Self-Presentational Analysis of the Effects of Incentives

Journal of Personality and Social Psychology1980, Vol. 39, No. 4, 553-577

Self-Presentational Analysis of the Effects of Incentiveson Attitude Change Following Counterattitudinal Behavior

Barry R. SchlenkerUniversity of Florida

Mark R. LearyDenison University

Donelson R. ForsythVirginia Commonwealth University

Rowland S. MillerSam Houston State University

A self-presentational analysis of the forced compliance setting focuses on thepredicament-creating aspects of Counterattitudinal actions and on the types ofaccounting tactics people use to excuse or justify events that threaten theiridentities before real or imagined audiences. The social meaning of monetarypayments that are offered for the performance of Counterattitudinal actionsshould affect the nature of the predicament subjects confront. It was hypothe-sized that when payment is introduced in a context that increases subjects'concerns about moral evaluation relevant to bribery, a direct relationship shouldoccur between magnitude of payment and attitude change. If payment is in-troduced in a context that minimizes moral evaluation relevant to bribery,however, an inverse relationship should occur. Three experiments provided sup-port for these hypotheses. In addition, attitude change was enhanced whensubjects thought they were presenting their accounts to an audience that hadobserved their actions. Finally, compared to observers, subjects who had re-ceived large payments attempted to redefine them to make them appear morelegitimate.

Despite the amount of theory and researchdevoted to understanding the impact of coun-terattitudinal action on attitudes, one par-ticularly enigmatic problem still exists—un-derstanding the effects of incentives on atti-tude change following Counterattitudinal be-havior. Dissonance theory (Festinger, 1957)originally generated the prediction that atti-tude change should be inversely related to themagnitude of payment offered subjects to per-form a Counterattitudinal action, a relation-

The present studies were supported by ResearchScientist Development Award K02-MH00183-02from the National Institute of Mental Health andNational Science Foundation Grant BNS 77-08182(both to the first author) and a National ScienceFoundation predoctoral fellowship to the thirdauthor. Thanks are extended to Howard J. Gold-man for his able assistance in the first two experi-ments and Marvin E. Shaw for his helpful com-ments .

Requests for reprints should be sent to Barry R.Schlenker, Department of Psychology, University ofFlorida, Gainesville, Florida 32611.

ship that has been observed in numerousstudies (see Collins & Hoyt, 1972; Wicklund& Brehm, 1976). In contrast, incentivetheory (see Elms, 1967) held that monetarypayments act as both incentives for superiorCounterattitudinal performance and reinforc-ers for that performance. The incentive posi-tion thus generated the prediction that atti-tude change should be directly related to pay-ment magnitude, a relationship that also hasbeen revealed in several studies (e.g., Elms,1967; Rosenberg, 1965).

After an initial confrontation between theseapproaches, several studies found both thedirect and inverse relationship in differentconditions of the same experiment (Calder,Ross, & Insko, 1973; Carlsmith, Collins, &Helmreich, 1966; Cooper & Worchel, 1970;Crano & Messe, 1970; Helmreich & Collins,1968; Holmes & Strickland, 1970; Linder,Cooper, & Jones, 1967) and led many to con-clude that dissonance predictions hold underone set of conditions, and incentive predic-tions hold under another set. Unfortunately,

Copyright 1980 by the American Psychological Association, Inc. 0022-3S14/80/3904-OS53$00.75

553

554 SCHLENKER, FORSYTE, LEARY, AND MILLER

a search for the specific boundary conditionsof each theory—the additional variables thatwill generate the incentive effect versus thedissonance effect—has proved unsuccessful,particularly with regard to producing thedirect relationship (cf. Collins & Hoyt, 1972;Schlenker, Note 1).

Both approaches acknowledge that mone-tary payment can take on somewhat differentsocial meanings depending on the context inwhich it is offered. For example, Rosenberg(1965) contended that large payments aresometimes viewed as bribes. Typically,though, researchers have regarded these socialmeanings as artifacts that should be elimi-nated rather than as an integral part of thephenomenon of interest. Consequently, someof the implications of these meanings havenot been adequately pursued, particularly asthey reflect on the social identities of sub-jects who participate in forced complianceexperiments. The concerns that differentmeanings of payment generate for subjectscould in part determine how they respond,and hence affect whether an inverse or directrelationship is obtained.

Self-Presentation and the ForcedCompliance Paradigm

It has been suggested that attitude changeoften represents an identity protecting orenhancing tactic used for self-presentationalpurposes (e.g., Alexander & Knight, 1971;Schlenker, 1973, 1978, 1980; Schlenker &Schlenker, 1975; Tedeschi, Schlenker, &Bonoma, 1971 ).1 Self-presentation is the con-scious or unconscious attempt to control self-relevant images that are projected in real orimagined social interactions (Schlenker,1980). Since people respond to others on thebasis of the identities that those otherscreate, everyone finds it to his or her ad-vantage to control the self-images that arepresented. Some such attempts are deliberatetactical gestures designed to create a desiredimpression, whereas others are well-ingrainedhabitual responses triggered by social cues.

The counterattitudinal advocacy paradigmstypically employed by social psychologistsinduce subjects to lie, cheat, harm others, re-frain from doing what they would prefer to

do, or otherwise make themselves appear tobe immoral, unattractive, incompetent, orirrational. In other words, most counteratti-tudinal advocacy paradigms place subjects ina predicament that could threaten their pro-jected identities and associate them with nu-merous undesirable images.2 When confront-ing predicaments, people seem to employself-presentational activities to remedy thesituation and restore (as much as possible)their desired identities. Accounting tactics area major type of remedial activity. Accountsare explanations or interpretations of a pre-dicament-creating event that are designed tominimize the severity of the predicament.

The two major classes of accounts are ex-cuses and justifications (Lyman & Scott,1970; Schlenker, 1980). Excuses are attemptsto minimize one's responsibility for the pre-dicament, and include such tactics as denyingone did it, citing external coercive pressures(e.g., "He made me do it"), or citing internalcoercive pressures (e.g., "I couldn't help it,I was . . . drunk, mentally ill, etc."). Justifi-cations allow a person to admit responsibilitybut attempt to minimize the negative conse-quences of the action. For example, onemight justify harming another person bynoting why he or she deserved it or by ex-plaining why the consequences were not reallyharmful. People who believe that marijuanais harmful but who are induced to persuadeothers to smoke it (e.g., Nel, Helmreich, &Aronson, 1969) might justify their actions byproclaiming that marijuana is not very harm-ful. By moderating their attitudes, theyjustify the action.

Following a predicament, accounting activi-ties should be more likely to occur and be of

1 Social identity is a composite picture of an in-dividual as a social being. Just as the self-conceptcan be denned as a theory about oneself (Epstein,1973), so can identity be denned as a theory abouta person that describes the way the individual isdenned and regarded in interaction (see Schlenker,1980).

- Predicaments are situations in which events haveundesirable implications for the images actors haveclaimed or desire to claim in front of real orimagined audiences, such as when one's actions areembarrassing, socially undesirable, antinormative,illegal, or disruptive (Schlenker, 1980).

SELF-PRESENTATION AND ATTITUDES 555

greater magnitude as both the undesirabilityof the predicament-creating event increasesand as the actor appears to be more respon-sible for causing the event. The more negativethe action appears to be and the greater theresponsibility that can be attributed to theactor, the more imperative it is to eliminate,or at least minimize, the potential for guilt,disapproval, and punishment.

In the typical counterattitudinal advocacyparadigm, subjects are given high or low de-cision freedom about performing a persuasivespeech that they expect will harm (mislead)or not harm (not mislead) another person.If no harm arises from the counterattitudinalspeech, no accounting tactics are necessary.If it is obvious that subjects have no re-sponsibility for the behavior (i.e., have lowdecision freedom), again no accounting isnecessary—the subjects already have a visible,ready-made environmental excuse. However,when a person appears to be personally re-sponsible for producing harm, accounting isnecessary. Since the experimenter has alreadyemphasized that the subjects have high de-cision freedom, the subjects cannot excusetheir actions—the experimenter-audiencewould not accept it. The only accountingtactic available is to try to justify the be-havior through attitude expressions. If sub-jects say that they do not strongly disagreewith the contents of their persuasive speeches,then no harm has been done and they haveaccounted for their predicament. Subjectscannot be faulted for trying to persuadeothers to adopt a position the subjects donot find personally offensive. Attitude changetoward the content of counterattitudinal ac-tions can thus be a justification tactic em-ployed when a person initially appears to bepersonally responsible for producing aversiveconsequences. Numerous studies have foundthat subjects must appear to be responsiblefor aversive consequences for attitude changeto occur following counterattitudinal behavior(e.g., Calder et al., 1973; Collins & Hoyt,1972; Hoyt, Henley, & Collins, 1972; Ver-haeghe, 1976).

Self-Presentation and Monetary Payment

Although monetary payments can be in-terpreted in many different ways, including as

rewards for achievement, in most situationsthat arouse a predicament they are perceivedas: (a) environmental variables that affectlevels of responsibility in a manner similar toa decision-freedom manipulation, or (b) vari-ables that raise questions about possible il-legitimate aspects of the action and thepayee's moral values (e.g., the payment couldbe a bribe to perform a morally questionabletask). Whether an inverse or direct relation-ship between payment magnitude and atti-tude change occurs should be partly deter-mined by which of these two identity con-cerns is salient in the situation.

Self-presentation and the inverse relation-ship. Numerous theorists have hypothesizedthat monetary payment acts as an environ-mental variable giving people either high orlow responsibility for their actions (e.g.,Bern, 1972; Collins & Hoyt, 1972; Kelley,1967; Steiner, 1970). Under conditions inwhich monetary payment is seen simply as anenvironmental variable that affects personalresponsibility, the effects of payments shouldbe comparable to manipulating high or lowchoice for performing the behavior. That is,high decision freedom plus low monetarypayment gives subjects high personal re-sponsibility; if aversive consequences are alsopresent, justification through attitude changeoccurs. If decision freedom is low or paymentis high, an environmental excuse is providedfor the action and no accounting is necessaryirrespective of the consequences. These areprecisely the conditions under which the in-verse relationship has been consistently ob-tained (Calder et al., 1973; Collins & Hoyt,1972; Hoyt et al., 1972; Linder et al., 1967;Sherman, 1970).

Self-presentation and the direct relation-ship. A direct relationship should occurwhen people believe that payment representsan attempt to examine their moral turpitude.If cues lead people to believe that a paymentis illegitimate (e.g., a bribe or an attempt toimproperly discharge an obligation), thenthey should be concerned with not conveyingthe image of an immoral individual who willdo anything for money. Take the case of sub-jects who accept $10 to write a short essayon a topic they and most others oppose and,for whatever reason, want to convey the im-

556 SCHLENKER, FORSYTH, LEARY, AND MILLER

pression that they were not bribed. Rosen-berg (1965) suggested that subjects wouldtry to accomplish this goal by not changingtheir attitudes on the topic. Yet this tacticwould only succeed if it was clear exactlywhere the subjects stood on the counteratti-tudinal issue prior to the behavior. If theexperimenter knew they were vigorously op-posed to the essay topic, any subsequentchange in attitudes would probably be takento mean that the payments had influenced orbribed them. However, in virtually all coun-terattitudinal behavior studies, subjects be-lieve that their prior attitudes on a particularissue are unknown or only very generallyknown by the experimenter.

Given such prebehavioral attitude ambigu-ity, expressing no attitude change (i.e., takingan extreme negative position on the issue)after performing a counterattitudinal actiondemonstrates precisely what people hope toavoid. It indicates that the actor vehementlydisagreed with the contents of the essay andtherefore must have performed it to get thepayment—he or she was bribed. Instead, themost effective accounting tactic would be toexpress a less negative, moderate attitudetoward the issue. In this way, subjects showthat their attitudes are not drastically in-consistent with the essay and therefore theywere not bribed: They held the more mod-erate position prior to the performance andthe payment had no effect on them. Whenmoral concerns become salient in the situ-ation, justifications are needed followinglarge payments to demonstrate that the ac-tion was not harmful and not performedmerely to receive money.

These hypotheses are similar in many re-spects to those contained in Kelman andBaron's (1968, 1974) functional approach tocognitive inconsistency. They argue that in-consistency does not produce a drive state inits own right; it is not a "master motive" forbehavior. Instead, it is a signal to the indi-vidual indicating that his or her coping mech-anisms are not functioning properly. In situ-ations that involve a violation of an importantmoral precept, which they term "moral dis-sonance" situations, the signal produces aguilt reaction and efforts aimed at undoingthe "reprehensible" act. They suggest that

"in moral dissonance situations the greaterthe reward the greater the psychological dis-comfort. That is, the knowledge of havingallowed oneself to be 'bought,' of havingviolated one's values for a price, may actuallyincrease one's feelings of guilt" (p. 334). Incontrast, "hedonic dissonance" situations aremerely ones that are "boring, unpleasant,effortful, or nonsensical." Such situations arepresumed to generate more short-lived dis-comforts and produce the typical dissonance-type relationship of more tension followingsmaller payments for performance of thetasks.

According to the Kelman and Baron analy-sis, situations that involve freely telling aharmful lie would seem to be categorized asproducing moral dissonance. However, it isclear from past research that such situationsoften produce an inverse relationship betweenpayment magnitude and attitude change in-stead of the direct relationship predicted. Incontrast, we suggest that such a situationcould involve either high or low moral con-cerns depending on the context in which thepayment is introduced. If it is introduced ina context that increases subjects' concernsabout moral evaluation relevant to bribery, adirect relationship should occur; but if it isintroduced in a context that minimizes moralevaluation relevant to bribery, an inverserelationship should occur.

A variety of contextual factors should in-crease moral evaluation and focus people onthe illegitimate aspects of payment. Paymentsmight seem illegitimate for several reasons.It could become known that the payer wasprimarily interested in scrutinizing the per-son's moral values rather than simply havinga task performed; the payment could be de-scribed as an inducement to elicit behaviorsthat are socially undesirable in nature; thepayer might not really want the person toperform the task, but be under some ex-ternal pressure to make the offer; or an un-desirable task may have been performed as apersonal favor to another individual, and theactor subsequently have received a large, un-expected payment from the other dischargingthe obligation. Of course, moral evaluationcan also be decreased by a variety of factors.If, for example, the experimenter goes to


great lengths to legitimize the payment, orsubjects are distracted from moral concernsby some subterfuge (e.g., focusing them onthe intellectual aspects of a complex task),then the money should not take on connota-tions of illegitimacy.

When moral evaluation is salient, paymentmagnitude may be directly, rather than in-versely, related to responsibility. Kelley(1967, p. 218) proposed that "volition ishigh if the strength of illegitmate forces ishigh and you comply." Marvin E. Shaw(Note 2) found a. curvilinear relationshipbetween payment magnitude and responsibil-ity: People were perceived as most responsi-ble for their actions when payment was eitherinsufficient or oversufficient (illegitimate).Thus for any particular situation, it can besuggested that a curvilinear, U-shaped rela-tionship exists between payment magnitudeand responsibility. When moral evaluation isincreased in the situation, it should have theeffect of shifting the curve. An inverse rela-tionship between payment magnitude andattitude change obtained in the original situ-ation should change into no relationship oreven a direct relationship when moral evalu-ation is heightened. This is because a pre-viously small payment that was seen as in-sufficient (generating high responsibility) anda previously large payment that was seen assufficient (generating low responsibility) thenbecome viewed as sufficient (low responsi-bility) and oversufficient (illegitimate, highresponsibility), respectively.

Given a curvilinear relationship betweenpayment magnitude and responsibility, theself-presentation hypotheses converge on aresponsibility-for-consequences analysis of theforced compliance problem. Collins and Hoyt(1972) suggested that people must feel re-sponsible for producing negative consequencesbefore they will change their attitudes follow-ing counterattitudinal behavior. Their analy-sis was largely atheoretical in nature, but theself-presentational analysis arrives at a similarconclusion via a theoretical route. Given thenature of the forced compliance predicament,attitude change is most needed when peopleappear to be responsible for producing nega-tive consequences—when they are paid eithera small amount under conditions of low moral

evaluation relevant to bribery or a largeamount under conditions of high moral evalu-ation relevant to bribery.

A review of the studies that have found adirect relationship suggests that virtually allof them seemed to have established condi-tions that should have maximized concernsabout immorality and illegitimacy. In gen-eral, two techniques can be used to affectperceptions of illegitimacy: (a) hold the taskconstant and vary payments across a widerange from very small (insufficient) to ex-tremely large (illegitimate), or (b) hold pay-ments at only two levels, small and large,and vary situational cues that increase moralevaluation.

The first technique was used by Gerard,Conolley, and Wilhelmy (1974). They foundthat when payments for performing a coun-terattitudinal action were varied from an in-sufficient to sufficient range, subjects changedtheir attitudes more the smaller the amountof payment, but when payments varied froma sufficient to an oversufficient range, subjectschanged their attitudes more the greater theamount of payment.

The second technique, holding payment attwo levels while varying situational factorsthat affect perceptions of illegitimacy, hasalso been used, often unintentionally. Oneimportant factor that produces differentialperceptions of bribery is the amount of de-cision freedom provided by the experimenter.Kaufmann (1971) found that observers of acounterattitudinal behavior situation per-ceived a monetary payment to be more of abribe when participants were given low ratherthan high freedom to refuse participation.This rather paradoxical result might makesense from a moralistic point of view. Sincean experimenter cannot "force" a subject todo something against his or her will in atypical laboratory study, the failure to men-tion that the subject can decline participa-tion may highlight the fact that the experi-menter believes the payment is "buying" par-ticipation. When a high amount of decisionfreedom is given, the payment may retreatfurther into the background; the paymentthen seems to be only one of the many fac-tors that could affect a decision to participate.

Consistent with this explanation, most (al-


though not all) studies that have found thedirect relationship have done so when sub-jects were given low decision freedom aboutperforming the behavior (Calder et al., 1973;Gerard, 1967; Linder et al., 1967; Rosen-berg, 1965; Sherman, 1970). For example,Calder et al. (1973) had subjects complete aboring task and then tell a confederate thatthe task was interesting. They found thedirect relationship only when subjects weregiven low choice about telling the lie and theconfederate was harmed by believing the lie.Thus it seems that once subjects are givenlow choice about agreeing to do somethingantinormative for payment, the remainder ofthe experiment must be relatively unevent-ful to wipe out the perception of bribery andmoral evaluation.

The direct relationship can still be ob-tained when subjects are given high decisionfreedom, although it seems to take more sig-nificant events to heighten moral evaluationto the point at which it emerges. Janis andGilmore (1965) gave subjects high choiceand either $1 or $20 to write a brief counter-attitudinal essay. They found that the $20subjects were suspicious of the inordinatelylarge amount of money offered to them, andthat they tended (p < .10) to change theirattitudes more than the $1 subjects. Theattitude change trend that accompanied theincreased suspicions is consistent with theabove prediction but is contrary to the effectsof suspicion hypothesized by Rosenberg(1965).

Nuttin (1975, Experiment 9) had subjectswrite a counterattitudinal essay under mod-erate choice conditions and surreptiously of-fered some of them a highly illegitimate re-ward of exam "rescue" points. Compared tosubjects who were not offered the blatantbribe, the ones who received points showeddramatic attitude change effects.

Finally, Elms and Janis (1965) paid sub-jects a small or large amount to write essaysadvocating the sending of American studentsto universities in the Soviet Union for their4 years of college. It was explained that theAmerican government was "skeptical" aboutthe plan and reluctant to support it, althoughthe Soviet government was enthusiastic aboutit. Subjects changed their attitudes in the

direction of the essay only when they werepaid a large amount by a "favorable sponsor"—an interviewer hired by the U.S. State De-partment. An "unfavorable sponsor"—an in-terviewer hired by the Soviet embassy—produced no change. Although these resultsare often interpreted as supporting incentivetheory and disaffirming dissonance predic-tions, they are consistent with the self-presentation position. In the favorable spon-sor condition, subjects were being evaluatedby someone who should have induced highmoral evaluation: The interviewer was pre-sented as an extension of the government'sattitude that the plan was "suspect." Hence,with moral evaluation high, subjects in thelarge payment condition justified their ac-tions to appear as if they did not do some-thing they and the interviewer would regardas reprehensible solely to obtain the money.In the unfavorable sponsor condition, though,the interviewer was hardly in a position tocreate high moral evaluation: He both agreedwith the position of the essay and was him-self hired or bought by the Soviet embassy.Moral evaluation, at least vis-a-vis the im-mediate interviewer-audience, should belower, although perhaps not to the point atwhich the inverse relationship emerges.

Experiment 1

Three experiments were conducted to di-rectly examine the adequacy of the hypothesesregarding the meaning of monetary payment.In the first experiment, subjects agreed to bevideotaped performing a counterattitudinalspeech that would persuade an audience andreceived either a small or large payment.After they made the speech, the experimenterintroduced information that emphasized mo-rality (by stating that the relationship be-tween moral values and behavior was beingexamined) or minimized moral concerns (bystating that the relationship between topicbeliefs and behavior was being examined).According to the above hypotheses, an inter-action of incentive magnitude and moralemphasis should be obtained under these con-ditions. An inverse relationship should occurbetween payment magnitude and attitudechange when moral evaluation is minimized,


whereas a direct relationship should occurwhen it is maximized.

Method

Subjects. Eighty introductory psychology studentspartially fulfilled a course requirement by partici-pating in the study. Sixteen subjects (eight malesand eight females) were included in each cell of thetwo ($.50 vs. $2.50 payment) by two (high vs.low moral evaluation) posttest-only factorial de-sign with an offset control condition (attitude mea-surement only). An additional three subjects refusedto participate, two complaining of "camera-shyness,"and one refusing to make a speech against seat beltusage. Two experimenters, one female and onemale, tested an equal number of subjects in all con-ditions.3

Procedure. Subjects were tested individually ina laboratory room that contained a table; chairs;a videotape camera, recorder, and monitor; and alarge one-way mirror on one wall. The cover storywas a modification of the instructions used by Riessand Schlenker (1977). Subjects were told that thestudy was investigating the relative efficacy of dif-ferent patterns of persuasive messages arguing eitherfor or against a specific issue—in this case, the useof seat belts in automotive vehicles. Videotapedmessages of various types, prepared by college stu-dents, supposedly would be shown to groups ofjunior high school students to determine messageeffectiveness. Measures of subsequent attitudestoward seat belts and the actual use of seat beltswere expected to demonstrate that the junior highschool students would be quite persuaded by themessages. Thus an anti-seat-belt speech had fore-seeable negative consequences for the audience. Theuse of college students as communicators was justi-fied by explaining that the generalizability of theresults would be enhanced through the use of avariety of communicators and messages.

Almost as an aside, the experimenter mentionedthat some other researchers were developing a newtechnique for coding and analyzing live speeches.When they heard about the present study, theythought it would be a perfect opportunity to usethe new rating scales. They had therefore requestedand received the cooperation of the experimenterin allowing raters to observe the speeches and tryout their scoring techniques. These raters would bewatching the proceedings from behind the one-waymirror, where they could work without disturbingthe participants. It was stressed that the observerswere not part of the "opinion research study" forwhich subjects had signed up. No observers wereactually present, but use of this cover story pro-vided a justification for obtaining attitude mea-sures from a person other than the experimenterand allowed for the introduction of the moralevaluation manipulation later.

The experimenter emphasized that the subjectscould choose to write an essay arguing for or

against the use of seat belts. However, to keep theprocedure somewhat uniform, a predesignated se-quence for completing the tapes supposedly hadbeen prepared and was being followed as closely aspossible. The experimenter stated that it would behelpful if the subjects would select the speech thatwas next in the sequence, which was "Seat beltscan be dangerous and should not be worn by oper-ators and passengers in automotive vehicles," butthat the choice was theirs. The experimenter alsomentioned that the project was being funded by agrant, so they could be paid $.50 ($2.50) for theirparticipation. This payment was in addition to thenormal experimental credit slip that they would re-ceive and would give to their instructor to indicatethat they had participated. After the choice instruc-tions and payment were described, subjects wereasked if they would agree to participate using theanti-seat-belt topic, and to indicate their agreementby signing a consent slip.

A short list of sample arguments against the useof seat belts was given to help stimulate thought,and subjects were allowed 15 minutes to prepareand practice their presentation before delivering itinto a videotape camera operated by the experi-menter. The speeches were to be "about a page orso in length, so that it will come out to be about2 minutes or under on the videotape." They wereasked to identify themselves on the tape as uni-versity students but not to give their names. Aftercompleting the tape, subjects were paid and givena credit slip; while doing so, the experimenterreinforced the negative consequences of the speechby stating that it should be quite effective for per-suasive purposes.

The experimenter then stated that the study wascompleted. However, the observers in the next roomsupposedly had asked if subjects could spare a fewminutes to assist them in their work by completinga brief questionnaire. The moral evaluation manipu-lation was introduced at this point. In the highmoral evaluation condition, the experimenter ex-plained that the observers were interested in develop-ing a reliable measure that related behaviors tomoral values. The experimenter stated that

They think, and you would probably agree, thatthe way people act is usually related to theirmoral and ethical values. So maybe by looking atthe way a person talks or acts in a certain situ-ation, an observer can tell the way that personfeels. Simply, to what degree do a person's be-haviors correspond to their feelings in situationslike this, and to what degree can observers of

3 Sex of subjects and experimenter was included inthe analyses of all three experiments. No significanteffects that qualified the reported results were re-vealed on attitude change, and no consistent patternswere obtained on the other dependent variables.Consequently, these factors will not be mentionedfurther.


this situation determine what these moral andethical values are?

In the low moral evaluation condition, subjects weretold that the observers were interested in develop-ing a reliable measure that related behaviors to anindividual's own attitude on a topic. The experi-menter stated that

They think, and you would probably agree, thatthe way people act is usually related to theiropinions and beliefs. So maybe by looking at theway a person talks or acts in a certain situation,an observer can tell the way that person feelstoward an issue. Simply, to what degree do aperson's behaviors correspond to their attitudes insituations like this, and to what degree can ob-servers of this situation determine how con-sistently a person is behaving with his own be-liefs on a topic?

The manipulation of moral evaluation was intro-duced after subjects had completed the speech sothat it would not affect either their decision tomake the speech or their actual speech behaviors.

In all experimental conditions, the experimenterthen stated that the speech raters had developed aquestionnaire to get the type of information theyneeded. All subjects agreed to complete the ques-tionnaire. The experimenter left the room, sup-posedly to get a copy from the observer, and re-turned with it and an envelope. Subjects were toldto seal the questionnaire in the envelope when fin-ished and that it would be delivered to the speechraters. The experimenter left the room while thesubjects completed the questionnaire.

The questionnaire included (a) 24 7-point bipolarattribute scales to be used in responding to thestatement "The use of seat belts is " (ex-amples of endpoints were good-bad, helpful-harm-ful, a waste of time-time well spent); (b) the item"Seat belts can be dangerous and should not beworn by operators and passengers in automotivevehicles," to which subjects were to express theextent of their agreement or disagreement on a 13-point scale; (c) 7-point bipolar attribute scales tobe used to rate the quality of the speech (IS items,e.g., logical-illogical, interesting-uninteresting) ; and(d) five items on 18-point scales that asked sub-jects to rate the amount of choice they had had inagreeing to do the speech, the amount of responsi-bility they had had in agreeing to do the speech,the amount of responsibility they had for the con-tents of the speech and for any persuasive conse-quences of the speech, the degree of persuasivenessof the speech, and the quality of their speechperformance. When subjects had completed the ques-tionnaire and placed it in the envelope, the experi-menter returned and handed them a page contain-ing manipulation checks and other ancillary mea-sures (on 18-point scales).

The 16 subjects in the control condition did notprepare or deliver a speech. They reported to thelab individually and completed the measures of at-

titudes toward seat belts (parts a and b above) aspart of an "opinion survey."

The self-presentation approach predicts thatpeople will try to escape from predicaments by re-defining the situation to make it appear to besocially acceptable. Consequently, "manipulationchecks" that ask subjects who are placed in a pre-dicament about possible bribery, moral evaluation,responsibility, and illegitimacy might not be ex-pected to reveal subjects' actual feelings. For exam-ple, subjects paid $2.50 under high moral evalu-ation conditions should want to avoid bringingfurther attention to such evaluation and shouldnot want to admit being bribed, morally evaluated,or guilt-ridden about their plight. Indeed, Calderet al. (1973) found that subjects in conditions thatdisplayed the greatest attitude change (i.e., lowincentive/high choice/high consequences and highincentive/low choice/high consequences) also re-ported the fewest misgivings about their behavior.To obtain information about how the manipula-tions were perceived by subjects who were notplaced in the counterattitudinal-behavior predica-ment, the experimenter's script was used to write adetailed scenario describing the exact procedure.The scenario described the experience of a typicalsubject, named Pat, who signed up for the study,reported to the laboratory, was told exactly whatthe subjects were told, and then performed thespeech. Four different booklets were constructedthat corresponded to the cells of the 2 (PaymentMagnitude) X 2 (Moral Evaluation) factorial de-sign. Except for the variations of the independentvariables, the booklets were identical and faithfullyduplicated the script. The booklets were randomlydistributed to 27 male and 16 female undergraduatesas part of a "person perception" study; each subjectread one scenario describing one of the four cells.After reading the scenario, these "observer" sub-jects were asked to record their reactions on 15-point scales with labeled endpoints. The itemsassessed the subjects' own moral reactions to Pat'sbehavior, their perceptions of how Pat felt aboutthe behavior, their perceptions of how the observersbehind the one-way mirror felt about Pat's be-havior, and ratings of the monetary payment.

Results

Payment, legitimacy, and moral concerns.Subjects perceived the money to be worthmore to them in the $2.50 than in the $.50condition, F(\, 56) = 4.96, p < .03 (Ms =6.4 and 3.5, respectively). In addition, theresponses of the uninvolved observers indi-cated that the moral evaluation manipula-tion was perceived appropriately. They be-lieved that greater stress was placed on theconsistency between moral values and behav-iors in the high moral evaluation condition,


whereas greater stress was placed on the con-sistency between topic beliefs and behaviorsin the low moral evaluation condition, F(l,35) = 6.64, p < .01 (Ms = 3.8 and -2.0,respectively).

To test the hypothesis that the magnitudeof payment is directly related to moral con-demnation when moral evaluation is high, theobservers were asked to respond to threeitems that assessed moral and ethical evalua-tions. Payment Magnitude X Moral Evalua-tion interactions were obtained on each of thefollowing questions: "How ethically accepta-ble do you think Pat's actions were?", F(l,35) = 3.85, p < .06; "How do you thinkPat felt about his/her actions (consideredthem moral-considered them immoral)?",F ( l , 35) = 4.42, p < .05; and "How do youthink the observers (speech raters) reacted toPat's behavior (considered it moral-consid-ered it immoral)?", F(l, 35) = 4.71, p <.04. As shown in Table 1, Pat's behavior wasperceived as least moral when paid $2.50 inthe high moral evaluation condition, provid-ing strong support for the underlying assump-tions of the study.

Observer-subjects were also asked to ratehow earned-unearned, deserved-undeserved,and honest-dishonest (on 7-point scales) wasthe money paid to Pat. A Payment Magni-tude X Moral Evaluation interaction wasobtained on the sum of these items, F(l, 35)= 5.77, p < ,02. Observers viewed the moneyleast positively, p < .05, when Pat was paid$2.50 in the high moral evaluation condition(M = 12.2). The other conditions did notdiffer amongst themselves (means for the$.50/high evaluation, $.50/low evaluation,and $2.50/low evaluation conditions were15.4, 14.1, and 15.3, respectively). Onceagain, the money evoked less positive reac-tions related to illegitimacy when paymentwas high and moral evaluation was salient.

Attitudes toward seat belts. The predictedPayment Magnitude X Moral Evaluationinteraction was obtained on subjects' agree-ment with the item "Seat belts can be dan-gerous and should not be worn by operatorsand passengers in automotive vehicles," F(l,56) = 5.45, p < .02. As shown in Table 2,a direct relationship between payment magni-tude and attitude change was suggested in

the high moral evaluation condition, whereasan inverse relationship between these varia-bles was obtained in the low moral evaluationcondition. Planned comparisons with the con-trol condition revealed significant attitudechange in the $.50/low evaluation condition,t ( 7 5 ) = 2.05, p < .05; however, the $2.50/high evaluation condition failed to differsignificantly from the control condition, t (75)= 1.61, p < .12 (all tests were two-tailed).The inverse relationship in the low moralevaluation condition was significant, £(75) =2.05, p < .05, whereas the direct relationshipfailed to reach significance, £ (75 ) = 1.39, p< .20.

An identical pattern was revealed on af-fective ratings of "the use of seat belts"derived from the semantic differential scales.Factor analysis was performed on the 24bipolar adjective ratings to extract a singlefactor representative of evaluative, affectivefeelings toward seat belts; most theoristsequate only the evaluative dimension of thesemantic differential with the concept of atti-tudes (Fishbein & Ajzen, 1975).4 The factoranalysis was done using principle axes andorthogonal varimax rotations and revealed onemajor factor that accounted for 50.7% of thetotal variance (eigenvalue = 12.17); no otherfactor accounted for more than an additional5% of the total variance. Sixteen of the 24items loaded heavily on the evaluative factor,with the highest loadings derived from wise-foolish (.92), good-bad (.89), beneficial-detrimental (.82), and safe-dangerous (.76).Analysis of variance performed on the factorscores revealed a significant Payment Magni-tude X Moral Evaluation interaction, F(l,56) = 6.87, p < .05. As shown in Table 2,

* The factor analysis was performed using thecombined data from Experiments 1 and 2, both ofwhich used the same scales to obtain ratings of seatbelts, and included both control groups. This wasnecessary to allow for comparisons across groupsand studies and to increase the number of subjectsincluded to produce greater stability of the analysis.Separate factor analyses were done within each studyand provided support for the compatibility of thetwo data sets. These analyses were done using thepooled within-cell correlation matrix, thereby cor-recting for possible treatment effects within eachexperiment.


Table 1Payment Magnitude X Moral EvaluationInteractions on Observer's Moral Evaluations

Condition

Dependentvariable

High moral Low moralevaluation evaluation

$.50 $2.50 $.50 $2.50

Observer'sevaluation 13.2n 9.4b 12.3a.b 12.1n , i>

Pat'sevaluation 12.6»,b 10.2b 11.9a,b 13.3n

Rater'sevaluation 12.Oa 8.9b 10.3,,b 10.1a.b

Note. Higher means indicate that the behavior wasperceived as more moral. Means without a commonsubscript differ by at least p < .05 using Duncan'smultiple-range test.

the direct relationship was again suggested inthe high moral evaluation condition, and theinverse relationship was evidenced in the lowmoral evaluation condition. The inverse rela-tionship was significant, £ ( 7 5 ) = 2.91, p <.05, but the direct relationship was not, p <.20. Also, comparisons with the control grouprevealed significant attitude change only inthe $.50/low moral evaluation condition,< ( 7 5 ) = 2.22, p < .05.

Subjects' perceptions. Although the directrelationship between payment magnitude andattitude change was suggested on both atti-tude measures in the high moral evaluationcondition, it failed to reach conventional levelsof statistical significance. One possible reasonfor the failure could be that the subjectsattempted to redefine the larger payment asbeing relatively legitimate, thereby reducingtheir predicament and the need for justifica-tory attitude change. Analyses of severalmeasures support this interpretation. Twoitems asked subjects to rate why the moneywas given to them ("to induce you to makethe speech" and "to encourage you to do agood job" on the speech, each followed byan 18-point scale). No effects were found onthe inducement item, as subjects responded onthe "not the reason" side of the scale's mid-point (M = 7.1). However, a main effect ofpayment magnitude revealed that subjectsmore often said that the money was to get

them to do a good job in the $2.50 than inthe $.50 condition, F(l, 56) = 4.15, p <.04 (Ms = 9.3 and 6.2, respectively). Thussubjects avoided describing the payment as aninducement and preferred, when payment waslarge, to label it as an incentive for su-perior performance.

Similarly, when responding to two itemsthat assessed the degree to which they agreedto do the speech to accommodate the experi-menter and to get the money, subjects em-phasized that their agreement was almosttotally to accommodate the experimenter (M= 15.7, no effects were obtained). However,main effects of moral evaluation, F(l, 56) =6.00, p < .02, and payment magnitude, P(l,56) = 6.01, p < .02, showed that subjectssaid they agreed less for the money in thehigh than low moral evaluation condition(Ms = 1.6 and 3.0, respectively), and lessfor the money in the $.50 than in the $2.50condition (Ms = 1.6 and 2.9, respectively).These means fall close to the "not at all forthe money" side of the scale, but indicatethat when moral concerns are high, the im-portance of the money as a decision-affectingfactor is further denied. In addition, a maineffect of moral evaluation showed that whenasked how they interpreted the money, sub-jects said it was less of a bribe in the highthan low moral evaluation condition, F(l,56) = 11.61, p < .01 (Ms = 3.8 and 8.0,respectively). Finally, no effects were obtainedwhen subjects were asked how much stresswas placed on moral evaluation (M = 9.7).

In sum, unlike the uninvolved observers,who disparaged the $2.50 payment whenmoral evaluation was salient, the experi-mental subjects did not admit high moralevaluation, stated that they agreed to do thespeech almost totally to accommodate theexperimenter, and stated that the large pay-ment was primarily to get them to do a goodjob on the speech rather than to induce themto make the speech. Subjects in the highmoral evaluation condition also denied thatthey agreed to perform the speech in order toget the money and rated the money as less ofa bribe. The pronounced differences betweenthe patterns for the uninvolved observers andthe experimental subjects suggests that thehigh moral evaluation condition subjects were


Table 2Attitudes Toward Seat Belts: Experiment 1

Condition

Item and condition

Agree-disagree item$ .50$2.50

Semantic differential items$ .50$2.50

High moralevaluation

3.65.2

-.06.34

Low moralevaluation

5.83.5

.53-.43

Control

3.4

-.05

Note. Higher positive scores indicate greater opposition to seat belt usage, or less favorable evaluations ofseat belts (i.e., more attitude change).

attempting to account for the predicamentby emphasizing situational factors that wouldmake the payment appear to be more legiti-mate.

Perceptions of speech quality, choice, andresponsibility. No independent variable ef-fects were obtained on subjects' ratings ofthe quality of their performance or the qual-ity, persuasiveness, or consequences of thespeeches (the consequences were rated as be-ing slightly on the "detrimental" as opposedto the "beneficial" side of the scale's neutralpoint, M = 9.2). They also stated that theyhad "moderate choice" in deciding on whichside of the topic to write an essay (M =8.6), "some responsibility" for any persuasiveconsequences of the speech (M = 7.2), and"moderate responsibility" for the content oftheir speech (M = 9.1).

Discussion

When moral evaluation was deemphasized,a significant inverse relationship was obtainedbetween payment magnitude and attitudechange on both attitude measures, and sig-nificant attitude change compared to the con-trol condition occurred in the $.50/low moralevaluation condition. By moderating theirattitudes, these subjects apparently justifiedtheir behaviors. The inverse relationship isconsistent with the predictions made by theself-presentation approach, as well as thosemade by dissonance theory and self-perception

theory. All assume that the subjects' respon-sibility for the aversive consequences (or, forself-perception theory, the apparent internalorigin of the action) was greater in the smallrather than the large payment condition. Thefailure to find differences in perceived respon-sibility for the action does not seem criticalto the validity of any of the theories. Eitherthe term "responsibility" may carry differentconnotations for theorists than for subjects,producing an insensitive measure, or as aconsequence of justifying the behavior orreducing dissonance, subjects may not objectto admitting moderate responsibility for theaction under both payment magnitude con-ditions.

When moral concerns were accentuated,however, the inverse relationship disappearedand a slight but nonsignificant direct rela-tionship between payment magnitude andattitude change occurred. The uninvolved ob-servers stated that moral evaluation wasgreater in this condition and condemned thebehavior of subjects paid $2.50. The apparentillegitimacy of the larger payment should haveproduced attitude change designed to showthat the action was legitimate and not influ-enced by monetary concerns. Indeed, subjectsin the high moral evaluation condition didshow such an attitude change trend, and alsoattempted to account for the predicament byemphasizing factors that made the paymentappear to be more legitimate. These directlegitimizing tactics minimized the need forjustificatory attitude change.


Experiment 2

The major limitation of the data from Ex-periment 1 was the failure to obtain a strongerdirect relationship between payment magni-tude and attitude change in the high moralevaluation condition. Rather than demon-strating significant attitude change thatwould justify and legitimize the situation,these subjects apparently legitimized the pay-ment directly through their interpretations ofwhy payment was offered and what effects ithad on them. If such tactics could be blocked,then the direct relationship should emerge.

Experiment 2 allowed a means of blockingthese alternative accounting tactics by explic-itly defining the reason why the money wasbeing paid. Prior to agreeing to make thecounterattitudinal speech against seat belts,subjects were told that the money was beingoffered either as an explicit inducement toget them to agree to make the speech (agree-ment condition) or as an incentive to getthem to do a good job on the speech (per-formance condition).

The agreement condition should block at-tempts by subjects to redefine the large pay-ment as a legitimate incentive and shouldheighten moral concerns. Prior counteratti-tudinal behavior studies have typically intro-duced payments in an ambiguous way, suchthat they could be perceived as inducements,incentives, or both. The present focus ononly the inducement aspect should increasepossible illegitimacy and hence moral evalua-tion. Consequently, a direct relationshipshould be obtained between payment magni-tude and attitude change in the agreementcondition. Subjects in the performance condi-tion, however, do not have to do anything tolegitimize the payment, because the experi-menter has legitimized it for them—it is notto get them to agree to do the speech, only toget them to do a high caliber performancethat will benefit the research project. Addi-tionally, initial responsibility might be roughlyequated in the performance condition irre-spective of payment magnitude by virtue ofthe fact that payment is presumably notoffered to affect decisions to perform the task.If payment is supposedly unrelated to agree-ment, it should be unrelated to responsibility;

consequently, magnitude should be unrelatedto justificatory attitude change.

Method

Subjects. Eighty introductory psychology stu-dents partially fulfilled a course requirement throughparticipation. Sixteen subjects (eight males andeight females) were included in each cell of the 2(Payment Magnitude) X 2 (Contingency) post-test-only factorial design with an offset controlcondition (attitude measurement only). Five addi-tional subjects refused to make a speech againstseat belts, with attrition distributed throughout thecells. Four experimenters, two males and two fe-males, tested equal numbers of subjects in all condi-tions.

Procedure. The procedure was identical to thatused in Experiment 1, with two exceptions: Thereason for the monetary payment was explicitlymanipulated, and the experimenter did not mentionthe topic under study by the speech raters prior todistribution of the questionnaires. As in Experiment1, subjects were asked to make a potentially harm-ful anti-seat-belt speech under conditions of highchoice for either $.50 or $2.50. Subjects in theagreement condition were told, after the experi-menter had noted that money was available, thatthe money was "intended to provide you with anextra reason for going ahead and agreeing to makethe speech." Subjects in the performance conditionwere told that the money was "intended to provideyou with an extra reason for doing a good job onthe speech. We want you to make your arguments,examples, and presentation as persuasive as possible,so we decided that perhaps people would do abetter job if we paid them $.50 ($2.50) to write theessay and deliver the speech for us." These state-ments were reemphasized after the subjects hadcompleted the videotapes. As the experimenterthanked subjects for their participation, they werehanded their class credit slips and the appropriateamount of money and told that the money waseither for "agreeing to make the speech for us" or"for doing the most persuasive and forceful speechpossible." All subjects were then told, "Good job.Your essay should be quite persuasive," thus re-emphasizing the negative consequences.

As in Experiment 1, the experimenter noted thatthe speech raters were interested in obtaining"background information about the communicatorsthey observe," but no mention was made of theexpress purpose of the information. The question-naire was similar to that used in Experiment 1,with the addition of appropriate manipulationchecks and the deletion of irrelevant items.

As in Experiment 1, uninvolved observers (20males and 26 females) were given booklets contain-ing a scenario written by using the experimenter'sscript and including instructions from only one ofthe four key cells in the design.


Results

Payment, legitimacy, and contingencies.Once again, subjects perceived the money tobe worth more to them in the $2.50 than$.50 condition, F ( l , 56) = 9.51, p < .01 (Ms= 9.1 and 4.9, respectively).

Before any attitude change displayed bysubjects can be interpreted, it is necessary todetermine if the experimental subjects ap-propriately perceived the contingencies and ifthe contingencies affected uninvolved observ-ers' evaluations of the morality of the situa-tion. The experimental subjects were askedto rate why the money was given to them ontwo scales—"to induce you to make thespeech" and "to encourage you to do a goodjob" (18-point scales with endpoints labeled"definitely not the reason" and "definitely thereason"). A main effect of contingency wasobtained on the good job item, F(l, 56) =11.35, p < .001, with subjects in the per-formance condition stating that the moneywas to encourage them to do a good job, andsubjects in the agreement condition respond-ing on the "not the reason" side of thescale's midpoint (Ms = 13.1 and 8.4, respec-tively). Thus subjects in the agreement con-dition could not legitimize the payment byemphasizing the performance aspects of themoney. The contingency manipulation had noeffects on the "inducement" item, as subjectsin both conditions responded on the "not thereason" side of the scale (M = 7.8). As mightbe expected, subjects in the agreement con-tingency condition did not want to label thepayment as an "inducement" (bribe), eventhough they did have to admit that the pay-ment was not offered to obtain a good per-formance (which presumably would be theonly other reason that payment would beoffered). Considering both items, the manipu-lation appears to have been successful.

The contingency manipulation also affectedthe observers' ethical reactions. A PaymentMagnitude X Contingency interaction, F(l,38) = 4.12, p < .05, was obtained on ratingsof "How ethically acceptable do you thinkPat's actions were?" When paid $2.50, Patwas evaluated as less moral in the agreementcondition than in the performance condition,p < .05 (Ms = 9.1 and 11.4, respectively).

The means for the $.50 conditions were inter-mediate (Ms = 10.3 and 9.4 for the agree-ment and performance conditions, respec-tively). A comparable, although less pro-nounced, interaction was obtained on theitem "How do you think Pat felt about his/her action?", F ( l , 38) = 2.90, p < .10, al-though no effect was obtained on the ques-tion, "How do you think the observers(speech raters) reacted to Pat's behavior?"

Attitudes toward seat belts. Analysis ofvariance performed on the factor scores fromthe evaluative factor of the semantic differen-tial ratings of seat belts revealed a PaymentMagnitude X Contingency interaction, F(l,56) = 7.65, p < .01. As shown in Table 3,a significant direct relationship was obtainedbetween payment magnitude and attitudechange in the agreement condition, t(75) =4.11, p < .01, whereas no effect was obtainedin the performance condition. Subjects in the$2.50/agreement condition were the only onesto show significant attitude change as com-pared to the control condition, *(75) = 2.95,p < .01. Agreement with the item "Seat beltscan be dangerous and should not be worn byoperators and passengers in automotive ve-hicles" revealed an identical interaction, F(l,56) = 3.27, p < .07. Planned comparisonsperformed on the latter item, means forwhich are presented in Table 3, revealed asignificant direct relationship between pay-ment magnitude and attitude change in theagreement condition, t ( 7 5 ) = 3.08, p < .02;again, the $2.50/agreement condition was theonly one to differ significantly from the con-trol condition, < ( 7 5 ) = 2.90, p < .05. Onboth dependent measures, subjects paid $2.50were significantly more negative toward seatbelts when agreement rather than performancewas stressed, ps < .05.

Subjects' perceptions. No effects of pay-ment magnitude or contingency were ob-tained on ratings of choice in deciding onwhich side of the topic to make the speech(M = 7.4) and in deciding responsibility forthe content (M = 9.0) or consequences (M= 7.4) of the speech. However, subjects paid$.50 reported that the consequences of theirspeeches were slightly more harmful thansubjects paid $2.50 did, F(l, 56) = 5.83, p< .02 (Ms = 8.2 and 10.1, respectively).


Table 3Attitudes Toward Seat Belts: Experiment 2

Condition

Agreement PerformanceItem and condition

Agree-disagree item$ .50$2.50

Semantic differential items$ .50$2.50

contingency

3.46.6

- .49.45

contingency

4.04.4

.15- .21

Control

3.6

-.23

Note. Higher positive scores indicate greater opposition seat belt usage, or less favorable evaluations of seatbelts (i.e., more attitude change).

Ratings of the quality of the speech per-formances revealed a Payment Magnitude XContingency interaction, F(l, 56) = 4.80, p< .03. When paid $2.50, subjects said theydid a better job in the performance conditionthan in the agreement condition, p < .05 (Ms= 11.1 and 7.6, respectively), with the $.50conditions being intermediate (Ms = 9.7 and9.8, respectively). No effects were obtainedon an item asking how persuasive the speechwas or on the average rating of the speech onthe semantic differential scales.

Discussion

The results of Experiment 2 support theprediction of a direct relationship betweenpayment magnitude and attitude change whenmoral evaluation is salient and payments can-not be legitimized by being interpreted as in-centives for superior performance. The unin-volved observers were least favorable towarda person paid $2.50 in the agreement condi-tion; thus the situation effectively evokedmoral concerns related to illegitimate pay-ment, and legitimizing payment as a meansof reducing the dilemma was blocked. Accord-ingly, the $2.50/agreement condition subjectschanged their attitudes toward seat belts,demonstrating that they did not perform thetask solely because of the payment. By recog-nizing the dangerous aspects of seat belts,they could be seen as acting in good con-science when they made the speech.

When payment was legitimized by makingit appear to be an incentive for a quality per-

formance rather than for agreement, no atti-tude change occurred. With subjects' atten-tions directed away from both moral concernsand the role of payment in getting them toagree to make the speech, no justificationwas apparently needed.

Experiment 3

In the first two experiments, moral evalua-tion was increased by informing subjects thateither their moral values were under scrutinyor the payment for agreeing to perform thepotentially harmful act was to be viewedsolely as an inducement. To further reducealternative interpretations of the effects ofthese manipulations, an even more directmeans of raising moral evaluation was em-ployed in Experiment 3. Some of the subjectswere accused by the observers of having beenbribed by the payment. Such an accusationshould leave little doubt in the minds of thesubjects about how their actions are beinginterpreted by an audience. In contrast, othersubjects received a note from the observersindicating that the latter thought the pay-ment was legitimate. If a direct relationshipis again obtained between payment magnitudeand attitude change in the bribe condition, itwould provide even stronger support for thehypothesized effects of moral evaluation.

If attitude change in a forced compliancesetting is a self-presentational tactic usedwhen subjects confront a predicament, thenthe degree to which actors are concernedabout the impression they are making on a


particular audience should affect the amountof attitude change that occurs. To manipulateself-presentational concern, half of the actorswere led to believe that the attitude question-naires were being collected for the observerswho had just commented on the payment(high self-presentational concern). The re-maining actors were led to believe that thequestionnaire was part of a psychology de-partment survey to assess subjects' reactionsto studies performed under its auspices (lowself-presentational concern). To further dis-sociate the latter questionnaire from theexperimenter and observer audiences, thesesubjects completed it on another floor of thebuilding and were told that neither the ex-perimenter nor the observers would see theirresponses. It was hypothesized that the inter-action between payment magnitude and moralevaluation would be more pronounced in thehigh than in the low self-presentational con-cern condition. Further consideration of theeffects of this manipulation as it bears on"genuine" attitude change will be reservedfor the discussion.

Finally, subjects participated in pairs. Onesubject was assigned to perform the counter-attitudinal task while the other unobtrusivelyobserved the former's actions from behind aone-way mirror. Headphones permitted theobserver to hear everything that transpiredin the adjacent room. The use of actor-ob-server pairs allowed examination of possibledifferences in perceptions as well as examina-tion of the degree to which the observerscould correctly predict the actors' attitudes(cf. Bern, 1972).

Method

Subjects. Two hundred twenty-four introductorypsychology students (seven males and seven femalesper cell) served in the 2 (Actor or Observer) X 2($.50 or $2.50 Payment) X 2 (Bribe or LegitimateComment) X 2 (Low or High Self-PresentationalConcern) posttest-only factorial design. An addi-tional 14 subjects (7 males and 7 females) weretested in the offset control condition, where theyresponded to the attitude items as part of an"opinion survey." Subjects were tested by one oftwo male or two female experimenters.

Procedure. The procedure closely followed thoseof the first two experiments with the followingexceptions. Two same-sexed subjects reported indi-

vidually to different waiting rooms, where they weregreeted by the experimenter and led to adjoininglaboratory rooms that were separated by a one-waymirror. The observer was seated at a table behindthe mirror, provided with headphones that wereconnected to a microphone in the actor's room, andgiven a page of printed instructions. The instruc-tions explained that he or she would be observinganother subject participating in a study of per-suasive speech making. They continued:

We are interested in examining how outside ob-servers (like yourself) perceive the situation andthe other subject's behavior. You will observe theother person through the one-way mirror andcomplete a questionnaire at the end of the studyasking for your perceptions . . . of the subject,his or her speech, the experimenter, and the ex-periment in general.

It was explained that to protect the observer'sanonymity, the other subject would not know thathe or she was watching, although the actor wouldknow that other observers from the speech depart-ment were watching from behind a second one-waymirror in another adjacent room (once again, noobservers from the speech department were actuallypresent). The instructions then reiterated the im-portance of paying close attention to the pro-ceedings.

The actor was seated in the main laboratory roomand given general tape-recorded instructions com-parable to those of the prior studies. After notingthe actor's choice in deciding to make the anti-seat-belt speech (but prior to the subject's agreement),the experimenter said, "I also need to mention thatthis project is being funded by a grant, and if youagree to do the speech we can pay you $.50 ($2.50)in addition to the regular 1-hour credit that youwill receive for your participation." Thus thepayment was introduced in a way that implied anagreement contingency, but did not make it explicitas had been done in Experiment 2.

After delivering the speech into the videocamera,the actor was paid and informed that the study wasover. The experimenter then remarked in an off-hand manner, "The observers from the speech de-partment have rated your speech, which was. thepoint of their watching, and since most peoplereally wonder what the observers thought, I usuallygo get a copy of their rating form for them." He orshe then left to get the form.

The form contained two general categories forratings of the content of the speech and the presen-tation of the speech; each category was followedby five rating dimensions (e.g., organization, logicof arguments, persuasiveness, voice clarity, use ofgestures). These were filled out to indicate that thespeech and presentation had been evaluated asslightly above average, with the highest rating givenfor persuasiveness (thus reinforcing the belief thatthe speech would be effective in persuading thejunior high school students). In addition, the bot-


torn of the form contained a space for "Notes andComments," and included a handwritten statementthat introduced the bribery manipulation. The state-ment read: "Subject received payment for agreeingto participate—," and continued with either"Bribe?!?" (bribe condition) or "legitimate com-pensation; good idea to pay subjects" (legitimatecondition). The experimenter gave one copy of theform to the observer, reiterating that the ratingswere those of the speech observers, and then re-turned to give another copy to the actor.

After giving the subjects time to scrutinize theform, the self-presentational concern manipulationwas introduced. In the high self-presentational con-cern condition, the actor was told that the speechraters were interested in his or her own perceptionsof the speech, which would presumably assist themin refining their techniques. The actor was given aquestionnaire, supposedly devised by the speechraters, asked if he or she would assist them bycompleting it, and told to seal the questionnaire inan envelope when finished so that it could bedelivered to the raters. In the low self-presentationalconcern condition, the actor was told that thepsychology department was interested in obtainingsubjects' opinions about certain randomly selectedstudies as part of a process of evaluating studiesconducted under its auspices. The actor was givendirections to the main psychology office, which wason a different floor, and told that the questionnairehe or she would complete would not be shown tothe experimenter or the speech raters. On arrivingin the main office, the actor was greeted by a con-federate who posed as a secretary-receptionist, wasasked what study he or she had just participated in,and was given a questionnaire that was supposedlytailored for that study. After completing the ques-tionnaire, the actor was asked to return to the lab-oratory room, where he or she was fully debriefed.

The questionnaire was similar to those used inthe first two experiments with the addition ofappropriate manipulation checks and other items(unless otherwise noted, ratings were on 18-pointscales). The observer always completed the ques-tionnaire in the original observation room. The itemson the observer's questionnaire were phrased toparallel those on the actor's.

Results

Manipulation checks. Responses to theitem, "How much was the money paid to you(the other subject) worth to you (him/her)?"revealed main effects of payment magnitude,F(l, 208) = 37.21, p < .001, and actor-observer role, F(l, 208) = 6.59, p < .03.Both actors and observers acknowledged thatthe $2.50 payment was more valuable thanthe $.50 payment, ps < .01, but observersbelieved that the $2.50 was even more valua-ble to the actors than the actors said it was,

p < .001. Means for the actors and observers,respectively, were: small payment—3.7 and3.8; large payment—5.9 and 8.9. The actors'relative denigration of the worth of thelarge payment may have represented an effortto counteract the appearance that they hadperformed the behavior to obtain a valuablereward.

The item, "How do you think the . . .speech raters reacted to your (the other sub-ject's) behavior?" (scale end points werelabeled "considered it moral" and "consid-ered it immoral") revealed a main effect ofbribery emphasis, F(l, 208) = 11.39, p <.001, and a Role X Bribery Emphasis inter-action, F ( l , 208) = 4.90, p < .03. The maineffect revealed that the manipulation wassuccessful overall, with subjects stating thatthe behavior was less moral in the briberythan legitimacy conditions (Ms = 7.7 and5.8, respectively). The interaction indicatedthat although the means for both actors andobservers were in the appropriate direction,observers (Ms = 8.1 and 4.8, respectively)demonstrated a greater differentiation betweenthe bribery and legitimacy instructions thandid the actors (Ms = 7.3 and 6.7, respec-tively). Indeed, some actors anecdotallyreported during the debriefing that if thespeech raters had felt that the payment wasperfectly legitimate, they would not havementioned anything at all about it on therating sheet. Thus the actors' involvement inthe situation and possible concern about howtheir actions were viewed by the note-sendingaudience may have prompted them to besomewhat sensitized to the morality issue byeven the legitimate statement; naturally, theuninvolved observers would not share theseconcerns. (Indeed, the observers rated thelegitimate note as being more moral than didthe actors, p < .04).

Attitudes toward seat belts. Actors wereasked to express their own attitudes andobservers were asked for their perceptions ofthe actors' attitudes on (a) the 13-pointagree-disagree item used in the first two ex-periments and (b) the 16 bipolar adjectiveitems that showed the strongest evaluativeloading in the first two experiments. Bothmeasures revealed a main effect of role, Fs(l,208) = 73.42 and 93.14, respectively, ps <


Table 4Actors' and Observers' Perceptions of Actors' Attitudes Toward Seat Belts:Experiment 3

Condition

Item and condition

High self-presentationalconcern

Legitimate Bribe

Low self-presentationalconcern

Legitimate Bribe Control

Agree-disagree itemActor

$ .50$2.50

Observer$ .50$2.50

4.34.6

9.49.1

3.66.5

8.28.9

4.25.6

8.58.4

5.53.9

5.19.1

4.4

Semantic differential itemsActor

$ .50$2.50

Observer$ .50$2.50

1.72.0

4.03.7

1.22.1

3.73.9

1.41.8

4.13.0

2.01.6

1.83.7

1.2

.001, and a four-way interaction of role, pay-ment magnitude, bribery emphasis, and self-presentational concern, Fs(\, 208) = 6.6Sand 6.8S, respectively, ps < .02. Means forthe items are presented in Table 4.

Considering the actors first, there was a sig-nificant simple three-way interaction betweenpayment magnitude, bribery emphasis, andself-presentational concern on the agree-dis-agree item, F(l, 208) = 4.32, p < .04. Dis-secting the interaction further, it was foundthat payment magnitude and self-presenta-tional concern interacted for actors who re-ceived the bribery statement, F(l, 208) =S.SO, p < .02, but not for actors who re-ceived the legitimate statement. As predicted,a direct relationship was found between pay-ment magnitude and attitude change whenbribery was emphasized and self-presenta-tional concerns were high, F(l, 208) = 2.96,p < .05. Thus when actors had to presenttheir attitudes to the speech raters who hadaccused them of being bribed, those receiving$2.50 espoused relatively neutral attitudestoward seat belts, whereas those paid $.50remained staunchly in favor of seat belts.Moreover, the actors in the $2.50/bribe/highself-presentational concern condition were the

only actors whose attitude statements dif-fered significantly from those of the controlsubjects, *(123) = 2.59, p < .OS.5

Contrary to initial expectations, no signifi-cant effects were obtained for actors in thelegitimate payment conditions. Despite thepresence of high decision freedom, negativeconsequences, and small payment (Calder etal., 1973; Collins & Hoyt, 1972), no signifi-cant inverse relationship was obtained. Asevidenced by the actors' responses on themorality manipulation check, it is possiblethat the legitimacy note did not clearly dispelthe actors' concerns about how their actionswere viewed by the speech raters. Actors inthe legitimacy condition may have been suf-ficiently apprehensive about moral evaluationto wipe out an inverse relationship but notsufficiently apprehensive to generate a directrelationship.

No evidence of attitude change was ob-tained when self-presentational concern was

s Analyses performed on only the data from theactors again revealed the three-way interaction,F(l, 104) = S.47, p < .03, along with each of thereported significant simple effects. Thus deleting theobservers' responses has no effect on the results.


Table SInterpretations of the Monetary Payment

Condition

Actors Observers

Dependent variable $.50 2.50 $.50 $2.50

Performed speech for the experimenterPerformed speech for the moneyMoney offered as a bribeMoney offered to do a good jobEvaluation of the money

15.1.3»,b

3.51.4»3.1

14.30

2.3n.c

4.7a

6.5,,2.8

14.2b

2.2b,d

4.2b

3.6b

4.0

H.4.,b6.20ld

81

"J».b6.9b3.5

Note. Means within each row with a common subscript represent significantly different orthogonal com-parisons via simple effects tests, p < .05. Higher means indicate greater amounts of the dependent variable(i.e., the speech was performed more for the experimenter or more for the money, the money was offeredmore as a bribe or more to do a good job, or there was a more positive evaluation of the money).

low. Although a slight inverse relationshipwas suggested when self-presentational con-cerns were low and bribery was emphasized, itdid not approach significance (p > .15).Such an effect, had it been significant, wouldhave supported Rosenberg's (1965) hy-pothesis.

The actors' average ratings of seat belts onthe bipolar adjectives revealed a pattern sim-ilar to that described above, but the effectswere not as pronounced. None of the actorconditions differed significantly from the con-trol group or from any other condition.

The observers' perceptions of the actors'attitudes failed to track the actors' reportedattitudes. The main effect of role obtainedon both attitude measures indicated that theobservers thought the actors were in muchgreater agreement with the anti-seat-beltspeeches than the actors actually were. Byand large, the actors' behavior engulfed thefield (Heider, 1958; Jones & Harris, 1967),an effect that may have been aided by thehigh choice that was provided prior to agree-ment. The exception to this effect occurredfor the observers in the $.50/bribe/low self-presentational concern condition, who closelymatched their yoked actors' attitudes. Theseobservers differed on every orthogonal com-parison from other observer conditions (ps <.01), and the simple two-way interactions inwhich the condition was embedded were sig-nificant (ps < .03). These observers mayhave believed that accusing the actors of

being bribed by a niggardly $.50 was anaffront the actors would react against by ex-pressing negative attitudes toward the topic,at least when the actors were provided withanonymity in the low self-presentational con-cern condition. No other significant effectswere obtained for the observers.

Interpretations of the payment. Fiveitems assessed interpretations of the payment:(a) "To what degree did you (the subject)agree to do the speech to accommodate theexperimenter?", (b) "To what degree didyou (the subject) agree to do the speech forthe money?", (c) "How did you interpret theoffer of the money?" (scale end points werelabeled "totally as a bribe" and "not at all asa bribe"), (d) "To what degree do you be-lieve that the money was given to you (thesubject) to encourage you (him/her) to do agood job on the speech?", and (e) the aver-age of three 7-point bipolar adjective scalesthat asked for ratings of the money: de-served-undeserved, honest-dishonest, andearned-unearned. Main effects of role, Fs (1,208) = 13.09, 39.98, 8.79, 4.11, and 17.15,respectively, ^s < .05, were obtained on eachof the items, and a Role X Payment Magni-tude interaction was obtained on the firstthree items, Fs(l, 208) = 3.46, p < .06;5.68, p < .05; and 4.35, p < .05, respectively.No other effects were significant. Item meansfor the four Role X Payment Magnitudeconditions are presented in Table 5.

As can be seen in the Table 5, compared to


their observers, actors who received $2.50said they did the speech more for the experi-menter, less for the money, believed themoney was less of a bribe, and rated themoney slightly less positively. The most strik-ing difference between actors and observerswas obtained on the bribe item, on whichactors who received $2.50 did not differ fromeither the actors or the observers in the $.50conditions, but differed dramatically from theobservers in the $2.50 condition. These re-sults, like those of the first two experiments,again suggest that the actors were motivatedto interpret the large payment in socially de-sirable ways and to avoid the onerous impli-cations of having been bribed to perform atask that had negative consequences.

Harm. If actors were concerned about theappearance of producing negative conse-quences, it could be expected that they shouldattempt to minimize the potential harmful-ness of their speeches, at least relative to ob-servers. Indeed, a role main effect, F(l, 208)= 5.71, p < .02, on the item, "How beneficialor harmful do you think your (the subject's)speech could be for listeners?" indicated thatactors stated their speeches were less harmfulthan the observers believed them to be (Ms =9.8 and 10.8, respectively).

Choice and responsibility. Responses tothe item, "How much choice did you (thesubject) have in deciding on which side ofthe topic, pro or con, to write an essay anddeliver a speech?" revealed a main effect forrole, F(l, 208) = 5.56, p < .02, and a Pay-ment Magnitude X Bribery Emphasis inter-action, F(l, 208) = 6.16, p<.02. Actorsreported greater choice than the observersascribed to them (Ms = 7.6 and 5.9, respec-tively). The interaction supports Kelley's(1967) analysis of attributions under vary-ing payment magnitude and legitimacy con-ditions. An inverse relationship appeared be-tween payment magnitude and choice whenthe legitimate comment was made (Ms =8.1 and 6.3 for the small and large paymentconditions, respectively), but a direct rela-tionship occurred when the bribery commentwas made (Ms = 5.5 and 7.2, respectively).Large payments accepted when illegitimateforces were salient permitted high choice tobe ascribed to the actor (Kelley, 1967).

As in the first two experiments, an itemthat asked how responsible the actors werefor any persuasive consequences of theirspeech failed to reveal significant effects.Given the above pattern on the choice item,it seems likely that subjects did not directlyequate choice with the term responsibility,perhaps because they did not attach the samemeaning to the latter term that theorists do.

Speech performance. Ratings of how wellthe actors performed the speech revealed amain effect of role, F(l, 208) = 31.12, p<.001, and a Role X Payment Magnitude XSelf-Presentational Concern interaction, F(l,208) = 4.86, p< .03. Actors believed thatthey did a poorer job on the speech than didobservers (Ms = 8.3 and 10.6, respectively).The interaction, as well as a comparable inter-action obtained on the average bipolar adjec-tive ratings of the quality of the speeches,F(l, 208) =4.02, p < .05, does little toclarify any of the preceding effects. Themajor finding was that actors in the $.50/lowself-presentational concern condition ratedtheir speeches and their performances morefavorably than other actors rated theirs. De-spite these high ratings, however, these actorsdid not demonstrate significant attitudechange: Perceptions of superior speech qual-ity were not sufficient to produce attitudechange.

Discussion

As predicted, a direct relationship was ob-tained between payment magnitude and atti-tude change when moral evaluation washeightened by the bribery accusation andactors had to report their attitudes to theaccusing audience. The effect supports theinterpretation that subjects whose identitiesare threatened by the appearance of briberywill account for their actions by attemptingto show that they are not in disagreementwith the nature of the action and did notperform it solely to get a large payment. Inaddition, actors who received the large pay-ment, as compared with their observers, saidthat the payment was less valuable, empha-sized that they did the speech more for theexperimenter and less for the money, andproclaimed that the money was less of a


bribe. These actions suggest the attempt tointerpret the large payment in a way thatavoids the appearance of monetary gain as aprimary motive for their actions.

Although the legitimacy condition was in-tended to dispel the negative aspects of moralevaluation for the actors, it was apparentlynot completely successful in doing so. Theobservers believed that the actors were morefavorably evaluated by the speech raters inthe legitimacy condition than in the bribecondition, attesting to the success of themanipulation when viewed by uninvolvedbystanders. But the involved actors did notdisplay as pronounced a difference betweenthese conditions on the manipulation check.Perhaps the combination of the extreme na-ture of the actors' actions (persuadingyounger pupils to avoid the use of seat belts)and the fact that they received an audience'snote mentioning their monetary payment pro-duced an intermediate degree of moral evalu-ation for actors in the legitimacy condition.That is, it may have been high enough towipe out an inverse relationship, but not suf-ficiently high (at least as compared withactors who were explicitly accused of beingbribed) to generate a direct relationship. InExperiment 1, the inverse relationship wasclearly obtained when moral evaluation waslow, but those subjects did not receive an ex-plicit note from the audience that mentionedmoney and were further distracted from moralevaluation by the statement about the ob-server's different focus.

Attitude change was obtained only whenself-presentational concerns were high. Al-though it was predicted that the effects of theindependent variables would be more pro-nounced under high than low self-presenta-tional concern conditions, it was not neces-sarily anticipated that low concerns wouldwipe out effects. Symbolic interactionists pro-pose that once people learn and incorporatethe rules society applies when judging be-haviors, they use these "personal" standardsto judge their own actions irrespective ofwhether a real audience is present (althoughaccounting should be greater before a realaudience than when one is alone). Inappro-priate behaviors produce feelings of guilt,self-doubt, anxiety, or embarrassment, all of

which people attempt to reduce. In this sense,people can be said to be concerned with self-presentations even in private, so long as theirattention is focused on themselves as socialobjects vis-a-vis an imagined audience(Schlenker, 1980). Such private self-presen-tation would generate self-serving interpre-tations of situations and should produce rela-tively enduring attitude change.

In the present study, though, genuine atti-tude change did not seem to occur, at least inthe sense that no attitude change effects werefound when the actors were free to anony-mously fill out the questionnaire for an audi-ence that supposedly was unaware of thenature of the predicament. On the surface,then, the results might appear to support thearguments of those who contend that nogenuine attitude change occurs in the forcedcompliance paradigm (Gaes, Kalle, & Tedes-chi, 1978).

Such a conclusion, however, would be apremature overgeneralization. There are theo-retical reasons to assume that genuine atti-tude change sometimes occurs and sometimesdoes not occur in predicaments (Schlenker,1980). People sometimes tell blatant lies toextricate themselves from predicaments, pri-vately label the account a lie, and never cometo believe a word of what they say. On theother hand, less blatantly deceptive accountsmight be employed for self-presentationalpurposes and end up being believed by theactor because they fit, at least roughly, theexisting facts as the actor knows them (Jones& Wortman, 1973; Schlenker, 1980).

Although the precise conditions that governthe occurrence of such genuine attitudechange are not completely understood, a hy-pothesis can be offered. Any factor that pre-vents the actor from labeling a public state-ment a lie should maximize the likelihood ofinternalization (Schlenker, 1980). For in-stance, if a self-serving account of a predica-ment is greatly discrepant from actors' initialprivate beliefs about themselves and theevent, they should label the account as a liedesigned to fool a real audience: It shouldnot be internalized. However, if a self-servingaccount is only slightly to moderately dis-crepant from actors' initial beliefs, theyshould label it as the truth and privately


believe that it accurately describes the event.Similarly, if a knowledgeable audience fa-miliar with the event accepts the account astruthful, the actors should be more likely tointernalize it.

In the present study, arguing against theuse of seat belts might have been sufficientlydiscrepant from subjects' private beliefs thattheir accounts were labeled as necessary butinaccurate. Likewise, Gaes et al. (1978) hadsubjects argue against toothbrushing, a topicthat also should be greatly discrepant fromprivate beliefs. Future research might bebetter directed toward the examination of theconditions under which genuine attitudechange occurs, instead of attempting to showthat genuine attitude change either does ordoes not always occur in forced compliancesettings.

General Discussion

The data from the three studies suggestthat interpersonal concerns can be quitesalient to subjects who are placed in a typicalforced compliance situation. Subjects do seemto use attitude statements as a means ofjustifying and legitimizing their actions whenthey confront social predicaments in whichtheir desired identities are threatened.Whether such accounting tactics result in aninverse or direct relationship between pay-ment magnitude and attitude expressions de-pends on whether the individual perceivesmoral evaluation in the situation to be lowor high. When moral evaluation is seen ashigh, a direct relationship is obtained, atleast when subjects must account to an audi-ence that observed the predicament. Such arelationship was obtained in Experiments 2and 3 and suggested in Experiment 1. Inaddition, subjects who receive large paymentsunder conditions of high moral evaluation at-tempt to interpret the payment in ways thatmake it appear more legitimate (Experiments1 and 3). In contrast, when moral evaluationis low, an inverse relationship occurs (Ex-periment 1).

The results of the three experiments canalso be examined within the theoretical frame-works of dissonance theory, self-perceptiontheory, and incentive theory. Considering

dissonance theory first, it appears that thedegree to which the theory can explain thedata depends largely on which interpretationof the theory is employed. According to theoriginal formulation of the theory, the coun-terattitudinal action should have been highlydissonant when payment (justification) wassmall. All of the factors implicated in thearousal of dissonance were maximized in the$.50 cells of Experiments 1 and 3 and in the$.50/agreement cell of Experiment 2: Sub-jects freely chose to make a potentially harm-ful speech, received minimal monetary justi-fication, and the money was introduced in amanner that should have affected their de-cisions (explicitly so in the agreement condi-tion of Experiment 2) . Yet a significant di-rect relationship was obtained in the agree-ment condition of Experiment 2 and in thebribery condition of Experiment 3, and thisrelationship was suggested in the high moralevaluation condition of Experiment 1.

For some dissonance theorists (e.g., Carl-smith & Freedman, 1968), factors introducedafter people freely decide to behave counter-attitudinally should not affect dissonancemagnitude. In this case, the moral evaluationmanipulations of Experiments 1 and 3 shouldhave had no effects on dissonance, yet thesemanipulations clearly were instrumental ingenerating a direct relationship. Aronson's(1968) revision of dissonance theory contendsthat dissonance is generated by discrepanciesbetween one's self-concept and behavior. Areasonable derivation of this approach mightsuggest that dissonance should be greatest(resulting in a strong inverse effect) whenpayment is small and moral evaluation ishigh, since discrepancies between importantaspects of one's self-concept and one's actionsare then most clearly evident. Yet a directrelationship appeared in these cases. Thus theabove "straightforward" derivations of dis-sonance theory were not supported.

However, there are two other possible dis-sonance interpretations that may account forthe obtained results. First, one might assumethat the experimenter's request was the majorforce for inducing the subjects' compliance(even though the money was offered as partof the request and subjects agreed to performthe counterattitudinal action only after it was


clear that they would receive the payment).Although such an assumption has usually notbeen made in prior dissonance studies, it is atleast plausible. One must also assume thatevents occurring after dissonance has beenaroused will affect dissonance magnitude.Given these assumptions, the introduction of"legitimate" money to the request consti-tutes a consonant cognition and reduces dis-sonance in relation to its magnitude. If, onthe other hand, the money is made to appearto be a bribe, it generates an additional dis-sonant cognition and increases dissonance inrelation to its magnitude.

Second, one could reconceptualize the pos-sible locus of dissonance in the presentstudies by focusing on different cognitions indifferent cells. One could say that dissonanceis generated by the cognitions, "I am a truth-ful, moral person who doesn't tell harmfullies," and "I just told a harmful lie" (Aron-son, 1968), and that these cognitions wereengaged in the low moral evaluation condi-tions of Experiment 1, producing an inverserelationship. But dissonance also might begenerated by the cognitions, "I am a moralperson who doesn't accept bribes" and "I justaccepted a bribe for an immoral action." Thispair of cognitions could lead subjects tochange their attitudes more the greater themagnitude of the payment (bribe) and couldhave been engaged under the conditions thatgenerated the direct relationships. Naturally,these derivations would produce another"evolved" form of dissonance theory (Green-wald & Ronis, 1978) that gets even furtheraway from the concept of cognitive incon-sistency. This evolved form would focus al-most exclusively on how people protect theirself-esteem and social identity when these arethreatened in real or imagined social inter-actions. In fact, the kinds of predictions typi-cally made by dissonance theorists can bereconceptualized as a "special case" of themore general social identity approach(Schlenker, 1978, 1980). Such a possibilityhas been indirectly suggested by Zimbardo(1969, p. 15), who argued that dissonancetheory "may be subservient to more basicphenomena which characterizes this approachas a 'face-saving theory.' "

In any event, it is clear that there are a

surprising number of possible dissonancetheory derivations that can be applied to thedata. At the very least, it can be said thatthere is nothing in dissonance theory to sug-gest that money or other inducing forces canbecome dissonant cognitions under certainconditions. Thus the present results also couldbe interpreted as extending basic dissonancetheory ideas.

Self-perception theory (Bern, 1972) wouldseem to have predicted an inverse relation-ship in all three experiments (particularly inthe agreement condition of Experiment 2, inwhich the payment was explicitly introducedin a way that was relevant to the descision toperform the behavior). The fact that thedirect relationship consistently appearedraises doubts that subjects were simply in-ferring their attitudes from their behavior.Additionally, it is not clear how self-percep-tion theory would explain the discrepanciesbetween the attributions made by the experi-mental subjects and the observers or the factthat the observers failed to track the actors'attitudes in Experiment 3. An extension ofself-perception theory that incorporates thenotion of self-awareness (Wegner & Finstuen,1977) predicts a dissonance-like inverse rela-tionship when subjects are objectively self-aware (focus on self) and an incentive-likedirect relationship when subjectively self-aware (focus on the environment), but thesepredictions were not supported either. Objec-tive self-awareness should have been high inall conditions, given the presence of the ob-servation mirror and videocamera, yet thedirect relationship was still obtained.

The data could be partially reconciled withself-perception theory if one accepts the hy-pothesis that choice should be inversely re-lated to payment magnitude when legitimateforces are salient and directly related to pay-ment magnitude when illegitimate forces aresalient. The choice data from Experiment 3support this hypothesis. When choice is high,subjects should infer their attitudes on thebasis of their behavior. Even so, it is difficultfor self-perception theory to explain whyattitude change occurred in Experiment 3only when subjects believed their responseswould be shown to the audience that accusedthem of being bribed. Further, the discrepan-


cies between the actors' and observers' attri-butions would still be unexplained by thetheory.

Finally, incentive theory predicted a directrelationship when payment was offered as anincentive for a high quality performance(performance condition of Experiment 2) .Indeed, subjects paid $2.50 in the perform-ance condition stated that they did a superiorjob on their speeches. Yet the direct relation-ship was obtained only in the agreement con-dition, where no performance quality differ-ences were obtained. Further, Rosenberg's(1965) prediction of an inverse relationshipwhen subjects experience high evaluation ap-prehension was not supported. Evaluationapprehension should have been particularlyhigh when subjects were accused of beingbribed in Experiment 3, yet the direct rela-tionship occurred.

It might be argued that the present experi-ments provided tests of the conditions thatgovern self-presentational behavior, but theexperiments are fundamentally different fromprior studies in the forced compliance arena,since it is not clear that genuine attitudechange was obtained. According to such anargument, the present studies would have fewor no implications for research and theory inthe area. There seem to be several flaws inthis argument. First, a surprising proportioneven of recent studies in the forced compli-ance area have employed measures of atti-tudes that could be readily used by subjectsfor self-presentation purposes. A survey of 25forced compliance studies that appeared inthe Journal of Personality and Social Psy-chology and the Journal of Experimental So-cial Psychology from 1970 through 1978showed that only seven clearly made an at-tempt to make subjects' postexperimentalattitude responses anonymous, thus reducingthe likelihood that subjects' responses wouldbe affected by concerns about their publicidentities. The rest either made no such at-tempt (many explicitly asked subjects tosign their names to the attitude question-naire) or did not provide sufficient informa-tion in the article to determine the condi-tions that surrounded the responses. Thusmuch of the existing forced compliance litera-ture does not allow one to differentiate clearly

between "genuine attitude change" and pub-lic self-presentation.

Second, the conditions that existing theories(dissonance, self-perception, incentive) pro-pose as necessary and sufficient for the oc-currence of attitude change were present inthe present studies. Further, effects occurredthat in isolation might have been labeled asmanifestations of "dissonance" or "self-per-ception." Given this fact, it is difficult toargue that the present results are irrelevantto the area without also calling into questionaspects of the existing theoretical knowledge.

Finally, it should be stressed that we arenot suggesting that all attitude change follow-ing the performance of counterattitudinal be-havior represents nothing but a public im-pression management strategy. As previouslymentioned, there are reasons to believe thatsubjects' attitude expressions in the forcedcompliance paradigm are sometimes genuinebut sometimes meant to influence or impressan audience. Future research should considerthe conditions under which the accountspeople use for their actions are privately rec-ognized as "con jobs" of others or are actualinternalized attitudes.

Reference Notes

1. Schlenker, B. R. Translating actions to attitudes,Grant proposal, 1977. (Department of Psychology,University of Florida, Gainesville, Fl. 32611.)

2. Shaw, M. E. Personal communication, November26, 1976.

References

Alexander, C. N., Jr., & Knight, G. W. Situatedidentities and social psychological experimentation.Sodometry, 1971, 34, 65-82.

Aronson, E. Dissonance theory: Progress and prob-lems. In R. Abelson et al. (Eds.), Theories of cog-nitive consistency: A sourcebook. Chicago: RandMcNally, 1968.

Bern, D. J. Self-perception theory. In L. Berkowitz(Ed.), Advances in experimental social psychology.New York: Academic Press, 1972.

Calder, B. J., Ross, J., & Insko, C. A. Attitudechange and attitude attribution: Effects of in-centive, choice, and consequences. Journal ofPersonality and Social Psychology, 1973, 25, 84-99.

Carlsmith, J. M., Collins, B. E., & Helmreich, R. L.Studies in forced compliance: I. The effect ofpressure for compliance on attitude change pro-duced by face-to-face role playing and anony-


mous essay writing. Journal of Personality andSocial Psychology, 1966, 4, 1-13.

Carlsmith, J. M., & Freedman, J. L. Bad decisionsand dissonance: Nobody's perfect. In R. Abelsonet al. (Eds.), Theories of cognitive consistency:A sourcebook. Chicago: Rand McNally, 1968.

Collins, B. E., & Hoyt, M. F. Personal responsibility-for-consequences: An integration and extension ofthe "forced compliance" literature. Journal oj Ex-perimental Social Psychology, 1972, 8, SS8-S93.

Cooper, J., & Worchel, S. Role of undesired conse-quences in arousing cognitive dissonance. Journalof Personality and Social Psychology, 1970, 16,199-206.

Crano, W. D., & Messe, L. A. When does dissonancefail? The time dimension in attitude measure-ment. Journal of Personality, 1970, 38, 493-508.

Elms, A. C. Role playing, incentive, and dissonance.Psychological Bulletin. 1967, 68, 132-148.

Elms, A. C., & Janis, I. L. Counter-norm attitudesinduced by consonant versus dissonant conditionsof role-playing. Journal of Experimental Researchin Personality, 196S, 1, 50-60.

Epstein, S. The self-concept revisited: Or a theory ofa theory. American Psychologist, 1973, 28, 404-416.

Festinger, L. A theory oj cognitive dissonance.Evanston, II.: Row, Peterson, 1957.

Fishbein, M., & Ajzen, I. Belief, attitude, intentionand behavior: An introduction to theory and re-search. Reading, Mass.: Addison-Wesley, 1975.

Gaes, G. G., Kalle, R. J., & Tedeschi, J. T. Impres-sion management in the forced compliance situ-ation. Two studies using the bogus pipeline.Journal of Experimental Social Psychology, 1978,14, 493-510.

Gerard, H. B. Compliance, expectation of reward,and opinion change. Journal of Personality andSocial Psychology, 1967, 6, 360-364.

Gerard, H. B., Conolley, E. S., & Wilhelmy, R. A.Compliance, justification, and cognitive change.In L. Berkowitz (Ed.), Advances in experimentalsocial psychology. New York: Academic Press,1974.

Greenwald, A. G., & Ronis, D. L. Twenty years ofcognitive dissonance: Case study of the evolutionof a theory. Psychological Review, 1978, 85, 53-57.

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

Helmreich, R. L., & Collins, B. E. Studies in forcedcompliance: Commitment and magnitude of in-ducement to comply as determinants of opinionchange. Journal of Personality and Social Psy-chology, 1968, 10, 75-81.

Holmes, J. G., & Strickland, L. H. Choice freedomand confirmation of incentive expectancy as deter-minants of attitude change. Journal oj Personalityand Social Psychology, 1970, 14, 39-45.

Hoyt, M. F., Henley, M. D., & Collins, B. E. Studiesin forced compliance: The confluence of choiceand consequences on attitude change. Journal ofPersonality and Social Psychology, 1972, 22, 1-7.

Janis, I. L., & Gilmore, J. B. The influence of in-centive conditions on the success of role-playingin modifying attitudes. Journal of Personality andSocial Psychology, 1965, 1, 17-27.

Jones, E. E., & Harris, V. A. The attribution of at-titudes. Journal of Experimental Social Psychology,1967, 3, 1-24.

Jones, E. E., & Wortman, C. Ingratiation: An at-tributional approach. Morristown, N. J.: GeneralLearning Press, 1973.

Kaufmann, D. R. Incentive to perform counter-attitudinal acts: Bribe or gold star? Journal ofPersonality and Social Psychology, 1971, 19, 82-91.

Kelley, H. H. Attribution theory in social psychol-ogy. In D. Levine (Ed.), Nebraska symposium onmotivation (Vol. 15). Lincoln: University ofNebraska Press, 1967.

Kelman, H. C., & Baron, R. M. Inconsistency as apsychological signal. In R. P. Abelson et al. (Eds.),Theories of cognitive consistency: A sourcebook.Chicago: Rand McNally, 1968.

Kelman, H. C., & Baron, R. M. Moral and hedonicdissonance: A functional analysis of the relation-ship between discrepant action and attitude change.In S. Himmelfarb & A. H. Eagly (Eds.), Readingsin attitude change. New York: Wiley, 1974.

Linder, D. E., Cooper, J., & Jones, E. E. Decisionfreedom as a determinant of the role of incentivemagnitude in attitude change. Journal of Per-sonality and Social Psychology, 1967, 6, 245-254.

Lyman, S. M., & Scott, M. G. A sociology of theabsurd. New York: Appleton-Century-Crofts,1970.

Nel, E., Helmreich, R., & Aronson, E. Opinionchange in the advocate as a function of the per-suasibility of his audience: A clarification of themeaning of dissonance. Journal of Personality andSocial Psychology, 1969, 12, 117-124.

Nuttin, J. M., Jr. The illusion of attitude change:Toward a response contagion theory of persuasion.New York: Academic Press, 1975.

Riess, M., & Schlenker, B. R. Attitude change andresponsibility avoidance as modes of dilemmaresolution in forced compliance settings. Journaloj Personality and Social Psychology, 1977, 35,21-30.

Rosenberg, M. J. When dissonance fails: On elimi-nating evaluation apprehension from attitude mea-surement. Journal of Personality and Social Psy-chology, 1965, 1, 28-42.

Schlenker, B. R. Self-image maintenance and en-hancement: Attitude change following counter-attitudinal advocacy. Proceedings of the 81st An-nual Convention of the American PsychologicalAssociation, 1973, 8, 271-272.

Schlenker, B. R. Attitudes as actions: Social identitytheory and consumer research. Advances in Con-sumer Research, 1978, 5, 352-359.

Schlenker, B. R. Impression management: The self-concept, social identity, and interpersonal rela-tions. Belmont, Calif.: Brooks/Cole, 1980.

Schlenker, B. R., & Schlenker, P. A. Reactions fol-


lowing counterattitudinal behavior which producespositive consequences. Journal of Personality andSocial Psychology, 19-75, 31, 962-971.

Sherman, S. J. Effects of choice and incentive onattitude change in a discrepant behavior situation.Journal of Personality and Social Psychology,1970, IS, 245-252.

Steiner, I. D. Perceived freedom. In L. Berkowitz(Ed.), Advances in experimental social psychology.New York: Academic Press, 1970.

Tedeschi, J. T., Schlenker, B. R., & Bonoma, T. V.Cognitive dissonance: Private ratiocination or pub-lic spectacle? American Psychologist, 1971, 26,685-695.

Verhaeghe, H. Mistreating other persons through

simple discrepant role playing: Dissonance arousalor response contagion. Journal of Personality andSocial Psychology, 1976, 34, 125-137.

Wegner, D. M., & Finstuen, K. Observers' focus ofattention in the simulation of self-perception.Journal of Personality and Social Psychology,1977, 35, 56-62.

Wicklund, R. A., & Brehm, J. W. Perspectives oncognitive dissonance. Hillsdale, N.J.: Erlbaum,1976.

Zimbardo, P. G. The cognitive control of motiva-tion: The consequences of choice and dissonance.Glenview, 111.: Scott, Foresman, 1969.

Received April 9, 1979 •

Documents

Self-Presentational Analysis of the Effects of Incentives