Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

PER SON AL ITYAND SO CIAL PSYCHOL OGYBUL LE TIN

When Be liefs Yield to Evidence: Re ducing


Biased Evaluation by Af firming the Self

Geoffrey L. Co hen
Yale University
Joshua Aronson
University of Texas, Austin
Claude M. Steele
Stan ford University

People often cling to beliefs even in the face of disconfirming evi- 1997) , much research has dem on strated th e power ful
dence and interpret ambigu ous information in a man ner that influ ence that beliefs can have on the inter pre ta tion of
bolsters strongly held attitudes. The authors tested a motiva- new information ( see Gilovich, 1991, for a review) .
tional account suggesting that these defen sive reactions would be This article begins with the assump tion th at beliefs
ameliorated by an affirmation of an alternative source of can con stitute valued sources of iden tity. They may thus
self-worth. Con sistent with this interpretation, participants were be given up on ly with great reluctan ce, and they may be
more persuaded by evidence impugn ing their views toward capi- embraced even when they con flict with the demands of
tal pun ishment when they were self-affirmed than when they fact, logic, or mate rial self-in ter est ( e.g., Abelson, 1986;
were not (Studies 1 and 2). Affirmed participants also proved Sears & Fun k, 1991) . Cap ital pun ish ment pro po nents,
more critical of an advocate whose argu ments con firmed their for exam ple, migh t clin g to a belief in th e death pen alty’s
views on abortion and less con fident in their own attitudes regard- deter rent efficacy in large part because it rein forces their
ing that issue than did unaffirmed participants (Study 3). iden tity as polit ical con ser va tives ( Ellsworth & Ross,
Results suggest that assimilation bias and resistance to persua- 1983) . The con flict ing attitudes that Blacks and Whites
sion are mediated, in part, by iden tity-main tenance motivations. h ad about th e O.J. Simpson trial migh t also h ave arisen ,
in part, from a desire to affirm racial iden tity and solidar -
ity. Eviden ce th at ch allenges th e valid ity of such cher -
On e of th e great est pains to hu man n a ture is the pain of ish ed beliefs presents a self-threat inso far as givin g up
a n ew idea. that belief would entail losin g a source of esteem or iden -
—Walter Bageh ot tity. To neu tralize that threat, peo ple are apt to evalu ate
evidence defen sively ( Ditto & Lopez, 1992; Dunning,
Leuenberger, & Sherman, 1995; Kunda, 1990; Munro &
Man y of us h ave h ad th e puzzling expe rien ce of pre - Ditto, 1997; Zuwerink & Devine, 1996). Scru tinizing belief-
sent ing argu ments that challen ge a friend’s endorse - discon firmin g evidence for fault or accept in g at face
men t of a presiden tial can didate or belief in ESP only to
find our appeals met with resistan ce. As social psycholo - Co h en et al. / WH EN BE LIEFS YIELD TO EV I DENCE

gists h ave lon g observed, peo ple tend to per sist in ch er - Authors’ Note: We thank Joshua Correll for his work on this pro ject
ish ed beliefs an d attitudes even wh en con fronted with and his in sight ful con tribu tions to th is ar ticle and Da vid Sh erman an d
clear and con tra dictory evidence ( Festinger, Riecken, & Sa rah Wert for com ments on ear lier drafts. In ad dition, we owe Lee
Schachter, 1956; Ross & Lepper, 1980) . Th ey also ten d to Ross an d Mark Lepper a large debt for th eir feed back on this re search
and for the primary title of th e present ar ticle, which we bor row from
evalu ate ambigu ous infor ma tion in a man n er that bol-
their 1980 ar ticle on be lief per se ver an ce. Cor re spon dence con cern ing
sters pre exist ing views ( Lord, Ross, & Lepper, 1979) . th is ar ticle sh ould be ad dressed to Geoffrey L. Co hen, De part ment of
From work on self-fulfilling proph e cies ( Snyder, Tanke, Psychology, Yale Univer sity, 2 Hillhouse Ave nue, Box 208205, New Ha-
& Bersch eid, 1977) an d in ter per son al expectan cies ven, CT 06520-8205; e-mail: co h en @psych .stan ford.edu.
( Darley & Gross, 1983) to investiga tions of ste reo types PSPB, Vol. 26 No. 9, September 2000 1151-1164
and preju dice ( Ham ilton & Rose, 1980; Munro & Ditto, © 2000 by the So ciety for Per son ality an d So cial Psychology, In c.

1151
1152 PERSO N ALITY AND SO CIAL PSYCH O LO GY BULLETIN

value belief-con firm in g evidence h elps to pro tect on e’s self-esteem individ u als, it h as been foun d, ten d to resist
belief and the iden tity it affirms. per sua sive messages more than do low-self-esteem indi -
Buffering peo ple against self-threat, however, should vid u als, pre sum ably because peo ple with h igh self-
atten u ate defen sive pro cessing. Accord ing to self-affir - esteem have greater con fidence in the valid ity of their
ma tion th e ory, the potency of a psych o logical threat lies beliefs ( e.g., Coh en , 1959; Jan is, 1954; Zelln er, 1970; cf.
in its capacity to imperil a global sen se of self-worth Rhodes & Wood, 1992) . Based on this find in g, it would
( Steele, 1988) . Because global self-worth derives from seem th at en h an cing self-worth with an affir ma tion pro -
man y sources, peo ple have much flexibility in h ow th ey ce dure should increase resistan ce to per sua sion rath er
cope with a par ticu lar self-threat. They can reaffirm th eir th an decrease it. On e rea son for th e differ en ce between
self-worth directly—in this case, by defen sively evalu ating ear lier results an d our own pre diction s involves the pres-
th e per sua sive evidence—or th ey can do so in directly— en t research ’s use of affir ma tions un re lated to the atti-
by affirm in g other equally valued domains of self-worth. tude issue. Th e effective ness of such self-affir ma tion s lies
In deed, research suggests th at such “self-affir ma tions” in their capacity to remin d peo ple that their self-worth
can reduce defen sive pro cessin g of h ealth risk in for ma - derives from sources other than the attitude issue. Al-
tion ( e.g., Reed & Aspinwall, 1998; Sherman, Nelson, & though high self-esteem might increase self-con fidence,
Steele, 2000) . They have also been shown to increase it would n ot n ecessar ily con fer th e flexibility th at an affir -
peo ple’s open n ess to in for ma tion that threat ens th eir ma tion does in terms of cop in g with self-image threats.
self-inter ests or motives ( Bastardi & Ross, 2000) . In deed, research suggests th at un less th ey are first re-
Accord in g to our analysis, peo ple sh ould prove less min ded of th eir esteem resources, h igh -self-esteem in di-
defen sive an d resistant in th e face of a coun terattitudin al vid u als will prove just as defen sive as their low-self-esteem
message wh en alter n a tive sources of self-worth are but - p eers in resp on se to a sp e cific self-th reat ( Steele,
tressed or activated. For exam ple, a cap ital pun ish ment Spencer, & Lynch, 1993; cf. Brown & Smart, 1991) .
pro po n en t should feel more open to evidence challen g - One other factor h elps to explain th e differ ence
ing the death pen alty’s effective n ess if h e or sh e feels between our pre diction s and find in gs obtained in past
affirmed as a good friend or valued employee. Self-affir - research. In gen eral, ear lier stud ies did n ot present par -
ma tions, we argue, trivialize th e attitude as a source of ticipants with threat en in g per sua sive messages—these
self-worth and thus make it easier to give up. stud ies typ ically used vague or oth er wise weak messages
The presen t con cep tual analysis may be com pared addressin g issues tan gen tial to par ticipants’ self-in ter ests
with that offered by research ers workin g in oth er th e o - or iden tities ( e.g., the effect of TV on the movie industry
ret ical tra ditions. Cogn itive disso n an ce th e o rists, for or whether pen icillin con stitutes a won der drug) . Such
exam ple, would suggest that infor ma tion in con sistent messages would trigger lit tle if an y defen sive pro cessin g
with stron gly h eld beliefs in duces an aversive state of because par ticipants were un likely to h ave a per sonal
arousal. Peo ple can reduce that arousal, and restore con - in vest men t in the issue. Even if they did have such an
so nance, by ch allenging the valid ity of th e disso nant in vest men t, th e messages often were insufficiently per -
infor ma tion, for exam ple, by den igrat ing its source as sua sive to pose a serious th reat to par ticipants’ beliefs.
untrust wor thy ( E. Aron son , Turn er, & Carlsmith , 1963; Accord in g to our an alysis, affirm ing self-worth should
Zimbardo, 1960) . Like disso n an ce th e ory, our an alysis atten u ate defen sive pro cessin g, an d to in duce such
implies th at an aversive drive state mediates, in part, defen sive ness it is essen tial to expose par ticipan ts to
resistan ce to coun terattitudin al m essages ( see also strong evidence th at con flicts with a ch er ish ed attitude.
Mun ro & Ditto, 1997; Zuwerin k & Devin e, 1996) . In con - Our three stud ies thus fea tured social-polit ical par ti-
trast to disso n an ce the ory, however, we argue that this san s resp on d in g to h igh ly p er su a sive evi d en ce. In
drive state reflects th e activa tion of ego-pro tective moti- Studies 1 an d 2, we expose cap ital pun ish ment par tisans
va tion s rath er th an con sistency-restor ing ones ( Steele, to a counterattitudinal scien tific report regard in g the
1988) . Accord ingly, disso n an ce may be atten u ated by death pen alty an d pre dict th at a self-affir ma tion will lead
addressin g th e pro vok in g in con sisten cy directly ( by them to be more positively in flu en ced by that report. In
attacking th e spe cific per sua sive appeal) or indirectly Study 3, we presen t prochoice and prolife advo cates with
( by reflect ing on alter n a tive sources of self-worth ) . Cog- a debate on abor tion. Here, we pre dict that a self-affir -
nitive disso n an ce the ory, at least in its most straight for - ma tion will atten u ate biased evalu a tion of the debate,
ward form, would n ot pre dict th at h eigh t en ed feelin gs of th at is, th e ten dency to rate th e like-min ded debater
self-worth in one domain would reduce disso nance ( an d more favor ably than the debater from the other side.
thus resistan ce to per sua sion) in an oth er. Alth ough th ese pre dictions follow from our con cep tual
The pat tern of fin d in gs that we pre dict also differs analyses, it is worth reit er at in g th at pre vious per sua sion
from th at pre viously obtained in research exam in in g the mod els would anticipate the oppo site pat tern of results—
role of self-esteem in mod er at ing persuasibility. High - th at self-affir ma tions sh ould raise self-con fidence an d
Co hen et al. / WH EN BELIEFS YIELD TO EVIDENCE 1153

thereby increase biased evalu a tion and resistan ce to per - At th e time th at th ey filled out th e preselection ques-
sua sion ( Coh en , 1959; Jan is, 1954; Zelln er, 1970) . In a tion n aire, par tic ipan ts also com pleted a ver sion of
sense, our stud ies pit the self-affir ma tion logic against Harber’s ( 1995) Sources of Valida tion Scale—responses
these ear lier the o ret ical frame works. to which would later be used in the prep a ra tion of the
self-affir ma tion man ip u la tion ( see the appen dix) . In the
STUDY 1
presen t study, th e question n aire asked stu den ts to rate
several traits and values in terms of th eir per son al impor -
Pro po n en ts and oppo n en ts of cap ital pun ish men t tan ce. Th e question naire included the values of sense of
were pre sen ted with a coun terattitudin al scien tific h umor, ath letic skills, and rela tion s with friends but did
report regard ing th e death pen alty ( see also Lord et al., n ot in clude items such as religion an d fam ily values,
1979) . Th ey were told th at th eir mem ory of the report which might be related to cap ital pun ish ment attitudes.
would be tested; thus, they pre sum ably felt motivated to
Procedure
read th e report system at ically rath er th an periph er ally.
Before doing so, however, par ticipan ts were ran domly Because research h as sh own th at warn ing of per sua -
assign ed eith er to an affir ma tion con dition or to a sive in ten t in creases resistance to per sua sion ( McGuire,
no-affir ma tion con dition. In the affir ma tion con dition, 1985) , a cover story was n ecessary. Stu den ts par ticipated
they wrote an essay about a per son ally impor tan t trait or in th e study in divid u ally, an d after bein g greeted by a
value un re lated to th eir views on cap ital pun ish ment male exper imen ter, th ey were told th at th e study con -
( Fein & Spen cer, 1997; Steele, 1988) . In the no-affir ma - cern ed m em or y. Th e research ers, th ey were told ,
tion con dition, th ey wrote an essay about a per son ally wan ted to exam in e the rela tion ship between sub jective
unim por tant topic. The depend ent mea sures in cluded mem ory—mem ory of per son al events as expe rien ced
both attitude change and question n aire items assessing first hand—and objective mem ory—mem ory of less per -
the gen eral positivity of response to the scien tific report. son ally rele vant stim uli. Par ticipants were in formed th at
th ey would first recall a few expe riences from their per -
son al life in a brief essay ( th e sub jective mem ory exer -
METH O D
cise) and that they would then read a scien tific report,
Design and Participants aspects of wh ich th ey would later try to remem ber ( the
objective mem ory exer cise) . After sign ing the con sent
Th e exper iment fea tured a 2 × 2 facto rial design, with form an d bein g assured of th e con fiden tiality of their
par tisan sh ip of the par ticipant ( pro po nent of cap ital responses, par ticipants were pro vided with wh at osten si-
pun ish ment or oppo nent) and affir ma tion con dition bly was the sub jective mem ory exer cise. They were given
(affir mation or no affir mation) the between-par ticipants a sheet of paper titled “Per son al Recall Exer cise” with
factors. in struction s printed under neath. In fact, this sub jective
Par ticipan ts con sisted of 36 male an d 41 female mem ory exer cise con stituted the self-affir ma tion manip -
under grad u ates wh o eith er received credit in an in tro - u lation. Par ticipants were ran domly assigned either to an
ductory psych ology course or were paid $6 for par ticipa - affir ma tion con dition or to a no-affir ma tion con dition.
tion . Stu den ts were recruited on th e basis of th eir Affirmation con dition. Respon ses to th e Sources of Vali-
responses to a preselection sur vey admin istered ear lier da tion Scale ( which, as noted ear lier, h ad been admin is-
in the aca demic quar ter. On e item in th is sur vey asked tered ear lier in the aca demic quar ter) were used to pre -
stu dents to indicate their attitude toward cap ital pun ish - p ar e th e m ate r i als in th e affir m a tio n co n d i tio n .
ment on a scale ranging from 1 ( very much in favor) to 13 Spe cifically, th e most h igh ly rated trait or value in each
( very much opposed) . Stu den ts who had indicated th at th ey par ticipan t’s Sources of Valida tion question naire h ad
either strongly favored cap ital pun ish ment ( a 1, 2, or 3) been iden tified an d embed ded in th e in structions con -
or stron gly opposed it ( an 11, 12, or 13) were tele phoned tain ed on th e Per sonal Recall Exercise ( when more than
an d in vited to par ticipate in the study. Data from 5 par tic- on e value h ad been rated h igh ly, the first one listed on
ipants ( 3 in the affir ma tion con dition, 2 in th e n o-affir - th e question n aire was used) . The instruction s on the
ma tion con dition) were discarded prior to an alyses, 4 per sonal recall sheet asked par ticipan ts to describe three
stu dents wh o expressed suspicion about our con cern or four per sonal expe riences in which their most highly
with per sua sion and attitude ch an ge an d 1 stu dent wh o rated ch ar acter istic from the Sources Scale had been
doubted the authen ticity of the scien tific report. Th is impor tant to them and had made them feel good about
left a total of 72 par ticipants—25 oppo nents an d 47 pro - th em selves. For exam ple, a par ticipant who h ad rated
po nents—ran domly assign ed to th e two exper imen tal sen se of h umor most h igh ly on th e Sources Scale was
con ditions. ( Fewer oppo nents were recruited sim ply be- instructed to write about “per sonal expe riences in which
cause of their dearth in our available par ticipan t pool.) your sense of humor was impor tant to you an d made you
1154 PERSO N ALITY AND SO CIAL PSYCH O LO GY BULLETIN

feel good about your self.” Par ticipan ts were fur th er Counterattitudinal scien tific report. Par ticipan ts were
instructed to pick one of these expe rien ces and to write a then given a 4 -page scien tific report addressin g the mer -
sh ort story describ ing th e even t an d th eir feelin gs at the its ( or lack th ereof) of cap ital pun ish men t, and they
time. Steele and his colleagues have found that reflect - were pro vided as much time as they wanted to read this
ing on a per son ally impor tant self-ch ar acter istic or value report care fully. The report appeared to be a lead article
is an effective means of inducin g self-affir ma tion ( Steele, from a recent issue of the Journal of Law and Human
1988) . Behav ior, writ ten by two research ers from Yale, titled
“Th e Death Pen alty: New Evidence Informs an Old
No-affirmation con dition. In this con dition, th e in struc-
Debate.” In fact, the report was fictitious. It con tain ed
tion s on th e Per son al Recall Exer cise asked par ticipants
facts, sta tistics, an d argu ments whose cumu la tive effect
to list, in as much detail as th ey could, everything th at
was to wage a per sua sive attack on par ticipan ts’ attitudes
they had eaten or drank in the past 48 hours. They were
toward cap ital pun ish ment. Pro po nents of the death
fur ther told “n ot to worry about th ose th in gs you fin d
pen alty th us read an an ti–cap ital pun ish men t report. By
your self un able to remem ber.” We chose this con trol
con trast, oppo nents read a pro–death pen alty report.
con dition ( in stead of on e th at asked par ticipants to write
The argu ments that each report pre sen ted addressed
about an unimpor tant value) because stu den ts ten d to
the effective ness of the death pen alty as a deter rent, its
turn almost any self-reflective writ in g task into a self-
affirm ing one. Nota bly, in our stud ies, th e affir ma tion eco nomic sound n ess, its mer its as a mean s of in capacitat -
ing known mur der ers, and its eth ical value. The argu -
effects were un re lated to th e ch ar acter istic ch o sen an d
n o stu den t wrote about his or her social-polit ical beliefs. ments were highly per sua sive and backed up, where pos-
sible, with rele van t sta tistics and research find ings. The
After th e exper imen tal man ip u la tion ( which took up
research pre sented was styled after auth en tic cap ital
to 13 min utes to com plete) , par ticipants were told th at
pun ish ment lit er a ture ( see Lord et al., 1979) . Impor -
they would now com plete the objective mem ory exer -
tan tly, the report that pro po nents read was iden tical in
cise, wh ose pur pose, th ey were remin ded, in volved
for mat to th e on e th at oppo n en ts read—both described
assessing th eir mem o ries of stim uli of a “more objective
rele vant research iden tical in meth od ology—but the two
an d less per son ally rele van t nature.” Par ticipants were
reports pre sented oppo site fin d ings and con clu sions.
th us in formed th at th ey would read a scien tific article
For exam ple, pro po nents of cap ital pun ish men t read
an d th en try to remem ber as much of its con ten t as they
th e following excerpt:
could.
To heighten the plau sibility of th is cover story, and to
New research tools have helped research ers to over come
fur ther allay suspicions regard ing the study’s con cern
so m e o f t h e sh o r t co m in gs t h at p lagu e d e ar lie r
with per sua sion an d cap ital pun ish ment attitudes, par - work. . . . Cran dall ( 1991) fin ished a 10-year study com -
ticipants were given a choice between two articles to par ing mur der rates for th e years before an d th e years
read, alth ough th is ch oice was forced. Th ey were told after adop tion of cap ital pun ish ment in 14 states. In 12
that the research ers wanted par ticipan ts to have at least a of the 14 states, mur der rates were significantly h igh er
min imal in ter est in the stim u lus article. Th e exper i- after th e adop tion of the death pen alty, in man y states by
men ter gave th e par ticipan t a sh eet of paper list ing two as much as 35%. This find ing h eld even wh en com pet ing
abstracts and asked par ticipan ts to pick an article to read factors, such as changes in a state’s social and eco n omic
on the basis of these abstracts. The abstracts were ficti- sta tus and in its prior mur der rate, were accounted for.
tious but each appeared to be an excerpt from an Finally, much evidence has shown that when an exe cu -
authen tic scien tific article. Th e first abstract described a tion is h igh ly pub licized . . . state and national mur der
rates increase dra mat ically ( Vidmar, 1991) .
dry article about research on structure-map ping an d
systematicity in lin guistics, whereas the second described
an article about the var ious issues in volved in cap ital By con trast, oppo nents of cap ital pun ish ment read
pun ish ment policy. All but 2 par ticipants ch ose to read the same para graph except th at th e ph rases “h igh er
the cap ital pun ish ment article. In response to the 2 less after” and “increase dra mat ically” were replaced with the
coop er a tive par ticipan ts, th e exper imen ter fum bled ph rases “lower after” an d “decrease dra mat ically.”
th rough papers an d fold ers, cursed a research assistan t Dependent measure. After th ey read th e scien tific
for havin g misplaced th e lin guistics article, an d asked report, par ticipants were given the depend ent measure
whether the par ticipan t would con sider read in g the cap - question n aire design ed to assess th e favorability of th eir
ital pun ish ment article instead. Both par ticipants were responses to the report. To probe for attitude change,
h appy to acquiesce, although postexperimental debrief- one item asked, “What is your attitude toward cap ital
ing indicated th at 1 of th ese par ticipants suspected the pun ish men t?” on a scale ran gin g from 1 ( very much in
true pur pose of the study; his data were discarded from favor) to 7 ( undecided) to 13 ( very much opposed) . Another
an alysis ( as noted ear lier) . item asked, “How much , if at all, did th e article affect
Co hen et al. / WH EN BELIEFS YIELD TO EVIDENCE 1155

your over all attitude toward cap ital pun ish ment?” on a toward cap ital pun ish men t, th eir rat in gs of its deter -
scale rangin g from 1 ( much more opposed) to 7 ( no change ren t efficacy and eco nomic sound n ess, an d th eir rat ings
in attitude) to 13 (much more in favor) . Two other items of the authors’ polit ical orien ta tion were also reverse-
asked, “In your opin ion, how effective a deter ren t is cap i- coded, as were pro po nents’ rat ings of th e exten t to
tal pun ish men t?” an d “In your opin ion , how eco n om i- wh ich th e article affected th eir over all attitude. The
cally sound is cap ital pun ish ment?” on sep a rate 13-point result ing postmanipulation items are con cep tu ally sim i-
scales ran ging from 1 (extremely) to 7 ( moderately) to 13 lar an d form a reliable in dex of th e favorability of par tici-
( not at all) . Finally, several items assessed evalu a tions of pan ts’ respon se to th e report ( Cron bach ’s alpha = .70) .
th e auth ors of th e article. One item asked, “How would Con se quen tly, these items were aver aged ( after being
you describe the polit ical orien ta tion of th e research ers stan dard ized to equate their varian ce) in to a sin gle
wh o wrote th e article?” on a scale ran gin g from 1 com posite.
( extremely liberal) to 7 (neu tral) to 13 ( extremely con serva- An index of attitude change also was com puted by
tive) . Th e oth er items asked, “In your opin ion, how rea - sub tract in g each par ticipant’s preman ipulation attitude
son able are th e research ers?” an d “How in formed do rat in g from h is or h er postman ipulation attitude rat ing.
you think the research ers are?” on sep a rate scales ran g - The result in g change score of oppo nents was th en multi-
ing from 1 (extremely) to 7 (moderately) to 13 ( not at all) . plied by –1 so th at positive num bers for oppo nents an d
These lat ter items con stituted estab lish ed mea sures of pro po n en ts alike would reflect greater attitude change
open ness to per sua sion ( e.g., E. Aronson et al., 1963; in the direction of the report.
Ross & Ward, 1995) . As past research suggests, polit ical
par tisans are apt to main tain th e san ctity of their beliefs Effects Along Favorability
by attrib ut in g opposing views to polit ical ide ology, ign o - of Response Composite
rance, or irra tio nality ( Ross & Ward, 1995) . Results sup ported th e pre diction that the affir ma tion
After com plet in g th e question n aire, par tic ipan ts wou ld p ro duce more favor able evalu a tion s of atti-
were probed for suspicion an d debriefed. Th e exper i- tu d e-d iscon firm in g evid en ce. Affirm ed par tic ipan ts
menter explained that the report they read was fictitious responded more positively to th e scien tific report (M =
and discussed the necessity of with hold ing the true pur - 0.18) than did unaffirmed par ticipan ts (M = –0.23) , F( 1,
pose of the study until its com ple tion. Par ticipants were 63) = 7.10, p = .01.
eith er paid $6 or pro vided with the rele van t signa ture to Of less impor tan ce, the analysis also yielded a main
obtain course credit. effect of par tisan ship; pro po nents of cap ital pun ish ment
responded more favor ably to the report ( M = 0.16) than
RESULTS AND DISCUSSIO N did oppo n en ts ( M = –.21) , F( 1, 63) = 5.93, p < .02. One
In Studies 1 and 2, data were ana lyzed using a two-way possible rea son for this find in g is th at th e an ti–cap ital
ANOVA, with par tisan ship ( oppo nents or pro po nents) pun ish ment report ( that pro po n en ts read) may have
and exper imen tal con dition ( affir ma tion or n o affir ma - been more per sua sive than the pro–cap ital pun ish ment
tion ) as between -p ar tic i p an ts fac tor s. Gen d er was report ( th at oppo nents read) . Th ose wh o oppose cap ital
included as a factor in all three stud ies an d, un less oth er - pun ish ment may regard scien tific evidence as less con -
wise noted, there were neither main effects nor inter ac- vincing because their oppo sition may derive from a
tion s in volving it. Also, some par ticipants failed to com - moral con viction rather than a scien tific ratio nale.
plete several question naire mea sures; as a result, degrees Effects Along Attitude
of free dom vary slightly. Change In dex
Creation of “Favorability of
The analysis yielded on ly a main effect of par tisan -
Response” Composite and In dex
ship. Pro po n en ts sh owed more attitude change in the
of “Attitude Change”
direction of th e report (M = 2.33) than did oppo nents
To obtain a gen eral in dex of th e exten t to wh ich par - ( M = 0.97) , F( 1, 64) = 6.34, p < .015. Con trary to pre dic-
ticipan ts were positively in flu enced by th e scien tific tion s, h owever, affirmed par ticipan ts did not change
report, we sim ply aver aged th e depen d en t mea sure their attitudes more ( M = 1.74) than did unaffirmed par -
items into a sin gle “favorability of response” com posite. ticipants ( M = 1.56) , F < 1. ( The pat tern of means, at
Several item s were first reverse-coded, h owever, to least, was in th e direction anticipated by our con cep tual
en sure th at h igh er rat in gs along all scales would reflect, an alysis rath er th an by mod els con cern ing the role of
for pro po n en ts an d oppo n en ts alike, more positive self-esteem in persuasibility [ e.g., Jan is, 1954] .)
responses to the scien tific report. Rat in gs of the rea son - In sum, the affir ma tion led par ticipants to respond
able ness an d in formedn ess of th e auth ors were th us more favor ably to the disconfirming eviden ce, although
reverse-coded for all par ticipants. Oppo nents’ attitude it did n ot prompt th em to give up th eir gen eral attitude
1156 PERSO N ALITY AND SO CIAL PSYCH O LO GY BULLETIN

toward cap ital pun ish ment. It seems likely that attitudes METH O D
about cap ital pun ish ment are sim ply more resistan t to
Design and Participants
change than are spe cific beliefs about its deter rent effi-
cacy or par ticu lar impressions of the authors of the Once again, the exper imen t fea tured a 2 × 2 facto rial
report. In a sense, one’s gen eral attitude toward cap ital design , with par tisan ship of th e par ticipan t ( pro po nent
pun ish ment is overdetermined, tied to rele van t values or oppo n en t of cap ital pun ish men t) an d affir ma tion
an d refer ence groups ( Katz, 1960) , an d groun ded in con dition ( affir ma tion or no affir ma tion) as between-
past beh avioral com mit ments ( Festin ger, 1957) . As dis- par ticipan ts factors.
so nance research ers have long noted, cen tral, self-defin - Par ticipan ts were 35 female an d 47 male Stan ford
in g cogn ition s prove more resistan t to ch an ge th an under grad u ates wh o received course credit for an in tro-
noncentral ones ( Coo per & Mackie, 1983; Pilisuk, 1968; ductory psychology course. Study 2 used the same gen -
Zu werin k & Devin e, 1996) . Th u s, th e d isso n an ce eral selection pro ce dure used in Study 1. However, in the
involved in givin g up a cen tral attitude may have been presen t study, a 7-point scale was used instead of the
too great, and par ticipants in Study 1 may h ave ch o sen 13-poin t scale described in Study 1 ( with par ticipants
other, less pain ful ave nues of ch an ge th at were assessed who cir cled a 1 or 2 qualifying as pro po n en ts and those
who cir cled a 6 or 7 qualifyin g as oppo n en ts) . Data from
by the question naire. In th is respect, writ ing about a per -
2 par ticipan ts—1 in the affir ma tion con dition, 1 in the
sonal trait may not have been affirm in g enough to neu -
no-affir ma tion con dition—were discarded prior to anal-
tralize the self-threat inher ent in givin g up one’s gen eral
yses because th ey expressed suspicion about our con -
attitude toward cap ital pun ish ment.
cern with per sua sion and attitude ch an ge amon g cap ital
pun ish ment par tisan s. Th is left a total of 80 par tici-
STUDY 2 pants—38 oppo n en ts of cap ital pun ish men t, 42 pro po -
Accord in gly, in Study 2, we used a more power ful n en ts—ran domly assign ed to th e two exper imen tal
self-affir ma tion p ro ce dure—par tic ipan ts were given con dition s.
positive feed back regard ing a per son ally impor tant skill. Procedure
Whereas in Study 1 par ticipan ts reflected on a prob a bly
Stu dents again par ticipated in the study individ u ally.
familiar event that had taken place in the past, in Study 2
They were greeted by a female exper imen ter who told
they expe rien ced a n ew sit u a tion designed to induce
them that the study con cern ed “social per cep tive n ess,”
self-affir ma tion. We also decided to focus our depend ent
th at is, “th e ability to read th e mean in g beh in d oth er
mea sure question naire almost exclu sively on attitude
peo ple’s physical gestures an d facial expressions.” Par tic-
change. We suspected th at multiple measures might ipants were told th at th is ability had proved highly cor re -
dilute the affir ma tion effects by pro vid ing alter na tive lated with career success and that the research ers were
out lets for par ticipants to relieve pressures to ch an ge. attempt in g to under stand the nature and origin of this
In deed, past research suggests th at disso nant infor ma - impor tant skill. Par ticipants were informed that they
tion will exert less in flu en ce on cen tral, resistan t would first take a well-validated test of social per cep tive -
cognitions if par ticipan ts believe th at th e question naire n ess. After ward th ey were told th ey would com plete
includes response items other than those rele van t to another exer cise in social per cep tive n ess—th ey would
change along this cen tral cognition ( Götz-Marchand, watch a vid eo taped pre sen ta tion made by a per son
Götz, & Irle, 1974) . speaking on a social issue an d th ey would th en try to
As in Study 1, Study 2 pre sen ted oppo nents an d pro - assess the pre sen ter’s true thoughts and feelings. After
po nents of cap ital pun ish men t with a scien tific report sign ing the con sen t form, an d bein g assured of th e con fi-
that challenged th eir views on th e death pen alty. As in den tiality of all th eir respon ses, par ticipants began the
Study 1, prior to read in g th is report, h alf of th e par tici- social per cep tive ness test.
pants were ran domly assigned to an affir ma tion con di- Test of social perceptiveness. The test was pre sented as th e
tion , wh ereas th e remain ing par ticipants were assigned “Archer Test of Social Per cep tive ness,” an d it con sisted
to a no-affir ma tion con dition. But, in the case of the of 25 pho to graph s of peo ple inter act in g in var ious sit u a-
presen t study, par ticipants in the affir ma tion con dition tion s ( taken from Arch er, 1980) . For each pho to graph, a
received positive feed back regard ing their per for man ce multiple-choice question was pre sen ted th at pre sum ably
on a test of their social per cep tive ness. Par ticipan ts in required par ticipants to infer the thoughts and feelin gs
th e n o-affir ma tion con dition com pleted the same test of th e ph o to graphed in divid u als. For exam ple, one pho -
but received no feed back. The depend ent mea sures en- to graph por trayed two men workin g side-by-side in a
com passed attitude change and one additional item store, and par ticipants were asked to assess which per son
assessing impressions of the con vin cingness of the article. was th e man ager and which the employee. Par ticipants
Co hen et al. / WH EN BELIEFS YIELD TO EVIDENCE 1157

marked th eir an swers on a Scan tron sh eet. Wh ile th ey polit ical issue, th at is, cap ital pun ish ment. Par ticipants
worked on th e test, th e exper imenter waited out side were told th at th ey sh ould try to assess th e speaker’s
and, by coin toss, assigned the par ticipant either to the th ough ts an d feelings about the issue. As she osten sibly
affir ma tion or to th e n o-affir ma tion con dition . Th e pre pared the TV and VCR, the exper imen ter explained
exper imen ter return ed to th e lab room wh en th e par tici- th at “to con trol for differ ences in prior knowledge, it was
pant fin ished th e test. n ecessary first to equalize all par ticipants in terms of
th eir kn owledge of cap ital pun ish ment.” Par ticipants
Affirmation con dition. In th is con dition , par ticipants
were then pre sented with th e appro priate coun terattitu -
were told th at cer tain eth ical con sid er ation s legally
dinal scien tific report regard in g th e death pen alty (as
obliged the research ers to sh are th e results of an y per -
described in Study 1) .
son ality test ing. Th e exper imenter th en left th e room to
grade the par ticipan t’s exam an d asked th at th e par tici- Dependent measures. Par ticipants were next given a ques-
pant com plete a demo graph ics question naire in the tionnaire that assessed cap ital pun ish ment attitudes.
m ean tim e. A few m in u tes later, th e exp er im en ter Again, the exper imen ter emph a sized the con fiden tiality
returned. Appearing slightly disap pointed, she asked, an d an o nym ity of th eir respon ses to th e question naire,
“Have you taken this test before?” When the par ticipant wh ich , wh en com pleted, was to be sealed in an enve lope.
said that he or she had not, the exper imen ter expressed Th e first question naire item asked par ticipants to indi-
relief and then explained, “Wow, I have to say you’ve cate h ow con vincing they found the article on a scale
don e extraor dinarily well; it’s almost as if you had the ranging from 1 (not at all con vincing) to 7 (extremely con -
answer sheet in front of you while taking the test.” She vincing) . The secon d item asked par ticipants to indicate
returned the par ticipan t’s answer sheet, where 22 of the their cur ren t attitude toward cap ital pun ish men t on a
25 the items had been marked cor rect. Th e exper i- scale from 1 ( very much opposed) to 7 ( very much in favor) .
menter showed the par ticipan t where his or her score Finally, as a manip u la tion check, par ticipan ts ( with the
fell in a dot-plot distribu tion of the scores of pre vious excep tion of an initial wave of 11 stu den ts) were asked by
test-takers, makin g it clear th at th e par ticipan t’s score th e exper imen ter to rate their social per cep tive ness on a
fell in the top 5%. The exper imenter th en said, “Because scale rangin g from 1 ( very bad) to 7 ( very good).
of th e quality of your scores, I h ave a few spe cial ques- After ward, par ticipants were debriefed ( in th e same
tions I’d like to ask you if that’s okay” and pulled out a man n er described in Study 1) , th an ked for th eir par tici-
sheet marked “Follow-Up Questions for High Scorers,” pa tion , an d pro vided with th e rele van t signa ture to
marking the sheet with the par ticipant’s iden tifica tion obtain course credit.
num ber an d per cen tile score. Read in g aloud from this
sheet, the exper imen ter asked par ticipants to describe RESULTS AND DISCUSSIO N
the expe rien ces and skills that they believed con trib uted
Manipu lation Check
to th eir social per cep tive ness skills an d took n otes on
th eir respon ses. Th is in ter view served to but tress the Affirmed par ticipants reported having stron ger social
plau sibility an d power of th e feed back man ip u la tion by per cep tive ness skills ( M = 5.49) than did unaffirmed par -
h avin g par ticipants gen er ate causal expla n a tion s for ticipants ( M = 4.43) , F( 1, 61) = 14.13, p < .001.
th eir per for man ce ( Ross, Lepper, & Hub bard, 1975) .
Effects Along Attitude
No-affirmation con dition. In th is con dition, par ticipants Change In dex
also were told th at cer tain eth ical con sid er ation s obliged
Results con firmed the hypoth e sis that the self-affir -
th e research ers to sh are th e results of per son ality test in g.
ma tion would make par ticipants more willin g to give up
But here, the exper imen ter explained that the par tici-
their attitude in th e face of discon firmin g eviden ce. We
pant would be shown his or her score at the end of the
used th e same pro ce dure reported in Study 1 to com -
exper imen t. The exper imen ter then left the room while
pute an in dex of attitude change. Affirmed par ticipants
the par ticipant filled out the demo graph ic question -
ch an ged th eir attitude significantly more in the direc-
naire noted ear lier. Par ticipan ts in this con dition were
tion of th e coun terattitudin al scien tific report (M = 1.93)
n ot pro vided with feed back regard in g th eir per for -
th an did un affirmed par ticipants (M = 1.25) , F( 1, 72) =
mance on the test. After com plet ing th e demo graph ics
4.12, p < .05. No other effects were significant.
question naire, they sim ply pro ceeded to the next phase
An illustra tive way to describe the data involves assess-
of the study.
ing the per cen t age of par ticipants who displayed sub -
Provision of scien tific report. After th e exper imen tal stan tial attitude change within each con dition . Sub stan -
manip u lation, par ticipants were told that they would now tial attitude ch an ge was defin ed as a sh ift of two or more
watch the vid eo taped pre sen tation and were in formed points in the direction of the report. (In addition to bein g
th at th e pre sen ter would speak about a pop u lar social- th e median ch an ge score, two poin ts reflects mean ingful
1158 PERSO N ALITY AND SO CIAL PSYCH O LO GY BULLETIN

attitude change—enough to move an extreme par tisan involved in a con flict see their side’s argu ments and con -
to n eu trality.) A minor ity of unaffirmed par ticipants cern s as more legit imate than those of the other side.
( 39.5%) showed sub stan tial attitude ch an ge, wh ereas a The net result of this biased evalu a tion is that attitudes
major ity of affirmed par ticipan ts ( 61.9%) did so. A per sist an d may even be stren gth en ed ( Lord et al.,
chi-square con tin gency table, com par in g th e observed 1979) .
with the expected count of par ticipan ts showin g sub stan - Th e stu d y r e p o r te d h e r e e xam in e d p ar ti san s’
tial ver sus unsub stan tial attitude change in the two con - responses to a polit ical debate and thus explored two
ditions, par alleled the results of ANOVA, indicat ing th at par ticu lar con se quen ces of such biased assim ila tion.
th e h ypoth e sis of equal attitude change in both con di- Th e fir st co n se q u e n ce is “b iased so u r ce p er ce p -
tions could be rejected, χ2(1, N = 80) = 4.02, p < .05. As in tion ”—th e ten dency to rate the debater rep re sent ing
Study 1, it was also foun d th at pro po nents were more on e’s own side more positively than the debater rep re -
likely to change their attitude in the direction of the sen t in g th e oth er sid e. After extract in g su ch atti-
report ( 64.1% did so) th an were oppo nents ( 39.0% did tude-con firm ing eviden ce, par ticipants also are apt to
so) , χ2(1, N = 80) = 5.03, p < .05. show “attitude polar iza tion”—the ten den cy to embrace
on e’s views with even greater con viction following expo -
Effects Along Ratings of
sure to mixed eviden ce or argu men ts ( Lord et al., 1979) .
Article’s Con vin cingness
In vestiga tors h ave lon g debated wh eth er th ese two biases
A main effect for exper imen tal con dition alon g th is have motiva tion al or purely cogn itive origins, an d recen t
mea sure again con firmed our pre dictions, although the research suggests that they have at least some motiva -
results proved some wh at weaker than those involving tion al basis ( Ditto & Lopez, 1992; Edwards & Smith ,
attitude change. Affirmed par ticipan ts rated the article 1996; Mun ro & Ditto, 1997) . Th e study to be reported
som e wh at m or e con vin c in g ( M = 4.70) th an d id here illustrates on e implica tion of the role of motiva -
unaffirmed par ticipants (M = 4.13) , F( 1, 72) = 3.79, p = tion al pressures in mediat in g these biases. If biased
.055. In ter est ingly, a mar ginal inter action with par tici- source per cep tion and attitude polar iza tion reflect a
pant gen der qualified th is con dition effect, suggest ing motiva tion to pro tect an esteem-bolster ing belief, then
th at, alon g th is mea sure at least, the affir ma tion effect th ese two b iases sh o u ld b e atten u ated by a self-
was con fined to women, F( 1, 72) = 3.39, p = .07. Whereas affir ma tion .
men reported bein g con vin ced by the article regard less Study 3 also began th e effort to disen tan gle th e effects
of whether they were affirmed ( M = 4.47) or n ot affirmed of th e affir ma tion that result from enhanced mood and
( M = 4.44), women showed the expected pat tern of bein g those that arise from height en ed self-regard. We asked
more con vin ced when affirmed ( M = 4.93) th an wh en par ticipan ts sim ply to indicate their cur rent mood and
un affirmed ( M = 3.82) . Alth ough th is Gen der × Con di- self-regard along two sin gle-item response mea sures. If
tion inter action is inter est in g, an inter pre ta tion would our th e o ret ical an alysis is accu rate, th en self-regard
be at best specu la tive in ligh t of its absen ce on th e pri- sh ould cor re late with our depen d en t mea sures but
mary mea sure of attitude ch an ge ( an d on eith er mea - mood sh ould n ot.
sure in Study 1) .
Wh at is clear in Study 2, h owever, is th e pre dicted Overview
main effect of th e affir ma tion on attitude ch an ge.
Prolife an d proch oice par tisan s par ticipated in a
Affirmed par ticipants proved significantly more likely
study osten sibly related to com mu nica tion and impres-
than unaffirmed par ticipants to ch an ge th eir attitudes in
sion for ma tion . Th ey were pre sen ted with a debate
th e face of th e cou n terattitu d in al rep ort. Su p p le -
between two opposing advo cates of th e abor tion issue.
menting the eviden ce pro vided by Study 1, Study 2 dem -
Before com plet in g a depend en t mea sure question naire,
o n st r a t e d a t t i t u d e c h a n g e a lo n g a n e n d u r in g
h alf of th e par ticipants were ran domly assign ed to an
social-polit ical attitude.
affir ma tion con dition where they wrote about a per son -
ally impor tant trait or value ( as in Study 1) ( see Fein &
STUDY 3 Spencer, 1997; Liu & Steele, 1986) . The remain ing par -
Wh ereas Studies 1 an d 2 exam in ed respon ses to ticipants were assign ed to a n o-affir ma tion con dition
disconfirming eviden ce, Study 3 exam ined respon ses to where they wrote about a per son ally unim por tan t trait or
mixed or ambigu ous evidence. As several classic stud ies value. We pre dicted that par ticipants would rate the advo-
attest, peo ple tend to find con fir ma tion of their pre exist - cate rep re sent ing their own side (the attitude-con firm ing
ing beliefs in such ambigu ous in for ma tion ( e.g., H astorf advo cate) more favor ably th an th e advo cate rep re sent -
& Cantril, 1954; Lord et al., 1979) . Scien tists ten d to ing the other side ( the attitude-disconfirming advo cate) .
believe that the stud ies th at con firm their the o ret ical We also pre dicted that the debate would cause them to
position are more valid than those that do not. Par tisan s feel even more con fiden t in their abor tion attitudes.
Co hen et al. / WH EN BELIEFS YIELD TO EVIDENCE 1159

Most impor tant, h owever, we expected that both of these polit ical issue. The second part of the study, par ticipants
ten den cies would be dimin ish ed in th e affir ma tion were told, would exam in e peo ple’s attempts to con vey
con dition. impression s of them selves. Par ticipants would thus be
asked to write a com mu nica tion about a per sonal value
METH O D or ch ar acter istic. In prep a ra tion for that exer cise, par tic-
ipan ts first com pleted a ver sion of Harber’s ( 1995)
Design and Participants Sources of Valida tion Scale, wh ere th ey ran ked a list of
Th is exper iment involved a 2 × 2 facto rial design, with 11 traits an d values in order of th eir per son al impor -
par tisan sh ip of the par ticipan t ( prolife or prochoice) tan ce ( see th e appen dix) . Th e list in cluded var ious
and affir ma tion con dition ( no-affir ma tion or affir ma - qualities but, as in Study 1, it excluded top ics th at migh t
tion ) as th e between -par ticipants factors. The depend - poten tially be asso ciated with th e attitude issue. Wh ile
ent mea sures com prised rat in gs of the two debat ers th e par ticipants ran ked th e list, th e exper imenter left
along several evaluative dimen sions an d par ticipants’ the room.
reports of how the debate had affected their con fidence Abortion debate. Par ticipants n ext read a th ree-page
in their own attitude toward abor tion. tran script of a debate between two opposin g advo cates of
A total of 30 male and 34 female Stan ford un der grad - abor tion righ ts. Par ticipants were given as much time as
u ates par ticipated. They were recruited from an intro - they needed to read the debate thor ough ly. Importan tly,
ductory psychology class or a pool of Stan ford un der - in craft in g th is debate, we reviewed rele vant prochoice
grad u ates who had expressed inter est in par ticipat in g in an d prolife lit er a ture an d in cor po rated into the debate
psych ology stud ies for paymen t. Par ticipan ts from the th e most per sua sive argu ments that we could find in sup -
psych ology class received course credit; all oth er par tici- port of each side of the issue. For exam ple, in one sec -
pants were paid $5. tion , th e debate pre sented the following exchange:
As in Studies 1 an d 2, stu den ts were selected for par tic-
ipa tion based on th eir respon ses to a preselection ques- Eric ( prochoice ad vo cate) : For me, on e of th e fun da men tal
issues is that only the woman should have con trol over
tion naire admin istered ear lier in th e quar ter. O n e ques-
her body. . . . Anti-abor tion laws un fairly legislate what a
tion naire item asked stu dents to in dicate their attitude woman can or can’t do with her body. Th ey’re essen tially
with regard to abor tion righ ts on a scale ran ging from 1 woman-con trol laws. . . . Sh ould th e state be allowed to
( extremely prolife) to 9 ( extremely prochoice) . A second item exer cise th at kin d of power over a per son’s most private,
asked stu den ts to indicate how per son ally impor tant the in timate affairs? . . . I think that un der an ti-abor tion laws,
women don’t re ally have the full hu man righ ts . . . guar -
abor tion issue was to th em on a scale from 1 (not at all
an teed th em un der th e Bill of Righ ts.
important) to 9 (extremely important) . This lat ter item was Mike (prolife ad vo cate): It’s not about choice or con trol. It’s
included because research suggests that attitude impor - about life. I un der stand the im por tance of the privacy and
tan ce mod er ates, in part, motivated biases in per sua sion sanctity of a per son’s body. But . . . how can you justify kill-
( e.g., Edwards & Smith , 1996; Zuwerin k & Devin e, 1996) ing an un born child to vin dicate a woman’s “privacy” and
“free dom of choice”? . . . We’re talking about two bod ies,
an d th is mea sure of attitude impor tan ce was used both
two sep arate lives. Should n’t the state assume pro tection
in our selection crite ria an d in our later an alyses, where of the un born baby—as it assumes the re spon sibility of
it proved a significant covariate. Stu den ts qualified for pro tect ing children from, say, abu sive par ents?
par ticipa tion if th ey rated th em selves extremely prolife
( a 1 or 2 on th e rele vant scale) or extremely prochoice Self-affirmation manipu lation. After read in g the tran -
( an 8 or 9 on the rele vant scale) and if they gave an atti- script, par ticipan ts were asked to write about one of the
tude impor tance rat ing at or above th e median score of per sonal ch ar acter istics or values th ey h ad ran ked ear -
4. A total of 38 prochoice and 26 prolife stu dents ulti- lier. Th is task con stituted the exper imen tal man ip u la -
mately par ticipated in th e study ( th ere were fewer tion and was sim ilar to the pro ce dure used in Study 1 an d
prolife par tisans in th e available par ticipant pool) . in past research ( e.g., Fein & Spen cer, 1997; Liu & Steele,
1986; Steele, 1988) . Par ticipan ts were ran domly assigned
Procedure
to one of two con ditions. In the affir mation con dition ,
Stu dents again par ticipated in the study individ u ally. par ticipan ts wrote about wh y th eir first-ran ked value or
On arrival, th ey were welcomed by a male exper imen ter char acter istic was impor tant to th em an d described a
who pre sen ted the study as a two-part investiga tion of time in their lives when it had proved mean in gful. In the
impression for ma tion an d com mu nica tion. As part of no-affir ma tion con dition, par ticipan ts wrote about why
th is cover story, it was explained that the first part of the their ninth most impor tan t value or ch ar acter istic might
study would exam ine peo ple’s impression of other peo - be impor tan t to th e typ ical Stan ford stu dent. All par tici-
ple and par ticipants would th us be asked to read a com - pan ts were in structed to write as much or as lit tle as they
mu n ica tion in volving two stu den ts debat in g a social- wanted.
1160 PERSO N ALITY AND SO CIAL PSYCH O LO GY BULLETIN

Dependent measures. Par ticipants then com pleted the cate. These differ en ce scores are con cep tu ally sim ilar
depen d en t measure question naire. To assess th e degree and, after being reverse-coded where appro priate, form
to wh ich par ticipants became more con fident in th eir an index of the extent to which par ticipan ts rated the
attitudes, th ey were asked to in dicate h ow th e debate h ad attitude-con firm in g advo cate more favor ably than the
affected th eir con fidence in th eir views con cern ing abor - attitude-disconfirming advo cate ( Cron bach ’s alph a =
tion on a scale from 1 ( made me much less con fident in my .73) . Th ey were aver aged in to a sin gle com posite ( after
views) to 5 ( did not affect my views at all) to 9 (made me much th e differ en ce scores were stan dard ized to equate th eir
more con fident in my views) —a self-report mea sure of atti- varian ce) , with h igh er n u m bers in d icat in g greater
tude polar iza tion sim ilar to that used in past research favorability toward the attitude- con firm ing advo cate
( e.g., Lord et al., 1979) . ( As Lord et al. [ 1979] note, this than the attitude-discon firmin g one.
mea sure is appro priate because par ticipan ts had been While our analyses use stan dard ized scores, we report
selected on th e basis of h aving extreme attitudes. Thus, com posite rat in gs based on aver aging unstan dard ized
th ey would h ave lit tle room to polar ize in their views fur - differ en ce scores. Th is way, the zero point reflects the
th er on a scale sim ilar to that used in preselection.) To absen ce of biased source per cep tion. ( Sta tistical sign ifi-
assess the extent to which par ticipants were biased in can ce is un affected by whether a stan dard iza tion pro ce -
favor of the attitude-con firm in g advo cate rela tive to the dure is used.) Th e differ ence score com posite had a nar -
attitude-disconfirming advo cate, they were asked to rate row spread ( the interquartile range was 0.67 to 2.38) ,
each of th e advo cates alon g several dimen sion s. Spe - reflect ing par ticipants’ ten dan cy to rate both advo cates
cifically, on sep a rate, appro priately labeled 9-poin t favorably.
scales, par ticipan ts rated h ow rea son able, how polit ically
extrem e, h ow close-m in d ed , h ow in tel li gen t, h ow Effects Along Biased Source
biased, an d h ow in formed th ey th ough t each of th e advo - Perception Composite
cates was.
The results indicated that the affir ma tion atten u ated
Afte r co m p let in g th e se m e a su r e s, p ar tic i p an ts biased source per cep tion, th at is, th e ten dency to rate
answered two question naire items design ed to assess th e attitude-con firm ing advo cate more favor ably th an
their cur ren t mood and state of self-regard. Following a the attitude-disconfirming on e. Th e differ en ce between
question n aire prompt request in g that they, “Take a par ticipan ts’ rat in g of the attitude-con firm ing advo cate
momen t to th in k about h ow you are feelin g,” they were an d th eir rat in g of th e attitude-disconfirming advo cate
asked, “How would you describe your mood right now?” proved smaller in th e affir ma tion con dition (M = 1.00)
on a scale from 1 ( extremely bad) to 5 ( neu tral) to 9 (extremely than in the no-affir mation con dition (M = 1.74), F(1, 55) =
good). Next, they were asked to rate their cur rent self- 4.85, p < .035.
regard by answer ing the question, “In gen eral, how do you Two other less impor tan t find in gs emerged. Among
feel about your self ?” on a scale ranging from 1 (extremely
prochoice par tisan s, men proved more likely than did
negatively) to 5 ( neu tral) to 9 ( extremely positively) .
women to en gage in biased source per cep tion, whereas
At the con clu sion of th e exper imen t, par ticipan ts
the reverse was true among prolife par tisan s, as reflected
were fully debriefed ( with a pro ce dure sim ilar to the one
by a mar ginal Par tisan ship × Gen der in ter action , F( 1,
used in Studies 1 and 2) , thanked for their par ticipa tion,
55) = 3.66, p = .06. In addition, prochoice par ticipan ts
an d eith er paid $5 or pro vided with the rele van t signa -
showed less biased source per cep tion when affirmed ( M =
ture to obtain course credit.
2.35) th an wh en un affirmed ( M = 1.03) , whereas the cor -
re spon d in g affir ma tion effect amon g prolife par tici-
RESULTS AND DISCUSSIO N pants was, at least along this mea sure, weaker (Ms = 1.13,
Analyses of th e primary depend ent variables were 0.97, respectively) , as indicated by a mar ginal Par tisan -
con d u cted u sin g a two-way an alysis of covarian ce sh ip × Affir ma tion in ter action , F( 1, 55) = 2.89, p < .10.
( ANCOVA) with par ticipant par tisan ship ( proch oice or
Does the Affirmation Reduce Bias in
prolife) an d exper imen tal con dition ( affir ma tion or n o
Favor of the Attitude-Con firming Advocate
affir ma tion) as in de pend ent variables and rat in gs of per -
or Bias Against the Attitude-Disconfirming One?
sonal impor tan ce of the abor tion issue ( as mea sured in
preselection) as the covariate. Sur prisingly, the affir ma tion reduced bias in favor of
the attitude-con firm ing advo cate. Affirmed par ticipants
Creation of Biased Source
rated that advo cate less positively ( M = 5.76) than did
Perception Composite
un affirmed par ticipants (M = 6.17) , F( 1, 55) = 4.04, p <
A differ ence score was com puted by sub tract ing, along .05. By con trast, th e affir ma tion h ad lit tle reliable effect
each dimen sion, ratings of the attitude-disconfirming on rat in gs of th e attitu d e-d iscon firm in g ad vo cate,
advo cate from rat ings of the attitude-con firm in g advo - although the pat tern is such that ( as would be expected)
Co hen et al. / WH EN BELIEFS YIELD TO EVIDENCE 1161

affirmed par ticipants rated the attitude-discon firmin g GEN ERAL DISCUSSION
advo cate more positively (M = 4.76) than did unaffirmed
Beliefs can con stitute impor tan t sources of iden tity.
par ticipan ts ( M = 4.43) , F( 1, 55) = 1.56, p = .22.
Th is n otion h elps to explain wh y peo ple resist evidence
Effects Along Attitude th at ch allenges th e valid ity of strongly held beliefs, as
Polarization In dex th ey did in Studies 1 an d 2, an d wh y th ey in ter pret
ambigu ous infor ma tion in a man ner th at rein forces pre -
The results also indicated that affir ma tion atten u ated
exist in g attitudes, as they did in Study 3. In each study,
attitude polar iza tion. Con sisten t with Lord et al. ( 1979) ,
pressures to main tain a valued self-image impeded a bal-
par ticipan ts over all reported that the debate made them
an ced con sid er ation of the evidence ( see Ellsworth &
more con fiden t in their views on abor tion, as in dicated
Ross, 1983) . However, an affir ma tion of an alter na tive
by a com par ison of the mean con fidence rat ing (M =
source of iden tity both atten u ated resistan ce to per sua -
5.94) with the point of neu trality on the scale, t( 50) =
sion an d pro duced a more even -h an ded evalu a tion of
5.24, p < .001. However, this height en ed con fidence was
evidence. Shoring up global self-worth, it seems, takes
sign ifican tly lower in th e affir ma tion con dition ( M =
th e stin g out n ew ideas, makin g them less pain ful to
5.53) th an in th e n o-affir ma tion con dition ( M = 6.35) ,
accept as true. We have dem on strated th e effective ness
F( 1, 50) = 5.49, p < .025.
of self-affir ma tions in ame lio rat ing defen sive reaction s
Disen tan gling Effects of to per sua sion in two social-polit ical domains and with
Mood and Self-Regard two man ip u la tions of self-affir ma tion . Taken togeth er,
th e results illustrate th e valid ity and gen er ality of our
We tried to assess th e exten t to wh ich th ese effects con cep tual frame work.
resulted from ele vated mood an d en h an ced self-regard.
O n e obviou s question is wh y th e affir ma tion in
Recall th at par ticipants h ad in dicated both their cur ren t
Studies 1 an d 2 led to more positive evalu a tions of the
mood and the level of self-regard that they felt at the
attitude-disconfirming infor ma tion, wh ereas in Study 3
presen t moment. As might be expected, these two items it led to more n ega tive evalu a tions of th e attitude-con -
were sign ifican tly cor re lated, r( 62) = .37, p < .005. Con sis-
firm in g infor ma tion. It seems that the affir ma tion atten -
tent with the find ings of Liu and Steele ( 1986) and Fein
u ated a disconfirmation bias in the first two stud ies but
an d Spen cer ( 1997) , th e man ip u la tion of self-affir ma -
ame lio rated a con fir ma tion bias in the third. One possi-
tion h ad n o sig n ifican t effect on mood ( p > .20) .
ble rea son for this pat tern sim ply involves the availability
Affirmed par ticipants did, h owever, report some what
of counterattitudinal and proattitudinal evidence in the
h igh er feelin gs of self-regard ( M = 7.15) th an did
th ree stud ies. In Studies 1 and 2, par ticipants could pro -
unaffirmed par ticipants (M = 6.79) , but th e distribu tion
tect th eir attitude only by den igrat in g th e coun terattitu -
of th is mea sure was severely skewed to th e left, posing
dinal eviden ce. In Study 3, however, par ticipants had the
inher ent sta tistical difficulties for para met ric tests. Con -
addition al option of exalt ing th e merit of th e proattitu -
se quently, a Kruskal-Wallis n on parametric test was per -
dinal in for ma tion. We can only specu late as to why they
formed. It yielded a mar ginal effect of con dition ( H =
ch ose a con fir ma tion bias rather than a disconfirmation
2.91, p = .088) . ( Th e same n on parametric an alysis
bias in th at study, but we suspect the rea son involves the
yielded no such effect along the mood mea sure.)
per sua sive nature of the evidence they read ( cf. Edwards
If th e reduction in biased assim ila tion and attitude
& Smith, 1996) . Praising a per sua sive ally may be less
polar iza tion in the affir ma tion con dition arose from ele -
effortful th an crit icizin g a per sua sive adver sary.
vated mood, th en self-reported mood sh ould cor re late
with that depend en t mea sure; it did n ot, r( 62) = –.12, ns. Con siderations of
More over, the cor re la tion between mood and rat in gs of Un derlying Process
attitude con fidence was slightly positive, r( 57) = .14, ns,
oppo site of what a mood-based expla n a tion would pre - Par ticipan ts felt less th reat en ed by eviden ce th at
dict. Con sistent with our the o ret ical per spective, h ow- impugn ed th eir attitudes, it seems, wh en th ey received
ever, higher rat in gs of self-regard proved significantly an affir ma tion of an alter na tive source of self-worth. As a
cor re lated with less biased source per cep tion , r( 62) = result, th ey en gaged in fewer defen sive man eu vers
–.31, p < .02. Higher self-regard also tended to be cor re - aimed at pro tect ing an iden tity at the expense of a lost
lated ( albeit n on sign ifican tly) with lower rat in gs of con - oppor tu n ity to learn. Future research will exam ine the
fidence, r( 57) = –.15, ns. Self-regard th us cor re lated with spe cific mech a nisms by wh ich self-affir ma tion attenu -
th e depen d ent mea sures more system at ically than did ates bias. It is possible, for exam ple, th at th e self-affir ma-
mood, pro vid ing fur th er evidence that mood does not tion reduced resistance to per sua sion by trivializing the
pro vide a sufficien t expla n a tion for th e affir ma tion impor tance of the attitude as a source of iden tity or self-
effects. Higher feelings of self-regard—not better mood— worth ( see Simon, Greenberg, & Brehm, 1995) . Indeed,
pre dicted less bias. per son ally unim por tant attitudes have been shown to be
1162 PERSO N ALITY AND SO CIAL PSYCH O LO GY BULLETIN

less resistan t to ch an ge th an per son ally impor tan t on es on -lin e an d m em ory-based p ro cessin g d eter m in ed
( Zu we r in k & De vin e , 1996; se e also Po m e r an tz, responses to the infor ma tion. ( In deed, we would sub mit
Chaiken, & Tordesillas, 1995) . It is also possible that the th at much attitude change in the real word occurs not
affir ma tion sim ply made par ticipants less extreme an d dur in g expo sure to disconfirming evidence but in later
con fident in th eir attitudes. moments of calm reflection.) Th e affir ma tion may thus
Two alter n a tive expla n a tion s for th e present find in gs reduce on -lin e defen sive pro cessin g, at th e time of
war ran t discussion. Per h aps the affir ma tion pro duced encod ing, as it pre sum ably did in Studies 1 and 2, or it
the effects it did by inducing self-focus. But past research may atten u ate mem ory-based defen sive pro cessing, as it
suggests that self-focus does not increase persuasibility p re su m ably d id in Stu d y 3. Past research p ro vid es
but decreases it ( Hutton & Baumeister, 1992) . A more su p p ort for th is rea son in g by d em on strat in g th at self-
plau sible alter na tive expla na tion involves the possibility affir ma tions may both buffer against a future threat
th at th e affir ma tion induced positive mood and thus (Steele, 1988) and dispel the effects of a past one (Tesser &
caused par ticipants to evalu ate th e evidence less crit i- Cor nell, 1991) .
cally or system at i cally ( see McGu ir e, 1985; Petty, An addition al question that future research could
Sch uman n , Rich man , & Strath man , 1993) . Four argu - exam ine involves dispositional self-esteem as a poten tial
men ts, h owever, cast doubt on this possibility. First, past mod er a tor of self-affir ma tion effects. Peo ple low in
research suggests th at positive mood does n ot reduce sys- self-esteem, it could be argued, would ben e fit most from
tem atic pro cessin g when peo ple have ample time, as a self-affir ma tion pro ce dure because after read in g a
th ey did in our stud ies, to read an d respon d to th e rele - th reat en in g message, they have fewer favor able self-con -
vant mate rials ( Mackie & Worth, 1989) . Secon d, pre vi- cepts with which to affirm and thus restore self-worth on
ous stud ies suggest that the effects of positive mood on their own ( Steele et al., 1993; see also Greenberg et al.,
per sua sion are dimin ish ed wh en ( as in our stud ies) peo - 1993) . It is also possible, however, th at peo ple with low
ple pre sum ably h ave stron g prior opin ions or exten sive self-esteem would ben e fit less from an affir ma tion pro ce -
kn owledge about th e attitude topic ( Bless, Boh n er, dure than peo ple with high self-esteem. Low-self-esteem
Sch warz, & Strack, 1990; Worth & Mackie, 1987; see also par ticipan ts might find affirm in g feed back ( of the sort
Bless, Sch warz, & Mackie, 1992; Mackie & Worth, 1989; used in Study 2) less plau sible than their high-self-esteem
see also Petty et al., 1993) . Third, con sisten t with find - peers, or th ey migh t h ave greater difficulty remem ber -
ings obtained in ear lier stud ies, th e affir ma tion pro ce - ing self-affirm in g expe rien ces ( as the pro ce dures in
dure used in the first and third stud ies was foun d, in Studies 1 an d 3 required) . Ultimately, of course, the role
Study 3, to h ave n o effect on self-reported mood ( Fein & of disposition al self-esteem in mod er at ing the effect of
Sp en cer, 1997; Liu & Steele, 1986) . Fin ally, m ost self-affir ma tion is an empir ical question.
mood-based accounts, at least in their most straight for -
Theoretical and
ward form, could n ot par simo niously explain why the
Practical Implications
affir ma tion used in Study 3 made par ticipan ts more neg-
a tive in th eir evalu a tion of th e attitude-con firm in g O ur th e o ret ical per spec tive h as implica tion s for
advo cate. nego tia tion, edu ca tion , and ther a peu tic in ter ven tions.
Although mood is prob a bly in sufficient to yield the Wh en alter n a tive sources of iden tity are affirmed, nego -
effects obtain ed in th e th ree stud ies, it may n ever the less tia tors may more clearly see the mer its of the other side’s
prove necessary. Affir ma tion effects may require both a argu men ts and more readily con cede their own biases
self-per cep tion of per sonal worth and an ele vated mood ( Bastardi & Ross, 2000) . Stu den ts may prove more crit i-
state. In this sense, positive mood con stitutes less an cal of their long-held views and more open to infor ma-
alter n a tive expla n a tion of affir ma tion effects th an on e of tion that challenges th eir pre con cep tions. Clien ts in
several possible media tors of th em. ther apy may better recognize and change erro n e ous
An oth er issu e rel e van t to u n d er lyin g p r o cesses beliefs th at cause th em psycho logical distress.
regards the place ment of th e self-affir ma tion . In Studies Our results also add to a growin g lit er a ture on th e
1 an d 2, par ticipants were affirmed prior to the pre sen ta - role of self-image main te n an ce motiva tion s in mediat -
tion of th e scien tific report, wh ereas in Study 3, par tici- in g a wide ran ge of social psych o logical ph e n om ena
pan ts were affirmed after th e pre sen ta tion of the debate. ( J. Aronson, Cohen, & Nail, 1999) . Not only does this
If the affir ma tion oper ates through its effect on infor ma - motiva tion h elp to explain th e present find in gs but it has
tion pro cessing, th en it seems n ecessary to affirm par tici- also been implicated in cognitive disso n an ce pro cesses
pants prior to expo sure to the rele vant eviden ce. Th is ( Steele, 1988; Steele et al., 1993) , ter ror man age ment
argu ment would be rea son able if all infor ma tion pro - phe nom ena (Greenberg et al., 1993), low self-esteem an d
cessin g took place while par ticipants read th is evidence depression ( Brown & Smart, 1991; Lin ville, 1987) , preju -
an d n on e occurred after ward. But it is likely that both dice and ste reo typ ing ( Fein & Spencer, 1997; Green berg
Co hen et al. / WH EN BELIEFS YIELD TO EVIDENCE 1163

et al., 1993) , attributional analysis ( Liu & Steele, 1986) , REFERENCES


biases in social judgment ( Dunning & Cohen, 1992; Abelson, R. P. ( 1986) . Beliefs are like possession s. Journal for the Theory
Dunning et al., 1995) , behavioral in h ibition (Vohs & of Social Behav ior, 16, 223-250.
Heath er ton , 2000) , decision making ( Joseph s, Larrick, Arch er, D. ( 1980) . How to expand your S.I.Q. (social intelligence quotient).
Steele, & Nisbett, 1992) , an d man y oth er rich ph e n om - New York: M. Evans.
Aron son , E., Turner, J., & Carlsmith , J. M. ( 1963) . Com mu nica tion
ena. Our find in gs suggest th at per sisten t biases in social cred ibility an d com mu n ica tion discrep ancy as deter minants of
judgment arise from iden tity-main te nance motiva tions. opin ion change. Journal of Abnormal and Social Psychology, 63, 31-36.
Con sisten t with Kunda ( 1990) and Dunning and his col- Aronson, J., Cohen, G. L., & Nail, P. ( 1999) . Self-affir ma tion the ory: An
update and appraisal. In E. Harmon-Jones & J. Mills (Eds.), Cognitive
leagues ( Dunning et al., 1995) , we argue that such moti- dissonance theory: Progress on a pivotal theory in social psychology ( pp. 127-
va tion s pressure cognitive pro cesses to a desired en d. 147). Wash in gton , DC: Amer ican Psycho logical Asso cia tion.
Peo ple search for an inter pre ta tion of the evidence th at Bastardi, N., & Ross, L. ( 2000) . Main taining self-integrity: Effects of threat
to personal iden tity on evidence evalu ation and con cession making. Man u -
best sup ports the con clu sion th ey h ope to draw, much as
script sub mit ted for pub lica tion.
a lawyer spin s court room evidence to present the stron - Bless, H., Boh n er, G., Sch warz, N., & Strack, F. ( 1990) . Mood and per -
gest case. Relieving these pressures fosters a more ratio - sua sion: A cognitive respon se an alysis. Person ality and Social Psychol-
nal and even-handed evalu a tion of evidence. ogy Bulletin, 99, 229-246.
Bless, H., Schwarz, N., & Mackie, D. M. ( 1992) . Mood effects on atti-
Our find in gs also address an older ten sion in West ern tude judgmen ts: In de pen d en t effects of mood before an d after
art an d ph iloso phy con cern ing th e rela tion ship between message elab o ra tion. Journal of Person ality and Social Psychology, 63,
emo tion an d rea son . One artistic and philo soph ical tra - 585-595.
Brown , J. D., & Smart, S. A. ( 1991) . Th e self an d social con duct:
dition ascribes h uman folly to th e misch ie vous dance of Linking self-rep re sen ta tions to prosocial behavior. Journal of Person -
th e passions and sees emo tion as a con tam inant of rea - ality and Social Psychology, 60, 368-375.
son. The other roman tic tra dition cele brates the role of Coh en , A. R. ( 1959) . Some implica tions of self-esteem for social influ -
emo tion in imagin a tion an d rela tion ships. But both of en ce. In I. L. Jan is ( Ed.) , Person ality and persuasibility ( pp. 102-120) .
New Haven, CT: Yale Univer sity Press.
these tra dition s view rea son and emo tion as antago n is- Coo per, J., & Mackie, D. ( 1983) . Cognitive disso nance in an in ter group
tic. Rea son is but a car riage bein g pulled by th e wild con text. Journal of Person ality and Social Psychology, 44, 536-544.
horses of the passion s, or th e passions must be curbed by Darley, J. M., & Gross, P. H. ( 1983) . A hypoth e sis-con firm in g bias in
labeling effects. Journal of Person ality and Social Psychology, 44, 20-33.
a disciplin ed applica tion of rea son. In a sense, however, Ditto, P. H., & Lopez, D. F. ( 1992) . Motivated skep ticism: Use of differ -
our research sh ows th at th e two sides of h uman n ature— en tial decision crite ria for pre ferred and nonprefered con clu sions.
the emo tion al side an d th e ratio n al side—are in ter - Journal of Person ality and Social Psychology, 63, 568-584.
Dun n in g, D., & Coh en , G. L. ( 1992) . Ego cen tric definitions of traits
twined ( Palfai & Salovey, 1994) . When peo ple are in a and abilities in social judgment. Journal of Person ality and Social Psy-
good emo tional state, th ey are more ratio n al. chology, 63, 341-355.
Dunning, D., Leuenberger, A., & Sh erman , D. A. ( 1995) . A n ew look at
motivated infer en ce: Are self-serving th e o ries of success a prod uct
of motiva tional forces? Journal of Person ality and Social Psychology, 69,
APPEN DIX 58-68.
Sources of Validation Scale Edwards, K., & Smith, E. E. ( 1996) . A disconfirmation bias in the evalu -
a tion of argu ments. Journal of Person ality and Social Psychology, 71,
5-24.
RANKING OF PERSO NAL Ellsworth , P. C., & Ross, L. ( 1983) . Pub lic opin ion an d cap ital pun ish -
CHARACTERISTICS AND VALUES men t: A close exam in a tion of th e views of abo lition ists an d
retentionists. Crime and Delin quency, 29, 116-169.
Be low is a list of ch ar acter istics an d values, some of which may Fein , S., & Spen cer, S. J. ( 1997) . Preju dice as self-image main te n an ce:
be im por tant to you, some of wh ich may be un im por tan t. Affirming the self through der o gat ing oth ers. Journal of Person ality
Please rank these values and qualities in or der of their im por - and Social Psychology, 73, 31-44.
tan ce to you, from 1 to 11 ( 1 = most important item, 11 = least im- Festin ger, L. ( 1957) . A theory of cognitive dissonance. Evanston, IL: Row,
Peter son.
portant item) . Use each num ber only once. Festin ger, L., Riecken, H. W., & Schachter, S. ( 1956) . When prophecy
fails. Min ne ap o lis: Univer sity of Min ne sota Press.
 Ar tistic skills/ aesth etic ap pre cia tion Gilovich , T. ( 1991) . How we know what isn’t so: The fallibility of human rea-
 Sen se of h u mor son ing in everyday life. New York: Free Press.
Götz-Marchand, B., Götz, J., & Irle, M. ( 1974) . Prefer ence of disso -
 Re la tion s with frien ds/ fam ily nance reduction modes as a fun ction of th eir order, familiar ity an d
 Spon ta n e ity/ livin g life in the mo ment reversibility. European Journal of Social Psychology, 4, 201-228.
 So cial skills Green berg, J., Pyszczyn ski, T., Solo m on , S., Pin el, E., Sim on , L., &
Jor dan , K. ( 1993) . Effects of self-esteem on vuln er a bility denyin g
 Ath let ics
defen sive distor tions: Fur th er evidence of an anxiety-buffer in g
 Mu sical ability/ ap pre cia tion function of self-esteem. Journal of Experimen tal Social Psychology, 29,
 Ph ysical at tractive n ess 229-251.
 Cre ativity H am ilton , D. L., & Rose, T. L. ( 1980) . Illu sory cor re la tion an d th e
main te n an ce of ste reo typic beliefs. Journal of Person ality and Social
 Busin ess/ man a ge rial skills Psychology, 39, 832-845.
 Ro man tic values Harber, K. ( 1995) . Sources of Validation Scale. Unpub lish ed scale.
Hastorf, A., & Can tril, H. ( 1954) . Th ey saw a game: A case study. Journal
SOURCE: K. Harber ( 1995) . of Abnormal and Social Psychology, 49, 129-134.
1164 PERSO N ALITY AND SO CIAL PSYCH O LO GY BULLETIN

Hutton , D. G., & Baumeister, R. F. ( 1992) . Self-aware n ess and attitude Rhodes, N., & Wood, W. ( 1992) . Self-esteem and intelligence affect
change: Seeing one self on the cen tral route of per sua sion. Person al- influenceability: The mediat ing role of message recep tion . Psycho-
ity and Social Psychology Bulletin, 18, 68-75. logical Bulletin, 111, 156-171.
Jan is, I. L. ( 1954) . Per son ality cor re lates of suscep tibility to per sua sion . Ross, L., & Lepper, M. R. ( 1980) . Th e per se ver an ce of beliefs: Empir i-
Journal of Person ality, 22, 504-518. cal an d n or ma tive con sid er ations. In R. A. Shweder & D. Fiske
Joseph s, R. A., Larrick, R. P., Steele, C. M., & Nisbett, R. E. ( 1992) . Pro - ( Eds.) , New directions for methodology of behav ioral science: Fallible judg-
tectin g th e self from th e n ega tive con se quences of risky decision s. ment in behav ioral research. San Fran cisco: Jossey-Bass.
Journal of Person ality and Social Psychology, 62, 26-37. Ross, L., Lepper, M. R., & Hub bard, M. ( 1975) . Per se ver ance in
Katz, D. ( 1960) . The function al approach to th e study of attitudes. Pub- self-per cep tion and social per cep tion. Journal of Person ality and
lic Opin ion Quarterly, 24, 163-204. Social Psychology, 32, 880-892.
Kunda, Z. ( 1990) . The case for motivated rea son ing. Psychological Bulle- Ross, L., & Ward, A. ( 1995) . Naive realism: Implica tions for misun der -
tin, 108, 480-498. stand ing and diver gent per cep tions of fair ness and bias. In T. Brown ,
Lin ville, P. W. ( 1987) . Self-com plexity as a cognitive buffer again st E. Reed, & E. Turiel (Eds.), Values and knowledge. Hillsdale, NJ: Law-
stress-related illness an d depression. Journal of Person ality and Social ren ce Erlbaum.
Psychology, 52, 663-676. Sears, D. O., & Funk, C. L. ( 1991) . The role of self-inter est in social an d
Liu, T. J., & Steele, C. M. ( 1986) . Attributional analysis as self-affir ma - polit ical attitudes. In M. P. Zanna ( Ed.) , Advances in experimen tal
tion. Journal of Person ality and Social Psychology, 51, 531-540. social psychology ( Vol. 24, pp. 2-91) . San Diego, CA: Aca demic Press.
Lord, C. G., Ross, L., & Lepper, M. R. ( 1979) . Biased assim ila tion and Sh erman , D. A. K., Nelson, L. D., & Steele, C. M. ( 2000) . Do messages
attitude polar iza tion : Th e effects of prior th e o ries on sub se quen tly on h ealth th reaten th e self? In creasin g th e accep tance of th reat en -
con sid ered evidence. Journal of Person ality and Social Psychology, 37, in g h ealth messages via self-affir ma tion. Person ality and Social Psy-
2098-2109. chology Bulletin, 26, 1046-1057.
Mackie, D. M., & Worth , L. T. ( 1989) . Pro cessin g deficits and the medi- Simon, L., Greenberg, J., & Brehm, J. ( 1995) . Trivialization: The for -
a tion of positive affect in per sua sion . Journal of Person ality and Social got ten mode of disso nance reduction . Journal of Person ality and
Psychology, 57, 27-40. Social Psychology, 68, 247-260.
McGuire, W. J. ( 1985) . Attitudes and attitude change. In G. Lindzey & Sn yder, M., Tanke E. D., & Berscheid, E. ( 1977) . Social per cep tion and
E. Aronson ( Eds.) , The handbook of social psychology ( 3rd ed., Vol. 2, in ter per son al beh avior: On the self-fulfillin g nature of social ste -
pp. 233-346) . New York: Ran dom House. reo types. Journal of Person ality and Social Psychology, 35, 656-666.
Mun ro, G. D., & Ditto, P. H. ( 1997) . Biased assim ila tion, attitude polar - Steele, C. M. ( 1988) . Th e psych ology of self-affir ma tion: Sustaining the
iza tion, and affect in reactions to ste reo type-rele vant scien tific in tegrity of the self. In L. Berkowitz ( Ed.) , Advances in experimen tal
in for ma tion . Person ality and Social Psychology Bulletin, 23, 636-653. social psychology ( Vol. 21, pp. 261-302) . New York: Aca demic Press.
Palfai, T. P., & Salovey, P. ( 1994) . Th e in flu ence of depressed an d elated Steele, C. M., Spen cer, S. J., & Lyn ch , M. ( 1993) . Self-image resilience
mood on deductive and inductive rea son ing. Imagination, Cogni- and disso nance: The role of affirmational resources. Journal of Per-
tion, and Person ality, 13, 57-71. son ality and Social Psychology, 64, 885-896.
Petty, R. E., Schumann, D. W., Richman, S. A., & Strathman, A. J. Tesser, A., & Cor nell, D. P. ( 1991) . On th e con flu ence of self pro cesses.
( 1993) . Positive mood and per sua sion : Differ ent roles for affect Journal of Experimen tal Social Psychology, 27, 501-526.
un der h igh - an d low-elab o ra tion con ditions. Journal of Person ality Worth, L. T., & Mackie, D. M. ( 1987) . Cognitive media tion of positive
and Social Psychology, 64, 5-20. affect in per sua sion. Social Cognition, 5, 76-94.
Pilisuk, M. ( 1968) . Depth , cen trality, and toler an ce in cogn itive con sis- Zellner, M. ( 1970) . Self-esteem, recep tion, an d in fluen ceability. Journal
tency. In R. P. Abelson, E. Aronson, W. J. McGuire, T. M. Newcomb, of Person ality and Social Psychology, 15, 87-93.
M. J. Rosen berg, & P. H. Tannenbaum ( Eds.) , Theories of cognitive Zimbardo, P. G. ( 1960) . In volve ment and com mu n ica tion discrep ancy
con sistency: A sourcebook ( pp. 693-699) . Chicago: Ran d McNally. as deter min an ts of opin ion con for mity. Journal of Abnormal and
Pomeran tz, E. M., Ch aiken , S., & Tordesillas, R. S. ( 1995) . Attitude Social Psychology, 60, 86-94.
strength and resistan ce pro cesses. Journal of Person ality and Social Zuwerin k, J., & Devin e, P. G. ( 1996) . Attitude impor tan ce and resis-
Psychology, 69, 408-419. tance to per sua sion: It’s not just the thought that counts. Journal of
Reed, M. B., & Aspinwall, L. G. ( 1998) . Self-affir ma tion reduces biased Person ality and Social Psychology, 70, 931-944.
pro cessin g of h ealth -risk in for ma tion . Motivation and Emotion, 22,
99-131.
Re ceived July 10, 1998
Re vision accepted May 12, 1999

You might also like