Professional Documents
Culture Documents
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR to
digitize, preserve and extend access to Journal of Consumer Research.
http://www.jstor.org
basic claim that sex-typed individuals are more gender-schematic than non-sex-typed individuals (see
Bem 1985 for a review). For example, to test whether
gender-schematic individuals, as identified by the
Bem Sex Role Inventory (BSRI; Bem 1974), spontaneously encode and organize information on the basis of the sex-linked associations that constitute the
gender schema, subjects were asked to remember
words presented in random order (Bem 1981). These
words included animal names, verbs, and articles of
clothing; one-third of each category had been judged
by undergraduatesas masculine (e.g., gorilla, hurling,
trousers),one-third as feminine (e.g., butterfly,blushing, bikini), and one-third as neutral (e.g., ant, stepping, sweater). In their recall of the items, genderschematic subjects were more likely than genderaschematic subjects to cluster the words on the basis
of gender, i.e., after having recalled a masculine item
to recall another masculine item and after having recalled a feminine item to recall another feminine
item. A related study measured subjects' response latencies when asked whether a series of attributes was
self-descriptive (Bem 1981). Gender-schematic subjects were faster than gender-aschematic subjects
when endorsing sex-appropriate attributes (e.g.,
"forceful" and "dominant" for male subjects and
"gentle" and "warm" for female subjects) and when
rejecting sex-inappropriate attributes. Finally, gender-schematic subjectswere more likely than genderaschematic subjects to choose sex-appropriateactivities (e.g., using a drill if they are male and knitting
and ironing if they are female) and to avoid sex-inappropriatebehavior. Moreover, while performing sexinappropriate activities, gender-schematic individuals feel significantlyworse than gender-aschematicindividuals (Bem and Lenney 1976).
Recently, several studies have tested but failed to
confirm
vender schema
theorv
122
?) JOURNAL OF CONSUMER RESEARCH * Vol. 15 * June 1988
mercialthat is extremely salient, and offensive to people with a liberal gender schema" (1986, p. 190)-a
type of schema that does not exist in Bem's theory.
Similarly,Kahle and Homer hypothesized, on the basis of gender schema theory as they claimed, that
"males ought to value food more based on the 'size of
the kill' (e.g., cattle are harderto kill than chickens),
whereas the nutritional aspects of food ought to be
more important for women, who have historically
planned menus and prepared food" (1985, p. 243).
This hypothesis cannot be derived from gender
schema theory, because the theory does not make
claims about how role assignimientsat differenttimes
of human history influence the selection of lunch
items today. In sum, it is not clear whether gender
schema theory has been given a fair test in consumer
research.
This article's three experiments were designed to
test the theory extensively. In designing the three experiments, we tried to adhere as closely as possible to
experimental paradigms employed in previous psychological research that revealed positive results for
gender schema theory.
In Experiment 1, subjects judged jeans advertisements that depicted a man and a woman either in a
sex-role conformist or in a sex-role nonconformist
way. In a second part of the experiment, subjectsprovided appealratingsfor fragranceflacons (two masculine, two feminine). We predicted that gender-schematic subjects, compared to gender-aschematicindividuals, would show a more positive attitude toward
sex-role conformist advertisements than toward sexrole nonconformist advertisements. Additionally,
gender-schematic individuals should show stronger
' Bem (1985) has not excluded the use of other sex role inventories in a test of gender schema theory. However, Kelly, Furman,
and Young (1978) have shown that a large proportion of subjects
may be classified into differentsex-type categories, depending on
the inventoryused.
123
EXPERIMENT 1
Method
Experiment 1 was presented as a study on advertising and packaging, and 151 students from an undergraduatemarketingcourse at Cornell University participated. At the end of the study, subjects completed
the BSRI, which was used to determine subjects' sex
types.3 The sample contained 39 androgynous sub2
Recent researchin social cognition has shown that people are
often better at rememberingschema-consistentinformation (Taylor and Crocker 1981). The results of some studies, however, are
inconsistent with this prediction. For example, based on research
by Hastie and Kumar (1979), we would predict that gender-schematic individualswould be betterat rememberingsex-rolenonconformist rather than sex-role conformist advertisements,but gender-aschematicindividualswould be less likely to show this differential recall. In either case, we would predict differencesbetween
gender-schematicand gender-aschematicindividuals.
3 The BSRI is a paper-and-penciltest that asks subjects to indicate the extent to which each of 60 attributesdescribeshimself or
herself.The BSRIconsists of two 20-item scales that reflectAmerican culture's definition of masculinity (e.g., "self-reliant,""independent") and of femininity (e.g., "affectionate,""tender"), and
20 neutralitems (e.g., "happy,""sincere").In variouspsychometric analyses, Bem (1974) demonstratedthat the masculinity (M)
and femininity(F) scalesareorthogonal,that the BSRIis internally
consistent (coefficientalpha > 0.8) and test-retestreliable(r > 0.9),
and that high scores do not reflect a general tendency to respond
in a socially desirabledirection. Recently, Qualls (1987) provided
evidence forthe BSRI'sreliability(alpha> 0.85) and for its convergent and discriminantvalidity.
Each respondent'scomposite M and F scores were calculated,
124
jects (21 males, 18 females), 37 undifferentiatedsubjects (18 males, 19 females), and 56 traditionally sextyped subjects (26 males, 30 females). According to
Bem (1985), androgynous and undifferentiated subjects are gender-aschematic, and traditionally sextyped subjects are gender-schematic. Because gender
schema theory does not make specific predictions
about cross-sex-typed individuals (Bem 1985, p.
195), they were excluded from subsequent analyses.
In the first part of Experiment 1, subjects saw 12
jeans advertisements. Jeans were selected as a product category because they are popular among college
students and because they qualify as a neutral product (Gentry, Doering, and O'Brien 1978). Each advertisement consisted of a male or female describing
himself or herselfwith a masculine statement (e.g., "I
seek out positions of authority."), a feminine statement (e.g., "I try to get others to notice the way I
dress."), or a neutral statement (e.g., "I admire free,
spontaneous people."). The person in the picture was
always the same male or female. Each subject saw
each person in the advertisement in connection with
two masculine, two feminine, and two neutral statements, resultingin 12 advertisements.The masculine
and feminine statements in the advertisements were
selected from the ANDRO PRF, a sex role inventory
that includes self-descriptive statements of the Personality ResearchForm (PRF); items of the PRF that
are not part of the ANDRO PRF and that had been
found to be non-sex-typed (Berzins, Welling, and
Wetter 1978) were selected as neutral statements. The
order of presentation of the 12 advertisements was
randomized across subjects. Using seven-point Likert-type scales, subjects indicated for each of the 12
advertisementshow much they liked the type of person portrayedin the advertisement, how similar they
thought they were to that type of person, and how
effective the advertisement might be in convincing
people to buy the product. This part of the experiment took the form of a 2 (sex of subject) X 3 (sex
type) X 2 (sex of person in the advertisement) X 3
(type of statement) experimentaldesign. The firsttwo
factorswere between-subjectsvariables,and the latter
factorswere within-subjectsvariables.
The stimuli of the second partof Experiment 1 were
professionally rendered drawings of four fragrance
flacons of differentshapes. Subjects indicated on sevand the M and F medians were computed. The median of the M
scale was 103 and the median of the F scale was 98. Following Bem
( 1985), a median split procedurewas used to divide the sampleinto
four groups:those scoring above the median on both scales were
categorized as androgynous;those scoring below the median on
both scales were categorized as undifferentiated;those scoring
above the median on the sex-congruentscale and below the median
on the sex-incongruentscale were categorizedas traditionallysextyped; those scoring below the median on the sex-congruentscale
and above the median on the sex-incongruentscale were categorized as cross-sex-typed.
Results
Did gender-schematicand gender-aschematicsubjects differ in the way that they responded to the sextyped items of the PRF that were used as self-descriptive statements in the jeans advertisements?To provide such a manipulation check, subjects' responses
on the similarity measure were analyzed by a 2 (sex
of subject) X 3 (sex type) X 2 (sex of person in the
ad) X 3 (type of statement) ANOVA, which revealed a
significant three-way interaction of sex, sex typing,
and type of statement (F(6,270) = 76.92, p < 0.0001).
Although the schematic group, composed of traditionally sex-typed subjects, was sensitive to the sextyping of the masculine and feminine statements, the
gender-aschematicgroup, composed 'of androgynous
and undifferentiatedsubjects, did not respond differently to masculine and feminine statements. Specifically, schematic males identified more easily with
people in the advertisements who used masculine
ratherthan feminine statements (M = 3.99 versus M
= 2.82, Bonferroni-adjustedp < 0.0001), and schematic females identified more easily with people who
used feminine rather than masculine statements (M
= 4.1 1 versus M = 3.46, p < 0.0001). Aschematic subjects' similarity ratings, however, did not differ significantly for the masculine and feminine statements.
Thus, subjects responded differentlyto the sex-typed
statements in the advertisementsas a function of their
sex type: gender-schematic subjects, but not genderaschematic subjects, were sensitive to the sex-typing
of the statements and responded accordingly.
Yet, did gender-schematic subjects have a more
positive attitude toward sex-role conformist advertisements comparedto sex-role nonconformist advertisements, and did gender-aschematicsubjectsjudge
both advertisements similarly? To test this prediction, a composite score was calculated for all sex-role
conformist advertisements (i.e., advertisements in
which the male gave a masculine self-description and
the female gave a feminine self-description), for all
sex-role nonconformist advertisements (i.e., advertisements that showed the man's picture with a feminine statement and the woman's picture with a masculine statement), and for all neutral advertisements
125
was also significant in the subsequent ANOVAs conducted separately on the two dependent variables
(F(2,124) = 46.03,p<0.0001
andF(2,124)
= 25.78,
EXPERIMENT 2
Method
Forty-six subjects (20 males, 26 females) from the
sample of Experiment 1 volunteered to participate in
Experiment 2, which was conducted about a week
later. The responses of cross-sex-typed subjects and
the responsesof three subjectswho failed to follow the
instructions were excluded from subsequentanalyses.
The sample consisted of 17 gender-schematicsubjects
(seven males, 10 females) and 22 gender-aschematic
subjects (nine males, 13 females).
Eighteen popular American magazines were selected as stimuli for the experiment. The magazines
were classified as masculine (e.g., Car & Driver,
Sports Illustrated), feminine (e.g., Cosmopolitan,
Family Circle), or neutral (e.g., Newsweek, Tennis)
on the basis of two criteria:(1) the ratio of male versus
female readership according to the 1986 marketing
report of the Simmons Market Bureau, Inc., and
(2) the magazine pretest ratings of 20 undergraduates
who made their judgments on seven-point scales
ranging from "predominantly read by men" to "predominantly read by women"-the midpoint was labeled "equal number of male and female readership."
Magazines that, according to the Simmons report,
were predominantly (>70 percent) read by men (or
by women) and received extreme ratingsin the pretest
126
were categorized as masculine (or feminine); magazines were categorized as neutral if they had an approximately equal number (between 40 percent and
60 percent) of male and female readershipand if they
received ratings around the midpoint of the scale in
the pretest.
Subjects were run individually. They were first
asked to sit in front of an Apple McIntosh computer
and to read the following instructions presented on
the computer screen: "Imagine you are waiting in a
doctor's office and the following magazines are on a
table." At this point, the names of the magazineswere
presentedon the screen, one afterthe other in random
order, for a duration of 500 ms (interstimulus interval: 500 ms). After having been exposed to 18 magazines, subjects were asked to enter the names of the
magazinesthey could remember.
Next, subjects were shown each stimulus again.
They were asked to press a yes key if they would be
interestedin readingthe magazine and a no key if they
would not. The yes key was always assigned to a subject's dominant hand. Subjectswere unawarethat the
computer programtimed their responses.
Results
Free Recall of Magazines. Gender-schematic and
gender-aschematicsubjects displayed no differences
in terms of the number of ;magazines they recalled
within each magazine-type category. A 2 (sex) X 2
(sex type) MANOVA, conducted on the percentageof
masculine, feminine, and neutral magazines recalled,
revealed a significant sex effect (Wilks's X = 0.651,
F(2,32) = 8.59, p < 0.001); the main effect of sex typing and the interaction of sex and sex typing, predicted by gender schema theory, were not significant
(Fs < 1.00). In subsequent ANOVAs, the sex effect
was significant for the masculine magazines (F( 1,38)
= 9.94, p < 0.01) and for the feminine magazines
(F(1,38) = 16.85, p < 0.001), but not for the neutral
magazines (F < 1.00). Males remembered significantly more masculine magazines than females did
(M = 41.40 percent versus 28.86 percent), and females rememberedsignificantly more feminine magazines than males did (M = 37.70 percent versus
22.41 percent). There was no differencein the relative
number of neutral magazines that subjects recalled.
Clustering in Recall. Although gender-schematic
and gender-aschematicsubjects did not differin their
free recall of sex-typed magazines, they might nonetheless have remembered the magazines in different
orders. Specifically, gender-schematic individuals
should be more likely than gender-aschematic individuals to recall magazines by gender, i.e., to recall in
sequences of masculine and feminine magazines. To
test this prediction, we calculated an index of gender
Magazine Choice. A 2 (sex) X 2 (sex type) X 3 (magazine type) ANOVA was conducted on the number of
magazines subjectsindicated they might be interested
in reading. Magazine type and the interaction of sex
and magazine type were significant (F(2,66) = 16.00,
p < 0.0001 and F(2,66) = 19.84, p < 0.0001). Other
effects, including the interaction of sex type and magazine type, were not significant. Pairwise comparisons using Bonferroni'sprocedurerevealed that male
subjects were more likely to choose masculine and
neutral magazines rather than feminine magazines
(M = 3.73 and M = 3.87 versus M = 1.00; p
<
0.05),
and female subjectswanted to read feminine and neutral magazines rather than masculine magazines (M
= 3.21 and M = 3.83 versus M = 1.88;p < 0.05).
Response Latencies. The previous analysis indicated that sex type did not affect the overall tendency
to make sex-role conformist choices. However,
were gender-schematic subjects faster than genderaschematic subjects in making sex-role conformist
choices (e.g., females saying "no" to a masculine
magazine and "yes" to a feminine magazine) and
slower in making sex-role nonconformist choices
(e.g., females saying "yes" to a masculine magazine
and "no" to a feminine magazine)?Results indicated
that this was not the case. Instead, female subjects
tended to respond faster than male subjects. In the
two 2 (sex) X 2 (sex type) ANOVAs conducted on sexrole conformist and sex-role nonconformist responses, this sex effect was the only effect that approached significance (p = 0.12 and p = 0.07).
EXPERIMENT 3
Method
Out of 312 undergraduatesfrom an introductory
psychology class who completed the BSRI, 60 students were recruited to participate in Experiment 3
on the basis of their BSRI scores. Thirty subjects (15
males, 15 females) were gender-schematic and 30
subjects (15 males, 15 females) were gender-aschematic. The responses of one male subject had to be
eliminated because he did not complete the memory
test.
4 The index is based on a simple formulaby Shuell (1969), which
has been frequentlyused in researchon categoryclustering.
Results
Subjects' recall for nongender-relatedcharacteristics in each type of advertisement was assessed to test
whether gender-schematic and gender-aschematic
subjects differedin their overall memory capacity. A
MANOVA and three 2 (sex of subject) X 2 (sex type)
X 2 (sex of reader) X 3 (type of magazine) ANOVAs
revealedno sex-type effects.
However, gender-schematic and gender-aschematic subjects also did not differ in their memory for
gender-related information. In the 2 X 2 X 2 X 3
ANOVA conducted on the relative percentage of
male and female readers that subjects correctly remembered, sex type was not significant as a main
effect or in any interaction (all ps > 0.25). Instead,
the main effect of sex of reader (F(1,55) = 16.20, p
127
versus M = 72.2 percent); a female was best remembered when she read a feminine magazine ratherthan
a masculine or neutral magazine (M = 77.2 percent
versus M = 70 percent and M = 71.6 percent). That
GENERAL DISCUSSION
To examine the relevance of gender schema theory
to the study of consumer behavior, we conducted
three experiments that tested predictions related to
advertising, packaging, consumer choice, and consumer memory. The experiments used three different
product categories(jeans, fragrances,and magazines)
and several dependent measures (attitude measures,
preference judgments, memory measures, choice
measures, and response latencies). We thus provided
a good opportunity for positive results to emerge;
however, we did not find any support for gender
schema theory.5
Although we took great care to minimize methodological shortcomings, our null effects may be partly
due to our sex-type measure or to the stimuli used in
the studies. Specifically, the BSRI and Bem's categorization procedure have recently been criticized
(Spence 1984). Paulhus (1987) has arguedthat the orthogonal M and F scales of the BSRI are inappropriate measures of the concept of gender schematism.
Another possible explanation for some of our null results may be that in some of our experiments we employed highly sex-typed stimuli (e.g., the flacons in
Experiment 1 and the magazines in Experiments 2
and 3) that may be poorly suited to detect subtle sextype differences.However, it is unclear why it should
be easier in principle to obtain sex effects ratherthan
sex-type effects. For example, some of the stimuli that
5 BesidesBem'sgenderschematheory, Markus'sself-schematheory also makes predictionsabout individual differencesin the processingof gender-relatedinformation(Markuset al. 1982). The two
theoriesdifferabout who should be consideredas gender-schematic
and gender-aschematic.Self-schematheoryproposesthat androgynous individuals are gender-bischematic(i.e., they process both
masculineand feminine informationschematically),and traditionally sex-typedsubjectsare either masculine or feminine schematic.
The only truly gender-aschematicindividuals, according to selfschematheory, are the undifferentiated.Additionalanalysesof our
data, however,also revealedno supportfor self-schematheory.
128
REFERENCES
Allison, Neil K., Linda L. Golden, Gary M. Millet, and
Donna Coogan (1980), "Sex-typed Product Images:
The Effects of Sex, Sex Role Self Concept, and MeasurementImplications,"in Advancesin ConsumerResearch, Vol. 7, ed. JerryC. Olson, Ann Arbor, MI: Association for ConsumerResearch,604-609.
Bem, SandraL. (1974), "The Measurementof Psychological Androgyny," Journal of Consulting and Clinical
Psychology,42 (April), 155-162.
(1981), "Gender Schema Theory: A Cognitive Account of Sex Typing,"PsychologicalReview, 88 (July),
354-364.
(1985), "Androgynyand Gender Schema Theory:A
Conceptual and Empirical Integration," in Nebraska
Symposiumon Motivation:Psychologyand Gender,ed.
Theo B. Sonderegger,Lincoln, NE: University of NebraskaPress, 179-226.
and Ellen Lenney (1976), "Sex Typing and the
Avoidance of Cross-SexBehavior,"Journal of Personality and Social Psychology, 33 (January),48-54.
Berzins, Juris I., MarthaA. Welling, and Robert E. Wetter
(1978), "A New Measureof Psychological Androgyny
Based on the Personality Research Form," Journal of
Consulting and Clinical Psychology, 46 (February),
126-138.
Coughlin, Maureenand Paul J. O'Connor(1985), "Gender
Role Portrayalsin Advertising:An Individual Differences Analysis," in Advances in ConsumerResearch,
Vol. 12, eds. Elizabeth C. Hirschman and Morris B.
Holbrook, Provo, UT: Association for Consumer Research, 238-241.
Gentry,JamesW. and MildredDoering (1977), "Masculin-