Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

download Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

of 10

Transcript of Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    1/10

    Pavlovian ConditioningI t ' s N o t Wh a t You Th i n k I t I s

    Rober t A. R escor la University o f Pennsylvania

    ABSTRACT: Current thinking about Pavlovian condi-tioning differs substantially fro m that o f 20 years ago . Yetthe changes that have taken place remain poorly appre-cia ted by psychologists generally. Traditional descriptionsof conditioning as the acquired ability of one stimulu s toevok e the original response to anothe r because o f theirpairing are shown to be inadequate. Th ey fa il to charac-terize adequately the circumstances producing learning,the content of that learning, o r the mann er in which thatlearning influences performance. Instead, condition ing isnow described as the learning o f relations a mo ng eventsso as to allow the organism to represent its environm ent.Within this framew ork, the stud y o f Pavlovian conditioningcontinu es to be an intellectually active area, fu ll o f newdiscoveries and inform ation relevant to other areas o fpsy-chology.Pavlov ian condi t ion ing is one of the o ldest and most sys-temat ica l ly s tud ied phen omen a in psychology. Outs ide o fpsychology, i t is one of our best k now n findings. But atthe same t ime, wi th in psychology it i s bad ly misunder -stood and misrepresented. In the last 20 years, knowledgeof the associative processes underlying Pavlovian condi-t ion ing has expanded dramat ica l ly . The resu l t i s tha tmodern th ink ing about condi t ion ing is comple te ly d i f -ferent from the views psychologists held 20 years ago.Unfortun ately, these changes are very poorly appreciate dby psychologists at large. The last tim e m an y psychologistsread any th ing about Pav lovian condi t ion ing was beforethese changes took place. Even those more recently ed-ucated of ten rece ived tha t educat ion f rom tex tbooks andinst ruc tors tha t had la rge ly ignored the d ramat ic concep-tua l changes tha t had taken p lace . The resu l t is tha t m anyth ink of Pav lov ian condi t ion ing as an obso le te techn ica lf ield that is intellectually stagnant.My in tention in this ar t icle is to show t hat this viewis incorrect. First , I will review som e o f the c hanges tha thave occurred in Pavlovian conditioning in order to givethe f lavor o f i ts con tem porary fo rm. I wi l l a rgue tha t i tis an intellectually challenging f ield, in which substantialand excit ing progress has been made . Second, I will arguetha t condi t ion ing con t inues to have a cen t ra l p lace inpsychology generally. I will describe how it touches onand info rms several related f ields that are curr ently m orein vogue.To begin the discussion, consider how conditioningwas described 20 years ago, when those in m y generat ionwere s tuden ts . One popular in t rod uctory tex t pu t i t thus:

    The essential operation in conditioning is a pairing of two stimuli.One, initially neutral in that it elicits no response, is called theconditioned stimulus (CS); the other, which is one that consis-tently elicits a response, is called the unconditioned stimulus(US). The response elicited by the unconditioned stimulus isthe unconditioned response (UR). As a result of the pairing ofthe conditioned stimulus (CS) and the uncondition ed stimulus(US), the previously neutral conditioned stimulus comes to elicitthe response. Then it is called the conditioned response (CR).(Morgan & King, 1966, pp. 79-80)This descr ip t ion is typ ica l o f those fo und in b o th in t ro -ductory an d advanced tex tbooks 20 years ago .Unfo rtunately , i t is also typical of wha t one f inds intex tbooks today. One popular in t rodu ctory tex t pub l ishedin 1987 describes conditioning in this way: "The originallyneu tra l condi t ioned s t imulus , th rough repeated pa i r ingwith the unco ndi t ioned one , acqu ires the response o r ig -ina l ly g iven to the uncondi t ioned s t imulus" (Atk inson ,Atkinson, Smith, & Hilgard, 1987, p. 658). Students areexposed to similar descriptions in textbooks specializingin all ied f ields of psychology. In a cognitive textbook , onereads,We start out by taking an unconditioned stimulus (UCS) thatproduces the desired response without training . . . We pairthe UCS with a conditioned stimulus (CS) . . . This procedure,when repeated several t i m e s . . , will ultimately result in theoccurrence of the response following the C S alone. (Klatsky,1980, p. 281)A widely used developmental text agrees, call ing condi-t ion ing a " form of learn ing in which a neu tra l s t imulus ,when pa i red repeated ly wi th an un condi t io ned s t imulus ,even tua l ly comes to evoke the o r ig ina l response" (Gard-ner, 1982, p. 594). Similarly, a best-selling textb ook o fabno rmal psychology describes a condi t ioned s t im ulusas "a s t imulus tha t , because o f it s hav ing been pa i redwith ano ther s t imulus (uncond i t ioned s t imulus) tha t na t -u ra l ly p rovokes an uncond i t ioned response , i s even tua l lyab le to evoke tha t response" (Rosenhan & Sel igman,1984, p. 669).Of course, textbook descriptions va ry widely in theirprecision and sophistication, but these citations representa co mm on v iew. Indeed , these quota t ions wi l l cer ta in lysound so famil ia r tha t m any readers may wonder what i swrong wi th them. I want to suggest tha t the answer i s"a lm ost every th ing ." These descr ip t ions make asser t ionsabout w hat I take to be the prim ary issues to be addressedin the s tudy of any learn ing process: W hat a re the c i r-cumstances tha t p roduce learn ing? What i s the con ten t

    March 1988 9 Amer ican Psycholog is tCOl~ri~at 1988 by the Am erican Psychological Association, Inc. 0003-066X/88/$00.75Vol. 43, No. 3, 151-160

    151

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    2/10

    of the learning? How does that learning affect the behaviorof the o rgan ism? But they are mistaken or mislead ing invir tually every assertion they m ake abou t eac h of these.These descr ip t ions in fac t cap ture a lmost no th ing ofmode rn da ta and theory in Pav lov ian condi t ion ing .I want to i l lustrate this claim using some data col-lected in my own laboratory over the years, but f irst letme m ake an or ien t ing com men t . D escr ip t ions o f condi-t ioning, such as those jus t cited, com e from a long andhonorable tra dit ion in physiology, the ref lex tradit i on inwhich Pavlov worked and wi th in which ma ny ear ly be-haviorists thought. This tradit ion sees conditioning as akind of low-level mech anical process in which the controlover a response is passed from one stimulus to another .Much modern th ink ing about condi t ion ing ins tead de-r ives largely from the associative tradit i on originating inphilosophy. I t sees conditioning as the learning that resultsf rom exposure to re la tions amo ng even ts in the env iron-ment . Such learn ing is a p r imary means by which theorganism represents the structure of i ts world. Conse-quently, Pavlovian conditioning must have considerablerichness, both in the relations i t represents and in theways its representation influences behavior , a r ichness th atwas not envisioned within the ref lex tradit ion.Let m e n ow turn to i l lus t ra t ing the d i f ference tha tthis alternative view makes for each of three issues: thecircumstances prod ucing learning, the conte nt of learning,and the effects of learning on behavior .Circumstances Producing PavlovianConditioningEach of the descriptions given earlier cites one m ajor cir-cums tance as responsib le fo r p roducing Pav lovian con-dit ioning, the pair ing or contiguity of two events. To besure, con t igu i ty remains a cen t ra l concep t , bu t a m odernview of cond itionin g as the learning of relations sees con-tiguity as neither necessary nor suff icient. Rather , thatv iew emphasizes the in form at ion tha t one s t imulus g ivesabout ano ther . We now know tha t a r rang ing for two wel l-p rocessed even ts to be con t iguous need no t p roduce anassocia tion be tween them ; nor does the fa i lu re to a r rangecontiguity preclude associative learning.The insuf f ic iency of con t igu i ty fo r p roducing Pav-lov ian condi t ion ing can be i l lus t ra ted by resu l ts tha t havebeen available for almost 20 years (e.g., Rescorla, 1968)bu t tha t have apparen t ly fa i led to be in tegra ted in to theview of condition ing held by m an y psychologists. Considera learning situation in which a rat is exposed to twoprom inent events , a tone CS tha t occurs fo r two-m inuteperiods and a brief , mild electr ic shock US applied to agr id on which the an imal i s s tand ing . Suppose tha t thosetwo even ts a re uncor re la ted in t ime, such tha t the toneThis article is an adaptation of a Presidential Address given to the EasternPsychological Associati on in Arling ton, VA, in Ap ril 1987. The researchreported here was generously supported by grants from the NationalScience Foundation.Correspon dence concernin g this articl e should be addressed toRob ert A. Rescorla, Dep artm ent of Psychology, Univers ity of Pennsyl-vania, 3815 Walnut St., Philadelphia, PA, 19104.

    provides no in format ion about the shock . That re la t ionis schem atized in the top of Figure 1. Also schematiz edin that f igure is a variation on that treatment in whichonly those USs scheduled to occur dur ing the tone areac tua l ly app l ied to the an imal . The po in t to no t ice aboutthose two treatment s is that they share the same contiguityof the tone wi th the US, bu t they d i f fer in the am ou nt o fin format io n tha t the tone g ives about the US. In the f i r stt rea tment , the shock is equal ly l ike ly whether o r no t thetone is p resen t, an d so the tone prov ides no in format ion ;in the second t rea tment , the shock on ly occurs dur ingthe tone , and so the tone i s qu i te in format ive about shockoccur rence . I t tu rns ou t tha t in ma ny condi t ion ing s i tu-a t ions learn ing is de termined no t by what these t rea tmentsshare but rather by how they differ . The second groupwill develop an association between the CS and US, butthe first will fail to d o so. In effect, con ditio ning is sensitiveto the base ra te o f US occur rence against which a CS/US co ntiguity takes place.Indeed , systemat ic exper iments show tha t in man ysi tua t ions the a mo unt o f condi t ion ing is exquis i te ly a t-tuned to var ia t ions in the base ra te o f the US. An ear lyil lustration of that po int is shown in Figu re 2, which plotsasympto t ic leve ls o f fear condi t ion ing (measured by theabili ty of the CS to interfere with on going behavior) as afunct ion of the l ike lihood of the U S d ur ing the CS. Theparam eter in the f igure is the base-rate l ikelihood o f theUS in the absence of the CS. Eac h curve shows tha t con-dit ioning is indeed an increasing function of the l ikelihoodof the shock dur ing the tone . For ins tance , in the f re-quent ly studied case in whic h the sho ck l ikelihood is zeroin the absence of the CS, then condi t ion ing is g rea ter thegrea ter the p robabi l i ty o f the shock dur ing the tone . Th isis not a surprising result. W hat is more intere sting is theeffect of the base rate of US occ urren ce in the abs ence o fthe CS. At any g iven l ike l ihood of shock dur ing the CS,condi t ion ing is an inverse funct ion of the base ra te . W henthe CS/US cont igu i ty i s he ld constan t , condi t ion ingchanges from exce llent to negligible sim ply by increasingthe shock base rate. Indeed, whe n the l ikelihoo d of a USis the sam e in the presence an d absence of the CS (as istrue of the init ial poin t on eac h func tion), there is l it t leev idence of condi t ion ing a t a l l. O ne descr ip t ion o f theseresu lts i s tha t condi t ion ing depends no t on the con t igu i ty

    Figure 1Schematic of Two Conditioned Stimulus/Unconditioned Stimulus (CS/US) Relations ThatShare the Same Contiguity but Differ in theInformation the CS Gives About the UScs_.l--I I~1 I~1USl I II I I II I I II I I I I

    CS_l"'l [" '1 1"-1US I I I

    152 Ma rch 1988 9 Amer ic an Psycholog is t

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    3/10

    Figure 2Dependence of Condi t ioning on Both the Likel ihood ofthe US Dur ing the CS and the Base Rate of USOccur rence in the Absence of the CS

    P (us/cs)Note. Adapted from "Probabil i ty of Sho ck in the Presence and A bsence of CSin Fear Conditioning" by R. A. Rescorla, 1968, Journal of Comparative andPhya/olog/cal Psychology, 66, p. 4. Copyright 1968 by the American PsychologicalAssociation. Reprinted by permission. Asymptotic fear conditioning is plotted asa function of shock likelihood during the conditioned stimulus (CS). The parame teris the shock likelihood in the absence of the CS. When the CS/U S cont iguity[P(US/CS)] Is held constant, cond itioning varies from substantial to negligibleas a function of the US (unconditioned stimulus) base rate. Conditioning is indexedby a ratio comparing responding during the CS with the ongo ing response rate.With that ratio, 0.5 indicates no conditioning an d O ndicates excellent conditioning.

    between the CS and the US bu t ra ther on the in form at ionthat the CS provides about the US. These are early data,but the basic results have been observed repeatedly in avar ie ty o f condi t ion ing prepara t ions . They s t rongly sug-gest that sim ple contiguity of CS an d US fails to cap turethe relation required to produce an association.The same conclusion is suggested by various otherm o d e r n co n d it i o nin g p h en o m en a , su ch a s th e Kam in(1968) blocking effect . That effect has had a profoundimpact on con temporary th ink ing about Pav lov ian con-d i t ion ing , yet i t is unkn own to ma ny psycholog is ts . In as imple b lock ing exper iment , two groups o f an imals re -ceive a compound stimulus (such as a l ight and tone)signaling a US. Ev entually both grou ps will be tested fortheir conditionin g of one stim ulus, say the tone. However,one o f the groups has a history of the l ight alone signalingthe U S, whereas the other group lacks that history. Noticethat the two groups share the same contiguous occurrenc eof the US w i th the l igh t / tone compou nd , bu t they d i f ferin tha t fo r one the p r io r t ra in ing of the l igh t makes thetone redundant . The resu lt o f in teres t i s tha t the tonebecomes well conditioned in the f irst group but poorlyconditioned in the group with l ight pretraining. Condi-t ion ing is no t governed by the con t igu i ty tha t the g roupsshare bu t ra ther by the in form at ional re la t ion on w hichthey differ . Again, simple contigu ity of two events fails tocap ture the resu lts ; rather, in format ion seems im por tan t .

    This is a result that has been widely repeated in manycondi t ion ing s i tua tions .These two classic experiments i l lustrate that conti-guity is not suff icient to pro duce Pavlovian conditioning.But neither is contiguity necessary to produce Pavlovianassociations. This can be i l lustrated in a va riety of ways,bu t a s imple one makes reference to the t rea tments inFigure 1. Consider a variation on the f irst treatment inwhich , ins tead of omit t ing a l l o f the shocks in the absenceof the tone, we omit all those in i ts presence. This variationtakes away all of the CS/U S contiguities while ma intai ninga h igh base ra te o f US occur rence . Un der these c i rcum-stances, the organism does n ot sim ply fail to learn; rather ,i t learns that there is a negative relation b etween the to neand the US. In the ja rgon of the f ie ld, the tone becom esa condi t ioned inh ib i to r . Again , th is ou tcome is no t in -tu i t ive ly surpris ing , bu t ne i ther i s i t well acco mm oda tedby the descr ip tion of condi t ion ing in which the main c i r-cumstance producing learn ing is con t igu i ty . Yet condi-t ioned inh ib i t ion is a major par t o f modern th ink ing abo utPavlov ian condit ion ing . No theory of condi t ion ing wouldbe considered adequate if i t failed to explain a wide varietyof inhibi tory phe nom ena (cf. Miller & Spear, 1985).These kin ds of results clearly suggest that the simplepai r ing of two even ts cannot be taken as fundam enta l tothe descr ip t ion of Pav lovian condi t ion ing . Ins tead , the yencourage the p revalen t modern v iew tha t condi t ion inginvolves the learning of relations amo ng events. I t providesthe an imal wi th a muc h r icher represen ta t ion of the en-v i ronment than a re f lex t rad i t ion would ever have sug-gested. Of course, one cannot leave the analysis at thislevel; rather , one ne eds to prov ide theo ries o f how thesere la t ions are coded by the o rgan ism. Such theor ies a renow available, several of which are stated in suff icientquantitative detail to be taken seriously as useful accounts(e.g., Mackintosh, 1975; Pearce & Hall, 1980; Rescoda& Wagner, 1972). These theories emphasize the impor-tance of a d iscrepancy be tween the ac tua l s ta te o f thewor ld and the o rgan ism's represen ta t ion of tha t s ta te.They see learning as a process by which the two arebrough t in to l ine. In effect , they offer a sophistic ated re-formula t ion o f the no t ion of con t igu ity . A usefu l short -hand is that organism s adjust their Pavlovian associationsonly when they are "surpr ise d ." This i s no t the p lace todescribe these theories in detail , but they do an excellentjob wi th ph enom ena l ike those descr ibed in Figure 2 .The impor tanc e of re la t ions can be seen in ye t an-other way. I t is not only temporal and logical relationsamong even ts tha t a re impor tan t to condi t ion ing . Con-ditioning is also sensit ive to relation s involving the p rop-erties of the events themselves. The re is a kin d o f ab-stractness with w hich the descriptions o f conditioning areoften stated, an abstractness th at is characterist ic of afield seeking general principles. These descri ptions suggesttha t condi t ion ing occurs whenever one ar ranges a tem-poral relation am ong the events, regardless of the othe rproperties of the events. The claim in essence is tha t thean imal comes to condi t ion ing wi th no preconcep t ionsabout the s t ruc ture o f the wor ld , ready to accom mod ate

    March 1988 9 Am erican Psychologist 153

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    4/10

    itself tO any world that it faces. Pavlovian condit ioninghas, of course, served as one o f the pillars for radical em-piricism. But in modern times it has become clear thatthis pillar itself is partly built on the existing structure inthe organism. Not all stimuli are equally associable; in-stead, a stimulus may be easier to associate with somesignals rather than others. The most well-known dem-onstration of this, of course, is Garcia and Koelling's(1966) seminal work on the cue-to-consequence effect.They found that an internal distress was easier to associatewith a gustatory rather than an auditory-visua l stimulus,whereas a peripherally administered pain was morereadily associated with the auditory-visual rather thanthe gustatory stimulus.But this work is not alone in identifying instancesof preferential learning among stimuli bearing qualitativerelations to each other. For instance, spatial relationship,a variable importan t to philosophical associationism butneglected by the reflex tradi tion, is now known to affectPavlovian associations (e.g., Rescorla, 1980). Similarly,recent work shows that perceptual relations among events,such as similarity and the part-whole relation, also areimportant determinants of conditioning.Figure 3 shows an example of how one perceptualrelation (part to whole) affects the results of Pavlovianconditioning. Those results come from an autoshapingexperiment in pigeons. Autoshaping is one of the mostpopular modern Pavlovian preparations, so it is worthmentioning in its own right. In that preparation, birdsare exposed to a response-independent signaling relationbetween an illuminated disc (say, a red square or a redtriangle) and food. As the birds learn that relation, they

    Figure 3Ef fect o f a Par t -W hole Re la t ion onPavlov ian Cond i t ion ing

    BLOCKS OF 4 TRIALSNote. From Psvlovian Seco nd-O rder Conditioning: Studies in A s s o c i a t i v eLearning (p. 49) by R. A. Rescorla, 19 80, Hillsdale, NJ: Erlbaum. Copyrig ht 198 0by Lawrence Erlbaum Associates. Reprinted by permission. Responding is shownto two second-order stimuli, an outline triangle and outline square, that signaleda colored tdangle or a colored square. In the similar group, each outline formsignaled a colored form o f the same shape; in the dissimilar group, each signaleda colored form o f a different shape.

    come to peck the disc. That result is worth analysis of itsown, but for the present we will simply take it as an indexthat the birds have associated the red square with food.More important for our present purposes, they will pecknot only the red square but also localized stimuli that inturn signal the red square (producing so-called second-order conditioning). Figure 3 shows the development ofpecking at two stimuli, colorless outlines of a square anda triangle, when they signal a red square and a red triangle.For the animals having a similar relation, each coloredfigure was signaled by the same-form achromatic figure;each whole was signaled by one o f its parts. For the an-imals having a dissimilar relation, the colored figures werealso signaled by the achromatic figures except that theforms were mismatched so as to destroy the part-wholerelation. It is clear that condi tioning proceeded more rap-idly in animals who had the part -whole relation. Thatis, a perceptual relation influenced the formation of anassociation. This is a particularly interesting perceptualrelation because in the natural environment partial in-formation about an object frequently serves as a signalof the entire object. Apparently, Pavlovian conditioningis especially sensitive to that fact.One final comment needs to be made about the cir-cumstances that produce conditioning. It is a common lyheld belief that Pavlovian conditioning is a slow processby which organisms learn only if stimulus relations arelaboriously repeated over and over again. Several of thedescriptions cited earlier acknowledge this belief by usingsuch terms as repeatedly and eventually. However, thisview is not well supported by modern data. Althoughcondit ioning can sometimes be slow, in fact most modernconditioning preparations routinely show rapid learning.One-trial learning is not confined to flavor-aversionlearning, and learning in five or six trials is common. Infact, the data displayed in Figure 3 are a good exampleof learning that is excellent after eight trials. Notice thatthose data were obtained in a second-order conditioningparadigm, a procedure that itself has an undeserved rep-utat ion for being weak and t ransient (see Rescorla, 1980).The picture that emerges from this discussion of thecircumstances that produce conditioning is quite differentfrom that given by the classical descriptions. Pavlovianconditioning is not a stupid process by which the organismwilly-nilly forms associations between any two stimulitha t happen to co-occur. Rather, the organism is betterseen as an information seeker using logical and perceptualrelations among events, along with its own preconcep-tions, to form a sophisticated represen tation o f its world.Indeed, in teaching undergraduates, I favor an analogybetween animals showing Pavlovian conditioning andscientists identifying the cause of a phenomenon. If onethinks of Pavlovian condition ing as developing betweena CS and a US under just those circumstances that wouldlead a scientist to conclude that the CS causes the US,one has a surprisingly successful heuristic for remem-bering the facts of what it takes to produce Pavlovianassociative learning (see Dickinson, 1980; Mackintosh,1983).

    154 March 1988 9 American Psychologist

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    5/10

    Content of Pavlovian Conditioning:What Is LearnedThe descr ip t ions o f cond i t ion ing g iven ear l i e r imply ahighly res t r icted con tent in which a s ingle neutra l s t imulusbecom es assoc ia ted wi th one tha t evokes a response . Bu tmo de rn Pavlovian thinking suggests a picture that is r icherin two ways.

    Fi rs t , i t i s c l ear tha t in any Pav lov ian exper imentthe an imal l earns abou t many d i f fe ren t s t imul i . Associ -a t i on s a r e f o r m e d n o t j u s t b e t w e e n t h e p r i m a r y e v e n tspsycho log i s ts p resen t , t he CS and US. For ins tance , eachof those even t s a lso become s assoc ia ted wi th the con tex tin which they are presente d (e.g. , Balsam & To mie, 1985).Such assoc ia t ions a re one way tha t o rgan i sms use Pav-lov ian cond i t ion ing to code spa t i a l in fo rmat ion . More-over, as soc ia t ions fo rm no t on ly be tween even t s bu t a l sowi th in eac h o f the even ts tha t the t rad i t iona l descr ip t ionident i f ies (e.g . , Rescorla & Durlach, 1981). Indeed, con-s iderab le e f fo r t is go ing in to a na lyz ing the l a t t e r l earn ingbecause wi th in -even t as soc ia t ions may be one way tha tthe o rgan i sm represen t s ind iv idual even t s . Moreover ,ma ny examples o f Pav lov ian assoc ia t ions invo lve s t imul itha t do no t evoke an o r ig ina l response . Pav lov ian con-d i t ion ing a lso encodes the re l a t ions amo ng re la tive ly in -n o c u o u s e v e n t s . So , mo d e r n e x p e r i me n t a t i o n s u p p o r t sthe p ropos i t ion tha t the o rgan i sm concurren t ly fo rms abroa d range o f as soc ia t ions am ong a wide var i e ty o f s tim-u l i . Moreover , qu i t e powerfu l p rocedures have been de-ve loped to expose the ex i s t ence o f these assoc ia t ions a ndto ca r ry ou t an ana lys i s o f the i r p roper t ies .Second , modern Pav lov ian th ink ing does no t env i -s ion a ll o f th i s l earn ing t ak ing p lace am ong s imple pa i rso f e l ements a l l t rea ted a t the same l eve l o f ana lys is b y theorganism. Rather, as the Bri t ish associat ionis ts claimedyears ago, there is good reason to bel ieve that there is ah ierarch ica l o rgan iza t ion in which assoc ia t ions amongsome pa i rs o f it ems y ie ld new en t i t ies tha t themselves canen ter in to fu r ther as soc ia tions .One i l lus t ra t ion comes f rom a recen t second-ordera u t o s h a p in g e x p e r i me n t c o n d u c t e d i n my l a b o r a to r y , th eexper im enta l des ign o f which i s shown in Figure 4 . Inth i s exper iment , one s t imulus (X) s ignaled the occ urrenc eo f a c o mp o u n d s t imu l u s c o mp o s e d o f a k e y li g h t t h a t w a sred (R) on one ha l f and h ad hor izon ta l s t r ipes (H) on theo ther ha l f. The b i rds were in te res ted in tha t fac t becauseR an d H each had a separa te h i s to ry o f s ignal ing theo c c u r r e n c e o f fo o d . P r e vi o u s e x p e r i me n t s h a d d e mo n -s t ra t ed tha t the b i rds would come to peck X as a resu l to f it s se c o n d - o rd e r c o n d it i o n in g b y t h e RH c o mp o u n d .Th e quest ion of interest was what would be th e associat ives t ruc tu re tha t suppor ted tha t peck ing . One poss ib i l i ty i stha t the b i rd would fo rm two pa i rwise assoc ia t ions ,l earn ing the ind iv idual as soc ia tions o f X wi th R and wi thH. But a m ore in te res t ing poss ib i l ity i s tha t the o rgan i smw o u l d f o r m a r e p r e s e n t a t i o n o f t h e R H e v e n t ( p er h a p sus ing the assoc ia t ion we know fo rms be tween R and Hin such se t t ings ) and then use tha t rep resen ta t ion as ane lemen t to as soc ia te wi th X. E i ther as soc ia tive s t ruc tu re

    Figure 4Design of an Experiment DemonstratingHierarchical Organization

    H+ y~-I~R RH-, R+, H+H

    Note. Birds received f i rst-order Pavlovian condit ioning of two keyl ights (R andH) with a food (+) u ncondit ioned stimulus (US). Then one s econd -order st imulus(X ) signaled the RH compound, whereas another (Y) signaled the elements.Then the birds received one of tw o condit ional discr iminations bet ween the R Hcompound and i ts e lements and were tes ted for the response to X and Y. Thephysical identities of the X and Y stimuli were counterba lanced as a b lue keyl ightand a black X on a white background.

    would cause the b i rd to show cond i t ion ing to X, bu t thefo rm er so lu t ion invo lves two para l l e l as soc ia t ive connec-t ions , whereas the lat ter involves a hierarchical organi-za t ion .T h e t e c h n o l o g y o f mo d e r n Pa v lo v i an c o n d i t i o n i n gprovides a way to separate these tw o al ternat ives . In m an ycond i t ion ing p repara t ions , respond ing to a s ignal t racksthe cu rre nt s tate of its associate (e.g ., Rescorla, 1980). Ifthe va lue o f a re in fo r cer is changed af t e r cond i t ion ing hasbeen comple ted , subsequen t respond ing to i t s as soc ia tedCSs wil l also change accordingly. This fact can be usedto dec ide wi th which s t imulus X has b eco me assoc ia ted .In thi s ins tance , we de l ibera te ly gave the R H com pou ndand i t s e l ements d i f fe ren t va lues . For some an imals , weex t ingu i shed the separa te ly p resen ted R and H e lementsb u t r e i n f o r c e d t h e RH c o m p o u n d ; f o r o t he r s , we d i d t h econverse . Then we t es t ed respond ing to X. I f the an im alhas on ly separa te as soc ia t ions o f X wi th the R a nd Helements , respond ing to X shou ld t rack the va lue o f thosee lements, bu t i f X has an assoc ia t ion wi th RH , respond ings h o u l d t r a c k t h e c o mp o u n d ' s v a l u e r a t h e r t h a n t h a t o fthe R and H e lements . In o rder to compare the resu l t sf rom X wi th those f rom an assoc ia t ive s t ruc tu re tha t weknow to represen t s imple pa i rwise assoc ia t ions , we a l soused R and H to cond i t ion ano th er s t imulus (Y) . L ike X,Y was fo l lowed by R and H, bu t un l ike X, Y rece ived Rand H on separa te t r ia l s , t hereby ensur ing i ts hav ing sep-arate associat ions with those elements . As a resul t , re-spond ing to Y shou ld t rack the c ur re n t va lue o f the Ra n d H e l e me n t s , n o t t h a t o f t h e RH c o m p o u n d .The resu l ts o f var ious s tages o f th i s exper ime nt a reshow n in Figu re 5 . The f i rs t pane l shows the level of re-spond ing to X and Y a t the end o f the i r second-orde rc o n d i t io n i n g b y th e RH c o m p o u n d a n d t h e R a n d Helements . Those t rea tment s p roduced s imi lar l eve l s o fcond i t ion ing . On tha t basi s a lone , one can no t iden t i fy anyd i f ferences in the assoc ia t ions o f X and Y. The midd lepanel shows the course o f the d i scr imin at ions be tweenRH and i t s e l ements . The b i rds cou ld read i ly code a com-pou nd and i ts e l ements d i f fe ren ti a lly , a resu l t o f somein teres t in i tse lf . Bu t the da ta o f mos t in te res t a re thoseshown in the f inal panel, f rom the t es t ing o f the second-order X and Y s t imuli . Cons id er f i rs t the resu l t s f rom Y,

    March 1988 9 Am er ican Psycho log i s t 155

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    6/10

    Figure 5Results of an Experiment Demonstrating Hierarchical Organization

    D A Y SNote. The left panel shows asymptotic second-order conditioning of X and Y by the RH comp ound and the R and H elements, respectively. The m iddle panel showsa conditional discrimination of the form RH+ , R -, H - (solid symbols) or RH -, R+, H+ (open symbols). The right panel shows responding to X and Y as a functionof the most recent treatment of the RH compo und an d its elements. In both cases, responding to X and Y tracked the current value of the stimulus that it hadsignaled.

    which had signaled R and H separately. Responding tothat stimulus tracked the value of the individual R andH elements, not the value of the RH compound. Underthose conditions, individual associations are indeedformed. Quite different are the results of testing X, thestimulus that had signaled the RH compound. Respond-ing to that stimulus tracked the current value of the RHcompound rather than the value of its elements. Clearly,the animals had not simply coded the RH compound interms of parallel associations with its elements. Rather,they had engaged in some more hierarchical structuringof the situation, forming a representation of the com-pound and using it as an associate. This is the kind ofhierarchical organization envisioned by the British as-sociationists; it is extremely important because it mayprovide a means for an associative theory to build complexperformances by bootstrapping based on elementarymechanisms. Such hierarchical structures are often dis-cussed in various learning literatures, but they turn outto be very difficult to document definitively. One dem-onstration, however, can be given within the frameworkof Pavlovian conditioning.Another illustration of such a hierarchical structurecomes from recent demonstrations of a phenomenonvariously called "occasion-setting" and "facilitation"(Holland, 1983; Rescorla, 1985). That phenomenon arisesin situations in which a Pavlovian stimulus is deliberatelyarranged to signal not another stimulus but rather a re-lation between two other stimuli. Under proper condi-

    tions, such learning readily develops. Moreover, it can berelatively independent of the learning of separate asso-ciations to the elements. For instance, a stimulus thatsignals a positive relation between two other stimuli cansimultaneously have either excitatory or inhibitory as-sociations with the elements themselves.Several laboratories are currently actively engagedin analyzing this kind o f hierarchical relation. Their find-ings have important general implications for our under-standing of Pavlovian conditioning. They suggest that as-sociations may play a modulatory, rather t han an elicitive,role. And they are changing the way we think about ex-citatory and inhibitory associations. Moreover, thinkingabout this modulatory role is beginning to be brought tobear on the analysis of stimulus control in instrumentallearning.Modem thinking about Pavlovian conditioning viewsassociations as basic, but those associations are formedamong representations of multiple events. Moreover, thoserepresentations themselves are often complex and includerelations generated by other associations. Pavlovian con-ditioning does not consist simply of learning relationsbetween a neutral event and a valuable event. Many dif-ferent associations are formed, and the resulting content.of learning allows a rich representa tion o f the world.I n f l u e n c e s o n B e h a v i o rThe descriptions quoted earlier contain a highly restrictedview of how conditioning affects behavior. They envision

    156 March 1988 9 American Psychologist

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    7/10

    o n l y o n e w a y i n w h i c h p e r f o r ma n c e i s g en e r a te d : T h e CSbecom es capab le o f evok ing the response o r ig ina l lyevoked by the US. How ever , there a re very few s tuden t so f cond i t ion ing who w ould care to defend tha t c l a im.There a re th ree reasons why I be l ieve it shou ld be re j ec ted .Fi rst , man y o f the s t andard c ond i t ion ing p repara-t ions s imply do no t show th i s fea tu re . Cons ider , fo r in -s t ance , cond i t ioned suppress ion s i tua t ions such as thoseused to co l l ec t the da ta shown in Figure 2 . The responseto the shock US i s ab rup t ly increased ac t iv i ty , whereasthe response to a ton e s ignal ing tha t shock i s d ram at ica l lyreduce d ac t iv i ty.Second , there i s an impor tan t , bu t poor ly appre-c ia ted , fac t abou t cond i t ion ing tha t makes nonsense ou to f any c la im tha t a s ignal s imply ac qu i res the ab i l i ty toevoke the response to the US: The response observed toa CS of t en depends no t on ly on the US bu t a l so on theperce ptual prope rt ies of the CS i tself . Two different s ignalso f the sam e U S may ev oke qu i t e d i ffe ren t responses . Forinstance, for rat subjects a diffuse tone that s ignals a sho ckUS resu l t s in immobi l i ty , bu t a loca l i zed p rod s ignal ings h o c k r e s u lt s i n a t t e mp t s t o h i d e t h e p r o d f r o m v i e w b ycovering i t wi th an y avai lable material (e.g . , Pinel & Trei t ,1979). Sim ilarly , di fferent CSs s ignal ing fo od to a pigeo ncom e to p ro duce qu i t e d i ffe ren t response fo rms . As no tedabove , a loca l i zed v i sua l s igna l o f food evokes d i rec tedpeck ing . However , a d i f fuse aud i to r y s ignal o f tha t samef o o d d o e s n o t e v o k e p e c k in g b u t r a t h e r e n h a n c e s g e n e r a lact ivi ty . Figure 6 shows a relevant i l lus t rat ion from a re-c e n t e x p e r i me n t i n o u r l a b o r a t o ry . T h e l e f t- h a n d p a n e lo f tha t f igure shows the resu l t s o f g iv ing the sam e b i rdsseparate keyl ight and audi tory s ignals for food. I t i s cleartha t the b i rds came to peck dur ing the key l igh t bu t no tdur in g the tone . Bu t the absence o f peck ing does no tresu l t f rom a fa i lu re o f l earn ing abou t the tone . Di rec t

    Figure 6Dependence of the Form of the Co ndi t ionedResponse on the Ident i ty of the Signal

    Note. The left panel shows keypecking in birds for whom both a localized keylightand a diffuse ton e signaled food. The right panel shows keypecking to two oth erlocalized keyUghts (X and Y ) that signaled th e light and tone, respectively. Thephysical identities of X and Y were counterbalanced as red and green.

    observa t ion o f the b i rd shows tha t the tone p rod uces en-hanced genera l ac t iv i ty . Moreover , the r igh t -hand panelof Figure 6 suggests that al thoug h the ton e an d l ight evoked i f feren t responses , the b i rd has in some sense l earnedthe same th ing abou t the two s t imul i . That panel showsthe resu l t s o f a second s t age o f the ex per im ent in w hichthe l igh t and the tone were each s ignaled by an o the r key-l igh t (X and Y). Bo th the tone and the l igh t served asexce l l en t re in fo rcers , thereby d i sp lay ing tha t they had be-come assoc ia ted wi th food . There a re two po in t s to no tef rom th i s dem ons t ra t ion : F i rst , t he fo rm o f the cond i -t ioned response var i es f rom CS to CS, and so i t canno ta lways be l ike the response to the US . S econd , som et ime swe have d i f f i cu l ty see ing any ev idence o f l earn ing i f wes imply look a t the responses e l i c i t ed by the CS; ra ther ,other m easures , su ch as the abi l i ty to serve as a reinforcer,can o f t en p rov ide be t t e r ev idence o f learn ing .The th i rd rea son to re j ec t the c lass ica l no t ion o f howcondi t ion ing af fec t s per fo rmance i s tha t there i s a sensein which the response one sees to a Pav lov ian CS can bearb i t ra r i ly se lec ted by the expe r imenter . Th at i s poss ib leb e c a u s e o n e i m p o r t a n t f e a t u re o f Pa v l o v ia n c o n d i t i o n in gi s i t s invo lvement in goal -d i rec ted ins t rumenta l per fo r -mance . I t has been known fo r years tha t Pav lov ian con-d i t i o n i n g ma k e s i mp o r t a n t c o n t r i b u t i o n s t o t h e c o n t r o lo f e mo t i o n s a n d mo t i v a ti o n s . T w e n t y y e a r s ag o , o n e o fmy mos t respec ted p ro fessors , Frank I rwin , asked mehow I cou ld be in te res ted in Pav lov ian cond i t ion ing , ap rocess tha t he charac ter i zed as be ing "a l l sp i t andtwi tches" and o f l it t le genera l psycho log ica l in te res t . Bu ti t i s impo r tan t to unders t and tha t I rw in 's charac ter i za t ionwas wrong . Condi t ion ing i s in t imate ly invo lved in thecon t ro l o f cen t ra l psycho log ica l p rocesses, such as emo -t ions and mot iva t ions . In fac t , two-process theor ies o fi n s t r u me n t a l p e r f o r m a n c e a r e b u i l t o n t h a t p r o p o s i t i o n(e.g . , Mowrer, 1947; Rescorla & Solomon, 1967; Trapold& Overmier, 1972).In ou r l abora to ry , we rou t ine ly exp lo i t the e f fec t onins t rum enta l behav io r to de tec t the p resence o f a Pav-lovian associat ion. Figure 7 shows the resul ts of one rece ntexper im ent conducted in co l l abora t ion wi th R uth Colwi ll .T h e s e d a t a c o me f r o m r a t s u b j e c t s t h a t a r e ma k i n g a nins t rum enta l cho ice response , pu l l ing a cha in fo r a pe l l e to r p ress ing a l ever fo r l iqu id sucrose . Whi le they wereengag ing in those per fo rmances , we p resen ted a CS tha thad be en m ade a P av lov ian signal e i ther o f food o r o fsucrose . The resu l t o f in te res t is tha t p resen ta t ion o f thePav lov ian CS b iased the resu l t s o f the ins t r um enta l per -f o r ma n c e . W h e n t h e CS s i g n a le d t h e s a me r e i n f o r c e r a sd id the cha in pu l l , i t enhanced cha in pu l l ing re l a t ive tolever p ress ing ; on the o ther hand , when the CS sharedthe same re in fo rce r wi th the l ever p ress, i t enhan ced l everpress ing (cf . Kruse , Overmier , Konz , & Rokke , 1983) .The p o in t i s tha t we can m odula te an a rb i t ra r i ly se lec tedresponse (cha in pu l l ing and l ever p ress ing) by the p re-sen ta t ion o f a Pav lov ian s ignal. The same Pav lov ian con-d i t ion ing can show up in a b road range o f responses de-pend ing on the co n tex t in which i t is assessed . These re -sui ts are of interest for wha t they tel l us abo ut the a nima l 's

    March 1988 9 Am er ican Psycho logi s t 157

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    8/10

    Figure 7Exhibition of Pavlovian Conditioning in the Controlof Instrumental Behavior

    Note. Responding is shown dur ing the presentation of a Pavlovian conditionedstimulus (CS) that signaled the sa me reinforcer as that earned e i ther by a chainpull or a lever press.

    knowledge about the consequences of its instrumentalresponding (see Colwill & Rescorla, 1986)~ but in thepresent context they make the point that conditioningcan show up in arbitrarily selected behaviors, not just inthe response the US evoked.The implication is that describing Pavlovian con-ditioning as the endowing of a CS with the ability to evokethe same response as the US is a wholly inadequate char-acterization. Pavlovian condit ioning is not the shifting ofa response from one stimulus to another. Instead, con-dition ing involves the learning of relations among eventsthat are complexly represented, a learning that can beexhibited in various ways. We are badly in need of anadequate theory of performance in Pavlovian condition-ing, but the classical not ion of a new stimulus taking onthe ability to evoke an old response clearly will not do.Return for a moment to the definitions of condi-tioning with which we began. They emphasized repeatedpairing between two stimuli, one neutral and one valu-able, with the result that the neutral one comes to evokethe response of the valuable one. But we have seen thatpairing is not central, that all sorts of stimuli becomeassociated in a manner that goes beyond simple dyadicrelations, and that the Pavlovian associations influencebehavior in many ways other than by the transferring ofa response.Finally, it is worth noting that these changes in ourviews of Pavlovian condit ioning have been accompaniedand encouraged by changes in the laboratory preparationsused for its study. Many of the early observations in con-ditioning were made using the salivary preparation, of tenby Pavlov (1927) himself. But no contemporary Americanlaboratory makes extensive use of that technique. As canbe seen from the preceding discussion, modern studiesof conditioning use a much more diverse set o f procedures,

    involving a range of signals, consequences, and behavioralmeasures, in various species. The flexibility of contem-porary thinking is partly an adaptation to that diversity.The Place of Pavlovian Condit ioningin PsychologyIt is worth making some comments about the role of Pav-lovian conditioning in psychology in general because tha thas also changed. It is impor tan t to realize that those whostudy this elementary learning process are not nearly asimperialistic as the animal learning psychologists of the1940s and 1950s. In those days, conditioning was morethan a learning process. It was the centerpiece for a setof theories intended to explain all behavior. More thanthat, it represented a way of doing science. Because con-ditioning came to psychology at a t ime when psychologistswere working out scientific ways of studying behavior, itbecame bound up with considerable philosophical bag-gage. It stood not only for an explanation of psychologicalphenomena but also for a way of doing psychology alto-gether. One can still see some of the aftereffects of thisheritage in the conservative style of introducing newtheoretical concepts and in the commitment to elemen-tarism. But Pavlovian conditioning has largely shed itsphilosophical role. Those who study conditioning havelittle interest in recapturing all o f psychology in the nameof behaviorism. What then is the role of Pavlovian con-ditioning in psychology? I see three kinds of contributionsthat it continues to make.First, it continues to be a sample learning processthat admits of careful detailed analysis. It is, of course,only one of a possibly quite large number of learningprocesses. Few would claim tha t all improvements fromexperience are based on a single process. However, Pav-lovian conditioning is an importa nt learning process forwhich the analysis is proceeding apace. As I hope myprevious comments have illustrated, important questionsare being addressed about what produces learning, aboutwhat the products of learning are, and about how organ-isms can represent their world. Moreover, by working ina relatively constrained domain, we can often better char-acterize what would be adequate answers to questionsabout the nature o f learning and better develop techniquesfor providing those answers. So one role for Pavlovianconditioning is as a model for the study of modificationby experience generally.A second role for Pavlovian conditioning is to con-tinue to provide a body of data and developed theory thatinform adjacent areas of scientific inquiry. The study ofPavlovian conditioning provides information about alearning process of continuing interest to allied fields.Two of he most intensely pursued current areas of interestprovide examples: cognitive science and neuroscience.After a period in which it neglected learning processes,modern cognitive psychology has returned to their study;indeed, even the association has regained some respect-ability. This is especially obvious in the approach to cog-nitive processes currently called "parallel distributedprocessing" or "connectionism." According to this ap-

    158 March 1988 9 American Psychologist

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    9/10

    proac h (e .g ., McCle l l and & Rume lhar t , 1986 ; Rum elha r t& Mc Cl e l l a n d , 1 9 8 6 ) , ma n y p h e n o me n a c a n b e u n d e r -s t o o d i n t e r m s o f mu l t ip l e p a r al l el c o n n e c t i o n s b e t w e e ns t i mu l u s i n p u t a n d r e s p o n s e o u t p u t. A l t h o u g h f u e l e d b ya n a lo g i es t o n e u r a l s t r u c t u r e s a n d mo d e r n c o m p u t e r d e -s ign , these connect ion i s t i c theor ies c l ear ly harken backto c l ass i ca l as soc ia t ion i sm. They appeal to mul t ip le as -s o c i a t i o n s i n t e r a c t i n g t o p r o d u c e c o mp l e x o u t p u t s . I nsome cases , theor ies o f th i s so r t have a t t acked ap paren t lycomplex behav io rs wi th su rp r i s ing ly p romis ing resu l t s .Fo r i n s ta n c e , s o m e t h i n g o f s p e ec h p e r c e p t i o n a n d p r o -duct ion , o f ca tegory learn ing , and o f p lace recogn i t ioncan be cap tu red by such theor ies . I t i s s t i l l t oo ear ly toknow wheth er these in i t i a l resu l ts fo recas t u l t imate ly suc-cess fu l accoun t s . B u t they do be l i e some w idely accep tedassert ions tha t cer t a in c l asses o f psycholog ica l pheno me naare in p r inc ip le bey ond the rea ch o f inheren t ly as soc ia-t ion i s ti c theor ies .Co n n e c t i o n i s t ic t h e o r i e s o f t h is s o r t b e a r a n o b v i o u sresem blance to theor ies o f Pav lov ian cond i t ion ing . Bo thv iew the o rgan i sm as us ing mul t ip le as soc ia t ions to bu i ldan overa l l rep resen ta t ion , and bo th v iew the o rgan i sm asad jus t ing i t s rep resen ta t ion to b r ing i t i n to l ine wi th thewor ld , s t r iv ing to reduc e any d i screpancies . Indeed , i t i ss t r ik ing tha t o f t en such c om plex m odel s a re bu i l t on e le -me nts tha t are t ied qui te closely to Pavlovian associat ions .For ins tance , one o f the l earn ing p r inc ip les mo s t f re -quen t ly adop ted wi th in these model s , the so -ca l l ed de l t aru le , i s v i r tua l ly iden t i ca l to one popu lar theory in Pav-l o v i a n c o n d i t i o n i n g , t h e Re s c o r l a - W a g n e r mo d e l . Bo t hare e r ro r -cor rec t ion ru les , in which the an imal uses ev i -dence f rom a l l ava i lab le s t imul i and ad jus t s the s t reng thof each s t im ulus based on the to ta l e r ro r . Here , then , i sa s t r ik ing po in t o f con tac t be tween Pav lov ian cond i t ion ingand a por t ion o f cogn i t ive sc ience .The second area o f in tense ac t iv ity is neurosc ience .A l t h o u g h t h a t a r e a h a s mu s h r o o me d a n d c o n t a i n s ma n yp a r t s t h a t d o n o t b o r d e r o n p s y c h o l o g y , o n e i mp o r t a n tsubarea i s the s tudy o f the neu ra l bases o f l earn ing p ro -cesses . Neuroscient is ts have decided, qui te r ight ly I be-l ieve, that Pavlovian cond i t ioning pr ovides one o f the best-worke d-ou t l earn ing s i tua t ions fo r them to ana lyze . I t hasa wel l -developed da ta base tha t can be charac ter i zed qu i t esuccessful ly by avai lable theories . Th e hopefu l s ign is that ,increasing ly, neurosc ien t i s t s a re fami l i a r i z ing themselvesw i t h t h e c o n t e m p o r a r y s t at e o f Pa v lo v i a n c o n d it i o n i n gand a re a t t em pt ing to a cco un t fo r a hos t o f new resu l ts ,such as sens i t iv i ty to in fo rmat ion , inh ib i to ry l earn ing ,and so fo r th . Indeed , many neurosc ien t i s t s a re be t t e r ac-quain ted wi th the mo dern s t a te o f Pavlov ian cond i t ion ingthan are psycho log i s t s a t l a rge . I t i s par t ly th rough tha ta c q u a i n t a n c e t h a t g e n u i n e p r o g r e s s is b e in g ma d e i n t h eb io log ica l ana lys i s o f l earn ing .

    Pav lov ian cond i t ion ing s t ands be tween these twovery ac t ive a reas o f research . I t p rov ides a con tex t in w hichto assess some o f the assum pt ions abou t the e l ementscon t r ibu t ing to more complex cogn i t ive theor ies . I t a l soprov ides an o rgan ized da ta base and theore t i ca l s t ruc tu reto he lp in fo rm a nd gu ide the n eura l ana lys is o f l earn ing .

    The assoc ia t ion i s no t dead , bu t ra ther con t inues to be afundam enta l concep t in the ana lys is o f l earn ing p rocesses .Moreover , i t i s in Pav lov ian cond i t ion ing tha t many o fthe impo r tan t d i scover ies a re cur ren t ly be ing ma de a bou tassociat ive processes . As a resul t , al l ied areas will con t inueto tu rn to cond i t ion ing fo r da ta a nd theory .Final ly , Pavlovian cond i t ioning con t inues to pla y thero le o f genera t ing p rac t i ca l app l i ca t ions . O f course , ane a r ly e x a mp l e w a s th e d e v e l o p m e n t o f s o me a s p e c ts o fbehav io r therapy . Behav io r therapy was sp un o f f ear lyand has now developed i t s own m atur e l i t e ra tu re . In myview, an u nfor tun ate conse quence o f tha t ear ly emergencei s tha t some behav io r therap i s t s s ti ll v i ew cond i t ion ing inthe way charac ter i zed by the quo ta t ions tha t I have cr i t -i c ized . Bu t there con t in ue to be o ther ins tances o f app l i -c a t i o n s a n d p o t e n t i a l a p p l i c a t i o n s s t e mmi n g f r o m t h elabora to ry s tudy o f Pav lov ian cond i t ion ing . For ins tance ,recen t work sugges t s tha t the body 's reac t ions to d rugsand som e d i seases invo lve Pav lov ian cond i t ion ing me ch-an i sms . Phenomena such as d rug to lerance (e .g . , S iege l ,1983), s t ress-induced analgesia, and immunosuppression(e .g . , Ader & Cohen , 1981) seem to invo lve Pav lov iancond i t ion ing . Those observa t ions sugges t new ins tancesin which cond i t ion ing wi l l have re l a t ive ly d i rec t p rac t i ca lconsequences .Trends come and go in psycho logy . Top ics tha t a reho t today wi l l be co ld in 10 o r even 5 years , bu t somepar t s o f psycho logy con t in ue to bu i ld sys temat ic and im-p o r t a n t d a t a b a s e s a n d t h e o r i e s. T h e s t u d y o f s e n so r yme c h a n i s ms i s o n e e x a mp l e . I t h i n k t h a t t h e s t u d y o f t h eassoc ia t ive mechan i sms under ly ing Pav lov ian cond i t ion-ing is ano ther . The se f i e lds a re end ur ing and sys temat ic ,bu t I hope i t is now obv io us tha t they are a l so chang ingand exci t ing.REFERENCESAder, R., & Co hen, N. (1981). Conditioned mmunopharmacologic e-sponses. In R. Ader (Ed.),Psychoneuroimmunology. New Y ork: Ac-ademic Press.Atkinson, R. L., Atkinson, R. C., Sm ith, E. E., & Hilgard, E. R. (1987).Introduction to psychology(9th ed.). New York: Harco urt, Brace,Jov-anovich.Balsam, P. D., & Tomie,A. (Eds.). (1985).Context and learning. Hillsdale,NJ: Erlbaum.Colwill, R. M., & R escoda, R. A . (1986). Associativestructures in in-strumental learning. In G. H. Bower Ed.),The psychology of earningand m otivation (Vol. 20, pp. 55-103 ). New York: Academic Press.Dickinson, A. (1980). Contemporary anim al learning theory. London,England: Cambridge U niversityPress.Garcia, J., & K oelling, R. A. (1966). Relation of cue to consequence navoidance learning.Psychonomic Science, 4, 123-124.Gardner, H. (1982).Developmentalpsychology (2nd ed.). Boston: Little,Brown.Holland, P. C. (1983). "Occasion-setting" n Pavlovian eature positivediscriminations. In M. L. Commons, R. J. H errnstein,& A. R. W agner(Eds.), Quantitative analyses of behavior: Volume 4. Discriminationprocesses (pp. 183-206). Cambridge, MA: Ballinger.Kam in, L. J. (1968). Attention-likeprocesses n classical conditioning.In M. R. Jone s (Ed.),Mia mi symposium on the prediction of behavior:Aversive stimu li (pp. 9-32 ). C oral Gables, FL: University of M iamiPress.Klatsky, R. (1980).Human memory (2nd ed .). San Francisco: Freeman.Kruse, J. M., Overmier, . B., Konz,W. A., & Rokke, E. (1983). Pavlovian

    Ma r c h 1 9 8 8 9 Am er ica n Psycho log i st 159

  • 8/22/2019 Rescorla r 1985, Pavlovian Vonditioning, It's Not What Ypu Think It Is

    10/10

    conditioned stimulus effects upon instrum ental choice behavior arereinforc er specific. Learning and Motivation, 14, 165-181.Mackin tosh, N . J. (1975). A theory of attentio n: Variations in the as-sociability of stimuli with reinforcement. Psycholo gical Reviews; 82.276-298.Mackintosh, N. J. (1983). Conditioning and associative learning. Oxford,England: Oxford U niversity Press.McClelland, J. L., & Rumelhart, D. E. (1986). Parallel distributed pro-cessing (Vol. 2). Cambridge, MA: MIT Press.Miller, R. R., & Spear, N. E. (Eds.). (1985). Information processing inanimals: Conditioned inhibition. Hillsdale, N J: Erl baum .Morgan, C. T., & King, R. A. (1966). Introduction to psychology (3rded.). New York: McGraw-Hill.Mowrer, O. H. (1947). On the dual na ture o fle arn ing --a reinterpretationof "conditioning" and "problem solving." Harvard Educational Re-viert; 17, 102-148.Pavlov, I. P. (1927). Conditioned reflexes. London: Oxford.Pearce, J. M., & Hall, G. (1980). A model for Pavlovian conditioning:Variations in the effectiveness of conditioned but not of unconditionedstimuli. Psychological Review, 87, 532-552.Pinel, J. P. J., & Treit, D. (1979). C ondit ioned defensive burying in rats:Availability of burying materials. Anim al Learning and Behavior, 7,392-396.Rescorla, R. A. (1968). Prob ability o f shock in the presence an d absenceof CS in fear conditioning. Jour nal of Comparative and PhysiologicalPsychology, 66, 1-5.Rescorla, R. A. (1980). Pavlovian second-order ctmditioning: Studie s inassociative learning. Hillsdale, N J: E rlbaum .

    Rescorla, R. A. (1985). Conditi oned inh ibitio n and facilitation. InR. R . Mil ler & N. S. Spear (Eds.), Inform ation processing in animals:Conditioned inhibition (pp. 299-326). Hillsdale, NJ: Erlbaum.Rescorla, R. A., & Durlach, P. J. (1981). Within-event learning in Pav-lovian conditioning. In N. E. Spea r & R. R. M iller (Eds.), Informationprocessing in animals: M emory mechanism s (pp. 83-111 ). Hillsdale,N J: Ertbaum.Rescorla, R . A., & Solomon, R. L. (1967). Two-process learn ing theory:Relationships between Pavlovian condition ing and instrum ental

    learning. Psychological Review 74, 151-182.Rescorla, R. A., & Wagner, A. R. (1972). A theo ry of Pavlovian con di-tioning: Variations in th e effectiveness of reinforcem ent and no nrein -forcement. In A. H. Black & W. E P rokasy (Eds.), Classical condi-tioning II: Current research and theory (pp. 64-99). New York: Ap-pleton-Century-Crofts.Rosenhan, D. I,., & Seligman, M. E. P. (1984). Abnorm al psychology.New York: Norton.Rumelhart, D. E., & McClelland, J. L. (1986). Parallel distributed pro-cessing (Vol. 1). Cambridge, MA: MIT Press.Siegel, S. (1983). Classical conditioning, drug tolerance, and drug de-pendence. In R. G. Sm art, E B. Glaser, Y. Isreal, H. K alant, R . E.Popham, & W. Schmidt (Eds.), Research advances in alcohol anddrug problems (Vol. 7, pp. 207-246). New York: Plenum.Trapold, M. A., & Overmier, J. B. (1972). The second learn ing processin instrumental learning. In A. A. Black & W. E Prokasy (Eds.),Classical conditioning II: Current research and theory (pp. 427-452).New York: Appleton-Century-Crofts.

    1 6 0 M a r c h 1 9 8 8 9 A m e r i c a n P s y c h o l o g i s t