CURRENT RESEARCH IN SOCIAL PSYCHOLOGY
Volume 8,
Number 2
Submitted: July 14, 2002
First Revision: September 5, 2002
Second Revision: September 10, 2002
Accepted: September 10, 2002
Publication date: September 10,2002
UNCOVERING THE MULTIDIMENSIONAL
NATURE OF STEREOTYPE INFERENCES: A WITHIN-PARTICIPANTS STUDY OF GENDER, AGE,
AND PHYSICAL ATTRACTIVENESS
Malcolm J. Grant
Cathryn M. Button
T. Edward Hannah
Memorial University
of Newfoundland, Canada
ABSTRACT
Eighty undergraduates
guessed the attitudes of several people whose pictures they were shown. Within-participant
regression analyses were conducted to assess, at the individual level, the
influence of targets' gender, age, and physical attractiveness. Participants
expected men to adopt conservative positions on child discipline, feminism,
immigration, and homosexuality while women were expected to be conservative
on religion. Older people were expected to be more conservative on most issues
while attractive persons, independent of age, were expected to be more liberal.
In addition, examination of interaction effects revealed several instances
where gender stereotypes were moderated by either the age or attractiveness
of the targets. We conclude that stereotypes frequently combine in an interactive
fashion and that future investigations of these interactions would benefit
from the within-participant, multiple-target procedure used here.
[19]
---------------
[20]
In many studies of
stereotypes (e.g., Bassili & Reil, 1981; Kite, Deaux, & Miele, 1991;
Signori, Butt, & Kozak, 1982), participants have been asked to attribute
characteristics to targets who have been categorized or labeled by the
researchers. The results of these studies leave
little doubt that people possess a variety of consensual beliefs about
characteristics associated with categories of persons and that stereotypic
associations can be reliably and easily triggered. Encountering a person who
belongs to the category (Word, Zanna, & Cooper, 1974), viewing a picture of
such a person (Bargh, Chen, & Burrows, 1996), or even being presented
subliminally with words related to a stereotype (Devine, 1989) can evoke
associations that may then influence a wide range of behaviors and judgments.
In everyday life
the stimuli that trigger stereotypic associations are rarely unidimensional.
People do not encounter "a male," "an old person," or
"an Arab." Rather, they are confronted by a myriad of social
information, some of which is physically prominent (e.g., gender, age,
ethnicity, attractiveness, etc.) and some of which is more subtle (e.g.,
occupation, education level, social class). Thus, it is quite likely that the
perceiver uses more than a single target attribute to make a stereotypic
judgment. Although previous research tells us quite a bit about what people can do in making judgments about others using
single bits of information, it does not tell us what people actually do when
they encounter others in stimulus-rich social situations. The challenge for
researchers interested in the process of stereotyping is to capture the rich
complex of information available to the perceiver. The question then becomes
how do these various target attributes, available in everyday interactions, simultaneously influence the perceiver's judgment?
One possibility
is that stereotypes combine in some additive fashion to affect people's
judgments. People may take account of several of the target's category
memberships and arrive at an overall judgment based on the characteristics that
are thought to be most typical of each category. Additive or averaging models
of this sort, often including weighting factors for the information being
combined, have a long history in social psychology (see Eagly and Chaiken,
1993, for a review) and have frequently been shown to account for significant
portions of the variability in people's judgments (e.g., Himmelfarb &
Anderson, 1975).
A second possibility
is that when several stereotypes are evoked simultaneously, they combine in
an interactive fashion. Bassili and Reil (1981), for example, found that the
influence of gender, occupation, and ethnic stereotypes was greater for younger targets than older ones. Similarly,
Macrae, Bodenhausen, and Milne (1995) have argued that in many circumstances,
the activation of one stereotype actually inhibits the activation of other
less salient ones.
[20]
---------------
[21]
Examining the simultaneous
influence of several stereotypic dimensions can be difficult because the number
of possible combinations of characteristics quickly becomes very large. This is particularly true when one or
more of the stereotypic dimensions of interest are continuous in nature (e.g.,
age) rather than categorical (e.g., gender). In a between-participants design,
where each participant is exposed to only one target, the number of
participants needed to assess higher-order interaction effects can be
prohibitive. An alternative is to expose each participant to several targets
and carry out preliminary analyses on the resulting data at the individual
level. The preliminary analyses in these within-participant designs may take the form of simple comparisons
among within-participant treatment conditions or more complex correlational and
regression analyses carried out on each participant's matrix of data. Judd,
Kenny, and McClelland (2001), for example, have described procedures that allow
testing for complex, within-participant, treatment mediation and moderation
effects. Once summary indices have been calculated for each participant, one
can then look for regularities in the indices across participants. Thus,
idiographic and nomothetic approaches can be combined within a single study
(Michela, 1990).
Although the advantages of
the within-participant designs just described have been known for some time,
their application to the study of stereotypes has been relatively infrequent,
possibly because of concern over demand characteristics. When the same
participant is asked to judge a small number of targets (e.g., male versus
female or old versus young), the researcher's purpose may be obvious and the
resulting judgments may be different from those made under more natural
circumstances. In the present paper, we show how the problem of demand
characteristics can be reduced by asking each participant to make judgments
about a relatively large number of targets, shown in pictures. In the study we
describe, participants were asked to judge the social attitudes of the targets
and the pictures that each person saw were randomly selected from a large pool.
Although the targets varied in gender, age, and physical attractiveness,
nothing in the procedure specifically alerted participants to these dimensions.
There is evidence that people's
judgments about a target person's attitudes are strongly influenced by the
target's age, gender, and attractiveness. Old people are expected to have
more conservative attitudes than young people on a variety of social and political
issues (Grant, Ross, Button, Hannah, and Hoskins, 2001; Griffitt, Nelson,
and Littlepage, 1972). Women are expected to be more conservative than men
on issues of sexuality but more liberal than men on minority and environmental
issues (Grant, Button, Ross, and Hannah, 1997; Grant et al., 2001). Finally,
in the case of physical attractiveness, people exhibit what may be considered
a type of self-serving bias. They expect attractive targets, more than less
attractive ones, to share their own views (Marks and Miller, 1982; Mashman,
1978; Schoedel, Frederickson, & Knight, 1975). Although each of these
influences on people's judgments has been demonstrated in isolation, little
is known about how they operate together.
[21]
---------------
[22]
Undergraduates were asked to
guess the attitudes of 10 men and 10 women, whose pictures they were shown. The
age and attractiveness of the people in the pictures had been rated by
participants in an earlier study (Grant, Button, Hannah, & Ross, 2000). The
matrix of data produced by each participant in the present study was subjected
to regression analyses designed to determine the separate and combined
influence of the targets' age, gender, and attractiveness. Of particular
interest was the possibility that the inclusion of interaction terms in the
regression model might increase the predictability of participants' judgments
about target persons' attitudes.
METHOD
Participants
We tested 40 male
and 40 female undergraduates at Memorial University. Participants ranged in age
from 18 to 45 (M =
20.41, Mdn = 19.00, SD = 3.62). The procedure took about 20 - 30
minutes. Upon completion, each participant was paid $2.75.
Materials
Target
Pictures
Digitized pictures
of adult men and women (showing head and shoulders) were drawn from a variety
of sources. Some came from internet websites, some were taken from television,
and others were scanned from pictures in magazines and family pictures contributed
by colleagues. Half the pictures were of women and half were of men. Within
each gender, an attempt was made to include pictures of people who, in the
judgment of the researchers, ranged in age from late teens to late seventies.
All pictures were digitally cropped to a width of 172 pixels and a height
of 203 pixels and saved as 256-colour, bit-mapped images. When displayed on
a participant's computer screen, the images were approximately 4.7 cm. wide
and 5.4 cm. high.
[22]
---------------
[23]
In an earlier
study (Grant et al., 2000), 91 participants (39 men and 52 women) between the
ages of 19 and 58 rated either the attractiveness (1 = very unattractive, 10 =
very attractive) or the apparent age (in years) of the person in each of the
pictures. A mean attractiveness and mean age rating were calculated for each
picture. The 100 pictures used in the present study were chosen so that the
average age and attractiveness of the men and women were matched as closely as
possible (see Table 1). An indication of the reliability of the picture ratings
was obtained by correlating, across pictures, the mean rating by male
participants with the mean rating by female participants. The correlations (df = 98) for age and attractiveness were .99
and .94, respectively. Finally, and not unexpectedly, the correlation across
pictures between the mean age and mean attractiveness ratings was significantly
negative, r(98) =
-.58, p < .001.
Table 1: Age
and Attractiveness Ratings for the Pictures of Male and Female Targets
|
|
Age ratings |
Attractiveness ratings |
||||
|
|
Male targets |
Female targets |
t (df = 98) |
Male targets |
Female targets |
t (df = 98) |
|
Mean |
43.20 |
43.48 |
0.11(ns) |
4.66 |
4.86 |
0.89 (ns) |
|
SD |
10.51 |
15.40 |
|
0.84 |
1.31 |
|
|
N(targets) |
50 |
50 |
|
50 |
50 |
|
Attitude
Statements
Two statements on
each of five issues (discipline of children, homosexuality, feminism, immigration,
and religion) were used. For each issue, one of the statements expressed an
attitude in favor of the concept and one expressed an attitude opposed to
the concept. The complete set of attitude statements is shown in Appendix
A. Each statement was followed by a scale ranging from 1 (disagree strongly)
to 7 (agree strongly).
[23]
---------------
[24]
To simplify
presentation of the results we created a single score for each issue by adding
the response to the positively worded statement to the reverse-scored response
to the negatively worded statement. [1]
The result was a set of scores that could range from 2 to 14 with higher scores
reflecting more favorable attitudes toward the issue.
Procedure
Participants were
tested up to three at a time. Each person sat in a separate cubicle equipped
with a personal computer running a Visual Basic program. All instructions and
experimental materials were presented by the computer and participants
responded by pointing and clicking the mouse.
For
each participant, 20 pictures (10 of men and 10 of women) were randomly selected
from the pool of 100 pictures. Pictures of these target persons were displayed
one at a time in a random order on the participant's computer screen. While
each picture remained on the screen, the ten attitude statements were presented
one at a time in a random order. The participant was asked to estimate, using
a 7-point scale that ranged from (1) strongly disagree to (7) strongly agree,
how the person in the picture would respond to the attitude statement. When
all ten statements had been presented for a particular photograph, a new picture
appeared and the procedure was repeated. The procedure ended when participants
had made ten attitude inferences for each of the different target persons.[2]
[24]
---------------
[25]
RESULTS
The mean attitudes assigned
to male and female targets on each of the five attitude issues are shown in
Table 2. As can be seen, participants expected significant gender differences
on each of the five issues.
Table 2: Mean Attitudes
Attributed to Male and Female Targets on Five Attitude Issues
|
Issue |
Male targets |
Female targets |
t(79) |
p (two-tailed) |
|
Discipline Mean SD |
8.15 1.15 |
7.41 1.21 |
6.04 |
< .001 |
|
Homosexuality Mean SD |
7.75 1.50 |
8.61 1.54 |
-5.74 |
< .001 |
|
Feminism Mean SD |
7.99 1.26 |
10.81 1.21 |
-17.71 |
< .001 |
|
Immigration Mean SD |
8.17 1.18 |
9.07 1.19 |
-6.88 |
< .001 |
|
Religion
Mean SD |
8.96 1.18 |
10.12 1.27 |
-7.23 |
< .001 |
Note. All means are based on the data of the same 80 participants.
[25]
---------------
[26]
Table 3: Correlations
among Target Predictor Variables and Attitude Inferences on Five Attitude
Issues.
|
|
Target Age |
Target Attractive. |
Discipline |
Homosexuality |
Feminism |
Immigration |
Religion |
|
Target Gender |
-.01 |
.08 |
-.16 |
.14 |
.47 |
.19 |
.22 |
|
Target Age |
|
-.57 |
.39 |
-.34 |
-.32 |
-.14 |
.36 |
|
Target Attract. |
|
|
-.31 |
.32 |
.35 |
.21 |
-.21 |
|
Discipline |
|
|
|
-.43 |
-.44 |
-.32 |
.14 |
|
Homosex. |
|
|
|
|
.54 |
.47 |
-.08 |
|
Feminism |
|
|
|
|
|
.39 |
-.02 |
|
Immig. |
|
|
|
|
|
|
.16 |
Note. Correlation coefficients are averaged across participants. All but two of the mean correlations, gender-age and feminism-religion, differ significantly (p < .05) from zero by two-tailed t-test.
Five regression analyses were
conducted for each of the five issues. The first analysis for each issue assessed
the influence of target gender. The second analysis assessed the influence
of target age and target attractiveness. Finally, the third, fourth, and fifth
analyses for each issue assessed each of the three possible two-way interactions
among the target variables.[3]
[26]
---------------
[27]
Table 4: Means and
Standard Deviations for Regression Coefficients on Five Attitude Issues
|
|
Regression Coefficients for Target Variables and Interaction Terms in Equation |
|||||
|
Issue |
Target Gender (G) |
Target Age (Age) |
Target Attractiveness (Att) |
G x Age |
G x Att |
Age x Att |
|
Discipline (n = 69)
Mean SD |
-.78 1.13 |
.06 .06 |
-.28 .66 |
-.00 .10 |
.49 1.41 |
.01 .05 |
|
Homosexuality (n = 68)
Mean SD |
1.11 1.15 |
-.05 .08 |
.48 .81 |
-.01 .16 |
.13 1.84 |
.00 .05 |
|
Feminism (n = 65)
Mean SD |
2.91 1.41 |
-.03 .08 |
.78 .77 |
-.05 .11 |
-.14 1.51 |
-.01 .06 |
|
Immigration (n = 64)
Mean SD |
1.10 1.12 |
-.00 .07 |
.48 .74 |
.01 .11 |
-.41 1.47 |
-.00 .05 |
|
Religion (n = 66)
Mean SD |
1.22 1.46 |
.08 .07 |
.12 .82 |
.03 .11 |
-.64 1.88 |
-.01 .05 |
[27]
---------------
[28]
Each participant's inferences
about target attitudes on each of the five issues were regressed on target
gender. The regression coefficient was taken as an index of the participant's
gender stereotype. Mean regression coefficients (Mb's) for gender differed significantly from zero for
all five issues. Participants expected male targets compared to female targets,
to be more favorable toward strict discipline, Mb = -.78, t(68) = -5.71, p < .001, but less favorable toward homosexuality, Mb = 1.11, t(67) = 7.95, p < .001, feminism, Mb
= 2.91, t(64) = ‑16.61,
p < .001, immigration, Mb = 1.10, t(63) = 7.86, p < .001, and religion, Mb
= 1.22, t(65) = 6.78, p < .001.
Age and Attractiveness
Stereotypes
Attitude inferences were
regressed on target age and target attractiveness. The regression coefficients
were taken as indices of the participant's age and attractiveness stereotypes.
With both regressors in the equation, each regression coefficient reflects the
influence of that regressor with the influence of the other removed. Mean
regression coefficients for target age (with attractiveness controlled)
differed significantly from zero for four of the five issues. Participants
expected older targets compared to younger ones, to be more favorable toward strict
discipline, Mb = 0.06, t(68) = 8.02, p < .001, and religion, Mb
= 0.08, t(65) = 10.01, p < .001, but less favorable toward homosexuality, Mb = -0.05, t(67) = -5.31, p < .001, and feminism, Mb
= -0.03, t(64) = ‑3.02,
p < .001.
Mean regression coefficients
for target attractiveness (with age controlled) differed significantly from
zero for four of the five issues. Participants expected attractive targets
compared to less attractive ones, to be less favorable toward strict discipline,
Mb = -0.28, t(68)
= -3.50, p < .001, but more favorable toward homosexuality,
Mb = 0.48, t(67) = 4.84, p < .001, feminism, Mb
= 0.78, t(64) = 8.18, p < .001, and immigration, Mb = 0.48, t(65) = 5.18, p < .001.
[28]
---------------
[29]
Gender x Age Interactions
Attitude inferences were
regressed on target gender, target age, and the interaction cross-product of
these two factors. The regression coefficient for the interaction term was
taken as an indication of the extent to which the participant's age stereotype
was different for male and female targets. The mean regression coefficient for
the gender x age interaction was significant only for the feminism issue, Mb = -0.05, t(64) = -3.57, p < .001. As already noted, people expected support for feminism to
decline with age. The negative value for the interaction coefficient indicates
that this decline was expected to be steeper (i.e., more negative) for female
targets (coded 2) than for male targets (coded 1). Best-fitting regression
lines, aggregated across participants, are shown in Figure 1.
Figure 1: Aggregated Regression Lines Showing Attitudes Toward Feminism Attributed to Male and Female Targets of Different Ages
[29]
---------------
[30]
Gender x Attractiveness
Interactions
Attitude inferences were
regressed on target gender, target attractiveness, and the interaction
cross-product of these two factors. The regression coefficient for the
interaction term indicates the extent to which the participant's attractiveness
stereotype was different for male and female targets. The mean regression
coefficient for the gender x attractiveness interaction was significant for the
issues of discipline, Mb =
.49, t(68) = 2.92, p = .005, immigration, Mb = -0.41, t(63) = -2.26, p = .027, and religion, Mb
= -0.64, t(65) = -2.75, p = .008. People expected attractive persons to be less
favorable toward strict discipline and this was especially true for male
targets. They expected attractive persons to be more favorable toward
immigration and this was especially true for male targets. Finally, the
relationship between attractiveness and attitudes toward religion was expected
to be more negative for female targets than for male targets. Best-fitting
regression lines, aggregated across participants, are shown in Figures 2, 3,
and 4.
Figure 2: Aggregated Regression Lines Showing Attitudes Toward Discipline Attributed to Male and Female Targets of Different Levels of Attractiveness
[30]
---------------
[31]
Figure 3: Aggregated Regression Lines Showing Attitudes Toward Immigration Attributed to Male and Female Targets of Different Levels of Attractiveness
Figure 4: Aggregated Regression
Lines Showing Attitudes Toward Religion Attributed to Male and Female Targets
of Different Levels of Attractiveness
[31]
---------------
[32]
Age x Attractiveness Interactions
Attitude inferences were
regressed on target age, target attractiveness, and the interaction
cross-product of these two factors. The regression coefficient for the
interaction term indicates the extent to which the participant's age stereotype
was moderated by target attractiveness. On none of the five issues was there a
significant interaction between age and attractiveness.
Comparisons Between Male
and Female Participants
Male and female participants
were compared on each of the regression coefficients described above. No
significant differences were found.
DISCUSSION
The results reported here demonstrate pervasive stereotypes
in people's inferences about other people's attitudes. The stereotypes are
consistent with the perception that women, compared to men, are less disciplinarian,
more tolerant and accepting of homosexuals and immigrants, more supportive
of feminism, and more religious, that old persons are relatively conservative
on these issues, and that attractive people are relatively liberal. Like other
stereotypes, these perceptions appear to reflect consensual generalizations
about the characteristics of large categories of people.
More importantly, however, the results show that people do indeed take advantage
of the multiplicity of information available to them in forming their perceptions
of others. Perceivers' inferences based on gender, for example, were qualified
by age. The gender stereotype that women would be more supportive
of feminism than would men weakened as target age increased. People expected
relatively little support for feminism among older persons of either gender.
[32]
---------------
[33]
Gender stereotypes were also
influenced by the attractiveness of the targets. People generally expected
women to be more opposed to strict discipline and
more supportive of immigration but both these tendencies weakened and
eventually reversed with increasing attractiveness of the targets. The stereotype
that men are "red necked" in their attitudes was especially
influenced by the attractiveness of the target. Attractive men were perceived
to have attitudes that were as liberal or more liberal than those of women. In
the same way, the stereotype that women are more religious than men was
qualified by attractiveness. More attractive females were perceived to be less
religious than less attractive ones while attractiveness of males was virtually
unrelated to their perceived religious attitudes.
In all the findings just
described, perceivers were evidently using multiple trait information to form a
judgment that was qualified in comparison to judgments based on single
attributes. It is quite likely that perceivers
also use other target attributes, such as race, ethnicity and apparent
socio-economic status, and that contextual features will determine which
features are more salient. Future
research using procedures similar to those used here should pursue such
questions.
Two instances where
significant interactions were not found should also be noted. First, there was no evidence
of an interaction between target age and attractiveness. As noted earlier,
however, these variables did show a substantial negative correlation and this
may well have affected estimates of the interaction terms.[4] Second, there
was no evidence that any of the target variables examined in the present study
interacted with the gender of the participant. Feingold's (1990) finding,
for example, that physical attractiveness was more important in men's perceptions
of women than in women's perceptions of men was not replicated here. The present
null results are consistent, however, with those of both Eagly, Ashmore, Makhijani,
and Longo (1991) and Jackson, Hunter, & Hodge (1995).
[33]
---------------
[34]
Two aspects of
the methods used in this study are worth noting. First, we avoided drawing
participants' attention to the particular categories of target person under
investigation. Thus, the stereotypes that emerged may be more typical of those
that influence people's judgments in everyday life. A second important aspect
of the current procedure is that not every participant saw the same set of
targets. Each participant saw a set of targets randomly selected from a pool of
100. The use of multiple target sets gives the present results a
generalizability they would not otherwise have.
In conclusion,
the stereotypes about people's attitudes that were evident here appear to be
every bit as pervasive and robust as the well-documented stereotypes concerning
people's traits and abilities. Like other stereotypes, those concerning
attitudes may help people anticipate, organize, and interpret complex social
information. Moreover, it appears that
perceivers use more than single attributes to make inferences about others and
these attributes interact in the inference process. Research should be directed
toward more faithfully simulating the perceiver's rich social environment to
uncover the critical aspects of the inference process.
ENDNOTES
[1] Summing responses
on oppositely worded pairs of items seemed justified on the basis of data, not
presented here, showing that responses to the items within each pair were
negatively correlated.
[2] Participants in
this study also indicated their own attitudes on the issues, sometimes before
the inference process and sometimes after. Analyses of these data, in
conjunction with participants' attitude inferences, revealed several instances
of the false consensus effect (i.e., people expected targets to share their own
attitudes), and a tendency for this effect to be slightly stronger for more
attractive targets.
[3] The three-way interaction
among gender, age, and attractiveness, although theoretically of interest,
could not be reliably assessed in this study because of the relatively small
number of targets and the lack of independence, noted earlier, of two of the
three regressors, age and attractiveness.
[4] We thank an anonymous
reviewer for this suggestion.
[34]
---------------
[35]
REFERENCES
Bargh, J. A., Chen, M., &
Burrows, L. (1996). "Automaticity of social behavior: Direct effects of
trait construct and stereotype activation on action." Journal of
Personality and Social Psychology, 71: 230-244.
Bassili, J. N., & Reil, J. E.
(1981). "On the dominance of the old-age stereotype." Journal of
Gerontology, 36: 682-688.
Devine, P. G. (1989). "Stereotypes and prejudice: Their
automatic and controlled components." Journal of
Personality and Social Psychology, 56: 5-18.
Eagly, A. H.,
Ashmore, R. D., Makhijani, M. G., & Longo, L. C. (1991). "What is
beautiful is good, but ...: A meta-analytic review of research on the physical
attractiveness stereotype." Psychological Bulletin, 110: 109-128.
Eagly, A. H.,
& Chaiken, S. (1993). The psychology of attitudes. New York: Harcourt Brace Jovanovich.
Feingold, A.
(1990). "Gender differences in effects of physical attractiveness on
romantic attraction: A comparison across five research paradigms." Journal
of Personality and Social Psychology, 59: 981-993.
Grant, M. J.,
Button, C. M., Hannah, T. E., & Ross, A. S. (2000, July). "Perceptions
of age and attractiveness by men and women of different ages." Poster
session presented at the meeting of the XXVII International Congress of
Psychology, Stockholm, Sweden.
Grant, M. J., Button, C. M., Ross, A. S., & Hannah,
T. E. (1997). "Accuracy of attitude stereotypes: The case of inferences
based on gender." Canadian Journal of Behavioural Science, 29: 83-91.
[35]
---------------
[36]
Grant, M. J., Ross, A. S., Button, C. M., Hannah, T. E., & Hoskins, R. (2001). "Attitudes and stereotypes about attitudes across the lifespan." Social Behavior and Personality, 29: