.,seealsoNo.24.

170. MILLER, N. E., BROWN, J. S., and LIPOFSKY, H. (1943)A

theoretical and experimental analysis of conflict behavior: III.Approach-avoidanceconflictasafunctionofstrengthofdriveandstrengthofshock.ReportedinMiller(1944).—320

171. MILLER, N. E., and DOLLARD, J. (1941)Social learning andimitation. New Haven, Yale Univ. Press, xiv, 341 pp.— 196,361,362,399MILLER,S.,seeNo.139.

172. MITRANO,A. J. (1939)Principlesof conditioning inhumangoalbehavior.Psychol.Monogr.,51,230pp.—82MORGULIS,S.,seeNo.300.

173. MOSS, F.A. (1924) Study of animal drives. J. exp. Psychol., 7,165-185.—297

174. MOWRER, O. H. (1939)A stimulus-response analysis of anxietyand its rô1e as a reinforcing agent.Psychol. Rev., 46, 553-565.—311

175. MOWRER,O.H. (1940)Anexperimentalanalogueof"regression"with incidental observations on "reaction-formation."J. abn.(soc.)Psychol.,35,56-87.—61,81,306

176. MOWRER, O. H. (1947) On the dual nature of learning–areinterpretationof"conditioning"and"problemsolving."HarvardEduc.Rev.,17,102-148.—66

177. MOWRER,O.H.,andJONES,H. (1943)Extinctionandbehaviorvariability as functions of effortfulness of task.J. exp. Psychol.,33,369-386.—196

178. MOWRER, O. H., and JONES, H. (1945) Habit strength as afunctionofthepatternofreinforcement.J.exp.Psychol.,35,293-311.—100,207

179. MOWRER, O. H., and LAMOREAUX, R. R. (1942)Avoidanceconditioningandsignalduration–astudyofsecondarymotivationandreward.Psychol.Monogr.,54,No.247,iii,34pp.—251

180. MOWRER, O. H., and LAMOREAUX, R. R. (1946) Fear as aninterveningvariableinavoidanceconditioning.J.comp. Psychol.,39,29-50.—251

181. MUENZINCER, K. F. (1928) Plasticity and mechanization of theproblem box habit in guinea pigs.J. comp. Psychol., 8, 45-70.—167

182.MUENZINGER, K. F. (1934a) Motivation in learning: I. Electricshock for correct response in the visual discrimination habit.J.comp.Psychol.,17,267-277.—113

183. MUENZINGER, K. F. (1934b) Motivation in learning: II. Thefunctionofelectricshockforrightandwrongresponsesinhumansubjects.J.exp.Psychol.,17,439-448.—113

184. MUENZINGER,K.F.(1942)Psychology:Thescienceofbehavior.NewYork,Harper,xi,441pp.—14

185. MUENZINGER,K.F.,BERNSTONE,A.H.,andRICHARDS,L.(1938) Motivation in learning: VIII. Equivalent amounts ofelectric shcck for right and wrong responses in visualdiscriminationhabit.J.comp.Psychol.,26,177-185.—113

186. MUENZINGER,K.F.,andFLETCHER,F.M.(1936)Motivationinlearning:VI.Escapefromelectricshockcomparedwithhunger-foodtensioninthevisualdiscriminationhabit.J.comp.Psychol.,22,79-91.—113

187. MUENZINGER, K. F., KOERNER, L., and IREY, E. (1929)Variability of an habitual movement in guinea pigs. J. comp.Psychol.,9,425-436.—195

188. MUENZINGER,K.F.,andNEWCOMB,H. (1935)Motivation inlearning: III.Abell signal comparedwith electric shock for rightandwrong responses in the visual discrimination habit.J. comp.Psychol.,21,95-104.—113

189. MUENZINGER, K. F., and WOOD, A. (1935) Motivation inlearning: IV.The function of punishment as determined by itstemporal relation to theactofchoice in thevisualdiscriminationhabit.J.comp.Psychol.,20,95-106.—113

190. MUNN, N. L. (1946)Psychology: the fundamentals of humanadjustment.Boston,HoughtonMifflin,xviii,497pp.—14

191. MURCHISON, C. A. (ed.) (1935)A handbook of socialpsychology.Worcester,ClarkUniv.Press,xii,1195pp.—400MURPHY,G.,seeNo.147.

192. MURPHY, M. J. (1943) The differentiation of a response–Anexploratory study with human subjects.Abridgement of Ph.D.dissertation,NewYorkUniversity,23pp.—188NEWCOMB,H.,seeNo.188.

193. NISSEN,H.W. (1930)A study ofmaternal behavior in thewhiterat by means of the obstruction method.J.genet. Psychol., 37,377-303.—299

194. NISSEN, H. W., and JENKINS, W. O. (1943) Reduction andrivalry of cues in the discrimination behavior of chimpanzees.J.comp.Psychol.,35,85-95.—316

195. NOTTERMAN, J. M. (1950) The interrelationships amongaperiodic reinforcement, discrimination, learning, and secondaryreinforcement. Ph.D. dissertation, Columbia Univ.—235, 245,246

196. PAVLOV,I.P. (1923) InHolt,E.B. (1931) Animaldriveand thelearningprocess.NewYork,Holt,vii,307pp.—19

197. PAVLOV, I. P. (1927) Conditioned reflexes. (Trans, by G. V.Anrep) London, Oxford Univ. Press, xv, 430 pp.— 24, 34, 68,83,117,126,134,138,162

198. PAVLOV,I.P.(1928) Lecturesonconditionedreflexes. (Trans,by

W.H.Gantt)NewYork,International,414pp.—34199. PERIN,C.T.(1942)Behaviorpotentialityasajointfunctionofthe

amount of training and the degree of hunger at the time ofextinction.J.exp.Psychol.,30,93-113.—75,266

200. PIAGET,J.(1929)Thechild'sconceptionof theworld.NewYork,Harcourt,ix,397pp.—154

201. PLOTKIN, L. (1943) Stimulus generalization in Morse Codelearning.Arch.Psychol.,40,No.287.—124POMEROY,W.B.,seeNo.136.

202. RABEN,M.W.(1949)Thewhiterat'sdiscriminationofdifferencesin intensity of illumination measured by a running response.J.comp.physiol.Psychol.,42,254-272.—54,129

203. RASMUSSEN, E.W. (1940)The shock method as a measure ofhunger and of the intensity of a conflict.Acta. psychol., Hague,5,63-78.—321RAYNER,R.,seeNo.290.

204. RAZRAN, G. H. S. (1939a) The law of effect or the law ofqualitativeconditioning.Psychol.Rev.,46,445-463.—66

205. RAZRAN,G.H.S. (1939b)Aquantitative studyofmeaningbyaconditioned salivary technique (semantic conditioning).Science,90,89-90.—160

206. RAZRAN,G.H. S. (1939c) Studies in configural conditioning: I.Historical and preliminary experimentation.J.gen. Psychol., 21,307-330.—324

207. REICH, W. (1949)Character analysis. New York, OrgoneInstitute,xxvi,516pp.—324

208. REINWALD,F.L.(1938)Personalcommunication.—29209. REINWALD,F.L.(1941)Personalcommunication.—64210. REYNOLDS,B. (1949)The relationship between the strength of a

habit and the degree of drive present during acquisition.J. exp.Psychol.,39,296-305.—112RICHARDS,L.,seeNo.185.

211. RICHTER,C.P.(1922)Abehavioristicstudyoftheactivityoftherat.Comp.Psychol.Monogr.,1,No.2,55pp.—279,280

212. RIESS,B.F. (1940)Semanticconditioning involving thegalvanicskinreflex.J.exp.Psychol.,26,238-240.—160

213. RIESS, B. F. (1946)Genetic changes in semantic conditioning. J.exp.Psychol.,36,143-152.—160,161

214. RITTER,A.M. (1949) Some conditions influencing the incidenceofresponseduplicationofverbalstimuli.J.Psychol.,28,93-118.—393ROBINSON,E.E.,seeNo.36.

215. ROBINSON, E. S., and HERON,W.T. (1924)The warming upeffect.J.exp.Psychol.,7,81-97.—188

ROSANOFF,A.J.,seeNo.135.ROSENZWEIG,S.,seeNo.232.

216. RUCH,F.L.(1948)PsychologyandLife.NewYork,Scott,Fores-man,xvi,782pp.—14

217. SALTZMAN, I. J. (1949)Maze learning in the absence of primaryreinforcement: a study of secondary reinforcement.J. comp.physiol.Psychol.,42,161-173.—235

218. SANFORD,R.N.(1936)Theeffectsofabstinencefromfooduponimaginalprocesses:apreliminaryexperiment.J.Psychol.,2,129-136.—290

219. SCHJELDERUP, EBBE T. (1935) Social behavior of birds. InMurchison,C.(ed.),Ahandbookofsocialpsychology.Worcester,Mass.,ClarkUniv.Press.Pp.947-972.—356

220. SCHLOSBERG,H. (1936)Conditionedresponses in thewhite rat:II. Conditioned responses based upon shock to the foreleg.J.genet.Psychol.,49,107-138.—314

221. SCHLOSBERG, H., (1937)The relationship between success andthelawsofconditioning.Psychol.Rev.,44,379-394.—66SCHLOSBERG,H.,seealsoNo.109.

222. SCHLOSBERG, H., and KATZ, A. (1943) Double alternationlever-pressing in the white rat.Amer. J. Psychol ., 56, 274-282.—203SCHNEIRLA,T.C,seeNo.36.

223. SCHOENFELD,W.N.(1946)Unpublishedstudy.—304224. SCHOENFELD,W.N.(1947)Unpublishedstudy.—62225. SCHOENFELD, W. N. (1950) An experimental approach to

anxiety, escape, and avoidance behavior. InAnxiety: Proc.Amer.psychopathol. Ass ., New York, Grune and Stratton. In press.—313SCHOENFELD,W.N.,seealsoNos.131,132.

226. SCHOENFELD,W. N.,ANTONITIS, J. J., and BERSH, P. J.(1950a)A preliminary study of training conditions necessary forsecondaryreinforcement.J.exp.Psychol.,40,40-45.—236

227. SCHOENFELD,W. N.,ANTONITIS, J. J., and BERSH, P. J.(1950b) Unconditioned response rate of the white rat in a bar-pressingapparatus.J.comp.physiol.Psychol.,43,41-48.—113

228. SCOTT,W.E.D. (1901)Dataonsong inbirds.Science, 14, 522-526.—354SEARS,R.R.,seeNo.101.

229. SEARS, R. R., and HOVLAND, C. I. (1941) Experiments onmotor conflicts: II. Determination of mode of resolution bycomparative strengths of conflicting responses.J. exp. Psychol.,28,280-286.—317,321SEARS,W.N.,seeNo.94.

230. SEWARD, J. P. (1946) Aggressive behavior in the rat: IV.Submission as determined by conditioning, extinction, anddisuse.J.comp.Psychol.,39,51-76.—356

231. SHAFFER, L. E. (1936)The psychology of adjustment Boston,HoughtonMifflin,xix,600pp.—316

232. SHAKOW, D., and ROSENZWEIG, S. (1940) The use of thetautophone ("verbal summator") as an auditory apperceptive testfor the study of personality.Character & Pers., 8, 216-226.—395

233. SHAW,W.A. (1940)The relationofmuscularactionpotentials toimaginal weight lifting.Arch. Psychol., No. 247, 50 pp.—215,216

234. SHEFFIELD, F. D. (1948)Avoidance training and the contiguityprinciple.J.comp.physiol.Psychol.,41,165-177.—314

235. SHERMAN,M. (1928)The differentiation of emotional responsesin infants: I. Judgments of emotional responses from motionpicture views and from actual observation.J. comp. Psychol., 7,265-284.—338,339

236. SKINNER,B.F. (1930)On the conditionsof elicitationof certaineatingreflexes.Proc.nat.Acad.Sci.,Wash.,16,433-438.—42

237. SKINNER,B.F. (1932)On the rate of formationof a conditionedreflex.J.gen.Psychol.,7,274-285.—43

238. SKINNER, B. F. (1933) The rate of establishment of adiscrimination.J.gen.Psychol.,9,302-350.—99

239. SKINNER, B. F. (1935) Two types of conditioned reflex and apseudotype.J.gen.Psychol.,12,66-77.—47,66

240. SKINNER,B.F.(1936)Theverbalsummatorandamethodforthestudyoflatentspeech.J.Psychol.,2,71-107.—394

241. SKINNER,B.F.(1937)Twotypesofconditionedreflex:areplytoKonorskiandMiller.J.gen.Psychol.,16,272-279.—47

242. SKINNER, B. F. (1938)The behavior of organisms: anexperimental analysis. NewYork,Appleton-Century-Crofts, ix,457pp.—14,20,30,44,45,46,50,71,77,80,85,87,88,90,91, 92, 95, 98, 106, 107, 108, 119, 120, 121, 162, 171, 186,202,207,234,267,268,282

243. SKINNER, B. F. (1948a) "Superstition" in the pigeon. J. exp,Psychol.,38,168-172.—103

244. SKINNER, B. F. (1948b) William James Lectures on verbalbehavior.HarvardUniversity.—400

245. SKINNER,B. F. (1950)Unpublished study.— 100 SKINNER,B.F.,seealsoNo.63.

246. SMALL, W. S. (1901) An experimental study of the mentalprocessesoftherat.Amer.J.Psychol.,12,206-239.—56,57

247. SMITH, S., and GUTHRIE, E. R. (1921)General psychology in

termsofbehavior.NewYork,Appleton,xii,270pp.—35248. SMOKE, K. L. (1932) An objective study of concept formation

Psychol.Monogr.,42,No.4,46pp.—157249. SOLOMON,R.L. (1946)Time and effort factors in the avoidance

ofrepetitionofresponses.Amer.Psychologist,1,291-292.—177350. SOLOMON, R. L. (1948) The influence of work on behavior.

Psychol.Bull.,45,1-40.—177,196251. SPENCE, K.W. (1936)The nature of discrimination learning in

animals.Psychol.Rev.,43,427-449.—158252. SPENCE, K.W. (1947) The r61e of secondary reinforcement in

delayedrewardlearning.Psychol.Rev.,54,1-8.—209253. SPRAGG, S. D. S. (1943)The relative difficulty of Morse code

alphabetcharacterslearnedbythewholemethod.J.exp. Psychol.,33,108-114.—148

254. STARCH,D. (1910)Ademonstrationof the trialanderrormethodoflearning.Psychol.Bull.,7,20-23.—169

255. STONE, C. P. (1927)The retention of copulatory ability in maleratsfollowingcastration.J.comp.Psychol.,7,369-387.—301

256. STONE,C. P. (1937)A sand-tube obstruction apparatus. J.genet.Psychol.,50,203-206.—96

257. STONE, C. P., and C OMMINS, W. D. (1936) The effect ofcastrationatvariousagesuponthelearningabilityofmalealbinorats:II.Relearningafteranintervalofoneyear.J.genet.Psychol.,48,20-28.—301

258. STRASSBURGER, R. C. (1950) Resistance to extinction of aconditioned operant as related to drive level at reinforcement.J.exp.Psychol.(inpress).

259. STROUD, J. B. (1931) The role of muscular tensions in stylusmazelearning.J.exp.Psychol.,14,184-185;606-631.—196

260. SWINDLE,P.F.(1917)Thetermreactiontimeredefined. Amer. J.Psychol,28,508-518.—229

261. SWINDLE, P. F. (1919) Some forms of natural training to whichcertainbirdsaresubjected.Amer.J.Psychol.,30,165-172.—229

262. TARDE, G. (1903)The laws of imitation (Trans, by E. C.Parsons).NewYork,Holt,xxix,405pp.—361

263. THORNDIKE, E. L. (1898)Animal intelligence.An experimentalstudy of the associative processes in animals.Psychol.Monogr.,2,No.8,109pp.—36,38,40

264. THORNDIKE, E. L. (1911)Animal intelligence: experimentalstudies.NewYork,Macmillan,viii,297pp.—14,39,61

265. THORNDIKE, E. L. (1913)The original nature of man(Educational psychology, I.) NewYork, Teachers College, vii,327pp.—40

266. THORNDIKE,E.L.(1931)Humanlearning.NewYork,Appleton-

Century-Crofts,206pp.—104,172,173,176267 THORNDIKE, E. L. (1932)The fundamentals of learning. New

York,TeachersCollege,xvii,638pp.—14,40,61,105268. THORSEN,A. M. (1925) The relation of tongue movements to

internalspeech.J.exp.Psychol.,8,1-32.—230269. TITCHENER, E. B. (1908)Lectures on the elementarypsychology

offeelingandattention.NewYork,Macmillan,ix,404pp.—255270. TITCHENER, E. B. (1915)A beginner's psychology. NewYork,

Macmillan,xvi,362pp.—219271.TOLCOTT, M.A. (1948) Conflict: a study of some interactions

between appetite and aversion in the white rat.Genet. Psychol.Monogr.,38,83-142.—322

272. TOLMAN,E.C.(1937)Theacquisitionofstring-pullingbyrats—conditioned responseor sign-gestalt.Psychol. Rev., 44, 195-211.—96

273. TOLMAN,E.C.(1942)Drives towardwar.NewYork,Appleton-Century-Crofts,xv,118pp.—325

274. TSAI,C.(1925)Therelativestrengthofsexandhungermotivesinthealbinorat.J.comp.Psychol.,5,407-415.—297

275. TSAI, L. S. (1932) The laws of minimal effort and maximalsatisfaction in animal behavior.Monogr. Nat. Res. Instit.Psychol.,Acad.Sinica,No.I,47pp.—196

276. VERPLANCK,W.S.(1942)Thedevelopmentofdiscriminationinasimplelocomotorhabit.J.exp.Psychol.,31,441-464.—129

277. WADA,T. (1922)An experimental study of hunger in its relationtoactivity.Arch.Psychol.,8,No.57,65pp.—290

278. WALKER, K. C. (1942) The effect of a discriminative stimulustransferredtoapreviouslyunassociatedresponse.J.exp.Psychol.,31,312-321.—242

279. WANG, G. H. (1923)The relation between 'spontaneous' activityand oestrus cycle in the white rat.Comp. Psychol. Monogr., 2 ,No.6,27pp.—295,296

280. WARDEN,C. J. (1924)The relativeeconomyofvariousmodesofattack inmastery of a stylusmaze.J. exp. Psychol., 7, 243-275.—60

281. WARDEN,C. J. (1931)Animalmotivation. NewYork, ColumbiaUniv.Press,xii,502pp.—288

282. WARDEN, C. J., and AYLESWORTH, M. (1927) The relativevalue of reward and punishment in the formation of a visualdiscrimination habit in the white rat.J. comp. Psychol., 7, 117-128.—111

283.WARDEN, C J., JENKINS,T. N., andWARNER, L. H. (1935-1940)Comparativepsychology,3vols.NewYork,RonaldPress.—58,163

284. WARNER,L.H.(1927)Astudyofsexbehaviorinthewhiteratbymeans of the obstruction method.Comp. Psychol. Monogr., 4,No.22,68pp.—298

285. WARNER, L. H. (1932)The association span of the white rat. J.genet.Psychol.,41.57-90.—311WARNER,L.H.,seealsoNo.283.WASHBURN,A.L.,seeNo.31.

286. WATERS,R.H. (1937)Theprincipleof least effort in learning. J.gen.Psychol.,16,3-20.—196

287. WATSON, J. B. (1916) The place of the conditioned reflex inpsychology.Psychol.Rev.,23,89-116.—35

288. WATSON, J. B. (1919)Psychology from the standpoint of abehaviorist.Philadelphia,Lippincott,ix,429pp.—7,379

289. WATSON,J.B.(1924)Behaviorism.NewYork,People'sInstitute,248pp.—338,343WATSON,J.B.,seealsoNo.301.

290. WATSON,J.B.,andRAYNER,R. (1920)Conditionedemotionalreactions.J.exp.Psychol.,3,1-14.—347

291. WELLS, F. L. (1908) Normal performance in the tapping test.Amer.J.Psychol.,19,437-483.—188

292. WHEELER, R. H. (1929)The science of psychology. NewYork,Crowell,xvii,556pp.—196

293. WILLIAMS, K. A. (1929) The reward value of a conditionedstimulus.Univ.Calif.Publ.Psychol.,4,31-55.—234

294. WILLIAMS,S.B. (1938)Resistance toextinctionasa functionofthe number of reinforcements.J. exp. Psychol., 23, 506-522.—72,73

295. WINNICK, W. A. (1949) Response vacillation in conflictsituations.Ph.D.dissertation,ColumbiaUniv.(tobepublished).—323,324

296. WOLFE, J. B. (1936) Effectiveness of token-rewards forchimpanzees.Comp.Psychol.Monogr.,12,No.60,72pp.—247

297. WOLFE, J. B., andKAPLON,M. D. (1941) Effect of amount ofreward and consummative activity on learning in chickens.J.comp.Psychol.,31,353-361.—74WOOD,A.,seeNo.189.WOODBURY,C.B.,seeNo.87.

298. WOODWORTH,R.S.(1938)Experimentalpsychology.NewYork,Holt,xi,889pp.—163,226,230,351

299. WOODWORTH,R.S.,andMARQUIS,D.G. (1947)Psychology.NewYork,Holt,x,677pp.—14

300. YERKES, R. M., and MORGULIS, S. (1909) The method ofPawlowinanimalpsychology.Psychol.Bull.,6,257-273.—35

301. YERKES, R. M., and WATSON, J. B. (1911) Methods ofstudyingvisioninanimals.Behav.Monogr., I,No.2, iv,90pp.

—137302. YOUNG, P.T. (1936) Motivation of behavior. NewYork,Wiley,

xviii,562pp.—325303. YOUNG, P. T. (1943) Emotion in man and animal. NewYork,

Wiley,xiii,422pp.—344,351304. YOUTZ, R. E. P. (1938) Reinforcement, extinction, and

spontaneous recovery in a non-Pavlovian reaction.J. exp.Psychol.,22,305-318.—76

305. ZEAMAN,D.(1949)Responselatencyasafunctionof theamountofreinforcement.J.exp.Psychol.,39,466-483.—73

306. ZIPF, G. K. (1949)Human behavior and the principle of leasteffort.Cambridge,Addison-Wesley,xi,573pp.—196

*John Dewey (1858–1952), philosopher and educator; James R.Angel(1869–1949), psychologist and university president; and Harvey Carr(1873–1954),psychologist.

*The best recent book on Freudian theory, especially in its relation tothe general field of psychology, is Calvin S. Hall'sPrimer of FreudianPsychology.NewYork:WorldPubl.Co.,1954,xii,137pp.