Journal of Personality and Social Psychology197S, Vol. 31, No. 3, 430-441

Attribution Versus Persuasion as a Meansfor Modifying Behavior

Richard L. Miller, Philip Brickman, and Diana BolenNorthwestern University

The present research compared the relative effectiveness of an attributionstrategy with a persuasion strategy in changing behavior. Study 1 attemptedto teach fifth graders not to litter and to clean up after others. An attributiongroup was repeatedly told that they were neat and tidy people, a persuasiongroup was repeatedly told that they should be neat and tidy, and a controlgroup received no treatment. Attribution proved considerably more effectivein modifying behavior. Study 2 tried to discover whether similar effects wouldhold for a more central aspect of school performance, math achievement andself-esteem, and whether an attribution of ability would be as effective as anattribution of motivation. Repeatedly attributing to second graders either theability or the motivation to do well in math proved more effective thancomparable persuasion or no-treatment control groups, although a groupreceiving straight reinforcement for math problem-solving behavior also didwell. It is suggested that persuasion often suffers because it involves a negativeattribution (a person should be what he is not), while attribution generallygains because it disguises persuasive intent.

Despite the volume of research on attitudechange and persuasion, there is surprisinglylittle evidence that persuasion can be effective,particularly if a criterion of persistence ofchange over time is applied (Festinger, 1964;

This research was carried out as a masters' thesisby the first author under the supervision of thesecond author and was partially supported by Na-tional Science Foundation Grant GS-28178. We aregrateful to Donald Campbell and Thomas Cook forserving on the thesis committee and to LawrenceBecker and Glenn Takata for help with the dataanalysis. Thanks are also due to Susan Bussey andCharlotte Yeh, who served as experimenters, and toBobbe Miller, who did much of the initial codingand typed the early drafts of this article.

This work would have been impossible without theenthusiasm, cooperation, and enlightened support ofIrene Timko, principal of Anderson ElementarySchool. Great thanks are also due to the teacherswho participated in this project. In Study 1, thanksgo to Catherine Lebenyi, Wendy Weiss, and IraMizell for willingly helping to initiate a new behav-ioral modification technique. In Study 2, where anincredible number of experimental manipulationswere administered by the teachers, special thanksgo to Margaret Paffrath, Ronnie Briskman, JuneButalla, and Susan Levie. Thanks are also due tothe eighth-grade assistants who administered ourtests.

Requests for reprints should be sent to Richard L.Miller, Department of Psychology, Georgetown Uni-versity, Washington, D. C. 20007.

Greenwald, 196Sb; Rokeach, 1968; Zim-bardo & Ebbesen, 1969; Cook, Note 1). Thefailure of persuasive efforts to produce lastingchange may be taken as evidence that sub-jects have not integrated the new informationinto their own belief systems (Kelman, 1958)or taken it as the basis for making an attri-bution about themselves (Kelley, 1967). Wemight expect that a persuasive communicationspecifically designed to manipulate the attri-butions a person made about himself wouldbe more effective in producing and maintain-ing change. This research was designed to testthe relative importance of attribution manip-ulations to persuasive attempts by comparinga normal persuasion treatment with an attri-bution treatment.

The persuasion conditions of the presentresearch were designed to be maximally effec-tive through their use of a variety of tech-niques which have been found to be helpful,at least on occasion, in past research. Pastresearch has shown that an optimal persuasivemanipulation should involve a high-credibilitysource (Hovland & Weiss, 1951) delivering arepeated message (Staats & Staats, 1958)with an explicitly stated conclusion (Hov-land & Mandell, 1952) which is supported byarguments pointing out the benefits of change

430

ATTRIBUTION VERSUS PERSUASION 431

(Greenwald, 196Sa) and overlearned by theaudience (Cook & Wadsworth, 1972). Face-to-face communication by the source (Jecker,Maccoby, Breitrose, & Rose, 1964), reinstate-ment of the source at the time of attitudeassessment (Kelman & Hovland, 19S3), andactive role playing or participation by theaudience in the message (Janis & King, 1954)are also helpful.

The attribution techniques were also de-signed to be maximally effective through theiruse of all three factors specified by Kelley(1967) as conducive to making a stable attri-bution: consistency of the evidence over time,consistency of the evidence over modalities,and consistency or consensus across sources.

STUDY 1: LITTERING BEHAVIOR

Study 1 attempted to modify children's lit-tering behavior. Behavior was monitored be-fore and after treatment and again after a2-week period of nontreatment. It was hy-pothesized that both the attribution and thepersuasion conditions would result in initialposttreatment behavioral change but that theattribution condition would show greater per-sistence as a result of altering the basic self-concept of the subjects in a direction incon-sistent with littering.

MethodParticipants

The research took place in three fifth-grade class-rooms in an inner-city Chicago public school. Twofifth-grade classrooms were randomly assigned to theexperimental conditions, while a third was designateda control group. Three female experimenters, allundergraduate psychology majors at NorthwesternUniversity, were randomly assigned to a differentclassroom for each test.

Experimental Manipulations

There were a total of 8 days of attribution andpersuasion treatments dealing with littering, withdiscussion intended to average about 45 minutes perday.

Attribution condition. On Day 1, the teacher com-mended the class for being ecology minded and notthrowing candy wrappers on the auditorium floorduring that day's school assembly. Also on Day 1,the teacher passed on a comment ostensibly madeby the janitor that their class was one of the cleanestin the building. On Day 2, after a visiting class hadleft the classroom, the teacher commented that paperhad been left on the floor but pointed out that"our class is clean and would not do that." The

students at this point disagreed pointedly and re-marked that they would and did indeed litter. OnDay 3, one student picked up some paper discardedon the floor by another and after disposing it in thewastebasket was commended by the teacher for herecology consciousness. On Day 4, Row 1 was pointedout as being the exceptionally neat row in the roomby the teacher. Also on Day 4, the principal visitedthe class and commented briefly on how orderly itappeared. After the principal left the room, thestudents castigated the teacher for her desk beingthe only messy one in the room. On Day 5, a largeposter of a Peanuts character saying "We are An-dersen's Litter-Conscious Class" was pinned to theclass bulletin board. Also on Day 5, the teacher gavea lesson on ecology and talked about what we "theclass" arc doing to help. On Day 6, the principalsent the following letter to the class: "As I talkedto your teacher, I could not help but notice howvery clean and orderly your room appeared. A younglady near the teacher's desk was seen picking uparound her desk. It is quite evident that each ofyou are very careful in your section." On Day 7,the teacher talked about why "our class" was somuch neater. In the interchange the students made anumber of positive self-attributions concerning lit-tering. On Day 8, the janitors washed the floor andostensibly left a note on the blackboard saying thatit was easy to clean.

Persuasion condition. On Day 1 during a field trip,the children were told about ecology, the dangers ofpollution, and the contribution of littering to pollu-tion. They were then asked to role play being atrash collector and to pick up litter as they cameacross it. On Day 2, inside the school lunchroomthe teacher talked about garbage left by studentsand gave reasons why it should be thrown away: itlooked terrible, drew flies, and was a danger tohealth. On Day 3, the teacher gave a lecture onecology, pollution, and litter and discussed with theclass how the situation could be improved. Also onDay 3, the teacher passed on a comment ostensiblyfrom the school janitor that they needed help fromthe students in keeping the floors clean, implyinghere as elsewhere that nonlittering would lead toapproval and commendation by various adult au-thorities. On Day 4, the teacher told the studentsthat everyone should be neat, mentioning aestheticsamong other reasons for neatness. Also on Day 4,the principal visited the class and commentedbriefly about the need for clean and tidy classrooms.On Day 5, the teacher told the students that theyshould not throw candy wrappers on the floor orthe playground but should dispose of them in trashcans. Also on Day 5, a large poster of a Peanutscharacter saying "Don't be a Jitterbug" with "Beneat" and "Don't litter" bordering it was pinned tothe class bulletin board. On Day 6, the principalsent the following letter to the class: "As I talkedto your teacher, I could not help but notice thatyour room was in need of some cleaning. It is veryimportant that we be neat and orderly in the up-keep of our school and classrooms. I hope each ofyou in your section will be very careful about litter."

432 R. MILLER, P. BEICKMAN, AND D. BOLEN

On Day 7, the teacher appointed several children ineach row to watch and see if people were neat out-side the building as well as in the classroom. OnDay 8, a note was left on the board ostensibly fromthe janitors to remind the children to pick up papersoff the floor.

Measurement oj Littering

Pretest,. To discover any existing differencesamong the three classrooms with respect to theirtendency to litter, a specially marked reading assign-ment which had previously been turned in to theteachers was returned to the students 5 minutes be-fore the end of the school day. The students werethen instructed to throw the assignment away afterthe bell rang for dismissal. After school the experi-menters counted the number of assignments thrownin the wastebasket versus left on the floor or on theshelf under the students' seats. Less than 20% ofthe students in each class disposed of their assign-ments in the wastebaskets. The precise percentageswere 20% for the control group (» = 31), 16% forthe persuasion group (n = 26), and 15% for the at-tribution group (w — 2 7 ) .

Posttest. A two-part behavioral test was designedto tap the two aspects of the ecology-littering prob-lem, nonlittering and cleaning up the litter of others.

On the morning of the tenth day, 10 minutes be-fore the first recess, each teacher introduced the ex-perimenter for her classroom as a marketing repre-sentative of a local candy manufacturing firm andleft the experimenter in charge of the room. Theexperimenter explained that she was testing thetastiness of a new brand of candy and passed outone piece of candy to each student. The candy waswrapped in colored cellophane with a different colorused for each classroom. Following the taste test,the class was dismissed for recess. During recess theexperimenters counted the number of candy wrappersin the wastebaskets, on the floor, and in the deskscats. The experimenters then relittered the classroomentrance area with seven specially marked candywrappers. After recess the experimenters checkedthe hallway and playground for discarded candywrappers. During the lunch break, which came 1hour after recess, the experimenters reentered theclassroom and determined the disposition of the spe-cially marked candy wrappers.

The second posttest followed the first by a periodof 2 weeks. During this time no mention of ecologyor littering was permitted in any of the classrooms.The second test was very similar to the first exceptthat this time, the experimenters did not interactwith the students. Ten minutes before the afternoonrecess the teacher passed out toy puzzles as Christ-mas presents from the Parent Teachers Association.The students were asked to try to work the puzzlesbefore recess. Each puzzle was wrapped in a color-coded container with a different color assigned toeach class. During recess the experimenters enteredthe classroom and determined the disposition of thecontainers left there. They then relittered the en-trance way. After recess the experimenters searched

the hallway and playground for other containers.After school the experimenters reentered the class-room and determined the disposition of the relitteredcontainers.

Results and Discussion

Littering Behavior

Figure 1 charts the percentage of items ineach group which were discarded in the waste-basket on each test. A chi-square test wasused to compare frequency of littering in thethree groups on the immediate and delayedposttests. Although the measures directly re-flect items of litter rather than individual sub-jects, it was observed by the experimentersthat subjects independently discarded theirown candy wrappers in the wastebaskets. Thethree groups were significantly different atboth the immediate posttest, x2 (2) = 18.14,p < .001, and at the delayed posttest, x

2 (2)= 20.99, p < .001. The attribution group wassignificantly superior to the persuasion groupon both the immediate posttest, x2 (!) —7.19, p < .01, and the delayed posttest, x2 (1)= 16.15, p < .01. Although the persuasiongroup appeared to show an immediate increasein litter-conscious behavior, it was not signifi-cantly different from the control group evenon the immediate posttest, x2 (1) — 2.57, ns.

Cleanup Behavior

All seven items were picked up by membersof the attribution group on both the immedi-ate and the delayed posttest. Persuasion groupmembers picked up four items on the immedi-ate posttest and two on the delayed posttest,while control group members picked up twoitems on the immediate posttest and three onthe delayed test. Since the total number ofwrappers left was only seven and more thanone wrapper may sometimes have been pickedup by a single individual, a chi-square test isnot fully appropriate. Nonetheless, such atest would be significant at the .05 level ateach posttest, which supports the appearanceof differences favoring the attribution group.

After this study the attribution teacherwas advised that the nonlittering behaviorcould perhaps be maintained if the studentswere occasionally reminded of the attribution"You are neat." Three months later theteacher reported that her class was still sig-

ATTRIBUTION VERSUS PERSUASION 433

In |0°§ 90

UJ

§ 80S

2 70Q

I 50

o:y 40

AIM button

Pretest ImmediatePosttest

TIME

DelayedPostfost

FIGURE 1. Nonlittering behavior of the attribution,persuasion, and control groups over time (Study 1).

nificantly neater than it had been prior totreatment.

The results for both the littering test andthe cleanup test support the hypothesis thatattribution is a more effective technique thanpersuasion for inducing stable behavioralchange. We would like to show, however, thatthis effect holds for other kinds of behavior.Furthermore, it would be desirable to over-come a weakness in Study 1 that arose fromthe fact that treatments were nested withinclassrooms. It is possible, if relatively un-likely, that the differences emerging overtime were due to teacher differences ratherthan treatment differences. Study 2 avoidsthis by including all treatment and controlconditions in each classroom.

STUDY 2: MATH ACHIEVEMENT

The results of Study 1 are certainly en-couraging for an attributional approach tomodifying behavior. However, while ecologyconsciousness and nonlittering are of somesocial importance, they are not the primaryfocus of the schools, which is, at least theo-retically, to teach skills. Will attribution andpersuasion techniques show the same patternof effectiveness in generating a skilled be-havior, like math achievement, as they havein generating a socially desirable but unskilled

behavior, like disposing of trash in waste-baskets? Furthermore, littering behaviorwould certainly be only a weakly valencedaspect of self, while most of the skills taughtin schools would be highly valenced and ofconsiderable import for a student's self-con-cept. Will attribution be as effective for ahighly valenced aspect of self as it was forthe more peripheral aspect? The first purposeof Study 2 was to answer just these questions.It might also be noted that attributions inStudy 2, in addition to being more central toself-concept, are specifically directed to par-ticular individuals rather than addressed toa group as a whole.

The second purpose of Study 2 was to testthe relative effectiveness of attributions ofmotivation versus attributions of ability inchanging behavior. Both perceptions of abil-ity and motivation are essential to the beliefthat a person will attain a given goal (Heider,19S8). Study 1, however, would seem to haveinvolved primarily the attribution of motiva-tion, since the children presumably beganwith a common belief that they had the abil-ity to be neat. In the case of a skilled be-havior like arithmetic, however, it would seemmore likely that motivation and belief inone's motivation to do well is more commonthan ability and belief in one's ability to dowell (Katz, 1964), so that attributions ofability would be of greater value than attri-butions of motivation. Nonetheless, enhancingpeople's perceptions of their motivation for atask may also benefit their performance.Study 2 attempts to separate ability and mo-tivation as the bases of attribution and thetargets of persuasive appeal.

So far we have only considered cognitivestrategies for modifying attitudes and be-havior. Staats (1965) has shown that evenyoung children will engage in complex learn-ing tasks if they are simply given appropri-ate reinforcement. According to Bandura(1969), a successful reinforcement strategyfor behavior modification requires a valuedreinforcer which is contingent upon the de-sired behavior and a reliable procedure foreliciting the desired behavior. In the presentstudy both verbal praise and extrinsic re-wards were used as reinforcers for efforts atmathematical achievement, and a number of

434 R. MILLER, P. BEICKMAN, AND D. BOLEN

overlapping procedures were used to elicitthese efforts.

To compare the relative efficacy of the at-tribution and reinforcement techniques withstandard attitude change approaches, a per-suasive manipulation was devised similar tothe one used in Study 1. The only changeswere that audience participation and roleplaying were deleted, since neither were ap-propriate to the treatment conditions, whilepublic labeling was added. It appears fromthe study of deviance (Becker, 1963) thatpublic labeling of a person can lead thatperson to redefine himself along the lines ofthe label. While it was felt in Study 1 thatchildren had to be convinced of the benefitsof nonlittering in the persuasion condition,the advantages of math achievement as ameans for obtaining rewards in school seemedtoo obvious to need pointing out.

The present study attempted to modifychildren's math-related self-esteem and theirmath scores on skill tests. The six conditionswere attribution ability, attribution motiva-tion, persuasion ability, persuasion motiva-tion, reinforcement control, and a no-messagecontrol. It was hypothesized that all threebasic techniques (attribution, persuasion, andreinforcement) should have an initial positiveeffect on the self-esteem and math behaviorof the subjects but that attribution shouldhave the most enduring effect over time.

MethodParticipants

The research took place in four second-gradeclassrooms of the same inner-city Chicago publicschool involved in Study 1. Second-grade studentswere picked, since it was felt on the basis of Rosen-thai and Jacobson (1968) that their school-relatedself-concepts would be more malleable than at alater age. In all, 96 students took part in the study.All five experimental conditions and one control con-dition were present in each of the four classrooms.From each class list of approximately 30 students,24 were randomly assigned to the six possible condi-tions. Thus 4 students in each classroom appearedin each condition.

Overview of ProcedureAll subjects first received a mathematics pretest

and a self-esteem pretest. Subsequent treatmentsconsisted of 8 days of attribution, persuasion, orreinforcement. Immediately following the treatment,math and self-' ;steem posttests were administered.

Delayed posttests were given after a 2-week periodof no treatment. The control group received thepretests and the immediate and delayed posttests butno treatment. Student absences for both treatmentsand tests were made up on the day the studentreturned to school.

Experimental Manipulations

Five treatment techniques were used with allgroups: verbal comments, written comments, lettersfrom the teacher, letters from the principal, andmedals. The above order is followed in discussingthese techniques for each experimental condition. Itshould also be noted that in the attribution andpersuasion conditions, students were initially calledto the principal's office in groups of eight, wherethey received a treatment-related message. Theprincipal discontinued these treatments after thethird day of the experiment, however, on the groundsthat they were too time-consuming for her and thatshe found the false attribution treatments too diffi-cult to deliver.

All treatment techniques were prepackaged. Be-fore the experiment began, teachers' treatment pack-ages were prepared which listed the treatment tech-niques and their recipients for each day. The orderin which each subject received the treatments wasrandomized for classroom "A" and repeated in eachof the other classrooms.

Attribution ability. The general focus of thistreatment was attributing to the students skill andknowledge in mathematics. Three different verbalcomments were made by the teacher to each studenton different days: "You are doing very well inarithmetic," "You are a very good arithmetic stu-dent," and "You seem to know your arithmeticassignments well." Three different written notes weretied to assignments on different days and handedback to the students: "You're doing very well,""excellent work," and "very good work." The let-ters from the teachers and principal underscored thestudents' excellent work in math and were sent homeon days when a verbal or written note was notscheduled. The letter from the teacher included thephrases, "very good student," "does all his assign-ments well," and "excellent arithmetic ability." Theletter from the principal used the phrases "excellentability," "knows his assignments," and "very goodstudent." The medals awarded to the attributionability students featured the words "good student—math."

Attribution motivation. The general focus of thistreatment was attributing hard work and consistenttrying to the student. Three appropriate verbal com-ments were made by the teacher to each child pri-vately, and three written notes were appended to atest or assignment. The verbal comments were thefollowing: "You really work hard in arithmetic,""You're working harder in arithmetic," and "You'retrying more in arithmetic." The written commentswere as follows: "You're working harder, good!""You're trying more, keep at it!" and "Keep tryingharder!" The letters from the teacher and principal

ATTRIBUTION VERSUS PERSUASION 435

underscored the child's application in math. Theteacher's letter used the phrases "working hard,""trying," and "applying himself." The principal'sletter used the phrases "working harder," "applyinghimself," and "trying harder." The medals awardedto the attribution motivation students featured thewords "hard worker—math."

Persuasion ability. The general focus of this treat-ment was to persuade the student that he should begood in arithmetic and doing well in that subject.Three verbal comments and three written commentssummarizing that message were made in the samemanner as that for the attribution messages. Thethree verbal comments were "You should be good atarithmetic," "You should be a good arithmetic stu-dent," and "You should be doing well in arithmetic."The three written comments were "should be bet-ter," "should be good at arithmetic," and "shouldbe getting better grades." The letters from theteacher and principal used ostensible "aptitude testscores" to inform parents that their child should bemaking good grades in math. The teacher's letterincluded the injunctions "should be doing well,""should be getting high grades," and "should bebecoming a good arithmetic student." The princi-pal's letter used the phrases "should get goodgrades," "should do very well," and "should be agood arithmetic student." The medals awarded tothese students contained the words "do better—math."

Persuasion motivation. The general focus of thistreatment was to persuade the student that heshould be working harder and spending more time onmath. The three verbal and the three written com-ments asked the child to try more at math and weremade in the same way as they were in the othertreatment conditions. The three verbal commentswere "You should spend more time on arithmetic,""You should work harder at arithmetic," and "Youshould try more on arithmetic." The three writtennotes were "Try harder," "Work more on arithmetic,"and "Work harder." The letters from the teachersand the principal informed parents that their childshould spend more time on math and to pass thatidea along to him. The teacher's letter included theinjunctions "should be trying harder," "should spendmore time on arithmetic," and "should be applyinghimself." The principal's letter used the phrases"spend more time" and "try harder," The medalsawarded the persuasion motivation students con-tained the words "work harder—math."

Reinforcement. The reinforcement condition alsofollowed the same format as the attribution andpersuasion conditions except that it added two addi-tional methods and deleted the principal's com-ments. Three verbal comments indicating pride inthe student's good work were made by the teacher.They were as follows: "I'm proud of your work,""I'm pleased with your progress in arithmetic," and"very good." Three written comments of simplepraise were appended to the student's math work.They were, "excellent," "very good," and "I'm veryhappy with your work." The letters from the teach-ers and principal indicated pride and satisfaction in

the child's work to the parent. The teacher's letterused the phrases "proud of his work," "goodgrades," and "happy with his work." The principal'sletter used the phrases "excellent progress," "doingvery well," and "proud of him." The medalsawarded those students contained the words "mathaward." On Days 2, 5, and 7, students were praisedverbally if they chose to work on an extra mathproblem rather than a reading exercise. On twoother days the students received silver stars bysolving a problem from a math assignment.

Control. The control condition received no treat-ment whatsoever but took part in all tests of mathe-matical ability and self-esteem.

MeasurementsSelf-Esteem

The self-esteem pretest was an adaptation of thequestionnaire originally developed by Rogers andDymond (1954) that includes items which measureself-esteem with regard to peers and parents, schoolinterests, and personal interests. Four new itemswere added which dealt specifically with math-relatedself-esteem. All items were declarative sentences gen-erally of the "I am " form, from which thechildren were asked to say whether it was "like me"or "unlike me." The self-esteem pretest and posttestwere both administered individually to each studentby one of a dozen eighth-grade assistants, who askedeach question privately to the second graders andpersonally recorded the subject's answer. The self-esteem posttest was given in the morning on theday after treatment ended.

Mathematics

The pretest and all subsequent math tests were20-item tests consisting of 25% review questions,50% current material, and 25% new material. Themath pretest was administered by the teachers as aregular arithmetic quiz. The immediate posttest wasadministered in the afternoon on the day after treat-ment ceased. The final posttest was administeredafter a two-week period in which no treatment tookplace.

ResultsSelf-Esteem

The mean math self-esteem scores for allsix conditions on the pretest and the posttestare presented in Figure 2. Analysis of varianceof the pretest scores indicated that there wereno significant differences among the fivetreatment groups and the control group inmath self-esteem before treatment began,F(S,90) =.79, ns.

The major analysis of the math self-esteemscores was a repeated measures analysis ofvariance using the six experimental conditionsand the two times of measurement as factors.

436 R. MILLER, P. BRICKMAN, AND D. BOLEN

Attribution Motivation

Persuasion AbilityReinforcement Control'Attribution Ability

No Treatment ControlPersuasion Motivation

TIME

FIGURE 2. Self-esteem scores for the attribution,persuasion, and control groups over time (Study 2).

Results indicated both a main effect of time,F (1,90) = 10.84, p < .01, and a significantTime X Treatment interaction, F (5,90) =2.41, p < .05. As can be seen from Figure 2,all treatment groups show an increase in self-esteem, while the control group shows a de-crease, which is probably due to the lack ofany treatment for control students in the faceof a number of public treatments visible tothem in their classroom.

A subsidiary analysis of variance was per-formed using only the attribution and thepersuasion groups, with mode of treatment(attribution vs. persuasion), basis of treat-ment (ability manipulation vs. motivationmanipulation), and time as the three inde-pendent variables. Results again indicated amain effect of time, F ( I , 60) = 19.02, p <.001, and a significant Time X Mode of Treat-ment interaction, F ( 1 , 6 0 ) = 4.76, p < .05.The math self-esteem of the attribution groupsincreased more sharply from pretest (2.66) toposttest (3.50) than did the math self-esteemof the persuasion groups (2.84 at pretest, 3.13at posttest). Only the attribution ability andattribution motivation groups were signifi-cantly different from the no-treatment controlgroup in their change from pretest to posttest,F (1,30) = 7.58, p < .01 for attributionability, F (1,30) = 7.66, p < .01 for attribu-tion motivation. This difference approachedbut fell short of significance for the persuasionability group, F (1, 30) = 3.12, 0 < .11.

Similar analyses showed no significant effectson general school-related and non-school-re-lated self-esteem.

Mathematics

The mean total math scores for all sixconditions on the pretest, the immediate post-test, and the delayed posttest are presented inFigure 3. Analysis of variance of the pretestscores indicated that there were no significantdifferences among the five treatment groupsand the control group in their total mathscores before treatment began, F ( 5 , 9 0 ) —.54, ns.

The major analysis of the math test scoreswas a repeated measures analysis of varianceusing the six experimental conditions and thethree times of measurement as factors. Theresults show both a main effect of time, F (2,180) = 36.69, p < .001, and a significantTime X Treatment interaction, F (10,180) =5.57, p < .001. From Figure 3 it can be seenthat both attribution conditions show markedincreases on the immediate posttest followedby a slight tendency to enlarge that increaseafter the 2-week delay. Both persuasion con-ditions appear to show an increase on theimmediate posttest but fail to maintain thatincrease over the 2-week delay. The rein-forcement condition shows a pattern similar to

•Attribution Ability

8: 17

§X<[ 165

_l<fe 15

'Attribution Motivation

,Reinforcement Control'Persuasion Ability

- • o No Treatment Control~° Persuasion Motivation

ImmediatePastiest

TIME

DelayedFustiest

FIGURE 3. Total math scores of the attribution,persuasion, and control groups over time (Study 2).

ATTRIBUTION VESSUS PERSUASION 437

that of the attribution conditions but with alesser degree of improvement.

To assess the extent to which attributionand persuasion differed in their initial effec-tiveness versus the extent to which they dif-fered in their ability to maintain their effec-tiveness over time, separate analyses weremade of changes from the pretest to the im-mediate posttest and from the immediate post-test to the delayed posttest. Mode of treat-ment (attribution vs. persuasion) and basisof treatment (ability vs. motivation) wereused as factors along with time of test. Forchange from pretest to the immediate post-test, a significant Mode X Time interaction,F (1,60) = 11.97, p< .001, indicated thatattribution was significantly more effectivethan persuasion in inducing change. Only theattribution ability and attribution motivationgroups were significantly different from theno-treatment control group in their changefrom pretest to immediate posttest, F (1,30)= 14.75, p < .01 for attribution ability;F (1,30) = 11.42, p < .01 for attributionmotivation. This difference approached butfell short of significance for the persuasionability group, F (1,30) = 2.92, p < .10.

For change from the immediate posttest tothe delayed posttest, another significant ModeX Time interaction, F (1, 60) = 13.67, p <.001, indicated that the attribution treatmentswere also superior to the persuasion treat-ments in maintaining what change they pro-duced. The attribution treatments show anincrease of .50 from the immediate to thedelayed posttest, while the persuasion treat-ments show a decrease of .63.

Finally, to make a preliminary test as towhether attribution and persuasion were simi-lar in their effects on high- and low-abilitystudents, subjects in each condition were di-vided at the cell median according to theirmath pretest performance. An analysis ofvariance on posttest math scores was con-ducted using pretest ability (high vs. low),

. treatment (attribution vs. persuasion, ignor-ing basis of treatment), and posttest time(immediate vs. delayed) as factors. Initially,more able students continued to outperforminitially less able students, F ( I , 60) = 46.01,p < .001, on the posttests. Attribution wasmore effective than persuasion for both groups,

F ( I , 60) = 19.39, p < .001, but the attribu-tion-persuasion difference was significantlylarger for the low-ability students than forthe high-ability students, interaction F (1,60) = 5.66, p < .05. Average posttest scoresfor the high-ability students were 18.8 forattribution and 17.8 for persuasion, whilefor low-ability students they were 16.5 forattribution and 13.0 for persuasion.

Discussion

In both studies the attribution treatmentscaused a significant change which persistedover time. These treatments were strongenough to overcome counterarguing by sub-jects in Study 1 and a history of at best mod-est success among many subjects in Study 2.Both attributions based upon subjects' abil-ity to do something and those based upon sub-jects' motivation to do it appeared effective.The effects of persuasion were, in general, in-significant and dissipated over time.

The fact that the superiority of attributiontreatments over persuasion treatments wasdemonstrated in two different field experi-ments using different behaviors (nonlitteringand math problem solving), two different sub-ject populations (fifth graders and secondgraders), and two different designs (a be-tween-classroom design and a within-classroomdesign) gives us some confidence that theseeffects are real and generalizable. Neitherstudy is without weakness. The nesting oftreatments within classrooms in Study 1leaves teacher or group differences as a possi-ble, if unlikely, alternative explanation forthe treatment effects that emerged over time.The public nature of the treatments in eachclassroom in Study 2 means that treatmenteffects may have been aided by implicit com-parisons students were making between theirown condition and other treatments, a processwhich must lie behind the unexpected drop inmath self-esteem shown by the no-treatmentcontrol group. (This last result must also becounted among the ethical costs of a within-classroom design.) The weaknesses of thetwo studies, however, are quite different,while their effects are quite similar, which sug-gests that the results are not due to idiosyn-crasies of design.

438 R. MILLER, P. BUICKMAN, AND D. BOLEN

The present research provides a generalframework into which previous work concern-ing the effects of teacher expectancies onpupil performance (Beez, 1968; Rosenthal &Jacobson, 1968; Seaver, 1973) can be fit. Themeans by which teacher expectancies arecommunicated is at best a dependent variablein previous studies (Meichenbaum, Bowers,& Ross, 1969) and often mysteriously unob-served. The present study made the com-munication of expectancies, in the form ofattributions, its central manipulation. Thefact that this programmed communication ofexpectancies worked as it did provides somesupport for the essential validity of the oftenelusive teacher expectancy effect.

Effects of Attribution

That attributions based upon ability andattributions based upon motivation did notdiffer in their effectiveness implies that directlinkage of skill-specific attributions to theself-system is more important than the basison which this linkage is made. The message"You are a particular kind of person" is moreimportant than the specification of "why."It should be noted, however, that the presentresearch contrasted only two kinds of internalattributions, ability and motivation. Attribu-tions made to a person on the basis of externalfactors ("You are neat because I am watchingyou") would presumably be less effective inproducing lasting change.

Attribution can, of course, involve elementsof persuasion. As we have seen, the statement"You are a neat person" may be a most ef-fective means of persuading someone to beneat. Nonetheless, such attributional state-ments need not involve persuasive intent butmay instead be simple statements of fact. In-deed, their guise as truth statements may bethought of as their most effective advantage.Not only does this enable them to work di-rectly on a person's self-concept, as noted, butit may also enable them to slip by the de-fenses a person ordinarily employs againstpersuasive attempts that are recognized assuch. Attribution as persuasion may be furthereffective because it is less easily recognized aspersuasion, and hence less likely to arouseresistance, counterarguing, or reactance.

In Study 1 the attribution treatment did

elicit counterarguing by the students, whichsuggests a possible reactance (Brehm, 1972)or boomerang effect (McGuire, 1969), butthere was no evidence of such an effect bythe end of the experiment. It is possible thatreactance, like other attitude-change forces,dissipates over time as the issue is workedthrough and the treatment is maintained. Itis also possible that the elemenary school stu-dents who were subjects in the present studiesare less likely to perceive the manipulativeintent of an attribution treatment and lesslikely to show reactance than a comparableadult group.

Attribution and Persuasion

In accounting for the relative ineffectivenessof persuasion, we may note first of all thatpersuasive communications urging a person todo something do not necessarily tap the in-ternal self-concept of their target. Worse yet,to the extent that they do implicate self-con-cept, they may involve the negative attribu-tion that the person is not the kind of indi-vidual who engages in the recommended be-havior. An appeal to be neat or an appeal towork hard can involve the implicit attribu-tion that the person is not currently the sortof individual who is neat or works hard. Ifconvincing people that they are neat orhard working is the key to making them neator hard working, a naive persuasive attemptcan cancel out its own message. At best itattributes to its target the potential for be-coming the sort of person recommended, butthis is clearly much weaker than the attribu-tion that the person already embodies thedesired behavior.

In this study, moreover, the persuasivemessages implied something negative aboutthe subjects, while the attribution treatmentsimplied something positive. Although blameas well as praise has been shown to be effec-tive in eliciting improved performance (seeKennedy & Willcutt, 1964, for a review ofthe mixed results in this area), one of thebest designed studies (Hurlock, 1925) foundthat the improvement elicited by blame dissi-pated over time, while that elicited by praisepersisted. It may be speculated that persua-sion and punishment both remain effective inmotivating behavior only so long as the actors

ATTRIBUTION VERSUS PERSUASION 439

feel that they can accomplish what is beingcalled for. As they accept the implicit nega-tive attributions of a persuasive message, theeffectiveness of the message diminishes. Thismay explain why persuasion was relativelymore effective for high-ability students thanfor low-ability students. The high-ability stu-dents may have been better able to respond tothe appeal to do better without becoming dis-couraged by the implicit attribution (whichthey could to some extent discount) that theywere not currently demonstrating appropriateaccomplishment.

The implicit attributions of the persuasiontreatments in this study were negative becausethe behaviors they were calling for (whichthe subjects were presumably not emitting)were positive. Likewise, the attribution treat-ments were positive because the behaviorsthey attributed to the subjects were positive.For practical and ethical reasons, all of thetreatments in the present study were aimed atproducing positive or socially desirable behav-iors. It is our idea that the implicit or explicitlabeling itself, and not merely the rewarding-ness or punishingness of the labels, made thepresent attribution treatments effective andthe persuasion treatments ineffective. Thiscan only be known for sure, however, by astudy that aims at producing undesirable be-havior, in which case attributions of the be-havior would have negative implications,while persuasion to the behavior would haveimplicit positive implications for the presentself-concept. If people were responding tothe attributions in the present experimentonly because they were rewarding, we wouldexpect that attributions of an undesirablebehavior would have reverse effects and, in-deed, that persuasion under these circum-stances might even be more effective. Thesociological literature on deviance (e.g.,Schur, 1971), however, suggests that negativeattributions (e.g., labeling as delinquent) canindeed produce or support the attributed be-havior, as does a recent experimental study oflabeling "charitable" or "uncharitable" be-haviors (Kraut, 1973). We may suspect, then,that the positive implications of the attribu-tions in the present study were not the solekey to their effectiveness and that if a suit-able experiment could be designed in which

the target behavior were socially undesirable,attribution would continue to be more effec-tive than persuasion in generating that be-havior.

Attribution and Reinforcement

A straightforward reward contingency pro-gram seems to modify behavior because itmakes that behavior worthwhile to the sub-ject. However, the separation between rein-forcement and attribution seems somewhatconfounded. Symbolic reinforcement, as usedin the present study, has some attributionalaspects. Simple praise is often interpreted asa "You are X" statement. Furthermore, at-tribution can contain elements of reinforce-ment especially when socially desirable be-haviors are the focus of the attributionalprocess. Thus to some extent a reinforce-ment procedure that produces enduring changemay require elements of attribution, while asuccessful attribution treatment may involveelements of reinforcement.

The remaining feature distinguishing at-tribution from reinforcement would seem tobe the noncontingent nature of the attribu-tion. In this regard it is interesting to notethat Kazdin (1973) has recently found thatunder circumstances where the desired behav-ior was emitted at a fairly high base rate andsubjects believed that reinforcement was con-tingent, noncontingent reinforcements wereas effective as contingent ones in modifyingbehavior. Attribution treatments, however,may have the very important further advan-tage over simple reinforcement of serving toelicit behavior (like modeling; see Bandura,1969) as well as to maintain it.

Practical and Ethical Implications

The present study supports the idea thatan effort to improve the child's academic self-concept will help improve academic per-formance, if only because the improved self-image will make actual success less incon-sistent, less unexpected, and less likely to bediscounted or rejected (Brickman, 1972). Adistinguishing feature of the present attribu-tion treatments is that they focused on raisingself-esteem in a specific area of skill ratherthan raising global or general self-esteem. Thefailure of more general "cultural enrichment"

440 R. MILLER, P. BEICKMAN, AND D. BOLEN

or general "self-concept enhancement" pro-grams may be due in part to the fact thatthese global manipulations have only vagueand diffuse implications for particular areas ofacademic performance (but cf. deCharms,1972).

Nonetheless, there are a number of reasonsto be cautious about considering these resultsas the basis for a solution to any social prob-lem. It is unlikely that long-standing indi-vidual differences in accomplishment will beovercome by short-term manipulations ofmotivation, incentive, or self-regard. A gainof three problems solved after a week oftreatment is not very substantial in terms oflife chances, and there is no reason to assumethat 10 weeks of treatment will necessarilyresult in a gain of 30 problems. Second, theattribution treatments in Study 2 were diffi-cult to administer and, unlike the treatmentsin Study 1, did not produce any immediateand visible indications of success to sustainteacher enthusiasm. While the teachers in-volved in Study 1 were quite positive towardthe study, those involved in Study 2 haddecidedly mixed feelings about the value ofthe time and energy involved. More seriously,the false attributions came increasingly tobe felt by at least one teacher and by theprincipal as an intolerable risk to theircredibility. As indicated, the principal termi-nated her office meetings with the studentsafter the third day of treatment mainly onthese grounds. While future research couldtailor attributions not to be too discrepantfrom individual pretest baselines, the practi-cal and ethical difficulties involved in main-taining such attributions will not thereby beeliminated. All of these matters warn againstan uncritical application of the presentresults to matters of educational importance.

REFERENCE NOTE

1. Cook, T. D. The persistence of induced attitudechange: A critical review of pertinent theories.Unpublished manuscript, Northwestern University,1969.

REFERENCES

Bandura, A, Principles of behavior modification.New York: Holt, Rinehart & Winston, 1969.

Becker, H. S. Outsiders: Studies in the sociology ofdeviance. New York: Free Press, 1963.

Beez, W. V. Influence of biased psychological reportson teacher behavior and pupil performance. Pro-ceedings of the 76th Annual Convention of theAmerican Psychological Association, 1968, 3, 605-606. (Summary)

Brehm, J. Responses to loss of freedom: A theoryof psychological reactance. Morristown, N.J.:General Learning Press, 1972.

Brickman, P. Rational and non-rational elements inreactions to disconfirmation of performance ex-pectancies. Journal of Experimental Social Psy-chology, 1972, 8, 112-123.

Cook, T. D., & Wadsworth, A. Attitude change andthe paired-associate learning of minimal cognitiveelements. Journal of Personality, 1972, 40, 50-61.

deCharms, R. Personal causation training in theschools. Journal of Applied Social Psychology,1972, 2, 95-113.

Festinger, L. Behavioral support for opinion change.Public Opinion Quarterly, 1964, 28, 404-417.

Greenwald, A. G. Behavior change following a per-suasive communication. Journal of Personality,1965, 33, 370-391. (a)

Greenwald, A. G. Effects of prior commitment onbehavior change after a persuasive communication.Public Opinion Quarterly, 1965, 29, 595-601. (b)

Heider, F. The psychology of interpersonal relations.New York: Wiley, 1958.

Hovland, C. I., & Mandell, W. An experimentalcomparison of conclusion drawing by the com-municator and the audience. Journal of Abnormaland Social Psychology, 1952, 47, 581-588.

Hovland, C. I., & Weiss, W. The influence of sourcecredibility on communication effectiveness. PublicOpinion Quarterly, 1951, 15, 635-650.

Hurlock, E. B. An evaluation of certain incentivesused in school work. Journal of Educational Psy-chology, 1925, 16, 145-159.

Janis, I. L., & King, B. T. The influence of roleplaying on opinion change. Journal of Abnormaland Social Psychology, 1954, 49, 211-218.

Jecker, J., Maccoby, N., Breitrose, H. S., & Rose,E. D. Teacher accuracy in assessing cognitivevisual feedback from students. Journal of AppliedPsychology, 1964, 48, 393-397.

Katz, I. Review of evidence relating to effects ofdesegregation on the intellectual performance ofNegroes. American Psychologist, 1964, 19, 381-399.

Kazdin, A. E. The role of instructions and rein-forcement in behavior changes in token reinforce-ment programs. Journal of Educational Psychol-ogy, 1973, 64, 63-71.

Kelley, H. H. Attribution theory in social psychol-ogy. Nebraska, Symposium on Motivation, 1967,14, 192-238.

Kelman, H. C. Compliance, identification, andinternalization: Three processes of opinion change.Journal of Conflict Resolution, 1958, 2, 51-60.

Kelman, H. C., & Hovland, C. I. "Reinstatement"of the communicator in delayed measurement ofopinion change. Journal of Abnormal and SocialPsychology, 1953, 48, 327-335.

ATTRIBUTION VERSUS PERSUASION 441

Kennedy, W. A., & Willcutt, H. C. Praise andblame as incentives. Psychological Bulletin, 1964,62, 323-332.

Kraut, R. E. Effects of social labeling on giving tocharity. Journal of Experimental Social Psychol-ogy, 1973, P, 551-562.

McGuire, W. J. The nature of attitudes and attitudechange. In G. Lindzey & E. Aronson (Eds.),Handbook of social psychology (2nd ed.).Reading, Mass.: Addison-Wesley, 1969.

Meichenbaum, D. H., Bowers, K. S., & Ross, R. R.A behavioral analysis of teacher expectancy effect.Journal of Personality and Social Psychology,1969, 13, 306-316.

Rogers, C., & Dymond, R. Psychotherapy and per-sonality change. Chicago: University of ChicagoPress, 19S4.

Rokeach, M. Beliefs, attitudes, and values. SanFrancisco, Calif.: Jossey-Bass, 1968.

Rosenthal, R., & Jacobson, L. Pygmalion in the

classroom: Teacher expectation and pupils' intel-lectual development. New York: Holt, Rinehart &Winston, 1968.

Schur, E. M. Labeling deviant behavior. New York:Harper & Row, 1971.

Seaver, W. B. Effects of naturally induced teacherexpectancies. Journal of Personality and SocialPsychology, 1973, 28, 333-342.

Staats, A. W. A case in and a strategy for theextension of learning principles to problems ofhuman behavior. In L. Krasner & L. P. Ullman(Eds.), Research in behavior modification. NewYork: Holt, Rinehart & Winston, 1965.

Staats, A. W., & Staats, C. K. Attitudes establishedby classical conditioning. Journal of Abnormal andSocial Psychology, 1958, 57, 37-40.

Zimbardo, P., & Ebbesen, E. B. Influencing attitudesand changing behavior. Reading, Mass.: Addison-Wesley, 1969.

(Received July 30, 1973)