On the Nature of Prejudice

MARK P. ZANNA Conversely, if, for some other reason, I were

University of Waterloo to evaluate Afro-Canadians unfavourably, T
might then be motivated to believe that
Afro-Canadians are lazy and stupid.
This conceptualization, especially the
Abstract notion that prejudice is based on stereotypes,
It was proposed l.h;it, in addition (<> is consistent with the way attitudes, in gen-
stereolypic beliefs (beliefs (hat typical eral, have been conceptualized in recent
members of the oulgroup possess certain years by social psychologists. Perhaps the
characteristics or trails), prejudice (or most prominent theory of attitudes has been
negative attitudes toward oulgroups) is based that proposed by Martin Fishbcin arid Icek
on symbolic beliefs (beliefs that typical Aj/.en (Ajzen & Fishbcin, 1980). According
members violate cherished traditions, to Fishbcin and Ajzen, altitudes (which they
customs, and values) as well as on emotions define as our feelings or affective responses
and past experiences that are associated with to an attitude object) are based upon evalu-
the oulgroup. In several studies (the first of ative beliefs. Tn this view, our attitudes
which is presented in the present paper) we toward a group are based upon the evalu-
have found that (1) although related, stereo- ative implications of the characteristics or
types, symbolic beliefs, emotions, and past traits we attribute to the group, which seems
experiences are far from redundant, and (2) to me to be the same as saying that the
each factor is, indeed, an important predictor amount of our prejudice is based on the
of prejudicial altitudes. Thus, there would extent to which we endorse the stereotype of
appear to be more to prejudicial attitudes the group.
than stereotypic beliefs. We have also dis- Recently, several attitude theorists, myself
covered thai the relatively more negative included, have suggested a slightly modified
attitudes held by individuals high in authori- view of the attitude concept. For example, I
tarianism are predicted best by symbolic- (Zamia & Rcmpel, 1988) have proposed that
beliefs. The implications of these results for affective and evaluative responses should no
understanding the development and reduc- longer be considered synonymous (cf.
tion of prejudice are discussed. Abelson, Kinder, Peters, & Fishe, 1982) and
that altitudes should be viewed as evaluations
based on, or developed from, three general
Traditionally, social psychologists have classes of information: (I) affective informa-
believed there is a strong relation between tion or the emotions associated with the atti-
prejudice and stereotypes. Some have pro- tude object, (2) cognitive information or the
posed that prejudicial attitudes are based on, beliefs associated with the attitude object,
or arc the result of, stereotypical beliefs. and (55) information concerning past
Others have suggested that .stereotypical behaviours or behavioural intentions. |T've
beliefs justify, or rationalize, prejudicial atti- further suggested, again along with others
tudes. TFT (along with everybody else in my (e.g., Kagly & Chaiken, 199-5), that evalu-
reference group) believe Afro-Canadians to ations or attitudes influence three modes of
be lazy and stupid, it would make sense for response, including affective, cognitive and
me to evaluate Afro-Canadians unfavourably. behavioural responses.)
When evaluations are based primarily on
G:in;idi;in I'sychulogy/l'sycliolugie r.;in:ulienne, 35:1 (utilitarian or instrumental) beliefs about the
12 Zanna

altitude object, tliis view can be reduced to ing whether the attitude object helped or
something like the formulation proposed by hindered the attainment of important goals
Fishbein and Aj/en. When evaluations are or values.
based primarily on affects produced by or All this suggests (at least, to me) that preju-
associated with the attitude object, this view dicial altitudes are not merely a function of
can resemble the formulation proposed by stereotypes. Although prejudice toward a
Bob Zajouc (Zajonc, 1980). Finally, when group may, indeed, be based upon
evaluations are based on inferences from stereotypic beliefs (that, is, beliefs about the
past behaviour, this view can be similar to characteristics possessed by the group), such
Daryl Bern's theory of sell-perception altitudes may also be based upon afl'ecis or
(Bern, 1972). emotions associated with the group and/or
In any event, if attitudes are not always symbolic beliefs (that is, beliefs about,
entirely based upon the sort of utilitarian or whether the group helps or hinders the
instrumental beliefs suggested by the attainment of important values).
Fishbein and Aj/.eii model, then prejudicial The study 1 will present today was designed
attitudes are perhaps not always based entire- to assess the nature of prejudice in these
ly upon stereotypical beliefs. In fact, several terms and to determine whether the causes
years ago Bob Gardner and his colleagues (at (or, at least, the correlates) of prejudice vary
the University of Western Ontario) demon- across different target groups and for individ-
strated that English Canadian prejudice uals disposed or not to be prejudiced.
toward French Canadians was virtually unre-
lated to their consensual stereotype of The Study
French Canadians (Gardner, 199-1; Gardner, What I'd like to do first, is "walk you
Wonnacott, & Taylor, 1968). More recently, through" the questionnaire we administered
Chuck Siangor and his colleagues (at the to 71 students (in groups of 2 to 6) from an
University of Maryland) demonstrated that introductory psychology class at the Univer-
whites' prejudice toward Afro-Americans is sity of Waterloo.
based on the affects or emotions that they In this initial study our target groups
associated with Afro-Americans, in addition included four outgroups for these partici-
to their stereotypical beliefs (Stangor, pants - French Canadians, Native Indians,
Sullivan, & Ford, 1991). Pakistanis, and Homosexuals - as well as the
And, in recent years David Sears, John ingroup - English Canadians.
McConahay and their colleagues, have sug-
gested that prejudice toward Afro-Americans ASSESSMENT OF PREJUDICE
(in addition, perhaps, to being based to a As can be seen in Table I, to assess preju-
greater extent upon affect) is based upon (in dice, by which I mean a negative (or hostile)
my terms) symbolic beliefs such as "Blacks attitude toward an outgroup, we simply asked
arc getting more than they deserve," "the our participants to evaluate the typical mem-
streets are unsafe today," and so on, rather bers of the outgroups (in this example,
(or to a greater extent) than on traditional typical French Canadians) on a 101-poiiit
stcreotypic beliefs (e.g., McConahay & "evaluation thermometer," which ranged
Hough, 1976; Sears, 1988). from 0, labelled "extremely unfavourable,"
Interestingly, the Sears and McConahay to 100, labelled "extremely favourable."
notion of symbolic racism, with its suggestion 1 should note that attitudes assessed wilh
that prejudice is based on threats to cher- our evaluation thermometer have a high
ished values, is reminiscent of a model of lesi-relesl reliability and correlate highly with
attitudes mat was proposed several years ago a more standard, multiple-item semantic
by Milton Rosenberg (1956), who suggested differential measure (Haddock, Zanna, &
that attitudes were based on beliefs concern- Esses, 1993). 1 should also note that whereas
 On the Nature of Prejudice IS

TABU-: I portion by valence," or, in Fishbcin and

Evaluation Thermometer Aj/en terms, "belief by evaluation," scores,
Please provide :i number between 0° and WOo f<>
and, finally, we divided the sum by the num-
indicate your overall evaluation of: ber ol characteristics listed.
Thus, our stereotype measure assesses the
Typical French Canadians average evaluation of characteristics t h o u g h t
I'ositive 10(K> Kxireiuely favourable to be possessed by typical members of a
g r o u p weighted by the proportion of typical
9<)o Very favourable members believed to possess the characteris-
K()o Quite favourable tic. (1 should note that our results are virtual-
ly identical when stereotype scores are calcu-
70° Fairly favourable lated as either the average or the sum of the
(i()° Slightly favourable valences p e r se.)

50° Neither favourable nor

ASSESSMENT OF AB'HITS
unfavourable
T o assess affect, participants were asked to
W° Slightly unfavourable list the feelings or emotions they experience
30° Fairly unfavourable
when they see, meet, or even think about
typical members of the target g r o u p (see
y()° Quite unfavourable middle panel ol Table 2). As with the stereo-
10° Very unfavourable type measure:, participants were also asked to
indicate (1) the valence of each emotion on
Negative 0° Kxtremely unfavourable a five-point scale ranging from "very nega-
tive" (-2) to "very positive" (+2), and (2) the
subjects always completed the attitude percentage of typical group members that
measure first, the order of the predictor produce or elicit each emotion.
variables (that is, the measures of stereotype, An affect score was calculated in the same
affect, and symbolic belief) were randomized, m a n n e r as the measure of stereotype and
and the order of the target groups were represents the average evaluation of emo-
counterbalanced. tions elicited by typical members of the
group weighted by the proportion of typical
ASSKSSMKNT Ol : STERKOTYPES members believed to elicit the emotion.
T o assess stereotypes, participants were asked
to list the characteristics or, if necessary, the ASSIiSSMKNT Ol- SYMBOLIC BELIU1S
short phrases they would use to describe T o assess symbolic beliefs, we asked o u r
typical members of the group (see top panel participants to list the values, customs, or
of Table 2). Having completed this task, they traditions that they believed are blocked o r
were asked to rate the valence of each char- facilitated by typical g r o u p m e m b e r s (see
acteristic o n a five-point scale ranging from bottom panel o f f able 2). U p o n completion
"very negative" (-2) to "very positive" (+2), of this task, they were asked to rate the
and, finally, to indicate the percentage of extent to which each value is blocked o r
typical g r o u p members who possess each facilitated by typical g r o u p members on a
characteristic. five-point scale, ranging from "almost always
We calculated a stereotype score tor each blocked" (-2) to "almost always facilitated"
group in the following manner: First, we (+2), and, finally, to indicate the percentage
multiplied the valence of each characteristic of typical g r o u p m e m b e r s whom they
by the proportion (percentage/100) of typi- believed block or facilitate each value.
cal group members believed to possess that A symbolic belief" score was calculated in
characteristic; next, we summed these "pro- the same m a n n e r as the previous measures
14 Zanna

TABLE 2
Instructions for Eliciting Stereotypes, Affects, and Symbolic Beliefs

Stcrcotypic Belief-Eliciting Instructions

We arc interested in the characteristics thai people use in describing members ol various groups in Canada.
For each of the following groups, please provide :i clesi riplion ol typical members of the group. Your descrip-
tion should consist of a list of characteristics or, if necessary, short phrases which you would use to describe
typical members of the group (e.g., "they are cheap", "they are intelligent"). Provide as many characteristics or
short phrases as you think are necessary to convey your impression of each group and to describe each group
adequately. Please be luiw.il- Your responses will be kepi strictly confidential.

Affect-Eliciting Instructions
We are interested in examining how members of various groups make you feel, that is the emotions you
experience when yon see, meet, or even think about lypiral members ol that group. For each of the following
groups, please provide a list of the feelings you experience (e.g., proud, angry, disgusted, happy) when you
think about typical members of that group. I'rovide as many feelings or emotions you believe are necessary to
accurately convey your impression of each group and to describe them adequately. I'II'IIII' be hone.il. Your
responses will be kept strictly confidential.

Symbolic Belief-Eliciting Instructions

We arc interested in looking at the extent to which you believe that different groups facilitate or block the
attainment of values, customs, or traditions that you cherish. Kor each of the following groups, please indicate
the values (e.g., freedom, a world of peace, a world of beauty), customs, and traditions (e.g., the Canadian
work ethic, respect for law and order, multicultiirulisin) whose attainment is either facilitated or blocked by
typical members of that group. Provide as many values, customs, or traditions thai yon feel are necessary to
convey your impression of each group. Please be honest. Your responses will he kepi strictly confidential.

and represents the average extent to which and respect for authority are the most
typical members of the group are believed to important virtues children should learn"),
block or facilitate the attainment of impori- conventionalism (e.g., Question d: "There is
ant values, customs and traditions weighted nothing wrong with premarital sexual inter-
by the proportion of typical members impli- course," reversed scored), and authoritarian
caied in hindering or helping. aggression (e.g.. Question (i: "Once our
government leaders and the authorities
ASSKSSMKNT OF RIGHT-WING AUTHORITARIANISM condemn the dangerous elements in out-
Finally, as can be seen in Table 3, we society, it will be the duty ol each patriotic
assessed participants' Right-Wing Authoritar- citizen to help stomp out the rot thai, is
ianism, by asking participants the extent to poisoning our country from within").
which they agreed (on 9-point scales, ranging Ill a series of fascinating studies, Allemcyer
from "very strongly disagree" to "very strongly has demonstrated thai high RWAs are,
agree") with the 10 best items from Bob indeed, more prejudiced individuals and thai,
Aliemeycr's RWA scale (Altcmeyer, 1988, the greater aggression and discrimination
1994). (In a separate sample, scores on the ihey direct toward various outgroups seems
10-item scale [with a Cronbach alpha of .7f)] to be due to the facl thai these exlremely
correlated .89 with scores on the complete self-righteous individuals feel threatened by
30-item scale.) the groups. If ever there were a group of
As many of you know, Altemeyer has devel- individuals inclined to base their prejudical
oped a balanced scale designed to assess the altitudes upon the sort of symbolic beliefs
three components of the original authoritar- suggested by Sears and McConahay, it would
ianism construct that he has identified as seem to be RWAs. It was to explore this possi-
being consistently related: authoritarian bility, then, that we included Aliemcycr's
submission (e.g., Question Tr. "Obedience measure of RWA in the present study.
 On llif Nature of Prejudice 15

TABLK3
10-lloin Kighi-Wing Authoritarian .Scale

1. The vv:iy things urc going in lliis coiiinry, ii's going '<> lake « lot of "strong medicine" to straighten mil the
troublemakers, criminals, and perverts.
Z People should pay less attention lo the Bible and the other liadilioJ1.il forms of religious guidance and
instead develop their own personal standards of what is moral aiul iniinoral.
3. It would he best for everyone if the proper authorities censored magazines and movies lo keep trashy
materials away from the youth.
-1. Then? is nothing wrong with premarital sexual intercourse.
f>. Obedience and respect for authority are the most important virtues children should learn,
(i. Once our government leaders and the authorities condemn (he dangerous elements in our society, it will
be the duty of each patriotic citizen to help stomp out the roi that is poisoning our country from within.
7. In these troubled times laws have to be enforced without mercy, especially when dealing with the agitators
and revolutionaries who are stirring things up.
8. Atheists and olliers who have rebelled against the established religions are no doubt every bit as good and
virtuous as those who attend church regularly.
9. The self-righteous "forces ol law and order" threaten freedom in our country a lot more than the groups
they claim are "radical" and "godless".
10. A lot of rules regarding modesty and sexual behaviour are just customs which are not necessarily any
better or holier than those which other people follow.

(Taken from Altemeyer, 1<RM)

Results the answer would appear to be qualified

Let ine now turn lo the results. "yes." For all but t h e ingroup, English Cana-
dians, there was a significant, if not. substan-
is THI-:RK PRKIUDICE AGAINST OUTGROUPS? tial, relation between stereotypical beliefs a n d
First, let me: present o u r subjects' overall prejudice. The m o r e negative the stereotype,
attitudes toward t h e live groups, as assessed the more: negative t h e attitude toward the
by o u r "evaluation thermometer". Not sur- oulgroup.
prisingly, subjects' attitudes were most
favourable toward English Canadians, the IS TIIKRK MORK TO PRIilUDICli THAN
ethnic g r o u p to which most of them SII:RI;OTYPKS?
belonged (M = 81.42), a n d least favourable These results were encouraging to us in two
toward Homosexuals (M - "M.13). Attitudes ways. First, they suggest that prejudice and
toward French Canadians (M = f>9.07), Native individual stereotypes arc, indeed, related.
Indians (M - (56.20), a n d Pakistanis (M - Second, and more important, the relatively
!)8.88) were intermediate a n d increasingly modest magnitude of these correlations
negative. This rank o r d e r of "group evalu- suggest that stereotypes do not entirely
ation" is generally consistent with past account for attitudes toward outgroups.
research. Further, each oulgroup was, indeed, Thus, there would seem to be more to preju-
evaluated m o r e negatively than t h e ingroup. dice than stereotypes.

IS TIIKRK A STKRKOTYPli-PRIiJUniCT; RELATION? IS THERE AN AH-iiCT-PRKJUDICK AND A

Let. m e now turn to the stereotype-prejudice SYMBOLIC BHI.IKI-PREJUDICE RELATION?

relation. Are prejudicial attitudes based upon Let me now move on to the affect-prejudice
sl.ereol.ypic beliefs? Alternatively, is prejudice and symbolic belief-prejudice relations. Does
justified by stereotypic beliefs? Or, because affect relate to prejudice and, if so, over and
we are dealing with correlations, is there above ibe stereotype-prejudice relation? Do
simply a strong stereotype-prejudice relation symbolic beliefs relate to prejudice and, if so,
across the various groups? over and above the stereotype-prejudice
As can be seen in the fust row o f ' f a b l e 4, relation?
 Zanna

TAB1.K 4
("orrelatioiis

Correlation Knglish French Nalivc

Between C '.an'.idiHu Canadian Indian Pakistani Homosexual
prejudice & .11 .49 .24 .30 .48
stereotype
sttTitfjIypo & <il M .72
alTcct
stereotype 8c .11 .44 .25 .20 .:«
symbolic
affect. & .20 .40 .25 .15 .23
symbolic
prejudice & .25 .(30 .44 .32 .43
allect
prcjiiflicc & .09 .49 .17 M ..10
symlxilic

AnivCT-STKKFXJTYPK, SYMBOLIC: BELIFI- of symbolic beliefs seems to be eliciting some-

STliKKOTYPE, AND AFFECT-SYMBOLIC KKLIHf thing different from our measure of stereo-
RELATIONS typic beliefs (as well as something different
Llut, before answering these- questions, I need from our affect measure).
to address the preliminary question of The relation between affect and stereotype
whether or not our affect and symbolic belief is clearly stronger. But, even here, 49 to 61
measures are completely redundant with, or percent of the variance in stereotypes is left
at least somewhat independent from, our unaccounted for. In any event, to the extent
stereotype measure. Just because we asked that the stereotype and affect measures are
our subjects three sets of questions, there is tapping the same information, affect will not
no guarantee that they will give us three, ai add to the prediction of prejudice over and
least partially, different answers. above stereotype - and our expectations that
The relations among our stereotype, afreet, it very well may do so will not be confirmed.
and symbolic belief measures are depicted in The different types of information elicited
rows two through lour of Table 4. Here it by our three questions is also revealed
can seen that although affect and symbolic through a content analysis of subjects' open-
beliefs are related to stereotypes, and to one ended responses. For example, subjects most
another, the relations are not consistently frequently characterized Native Indians as
substantial. alcoholic, lay.y, and uneducated in response
This is especially true for symbolic beliefs. to our stereotype question, as eliciting feel-
The fact that stereotypic and symbolic beliefs ings of anger, uneasiness, and pride in
hardly relate to one another is very encour- response to our affect question, and as pro-
aging. Admittedly, we were worried that our moting multiculluralism, not respecting
two belief-elicitation procedures could, iri Canadian law, and both promoting and
fact, be lapping the same beliefs. For violating peace in response to our symbolic
example, if a prejudice subject reported that belief question. As a second example, Homo-
typical members of a particular group were sexuals were characterized as effeminate,
lazy on our stereotype measure, he or she friendly, and normal, as eliciting feelings of
might also tell us that typical members of the disgust, discomfort, and confusion, and as
group hindered the attainment of the Protes- promoting freedom, blocking the attainment
tant or Canadian Work ethic on our symbolic of the traditional family, and promot-
belief measure. Clearly, however, our measure ing peace.
 On the Nature of Prejudice 17

TABLE F>
Multiple Correlations
a) Multiple Regression French Native
(Hierarchical) Canadian Indian Pakistani Homosexual

Block 1 (stereotype) .49 .24 .30 .48

Block 2 (airect .66 .45 .63 .61
and symbolic)

b) Multiple Regression French Native

(Hierarchical) Canadian Indian Pakistani Homosexual

Block 1 (stereotype) .49 .24 .30 .48

Block 2 (afreet .62 .44 .01 .60
or symbolic) (affect) (affect) (symbolic) (symbolic)

(symbolic)

c) Multiple Regression French Native

(Standard) Canadian Indian Pakistani I lomosexual

(Best unique .59 .44 .r>K .50

predictor) (affect) (aifcet) (symbolic) (symbolic)
(2nd best) .65 .63 .60
(symbolic) (afh-ct) (stereotype)

AFFECT-PRlilUniO: AND SYMBOLIC to two sets of multiple regression analyses,

BELIEI-PRIJUDICH RELATIONS designed to address the question of whether
Let me now turn to the affect-prejudice and there is more to prejudice than stereotypes,
symbolic belief-prejudice relations. As can be and, if so, whether it is affect and/or sym-
seen in row five of Table 4, the relation bolic beliefs that add to the prediction of
between affect and prejudice tends to be at intergroup attitudes over and above stereo-
least as strong as the relation between stereo- types, or, indeed, if affect and symbolic
type and prejudice (i.e., compare rows one beliefs are the primary determinants (or, at
and five of Table 4). In fact, for the least, correlates) of prejudice.
outgroups French Canadians and Native
Indians the affect-prejudice relation is THE PREDICTION OK PREJUDICE FROM STKRHO-
stronger than the stereotype-prejudice rela- TYPES, AFFECT, AND SYMBOLIC BEI.IKFS
tion, though only reliably stronger in the First, we conducted hierarchical regression
case of Native Indians. As can be seen in row analyses in which the stereotype score was
six of Table 4, the symbolic belief-prejudice entered in an initial block, followed by a
relations are also at least as strong as the second block in which the affect and sym-
stereotype-prejudice relations (i.e., compare bolic belief scores were entered simulta-
rows one and six of Table 4). And in the neously. These analyses allowed us to dis-
case of Pakistanis the symbolic belief-preju- cover (I) whether affect and symbolic beliefs
dice relation is significantly stronger than the together added to the prediction of attitudes
stereotype-prejudice relation. toward outgroups over and above the predic-
Thus, there is an affect-prejudice and a. tion of such altitudes from stereotypes, and
symbolic belief-prejudice relation. But, do (2) if so, whether affect, symbolic beliefs, or
these relations hold over and above the both affect and symbolic beliefs were respon-
stereotype-prejudice relation? The fact that sible for the increased predictive capacity.
our measures of affect, symbolic belief, and As can be seen in the top panel of Table
stereotype are not entirely redundant sug- 5 (where the results of the multiple
gests that it is possible. So, let me now turn regression analyses are depicted in terms of
18 Zanna

TABLE 6
Mean Altitudes
English I'VOIK l l Native
RWA Canadian Canadian Indian Pakistani Homosexual

Low 81.35 71.10 08.11 IW.Ki r>13.r.

HiKh 81.50 ()t..7(i W.I 2 5-1.12 36.2<i

Total Sample 81.42 60.07 M>.5>0 r>K.8H 44.13

multiple correlations for ease of presenta- Next, in order to determine (he unique
tion), when affect and symbolic beliefs were contribution of each predictor, we conducted
entered as a block, following slereotypic a series of standard multiple regressions in
beliefs, there was a substantial increase in the which the three predictors were entered
multiple correlation. The increase is signifi- simultaneously. As can be seen in the bottom
cant beyond the .005 level for each outgroup. panel of Table 5, affect provides the greatest
Which variable, however, affect and/or unique contribution to the prediction of
symbolic beliefs, is actually responsible for attitudes toward French Canadians and
this increase in predictability over and above Native Indians. In contrast, symbolic beliefs
stereotypes? As can be seen in the middle are most uniquely predictive of attitudes
panel of Table 5, where the results for affect towards Pakistanis and Homosexuals. Inter-
and symbolic beliefs are depicted separately, estingly, in these standard regression analyses
the answer seems to be affect, symbolic stereotypic beliefs uniquely contributed (over
beliefs, or both, depending upon the group and above symbolic beliefs) only to the
in question. prediction of altitudes toward Homosexuals.
Affective associates or emotion add, over
and above stereotypes, to the prediction of Authoritarianism and Prejudice
attitudes toward Native Indians. Symbolic Let me now turn to the second purpose of
beliefs add to the prediction of altitudes the study: to determine whether authoritar-
toward Pakistanis and Homosexuals. And, ianism relates to prejudice and, more inter-
finally, both affect and symbolic beliefs add estingly, whether attitudes toward outgroups
significantly to the prediction of attitudes are based on (or, at least, related to) differ-
toward French Canadians. ent sorts of information for high and low
Thus, in answer to the question of whether authoritarians. We wondered, for example,
there is more to prejudicial attitudes than whether the prejudice of authoritarian indi-
stereotypical beliefs, the answer is a resound- viduals is more likely to be based upon sym-
ing "yes." By adding affect and symbolic bolic beliefs.
beliefs, we increased the amount of variance First, are high KWAS, indeed, more preju-
accounted for in altitudes toward French diced than low KWAs? As can be seen in
Canadians from 24 to 44%; by adding affect Table 6, individuals above the median on our
to stereotypes, die comparable increase lor KWA measure (our High RWAs) do hold more
Native Indians was from 6 to 14%; by adding negative attitudes toward all the outgroups,
symbolic beliefs to stereotypes, the compar- especially Pakistanis and Homosexuals, than
able increases for Pakistanis and Homosex- do individuals below the median (our low
uals were from 9 to 40% and 23 to 38%, KWAs). Thus, our high RWAs appear, as in
respectively. So, although there is admittedly Altemeyer's research, to be "equal-opportun-
much variance in prejudice left unaccounted ity bigots," holding more negative (or
for, there was a substantial increase in pre- prejudicial) attitudes toward each outgroup.
dictability when affect and symbolic beliefs Second, are the attitudes of high and low
were taken into account. RWAs based on different sorts of information?
 On the Naiure of Prejudice 19

TABLE 7
Multiple Correlations

Low KWAs

a) Mullipk- Regression French Native

(Hierarchical) Canadian Indian Pakistani Homosexual
r
Block 1 (slcr™ty|ii:) .(VI .'J. . (if) .65
Block 2 (aH'cct and symbolic) 7<i , r .7 b'l .73
High kWAs
Block 1 (stereotype) ..'1(1 .OH yr.
Block 2 (afl'ecl and symbolic) ..r>8 .'1(1 .(>(> ,.r)l
Low kWAs
I)) Multiple Regression French Native
(Hierarchical) Canadian Indian Pakistani 1 lomosennal

Block 1 (stereotype) (VI .25 .(">() .()•>

Block 2 (all'ccl or symbolic) .73 .55 7:i

(ailed) (aflect) (allccl)

High RWAs

Block 1 (stereotype-) ;!(i .30 .OK Zr>

Block 2 (all'ect or symbolic) .55 — .IV) .50
(symbolic) (symbolic) (symbolic)
Low KWAs
c) Multiple Regression French Native
(Standard) Canadian Indian Pakistani 1 lomosexual

(Best, unique .70 M .(>!>

predictor) (iiircci) (alien) (stereotype) (ailed)
(2nd best) .70 _.. — .73
(symbolic) (stereotype)

High KWAs
(Best unique r
,. i0 39 .l>3 M
predictor) (symbolic) (ailed) (symbolic) (symbolic)
('2nd best) .57** .51*
(af (<•<() (ailed) (ailed)

*p< .18 **p < .OK

(For example, docs ihe prejudice of high cated Iwo interesting findings. First, compar-
RWAs tend to be based more on symbolic ing ihe results of block 1 for the low and
beliefs whereas the prejudice of low KWAs high RWAs (see lines I and 3 in lop panel of
tend to be based more on stcrcoiypic Table 7), it is clear that the stereotype-preju-
beliefs?) To answer this sort of question, we dice relation lends to be stronger for the low
repeated the two series of multiple regression than lor the high KWAs (at least for three of
analyses for the four outgroups, this time the four groups); second, examining the
separately for high and low RWAs. results for block 2 (see lines 2 and 4 in top
The results of the first set of hierarchical panel of Table 7), it is clear that ailed and
multiple regressions, in which stereotype is symbolic beliefs, as a block, account for
entered in a first block, followed by affect additional variation in altitudes for both
and symbolic beliefs in a second block, indi- groups of subjects, but especially for high
20 Zanria

RWAs (at least for three of the four out- Indians (cf. Haddock, Zanna, & Esses, in
groups). press, Experiment 2). All I can say today is
But, is it affect and/or symbolic beliefs that the replication study took place over the
that adds to the prediction of attitude over summer of 1990, during the so-called Oka
and above stcreotypic beliefs? When the crisis. As I'm sure you all recall, in response
results of the second block are depicted to the town's planned extension of a local
separately for affect and symbolic beliefs (see golf course, the paramilitary Mohawk Warrior
middle panel of Table 7), it is clear thai society erected a barricade in Oka, Quebec
affect adds to the prediction of outgroup (30 km northwest of Montreal). Alter a court
attitudes for low RWAs; in contrast, for high injunction to remove the barrier was
RWAs symbolic beliefs add to the prediction ignored, the Quebec Provincial Police
of outgroup attitudes. attacked the barrier, leading to the death of
Finally, we can ask straight out: What a police officer and the onset of a bitter
variables provide the largest unique contribu- standoff that lasted approximately two
tion to the prediction of altitudes toward die months. Perhaps the prominence of the
outgroups for low and high RWAs? The Mohawk Warriors in the media, a group
results of the standard multiple regressions, regarded as advocating rapid, even violent,
in which the three predictors are entered social change, increased the salience of
simultaneously, are presented in the bottom symbolic beliefs with regard to Native
panel of Table 7. Here it can be seen thai Indians.
affect, and, in the case of Pakistanis, stereo- Next, we conducted two studies in which
types, best account for variation in the we assessed male prejudice toward the target
outgroup attitudes of low RWAs. In contrast, group, "feminists," and discovered (1) that
symbolic beliefs, and, in the case of Native attitudes toward feminists are, indeed, very
Indians, affect, best predict prejudice for negative (entirely comparable to our
high RWAs. subjects' attitudes toward Homosexuals),
To sum up, it would appear that tradi- especially for High RWAs, and (2) that sym-
tional stereotypic beliefs are less relevant to bolic beliefs arc, once again, the strongest
high than to low RWAs. Further, for lows, unique predictor of prejudice for High RWAs
affect not only adds to the prediction of (Haddock & Zanna, 1994).
oulgroup attitudes over and above stereo- In another study (with "traditional Native
types, it is, by and large, the best unique Indians" as the target group) we added a
predictor of prejudice lor these relatively fourth predictor variable, past experience with
tolerant individuals. For highs, symbolic typical members of the group, and discovered
beliefs not only account for additional vari- (1) that past experience is not entirely redun-
ance over and above stereotypes, for these dant with our other predictor variables, and
relatively intolerant individuals they are, by (2) that the quality (not the frequency) of
and large, the best unique predictor of past experience adds to the prediction of
prejudice. prejudice over and above stereotypes, affect,
and symbolic beliefs (Haddock, Zanna, &
Follow-Up Research Esses, in press, Experiment 3).
Encouraged by these initial results, we have Finally, in two studies (one with "homosex-
conducted several follow-up studies that I uals" as the target group; the other with
would like to mention, if only very briefly. "feminists" as the target group) we created
First, we conducted an exact replication of a measure of perceived value dissimilarity (by
the present study. Interestingly, the results comparing subjects own values with their
were virtually identical except for the fact beliefs of the values of the outgroup), and
that symbolic beliefs (in contrast to affect) discovered (1) that perceived value dissimi-
best predicted attitudes towards Native larity is related most highly to our measure
 On ihe Nature ol" Prejudice

of.symbolic beliefs, especially for high KWAs, determine which source (or, at least, which
and (2) that perceived value dissimilarity correlate) of prejudice is influenced by these
accounts, in large part, for the RWA-prejudice variables. For example, does equal status
relation (Haddock, Ksses, & Zanna, 1993; contact reduce prejudice by changing stereo-
Haddock & /.anna, 1994). This research types, affect, and/or symbolic beliefs?
further supports the notion that threats to Further, we would like to determine
values are implicated in the relatively nega- whether the consequences of prejudice are
tive attitudes of authoritarian individuals. moderated by the source (or, at least, major
correlate) of prejudice. For example, would
Future Research prejudiced individuals be more likely to
In future research, we plan to determine oppose affirmative action policies if their
whether individual difference variables, other negative attitudes toward the outgroup were
than RWA, moderate the various predictor- based primarily upon symbolic belief's than
prejudice relations. Presently, Geoff Haddock if these altitudes were based primarily upon
is in the process of developing a scale, called stereotypic beliefs? Preliminary evidence docs
the Feeling-Belief Measure (IBM), designed suggest that discrimination is more likely to
to assess individual differences in the extent be the result of negative attitudes toward an
to which individuals believe their decisions, outgroup if such attitudes are based upon
preferences and evaluations are generally (or, at least, are more consistent with) sym-
influenced their emotions and/or their bolic than stereotypic beliefs (Haddock,
thoughts. Sample items from the scale Esses, & Zanna, 1993). Finally, if we can
include, 'The phrase 'follow my heart' is determine the primary source of prejudice,
important in making proper decisions," we would also likely to try to change the atti-
"When I buy a product, its usefulness is more tudes of prejudiced individuals by targeting
important than how it makes me feel" and the relevant information.
"Making a proper decision requires a long Of course, so far our research has been
period of" thought." Preliminary results sug- entirely correlational in nature. In the future
gest that individuals who, on the FBM, indi- we plan to take a more developmental
cated that they use their feelings in guiding approach, both in terms of investigating the
their preferences were, indeed, more likely development of prejudice toward established
to base (heir intcrgroup altitudes on affective groups in children and examining in adults
information (Haddock & Zanna, 1993). the formation of attitudes toward new
More generally, we would like to deter- groups, such as recently-arrived immigrant
mine how the "known" causes of prejudice, groups in the community.
both individual difference determinants In any event, by conceptualizing prejudice
(such as authoritarianism) and situaiional (and the determinants of prejudice) within
determinants (such as competition), influ- current models of attitudes, we believe that
ence the various sources of prejudice we social psychologists will have more than
have identified, both in terms of "main enough interesting questions to pursue in
effects" and, especially, in terms of moderat- the foreseeable future.
ing the relation between the source and
prejudice. Does competition, for example,
create more negative symbolic beliefs and/or This article is based on the D.O. Hebb award
increase the impact of existing symbolic address presented to the Canadian Psychologi-
beliefs on prejudice? cal Association in 1993. The research was
We would also like lo determine how the done in collaboration with Geoffrey Haddock
"known "causes of prejudice reduction, such and Victoria Esses (as the references to indi-
as superordinate goals and equal status vidual papers make abundantly clear) and was
contact, "work." That is, we would like to supported by a research grant from the Social
22 Zauna

Sciences and Humanities Research Council of la comprehension dt; la progression ou de

Canada to Victoria Esses and Mark Zanna. la regression du prtijugc est disciitee dans
Preliminary versions of this talk were pres- les pages qui suivenl.
ented at the Nags Head Conference on
Stereotyping and Intergroup Relations, Nags
Head Conference Center, Kill Devil Hills, References
North Carolina, June, 1990, at the annual Alleini-yer, B. (1988). Knemies of freedom:
meeting of the Society of Experimental Social Understanding right- wing authoritarianism.
Psychology, Buffalo, New York, October, 1990, San Francisco: Jossey-Bass.
and at (lie XXV International Congress of Altemeyer, B. (JO'.M). Reducing prejudice in
Psychology, Brussels, Belgium, |uly, 1992. A right-wing authoritarians. In M.P. Zanna &
more complete description of the research J-M. Olson (Eds.), The psychology of
program appeals in Esses, Haddock, and prejudice: The Ontario symposium (Vol. 7,
Zanna (1993). pp. 131-148). Hillsdale, NJ: Erlhaum.
Correspondence should be sent to Mark P. Abelson, R.P., Kinder, D.R., Peters, M.D., &
Zanna, Department of Psychology, University of Fiske, S.T. (1982). Affective and semantic
Waterloo, Waterloo, Ontario, Canada, N2l. 3(il. components in political person perception.
Journal of Personality and Social Psychology,
42, 619-030.
Resume Ajzen, 1., & Fisbbein, M. (1980). Understanding
Nous pretendons que le prcjuge (attitude altitudes and predicting social behavior.
negative envcrs le hors-groupe), en plus Englewood Cliffs, NJ: Prentice-Hall.
d'etre fondc stir des croyumes liees ;uix Bern, D.J. (1972). SeK-peiception theory. In L.
stereotypes (selon lcsquelles les mciubrcs Beikowil/. (Ed.), Advances in experimental
representatifs du hors-groupe posscdeni. social psychology (Vol. (i, pp. 1-02). San
certains traits on caracieristiqucs), est cga- Diefro, C"A: Academic Press.
Icment imputahlc a des croyances symboli- Eagly, A.H., & Chaiken, S. (1993). The psy-
ques (scion lcsquelles les membres repre- chology of altitudes. San Diego, CA: llaicomt
sentatiis violent des traditions, des coutu- Biace Jovanovicll.
mes ct des valeurs cheres), ainsi qu'a des Esses, V.M., Haddock, G., & Zanna, M.P.
emotions et a des experiences anterieures (1993). Values, stereotypes, and emotions as
en rapport aver le hors-groupe. Pliisieurs determinants of intergroup altitudes. In
eludes (la premiere est presentee ici) nous D.M. Mackie & D.I,. Hamilton (Eds),
out revele, d'une part, que inalgre les liens A//""'', cognition, and stereotyping: Interactive
unissaiil les stereotypes, les croyanccs syni- processes in group perception (pp. 137-loh).
boliques, les emotions et les experiences San Diego, <:A: Academic Press.
antciieures, il n'esi. pas supcrflu d'clablir Cardner, R.C. (1994). Steieolypes as
des distinctions einre tes entiles, et d'au- consensual beliefs. In M.P. Zanna & ].M.
trc pait, que chacun de ces factcurs consii- Olson (Eds.), The psychology of prejudice:
luc en fail line variable explicative impor- The Ontario symposium (Vol. 7, pp. 1-31).
lante des altitudes caracterisees par le Ilillsdale, NJ: Erlbaum.
prcjuge. I.e prejuge ne resullerait done
Gardner, R.C., Wonnacoll, E.]., & Taylor,
pas uniquemcni ties croyances liees aux
D.M. (I9()8). Ethnic stereotypes: A factor
stereotypes. Nous avons egalcment decou-
analytic investigation. Canadian Journal of
vert que les attitudes relaiivcmeni plus
Psychology, 22, 3.r.--14.
negatives maiiifestees par les personnes les
Haddock, C , & Zanna, M.P. (1993). Predict-
plus favomblcs a l'auloritarisine s'expli-
ing prejudicial altitudes: The importance of
queui davantage par des croyances symbo-
affect, cognition, and the feeling-belief
lifjue.s. I/incid«riice de (cs iesultal.s quant ;i
dimension. In I,. McAlisler & M.L.
 O n the Nuliiri: of Prejudice 23

Rothschild (Kris.), Advances in consumer Rosenberg, M.J. (195f>). Cognitive structure

research (Vol. 20, pp. 315-318). Piovo, UT: and altiluriinal affect. Journal of Abnormal
Association lor Consumer Research. and Social Psychology, 53, 3b7-372.
Haddock, C , & /anna, M.P. (1994). Prefer- Stangor, C , Sullivan, L.A., & Ford, T.E.
ring "housewives" to "feminists": (1991). Affective and cognitive rlelernii-
Categorization and the favorability of alti- nanls of prejudice. Social Cognition. 9,
tudes toward women. Psychology of Women 359-380.
Quarterly, 18, 25-52. Sears, D O . (1988). Symbolic racism. In PA.
Haddock, C , Zanim, M.P., & Esses, V.M. (in Kalz & D.A. Taylor (Kds), Eliminating rac-
press), 'l'he (limited) role of stereotypes in ism (pp. 53-84). New York: Plenum.
predicting altitudes toward Native Cana- Zajone, RR- (1980). Feeling and thinking:
dians, lirilish Journal of Social Psychology. Preferences need no inferences. American
Haddock, G., Zanna, M.P., & Ksses, V.M. Psychologist, 35, 151-175.
(1993). Assessing the structure of prejudi- Zanria, M.P., & Rcmpcl, JK. (1988). Attitudes:
cial attitudes: The case of altitudes toward A new look ill an old concept. In D. Bar-Tal
homosexuals. Journal of Personality and & A.W. Kruglanski (Eds.), The social psychol-
Social Psychology. 65, 1105-1108. ogy of knowledge (pp. 315-334). Cambridge,
McConahay, J.li., & Hough, |.C.,]r. (1976). England: Cambridge University Press.
Symbolic racism. Journal of Social Issues. 32,
23-45.