British Journal o f Social Psychology (200 3), 42, 257–280
©2003 The British Psychological Society
Thinking about gender types: Cognitive
organization of female and male types
* and Richard D. Ashmore
University of Nijmegen, The Netherlands
Rutgers University, USA
We examined the content and dimensional structure of a large and representative
sample of gender types. In Study 1, using an open-ended procedure, participants
generated 306 different labels for female types (e.g. housewife, feminist, femme fatale,
secretary, slob) and 310 for male types (e.g. workaholic, family man, sissy, womanizer,
labourer). In Study 2A, a multidimensional con guration of 229 of these male and
female types was derived from a free sorting task among a new set of participants. In
Study 2B, a subset of types was judged on several dimensions of meaning, which were
then tted into the con guration of types. The most important dimensions in
describing the structure of gender types were: young–old, masculine–feminine and
traditional–modern. The masculine–feminine dimension showed that the male and
female types were largely separated from each other; within each gender category,
the types were ordered by their position on the masculine–feminine dimension.
Several other aspects of current thinking about men and women are discussed.
Over the past decades, stereotype researchers have directed considerable attention to
perceptions of subcategories within the overarching social categories of race (Devine
& Baker, 1991), age (Brewer, Dull, & Lui, 1981), physical attractiveness (Ashmore,
Solomon, & Longo, 1996), and gender. Studies on gender types (Ashmore, Del Boca, &
Titus, 1984; Carpenter, 1994; Clifton, McGrath, & Wick, 1976; Deaux, Winton,
Crowley, & Lewis, 1985; Eckes, 1994a, 1994b; Edward, 1992; England, 1988, 1992;
England & Hyland, 1987; England & Manko, 1991; Haddock & Zanna, 1994; Holland &
Davidson, 1983; Holland & Skinner, 1987; Noseworthy, 1982; Noseworthy & Lott,
1984; Six & Eckes, 1991) have primarily addressed two questions: (1) What are the
female or male types that people distinguish, and what characteristics are associated
with these types? (2) What is the cognitive structure underlying perceptions of a set of
male or female types?
Regarding the first question, there is considerable agreement on the major types and
their stereotypic characteristics. The female types that are consistently named in
response to open-ended probes are: housewife/homemaker, career woman, woman
athlete and sex object (e.g. Clifton et al., 1976; Deaux et al., 1985, Study 3; Glick,
Diebold, Bailey-Werner, & Zhu, 1997, Study 1; Noseworthy, 1982). The housewife is
*Requests for reprints should be addressed to: Roos Vonk, Department of Social Psychology, PO Box 9104, 6500 HE
Nijmegen, The Netherlands (e-mail: vonk@ psych.kun.nl).
ascribed typically feminine attributes (e.g. nurturant, gentle), whereas career woman
and woman athlete are associated with traditionally masculine characteristics (e.g.
active, ambitious; England & Hyland, 1987). Among the male types, a similar distinc-
tion can be found. Businessman and macho man are masculine types that are fre-
quently mentioned, whereas homosexual,wimp and family man are familiar types
that have more feminine characteristics. This consensus regarding the contents of
gender types appears to exist across different nations (see, e.g., Chia, Moore, Lam,
Chuang, & Cheng, 1994; Eckes, 1994a; Kirchler, 1992) and regardless of subject sex
(Coats & Smith, 1999; Deaux & Lewis, 1984; Deaux et al., 1985).
The cognitive organization of gender types has been examined by means of
dimensional or clustering analyses of the perceived similarities among a selected set of
types (Ashmore et al., 1984; Eckes, 1994a, 1994b; Holland & Davidson, 1983; Six &
Eckes, 1991). For both male and female types, three dimensions of organization are
commonly found. The first is Evaluation (anchored by e.g. sweetheart vs. bitch for
females, and businessman vs. nerd for males).
A second dimension can be labelled as sexual–nonsexual (Schultz, 1975; see also
Glick et al., 1997), but it differs for male and female types. For male types, this
dimension is basically orthogonal to Evaluation and can be specified as sexual activity/
interest, with types such as Don Juan and playboy at one end, and scholar and
bureaucrat at the other. For female types, the literature reports both a negative
sexuality dimension in terms of respectability (e.g. whore vs. mother; Ashmore et al.,
1984; Six & Eckes, 1991) and a positive one in terms of sexual attractiveness (e.g. fox
vs. dog; Holland & Davidson, 1983; babe or chick vs. dyke or butch; Glick et al., 1997).
A third dimension can be described as traditional–non-traditional (see Glick & Fiske,
1996; Glick et al., 1997). For females, nurturing and submissive roles such as mother
and housewife are contrasted with modern and assertive types such as career woman
and feminist. For male types, businessman and many other occupational roles are in
line with traditional gender roles, whereas types such as wimp and homosexual reflect
deviations from the strong and dominant aspects of the male role.
The traditional–non-traditional distinction corresponds with the idea that
subcategorization of a group can serve two functions. First, it helps to preserve the
stereotype of a group by subtyping stereotype-inconsistent members (Weber &
Crocker, 1983), that is, by setting these members aside as atypical (see Johnston &
Hewstone, 1992). For example, the male type wimp may derive from experiences with
males who contradict the stereotype of men as strong and agentic. This refencing
mechanism serves cognitive consistency motives, and accounts for non-traditional
types such as career woman and homosexual (see Deaux & Lewis, 1984). Second,
subcategorization can serve utilitarian and sufficiently accurate perception functions
(Fiske, 1992; cf. Rosch, 1973) by refining the overarching category, accounting for the
variability with which group members manifest characteristics of the overall group
stereotype. This mechanism accounts for the traditional, stereotype-consistent types
such as macho man,businessman,Don Juan and adventurer, which all reflect
different aspects of the male stereotype (Brannon, 1976). In addition, refining may
serve to accommodate group members who partially deviate from the overall stereo-
type but who are not perceived as atypical of the group because they also possess
stereotypic characteristics. For instance, the princess type is associated with domi-
nance but also with stereotypically feminine attributes (e.g. nagging, complaining). In
sum, referencing leads to types that are mentally excluded from the overarching group
so that the global stereotype can be maintained, whereas refining produces types that
258 Roos Vonk and Richard D. Ashmore
are included as specifications or differentiations of the superordinate stereotype (see
Maurer, Park, & Rothbart, 1995, who make a similar distinction between subtyping and
subgrouping; see also Richards & Hewstone, 2001).
Given that most of the non-traditional gender types are psychologically excluded
from the overall stereotype and are associated with counter-stereotypic attributes (e.g.
career woman and businesswoman are associated with masculine attributes, such as
confidence and ambition), the question can be raised as to what extent these types are
perceived as similar to traditional types of the other gender category. For instance, the
career woman type may be psychologically closer to businessman than to mother. In
all previous work, male and female types have been considered separately. As a
consequence, it is not known how male and female types are perceived relative to one
another. The primary goal of our studies is to address this question by simultaneously
examining the pattern of perceived similarities among male and female types.
One possibility is that sex is a crucial variable in the cognitive structure of gender
types, so that male types and female types are organized separately. That is, men and
women are seen as opposite sexes, and the difference between them is categorical.
According to this possibility, for instance, businesswoman is more similar to house-
wife than to businessman simply because the latter has a different sex. In this case,
one of the core dimensions of a structural representation of subtypes will be the
contrast between male and female types. If people indeed think of gender types in this
way, the results of our study would legitimize the common practice of examining male
and female types separately.
There is, however, another possibility. In thinking about gender types, people may
focus on the characteristics of each type, rather than on sex. Indeed, we assume that
the very reason why these types have emerged is that people observe others’ charac-
teristics and notice the large differences between members of the same sex. In this
case, the traditional male types, such as macho man or businessman may be seen as
similar to non-traditional female types, such as dyke or businesswoman, and vice
versa. As a consequence of such cross-sex similarities, the outcome of our study would
be that male and female types are entirely mingled and are organized predominantly
by their characteristics, such as profession or personality. Instead of a male–female
dimension, the important dimension in this case is masculine–feminine in terms of
characteristics, with both macho man and dyke at the masculine end, and with sissy
and sweetheart at the feminine end.
Evidently, there are intermediate possibilities between these two. For instance,
people may perceive male versus female types as different and dissimilar on some
dimensions (e.g. sexuality), but not others (e.g. evaluation); or the perceived similari-
ties between types may reflect a compromise between similarity in terms of sex
(male–female) versus other characteristics (masculine–feminine).
In examining the cognitive structure of gender types, it is important that the types
under consideration are a sample of the gender types that people use in their everyday
thinking about men and women. Regardless of the particular technique used, the
output of the analysis is strongly affected by the input. For instance, a sexuality
dimension will not emerge in the results if one examines only ‘sexless’ types (or only
‘sexy’ types). A limitation of earlier studies is that most have not incorporated a
comprehensive set of types but, instead, a selected and limited sample. This means
that any conclusions about the relations among types within a gender category are
restricted by the particular set of types under consideration. For instance, Deaux et al.
(1985) concluded that male types were more homogeneous than female types, but
Thinking about gender types 259
they examined only four male types which were all traditional (blue-collar working
man,athletic man,businessman and macho man), whereas their female types were
two traditional (housewife,sexy woman) and two non-traditional types (athletic
In the present studies, we consider an extensive set of male and female types that is
representative of the types that people in our culture use in thinking about men and
women. In addition, the sample of participants in our studies contains not only college
students but also other adults with diverse occupations. Hence, the results are not
restricted to the college-student subculture and are representative in terms of subject
sample as well.
In sum, the goal of our study is to examine the content and the dimensional
structure of gender types, using a comprehensive set of types that reflects current
thinking in Western societies about men and women. Three studies were conducted.
First, participants generated labels for male and female types and attributes associated
with these types. Second, another group of participants performed a free sorting task
of both the male and female types generated in the first study. From the sorting data, a
multidimensional representation of all male and female types was derived. Finally, to
examine the dimensional structure in this representation, judges rated a selected set of
types on 10 dimensions (such as evaluation and sexuality) that were used to interpret
the pattern of similarities among the types. In this paper, we describe both the content
of the gender types (based on Study 1) and their structural organization in terms of
dimensions (based on Studies 2A and 2B).
Participants were 46 males and 48 females living in The Netherlands. They participated
voluntarily. The sample was quite diverse. About 25% of the participants were college
students; the others had various jobs (including homekeeper), were unemployed,
or were students at high schools or schools for professional education. Male and
female participants were evenly distributed across two age groups (17–34 and 35–53)
and across student/non-student groups. The interview, which lasted about 1 hr,
was conducted in each participant’s home. There were seven interviewers who each
interviewed about 50% male and 50% female participants.
The interviewer began by showing the participant a drawing of a hierarchical
classification of animals, with animals represented as the overarching category,
several kinds of animals at the level below that (e.g. mammals,fish,birds), and several
species at the subordinate level (e.g. dogs,horses,rabbits as subcategories of
mammals). Next, a similar drawing was shown with people as the overarching con-
cept and males and females at the level below. The participant was then asked to
complete the classification by naming as many labels as possible describing subgroups
of males and females. Half of the participants listed male subgroups first and then
female subgroups; the other half listed females first. This order variation was counter-
balanced across participant sex, age group and interviewer. The interviewer recorded
verbatim the gender types named.
After the male and female subgroups had all been recorded, the interviewer went
back through each type named and asked the participant to describe the characteris-
tics of each subgroup. These, too, were recorded by the interviewer and were used
260 Roos Vonk and Richard D. Ashmore
later to allow grouping of different types into larger categories (see Results section).
Finally, and for reasons irrelevant to the present purposes, participants rated each
group on a positive–negative scale and indicated which groups they themselves were a
member of (see Vonk & Olde-Monnikhof, 1998).
Results and discussion
Number of types and attributes
The mean number of male types listed by a participant was 8.65; for female types, the
mean was 8.52. Male participants mentioned more male than female subgroups (8.65
vs. 7.74), whereas females mentioned more female than male subgroups, 9.14 vs. 8.64;
Participant Sex×Type Sex interaction, F(1, 90) = 4.15, p< .05. This effect converges
with results obtained by Park, Ryan, and Judd (1992), who found that people distin-
guish more subgroups within their own university group (e.g. business vs. engineering
majors). As can be seen from the means, there was also a (non-significant) tendency for
female participants to list more types overall (see Deaux et al., 1985).
The mean number of attributes listed per type was 6.80. Correcting for the differ-
ential number of types mentioned, there were no differences between males and
females in the number of attributes mentioned for male and female types.
Content of the types
A list was made of all the types mentioned, after some preliminary groupings of
different labels that apparently referred to the same type (e.g. always-ill type/
chronically-ill type and smart woman/intelligent woman/woman with brains).
This resulted in 306 different female types and 310 male types. Subsequently, many of
these types were further grouped because of their high similarity or because one was a
more specific form of the other. For instance, we collapsed businessman/rich
businessman/successful businessman and wife of rich husband/wife of someone
important/woman who has status through her husband. This reduced the list to 206
female and 198 male types.
Table 1 provides a list of the types mentioned by at least four participants. Note that
the labels used are the best possible translations that capture the meaning of each type,
and that the number of different labels used per type does not always correspond with
the number of labels used by our Dutch participants: as will be discussed, some words
cannot be translated properly. The types are grouped as much as possible by the
categories we found in the existing literature. In many respects, our results are similar
to those obtained in American (e.g. Ashmore et al., 1984; Coats & Smith, 1999) and
German (Eckes, 1994a, 1994b; Six & Eckes, 1991) studies. For both male and female
types, many participants mentioned the same traditional (e.g. housewife,mother and
businessman,macho) and non-traditional types (e.g. career woman,businesswoman
and wimp,homosexual) as in the studies summarized in the Introduction. In
Table 1, the non-traditional types are distinguished into a predominantly positive
category (independent/liberated for females, nurturant for males) and a category
consisting of neutral or negative non-traditional types (e.g. bitch for females, wimp for
As in other studies, we also obtained sexual types for both males and females. For
females, we found the same distinction between sexually attractive (e.g. sexy woman)
and disrespectable female types (e.g. tramp). Note that the sexually promiscuous types
Thinking about gender types 261
Table 1. Female and male types mentioned by at least four respondents, and the number of times
the type was mentioned by female, male, and all respondents
Females Males Total
housewife 27 28 55
(house-)mother 20 15 35
submissive/obedient woman 5 6 11
granny/grandmother 4 2 6
wife-of-rich-husband/-businessman 3 3 6
dependent woman 4 1 5
wall ower 4 1 5
nurturant woman 2 2 4
woman athlete 14 8 22
snappish/bitchy woman, bitch 6 6 12
dominant/bossy woman 2 7 9
masculine woman/virago 4 4 8
lesbian 6 1 7
career woman 21 20 41
businesswoman 14 16 30
feminist 9 6 15
working woman 6 5 11
intelligent/smart woman/brains 2 7 9
liberated woman 2 6 8
independent/autonomous woman 4 0 4
con dent woman 2 2 4
ambitious woman 3 1 4
tramp/bimbo/slut/whore 7 12 19
femme fat ale/ irt 1 6 7
dumb blonde 5 4 6
sexy woman/sex bomb 2 3 5
beautiful/attractive woman 2 2 4
Traditional female jobs
secretary 1 5 6
cleaning lady/maid 1 3 4
nurse 0 4 4
snob/snobbish woman 14 8 22
cold/aloof/stuck-up/arrogant woman 3 6 9
prude/prissy girl 4 1 5
sorority type 3 1 4
princess 2 2 4
common/vulgar woman 8 6 14
trashy woman 5 3 8
slob/frump 3 1 4
262 Roos Vonk and Richard D. Ashmore
Table 1. Continued
Females Males Total
student 9 4 13
woman artist/artistic woman 8 3 11
hippie 6 2 8
socially conscious woman 4 2 6
young woman 2 2 4
business man 27 30 57
male chauvinist pig/macho man 30 17 47
male athlete/sportsman 18 15 33
ordinary man 8 7 15
yuppie 8 2 10
aggressive man 3 4 9
he-man/body builder 3 4 7
dominant man 3 2 5
intelligent/smart man 2 2 4
workaholic 1 3 4
(house-)father, family man 10 8 18
houseman/homemaker 8 9 17
sweet man 3 3 6
sensitive man 4 1 5
romantic man 5 0 5
softy/mother’s boy/sissy 8 11 19
homosexual/gay/fag 10 8 18
sucker/wimp/goofy guy/geek 6 1 7
feminine man 4 0 4
womanizer/charmer 4 3 7
pub crawler 4 2 6
adventurer 2 3 5
(eternal) bachelor 1 3 4
Traditional male jobs
labourer/blue collar worker 7 2 9
manager 6 1 7
civil servant 1 6 7
professor/scientist 3 3 6
construction worker 2 3 5
surgeon/physician 1 4 5
(car) salesman 2 2 4
farmer 4 0 4
snobbish/arrogant conservative 7 5 12
fraternity type 7 1 8
Thinking about gender types 263
tramp/bimbo/slut/whore and femme fatale/flirt were mentioned more frequently by
male than by female respondents (18 vs. 8). The sexual male type (womanizer/
charmer) was not mentioned very frequently (by 7 respondents). Thus, male types
referring to sexuality were less prominent than female types. We did obtain several
male types that have relationship non-commitment in common with the Don Juan
type (adventurer,bachelor and pub crawler/pub loafer), which is why we grouped
them into the same category.
In addition to the categories that are consistent with results from existing studies,
Table 1 also includes some additional categories. First, participants generated a wide
variety of occupational types, most of which were traditional (e.g. secretary and nurse
for females, manager and construction worker for males). The table presents only a
limited sample of this variety because it is restricted to occupations mentioned by at
least four participants. In total, 50 participants (19 females, 31 males) mentioned
female occupation types, and 115 participants (60 females, 55 males) mentioned
male occupations (including businesswoman and businessman). Thus, occupation
types were listed more frequently for males. Note also that the unemployed type we
obtained for males was never mentioned for females, suggesting that the ‘default’
option for males is that they have a job, whereas this is not the case for females.
An interesting addition is the snobbish category that emerged for male and
especially female types. This category includes familiar types such as princess and
prude (included because the Dutch noun for prude—‘tutje’ or ‘tut’—connotes
snobbishness in addition to sexual prudishness), but it also includes a wide variety of
labels that cannot be translated. For instance, in Dutch there are many ways to say that
a woman is snobbish, most of which involve the word ‘kak’ which can be used in the
form of a noun (e.g. ‘kouwe kak’, ‘kakker’), an adverb (e.g. ‘kakmadam ’, ‘kakwijf ’),
or an adjective (‘bekakte vrouw’, ‘bekakt type’). All of these occurred in our data.
Another example of this category concerns the fraternity type frat which emerged for
females (in Holland, all student associations are mixed, and there is no separate word
for male and female associations), but especially males. This type, mostly referred to by
the Dutch noun ‘bal’ or the adjective ‘ballerig’, is not necessarily a student, although it
includes a prominent subcategory (‘koor-bal’) of students belonging to fraternities that
Table 1. Continued
Females Males Total
anti-social man 2 6 8
(dirty) old man 5 2 7
criminal 3 5 5
bum 2 2 4
intellectual 3 9 12
granola/Birckenstocks type 8 3 11
student 8 2 10
artistic man/artist 5 3 8
unemployed man 2 4 6
socially conscious man 3 3 6
264 Roos Vonk and Richard D. Ashmore
engage in drinking matches and horseplay. Essentially, however, the type refers to
people, mostly males, who are arrogant, dominant, loud, politically conservative, wear
ties, and speak in a snobbish way.
Note that the snob type was also obtained by Glick et al. (1997, p. 1326), but it
appears to be more prominent and elaborate in this Dutch sample. In addition, it seems
that other languages lack the variety of words describing these types, so we may
conclude that the prominence and elaborateness of this category reflect a peculiarity
of the Dutch.
The types presented in Table 1 show evidence of the two mechanisms discussed
earlier, refining and refencing. Types such as intellectual,workaholic,labourer and
leader are specifications of the male stereotype, whereas types such as mother’s boy
and househusband contradict this stereotype. Similarly, the female types mother,
sex bomb,secretary and nurse specify variations of the female stereotype, whereas
businesswoman and bitch contradict it. Obviously, the stereotype-inconsistent male
types refer to males with typically feminine traits, whereas stereotype-inconsistent
female types refer to women with masculine traits. Thus, it is possible that these types
are perceived as relatively similar to traditional types of the other gender category.
For instance, family man might be seen as more similar to mother than to male
workaholic;businesswoman might be seen as more similar to workaholic than to
mother. As noted earlier, the existing literature does not shed any light on this
issue, because it has assessed the perceived relations among male and female types
separately. This issue was examined in Studies 2A and 2B. Study 2A was modelled after
Ashmore et al. (1984), Holland and Davidson (1983), and Six and Eckes (1991):
Participants sorted gender type labels on the basis of similarity but, deviating from
these other studies, they sorted labels of both male and female types. This resulted in a
spatial representation of all gender types. In Study 2B, a new group of participants
rated the types on several dimensions of meaning, so that the spatial representation
from Study 2A could be interpreted. Because the two studies are both required to draw
conclusions about the organization of gender types, we will first describe the method
of both studies and then present the results.
After Study 1, the list of types contained about 400 labels. To make the sorting task
manageable, the list was further reduced in two ways. First, we deleted all types that
had been mentioned by only one participant in Study 1. Second, we deleted most of
the types with labels that used a trait adjective to describe men or women, such as
‘nice (wo)man’, ‘honest (wo)man’, ‘rude (wo)man’. The reason was that, obviously,
any adjective can be paired with man or woman to form a description and, indeed,
hundreds of adjectives were used in Study 1, but most of them did not seem to refer to
cultural gender types. Therefore, type labels in the form of an adjective attached to the
word ‘(wo)man’ were deleted, unless two independent judges agreed that it was
Other example s of frequently mentioned types that we did not nd in the literature are: the artist type that emerged for
both ma les and females; the vulgar female type which is, in many ways, the opposite of snobbish (but negative as well,
e.g. slob), but also the opposite of sexually attractive; and the antisocial male type that parallels this category (e.g. bum);
these types were grou ped for both males and females because they a re all associated with lower socio-economic statu s.
Thinking about gender types 265
a culturally existing type (such as androgynous man,modern man,independent
After this reduction, we were left with 229 types; 115 for females and 114 for males.
The labels for these types were typed onto cards, which were presented to 15 partici-
pants; 8 women and 7 men. As in Study 1, these participants differed in age and
background, and they participated in the study individually in their own homes. They
were given all of the cards and were asked to sort them into piles, based on which
types belong together. They could create as many piles as they wanted to.
After the participant had completed the sorting task, the interviewer scored the
result by assigning a numeric code to each type: types that were sorted into the same
group were given the same code. Note that this ‘pile number’ is entirely arbitrary: what
matters in the analysis is whether any two cards are sorted into the same pile (i.e. have
been assigned the same number) or different piles; this information is contained in the
pile codes. Hence, the data are categorical.
The data formed a 229 (types) × 15 (participants) matrix, containing the codes that had
been assigned to each type on the basis of the sorting task. These data were analysed
by means of HOMALS
(SPSS procedure Categories; see Gifi, 1990; Heiser & Meulman,
1984), which stands for homogeneity analysis by alternating least squares. Essentially,
this procedure is similar to principal-components analysis, but it has been developed
specifically for nominal data. It analyses categorical data with the goal of grouping the
cases into homogeneous categories. This analysis was used because it is particularly
suitable for data sets with a large number of stimuli such as this one (Slooff & Van der
Kloot, 1985). In the present application of HOMALS (see Slooff & Van der Kloot, 1985),
the participants are treated as the variables, and the subtypes as the objects of obser-
vation (i.e. cases in SPSS-X). Each group of types sorted into the same pile by a
participant constitutes a category of objects. The basic goal of HOMALS is to obtain
scale values for the objects (the object scores) on a limited number of spatial dimen-
sions, such that objects that belong to different categories can be discriminated by
their object scores (i.e. are located far apart from each other in space). The solution is
found by maximizing, per dimension, the between-category sums of squares relative to
the total sums of squares.
The analysis thus produces a multidimensional representation of the stimuli. When
many participants have sorted two subtypes into the same pile (i.e. they are generally
perceived as similar to each other), HOMALS positions these types close together in
space (in the extreme case, if two stimuli are sorted into the same category by all
participants, they will obtain the same object scores on each spatial dimension); when
two stimuli are sorted into different categories by many participants, they will be
positioned at a large distance from each other.
One participant created a t otal of 81 piles; the others created between 22 and 56 piles, with a mean of 37.
Although the number of participants in this study is relatively small, it is not too small for the HOMALS analysis in w hich
‘a large number of probabilities are estimated simultaneously, and these in turn are weighted to arrive at a solution (i.e.
each coordinate value is a weighted average of 229 probabilities). HOMALS can be seen as an analysis on the Burt table,
which is the table consisting of 15 #15 blocks, for each pair of judges, and each block represents to wha t extent pile iof
judge A contains the same stimuli as pile jof judge B (in a regular HOMALS these are all cross-tables between the
variable s). Each of these blocks consists of N=229 observations, which is more than sufficien t for the stability of the
solution’ (W. J. Heiser, personal communication, September 2001).
266 Roos Vonk and Richard D. Ashmore
As with factor analysis and multidimensional scaling, a higher number of spatial
dimensions yields a better fit between the spatial solution and the data, but obviously
the goal is to reduce the entire pattern of data to a limited number of spatial dimen-
sions. The fit of the dimensional solution with the data is expressed in eigenvalues for
each dimension. The eigenvalue of a dimension is the ratio of the mean average (across
variables, i.e. participants) of the between-category sum of squares to the total sum of
For dimensions 1 to 10, eigenvalues in our analysis were: .72, .70, .69, .65,
.64, .63, .62, .59, .59 and .58. The first (albeit small) ‘elbow’ in this sequence is
between the three- and four-dimensional solution, suggesting that a 3-dimensional
solution is most parsimonious.
On the basis of this elbow criterion as well as interpret-
ability (see next section; it was not possible to find a meaningful interpretation of the
fourth spatial dimension), the 3-dimensional solution was selected as optimal.
We will later discuss our interpretation of the solution. At this point, it is important
to note that in HOMALS, as in multidimensional scaling, the spatial dimensions are not
fixed or meaningful; the solution can be rotated or reflected, as long as the mutual
distances between the objects (reflecting perceived similarities) remain the same. This
implies that dimensions of meaning may be identified in any direction, e.g. horizontal
and vertical, but also diagonal. In addition, when different dimensions are correlated,
one may identify more dimensions of meaning than there are spatial dimensions. (For
instance, a solution with two spatial dimensions may reveal at least four dimensions of
meaning, e.g. along both axes and along both diagonals.)
To interpret the dimensions of meaning underlying the configuration of types, we
relied on both a visual inspection of the HOMALS solution as well as on external data,
which we collected in Study 2B.
In this study, the gender types were judged directly on several bipolar dimensions of
meaning, called properties here, in order to examine whether these properties explain
the pattern of similarities among the types in the HOMALS solution. We selected 10
properties, mostly on the basis of existing literature, but also on the basis of our initial
interpretation. We included, first of all, three dimensions that have previously been
obtained in studies on gender subtypes (Ashmore et al., 1984; Eckes, 1994a, 1994b;
Six & Eckes, 1991) namely, (1) feminine–masculine, (2) modern–traditional, and
(3) non-sexual–sexual). In addition, because the present configuration of types
pertains to both males and females, we also included dimensions that are relevant in
thinking about people in general, namely, the three semantic-differential dimensions,
(4) Evaluation, (5) Potency and (6) Activity (Osgood, Suci, & Tannenbaum, 1957) as
well as (7) the incompetent–competent dimension (Rosenberg, 1977), which is often
used in thinking about minority groups (Fiske, Xu, Cuddy, & Glick, 1999; Glick &
Unlike the eigenvalues in factor analysis, there is no absolu te criterion (i.e. 1) where the eigenvalue of a dimension is
considered insufficient, but gen erally eigenvalues higher than .70 are seen as an indication that the cate gories can be
discriminat ed very well by the dimension (Slooff & Van der Kloot, 1985). Becau se the data are nominal, the percentage
of varian ce explai ned by a dimension cannot be computed.
Six types emerged as outliers in the initial analysis, namely, foreign woman, foreign man, black woman, black man,
immigran t woman and immigrant man. The majority of particip ants had sorted these types into the same pile and h ad
not included other types in this pile, so they were positioned far away from th e others. As a result, the dimensional
solution was dominated by the contrast b etween these six types and the other types, so that dissimilarities among these
other types were obscur ed. Therefore, th ese six types were deleted from the analyses re ported here.
Thinking about gender types 267
Fiske, 1998). We did not include Likeability because it is largely redundant with
Evaluation. Finally, we added three dimensions that seemed relevant on the basis of
our own visual inspection of the solution, namely, (8) Age, (9) Settled and (10) Choice.
The last two dimensions require clarification here because they are novel. We
included the Settled dimension because the types appeared to be ordered by the
extent to which the members are settled into their role in society (e.g. grandad,
mother,housewife) versus free to do what they feel like (e.g. eternal bachelor,
adventurer,party-crasher). The Choice dimension refers to whether one’s role is
consciously chosen (e.g. businesswoman,feminist,artist) or is something that one
has moved into without thinking about it too much—for instance, because of tradition
or because one does not have any opportunities to do something else (e.g. farmer,
labourer,dustman) or because one inevitably has to go through this role at some
point (e.g. first-year student,schoolboy). Thus, these dimensions were added because
our initial visual interpretation suggested that they were relevant).
Judges were 321 college students who participated in the study as part of a larger
series of studies. None of them had participated in Study 1 or Study 2A. They were
asked to judge the gender types on one of the 10 dimensions above. The number of
types was reduced to a representative subset of 34 (17 male and 17 female types). To
accomplish this, we divided the three-dimensional representation from Study 2B into
16 smaller sub-spaces of equal size, and we selected one or two types from each
sub-space (depending on how many types there were in that part of the overall
configuration; thus, we selected more types from the central part of the figure, which
is more ‘crowded’, than from the side parts).
Judges were randomly assigned to one of the 10 properties, so that there
were 31–33 participants for each property. Males (69) and females (252) were evenly
distributed across the properties. The judgment task was administered by means of
computers in individual cubicles. Judges were given a brief explanation of the meaning
of the property and were asked to judge each of the 34 types on that property, on a
7-point bipolar scale, by typing in their response. In all conditions, they were told to
think of the typical members of each category and judge their characteristics in
general. The order of presentation of the types was randomized.
For each combination of a gender type and a property, we computed the mean rating
across judges. These means were used to externally fit the properties into the HOMALS
solution from Study 2A by means of linear regression analyses (Rosenberg, Nelson, &
Vivekananthan, 1968; cf. Heiser & De Leeuw, 1981; see also Vonk, 1993) in which
the gender types operated as cases; thus, there were 34 cases in each analysis. The
properties (i.e. the mean ratings derived from Study 2B) were entered as the
dependent variable. The independent variables were the coordinate values of the types
as located in the HOMALS solution, i.e. the object scores. For each property, the
regression analysis locates an axis such that the linear correlation between the ratings
from Study 2B and the values of the objects along this axis is maximized. A separate
analysis was conducted for each of the 10 properties. For each analysis, the multiple
268 Roos Vonk and Richard D. Ashmore
Rvalue provides an estimate of the degree to which a property actually corresponds
with a direction in the subtypes space. That is, Rindicates whether the property
adequately explains the location of the types in the multidimensional configuration.
This analysis accomplishes three goals, all of which have to do with facilitating the
interpretation of the configuration of types. First, the multiple Rs allow us to examine
in a quantitative way which properties explain a large portion of variance in the spatial
location of the types. For instance, our visual inspection of the configuration suggested
that traditional–modern was an important property, accounting for a large part of the
distances between the types; this should be corroborated by a relatively high multiple
Rfor this property. Second, by computing multiple Rs for HOMALS solutions of
different dimensionality (from a 1-dimensional solution to a 10-dimensional one), we
were able to establish the optimal number of dimensions required to represent the
sorting data. After our visual inspection, we assumed that a three-dimensional
solution was optimal in terms of interpretability. This assumption can be validated by
examining the pattern of multiple Rs across solutions of increasing dimensionality:
Presumably, the Rfor any of the properties should not increase substantially by adding
more spatial dimensions beyond three.
Third, the predicted coordinate values
obtained from the regression can be used to position the properties as vectors in the
spatial representation of the types, as we will show below.
Results and discussion
Table 2 presents the Rvalues we obtained for each property in analyses with one, two,
three or four predictors, that is, using the types’ coordinate values from the one-, two-,
three- or four-dimensional solution. In the one-dimensional analysis, the best fitting
properties are Age (R= .61) and Choice (R=.64), with the latter becoming more
prominent in a three-dimensional solution. The other properties do not explain much
of the variance in the one-dimensional solution. In the two-dimensional analysis, the
feminine–masculine property emerges as an important one (R= .75). In addition,
the Settled dimension (R= .71) obtains a more substantial role in this analysis. In the
three-dimensional analysis, the traditional–modern dimension, whose role is negligible
in the first and second dimension, explains a large part of the variance (R=.83). These
results are all consistent with our initial interpretation based on a visual inspection.
The column for the four-dimensional analyses has been added to demonstrate that,
beyond three spatial dimensions, no other important properties emerged. This result
converges with our initial conclusion, that the fourth and higher dimensions were
uninterpretable. Note that, although Rdoes tend to increase across one to four dimen-
sions, this increase is caused primarily by the mere increase in predictors: There is no
substantial improvement in any of the Rs when the fourth dimension is added.
We will now describe the three-dimensional solution in more detail. The first spatial
dimension could be interpreted as young–old, with young types at one end (e.g.
schoolboy,adolescent,schoolgirl, ‘Yes’type—after a Dutch magazine targeted at
girls from 12 to 16 years) and older, mostly male types at the other (e.g. minister,
In HOMALS, the spatial dimensions are nested. That is, the coordinate values of the stimuli on the rst dimension are
the same in the one-dimensional and in the two- and hi gher-dimensional solution; similarly, coordinate values on the
second dimension remain the same in a three- or highe r-dimensional solution.
The only possible exception is competence, but we were unable to identify any competent–incompetent contrast in the
four-dimen sional solution.
Thinking about gender types 269
professor,old man,grandad). This interpretation was corroborated by the one-
dimensional analysis, in which we looked only at the locations of the types on the first
spatial dimension (Table 2, 1st column): In this analysis, the Age dimension has a
relatively high fit (R=.61).
Figure 1 presents the second and third dimension of the three-dimensional HOMALS
configuration. The figure does not include all 229 types, but a selection that shows the
most important results. In the HOMALS solution, parts of the space contain so many
types clustered together in a small space that they cannot all be presented; in these
cases, only a selection of types is mentioned in the figure. The 34 types that were used
in Study 2B are indicated with an asterisk. All types describing males are italicized, and
types describing females are in roman. In the figure, the second spatial dimension is
represented along the horizontal axis, the third dimension along the vertical axes, and
the first dimension (which is the easiest one to interpret because it largely coincides
with young–old) is represented by the small triangles printed above the type labels.
One end of the first dimension contains the types with the large open triangles (i.e. the
younger types such as adolescent and prissy girl), and the other end contains the types
with smaller black triangles (the older types such as grandad and old maid). Types
without any triangles are intermediate on this dimension. Thus, although the first
dimension is continuous, just as the others, it is summarized here in terms of three
categories (low, middle and high on the Age dimension) to simplify the presentation.
One may imagine this dimension as the depth dimension of the figure, with the types
with large open triangles up at the front (i.e. above the page) and the types with small
black triangles at the back (below the page).
In addition to the spatial locations of the types, the figure also shows the direction of
some of the properties. The vectors representing these properties were obtained by
drawing a straight line from the origin through the coordinate values predicted in the
three-dimensional regression analysis. In Fig. 1, these lines have been drawn for the
four properties that explain most of the variance in the second and third spatial
Table 2. Multiple Rvalues obtained in regression analyses with ratings on 10 properties as a
dependent variable, and coordinate values of the types on one to four spatial dimensions as
1 1–2 1–3 1–4
Age .61 .63 .64 .77
Settled .47 .71 .84 .85
Choice .64 .67 .75 .79
Masculine/feminine .27 .75 .76 .76
Traditional .10 .11 .83 .83
Sexual .03 .43 .45 .49
Evaluation .33 .45 .45 .52
Potency .39 .41 .59 .60
Activity .36 .47 .64 .69
Competence .42 .60 .65 .77
270 Roos Vonk and Richard D. Ashmore
dimension, i.e. feminine–masculine, traditional–modern, settled, and choice. For the
other properties, the direction is not informative because they do not explain the
pattern of distances in this figure (see Table 2).
One of the most striking findings in this representation is that the male and female
types are in different parts of the space. In fact, the male and female types can largely
be separated by drawing a curve in the figure from the upper left to the lower right;
that is, almost perpendicular to the feminine–masculine vector which separates the
masculine from the feminine types. This property represents masculine, aggressive,
physically strong types on one end (e.g. macho man,male chauvinist,body builder)
and female types on the other (e.g. mother,wife, but also modern woman and
feminist). Two kinds of male types encroach into the female side of the space, namely,
homosexual and the family-man types such as grandad,father and male homemaker.
Figure 1. Dimensions 1 (depth), 2 (horizontal) and 3 (vertical) of HOMALS solution of gender
types (Study 2A), including four externally tted properties with the highest t (obtained from
Study 2B). ;and 6= front vs. back of the depth dimension (Dimension 1).
Thinking about gender types 271
Aside from these exceptions, it seems that participants sorted the types predominantly
by their gender (i.e. they created separate piles for male and female types). Within the
two gender categories, the non-traditional types are closest to the opposite gender
category. That is, non-traditional male types such as male homemaker,sissy and
modern man are positioned close to the female types. Conversely, the female
type athlete is close to the male side of the figure, but the picture is slightly more
complicated here because several non-traditional female types are not so close to the
male side and are seen as relatively feminine (e.g. feminist and independent woman).
Whereas the masculine–feminine contrast is highly correlated with the second
spatial dimension, the third spatial dimension almost perfectly coincides with the
traditional–modern property, reflecting the stereotype- and role-consistent types
versus the inconsistent types. Note that this dimension has a low fit in a one- or
two-dimensional solution, meaning that it is unimportant in the first and second spatial
dimension, but has a high fit (R=.83) in the three-dimensional analysis (see Table 2).
For women, the traditional types (upper-left corner in Fig. 1) are housewife,mother,
bourgeois woman, and women who hold jobs that do not require a higher education
(e.g. cafeteria worker,cashier). There appears to be two distinct kinds of non-
traditional types. First, the lower-left corner in Fig. 1 presents women who have
careers and brains and are independent and liberated (e.g. businesswoman,liberated
woman,feminist), who violate traditional gender roles by building their own lives and
careers instead of depending on a husband. Second, on the right-hand side of the
traditional–modern vector, the sexual types are represented (e.g. man eater,femme
fatale,flirtatious woman), who violate traditional gender roles by being sexually
active. For men, the most traditional types, again, occupy lower-class jobs such as
farmer,labourer, and dustman (upper-right part of Fig. 1). Interestingly, the non-
traditional types are not primarily soft men such as wimp and sissy but, instead,
happy-go-lucky types who do not work, such as bon-vivant,adventurer and ‘profes-
sionally unemployed’, a category referring to people who voluntarily and chronically
live on welfare. These types violate the traditional male role of provider.
Close to this
category of types is the female type welfare mother, a type that would actually fit into
several categories, because it describes a single mother without a job (i.e. a woman
who does not depend on a man and who lives on welfare, like the professionally
unemployed, although it should be noted that the label welfare mother is usually
associated with a hard life, quite unlike that of the professionally unemployed).
In addition to the properties Age, masculine–feminine, and traditional–modern, two
other properties have a good fit in the HOMALS solution (see Table 2). One is the
Settled dimension: From the upper left to the lower right, the types differ in whether
the persons are being settled into their role in society (e.g. mother,bourgeois woman,
family man) versus being free to do as they please (e.g. eternal bachelor,adventurer,
professionally unemployed). This property is moderately correlated with traditional-
modern (in the three-dimensional solution, the angle between the two vectors is 44.1°)
and with Age (angle of 40.2°). Generally, the non-traditional types and the younger
types are seen as more free than the traditional and the older types.
A final property that is important in these data is the Choice dimension, which
describes whether one’s role is consciously chosen versus being given without
Note that the government worker/civil servant is also in this category, which may be explained by the Dutch
stereotype of government workers, i.e. tha t they are lazy and uninvolved with the job, and mainly read the paper and
drink coffee .
272 Roos Vonk and Richard D. Ashmore
thinking about it, i.e. achieved versus assigned. One pole of this dimension contains
primarily lower-class occupations (farmer and labourer for males, cafeteria worker
and cashier for females), the male type soldier (which, until a few years ago, was an
obligatory role for all young men in The Netherlands), as well as young types who
have not yet made any conscious choices. The other pole is represented by primarily
independent female types (e.g. businesswoman,working woman, and BOM—which
is a common Dutch abbreviation for intentionally unmarried mother). This property
is related to Age (angle of 26°), and slightly to the Settled dimension (61.4°). Note
that the direction of the correlation with Age is different for the two dimensions:
The younger types are seen as relatively closer to the ‘given/assigned’ pole on the
Choice dimension, as well as closer to the ‘free’ pole on the Settled dimension.
In other words, they are seen as potentially free to choose their role in society, but
unlike older people, they have not made this choice yet, and their present roles
(13-year-old,schoolboy,soldier) are mostly assigned, rather than chosen. Note also
that the ability to make such conscious choices is perhaps assumed to evolve only later
In sum, the structure underlying perceptions of male and female types can be
described by means of three major dimensions: Age, Gender, and traditional–modern.
It should be noted that the masculine–feminine dimension, although examined here as
referring to psychological characteristics of the subgroups, largely coincides with
another demographic variable, in addition to the Age dimension, because it distin-
guishes men from women.
Two other dimensions were found, which are partly
correlated with the three major dimensions, namely, Settled and Choice. These dimen-
sions may be seen as specifications of particular aspects of the three major dimensions.
The Choice, or achieved–assigned, dimension shows that traditional roles, especially
male traditional roles, are generally assigned, whereas non-traditional roles, especially
female roles, are more often the result of a conscious choice (see Fig. 1). The Settled
dimension shows that traditional females are usually settled into their role, whereas
non-traditional male types are usually free.
Out of the 10 properties we examined, five others turned out to be only of minor
importance in these data. Most remarkably, the single most prominent dimension in
social judgment, Evaluation (see Osgood et al., 1957; Rosenberg & Sedlak, 1972),
appears to be absent in the present representation of types: No matter in which
direction one looks, there is never one where the types are positive on one end and
negative on the other. An examination of higher-dimensional solutions indicated that
an evaluative distinction did not emerge in additional spatial dimensions either. This
could also be seen from the Rs obtained for this property, which remained small across
solutions of higher dimensionality. The same goes for the other general properties that
we selected: Potency, Activity, Competence, and also for the Sexuality dimension. The
regression analyses indicate that the judgments on these properties, as assessed in
Study 2B, cannot be accounted for by the spatial representation of the types derived
from Study 2A. Thus, we may conclude that these dimensions were hardly used by
participants in sorting the subgroup labels.
Note that a third important demograp hical distinction, based on race, did not emerge as a dimension in these data
because non-white subtypes were such a small minority that they had to be eliminated as outliers; see footnote 3.
Considering that these ty pes were in fact se parated from all the white ty pes during the sorting task, we m ay assume that
Race wou ld indeed have formed an other dimension if there had been more of these types.
Thinking about gender types 273
In Study 1, participants provided open-ended descriptions of male and female types. In
Study 2A, the types listed in Study 1 were sorted by similarity in a free sorting task.
Because the procedure in both studies was largely unstructured by the investigator,
the results from these studies may be considered as a representative reflection of the
contents (Study 1) and the dimensional structure (Study 2) of gender types in The
Consistent with previous work, our investigation of the contents of the types in
Study 1 showed that females are more often typed in terms of sexuality (especially by
male participants, see Table 1) and sexual attractiveness. Most of these sexual female
types were negative (e.g. bimbo,whore,dumb blonde). For males, we found that the
types were more differentiated in terms of occupation and that the ‘default’ male has a
job and a profession, whereas the non-traditional males do not work. Because these
results are consistent across multiple studies from different Western countries, we may
conclude that they reflect current stereotypes of men and women in Western society.
Our analysis of the contents of the types also revealed some additional categories
which, presumably, reflect typically Dutch or European cultural characteristics. Most
importantly, the snobbish category was rather prominent for both male and female
types. This is probably typical for the Dutch, who tend to feel that people should be
plain, without any frills, and inconspicuous (‘Behave normally and your behaviour will
be crazy enough already’) and who strongly disapprove of people who pretend to be
more interesting than they are (‘phony baloney’) or who flaunt their wealth or other
In the second part of the study, we examined the dimensional structure of the
perceived similarities among the types. Differing from previous studies, male and
female types were examined simultaneously. Our main question was: Are male and
female types organized separately, and are they distinguished primarily by their gen-
der, or are they jointly organized on the basis of other characteristics? In the latter case,
we would have found that the male and female types were mingled across the multi-
dimensional space; for instance, businesswoman would be very close to business-
man, and on the masculine–feminine dimension, it would be closer to the masculine
end than the male types sissy or mother’s boy. Instead, however, we found that the
male and female types were largely separated from each other. There were two
exceptions. First, the ‘homely’, family types, such as mother and father,housewife and
homemaker, and grandmother and grandad, were grouped close together, in the
feminine part of the space. This may be explained by assuming that the family role is
seen as nurturant and, hence, as feminine. Apparently, in thinking about the family
man, participants imagined a man at home (perhaps with slippers on, which does not
evoke a particularly masculine image), being involved in household chores and raising
children, that is, in traditionally feminine tasks. Note that merely being at home may
have a stereotypically feminine connotation. Thus, even though the male role of family
provider is traditionally masculine, the involvement in other functions of the family is
not: The masculine role ends on the doorstep of the family home.
The second exception to the male–female segregation concerns the male types
homosexual and, to a smaller extent, sissy. Note that these feminine male types
encroached further into the female side of the space than vice versa (i.e. masculine
female types hardly entered the male side of the space). In fact, some of the least
feminine female types, such as feminist and lesbian, did not even approach the
274 Roos Vonk and Richard D. Ashmore
boundary between male and female types. Because we used a large and representative
sample of gender types, this result cannot be attributed to a biased set of types.
Instead, it seems that males are more easily seen as trespassing the psychological
threshold for maleness (‘Avoid sissy stuff’, Brannon, 1976) than vice versa. In other
words, there may be a wider range of ways to be feminine than there are ways to be
This is shown not only by the boundary between males and females in the figure, but
also by the male and female extremes. The female extreme includes both traditional
nurturing types (mother,wife) and modern career types (two-piece suit type,modern
woman,liberated woman) as well as other women who are seen as having chosen
their roles (e.g. lesbian,intentionally unmarried mother). Thus, there are several
distinct extremes of femaleness, whereas the male end of the configuration is more
homogeneous. In our study, the male extreme (macho,chauvinist), largely coincides
with Brannon’s (1976) Blue Collar Brawler, a prototype that emphasizes the strong and
tough aspects of masculinity, not the instrumentality/agency aspects (see, e.g., show-
off ’, windbag ). Indeed, our feminine–masculine dimension does not coincide with the
typical agency versus communion continuum.
In addition to the masculine–feminine dimension, four other dimensions of meaning
are relevant in describing the structure of gender types. The first one is age, contrast-
ing the young versus old types. This dimension, obviously, is used quite generally in
categorizing people. Like gender, a person’s age is both easily visible and societally
important. Another important dimension, traditional–modern, has been obtained
consistently in research on gender types. For female types, this dimension describes
the contrast between sex-role-consistent versus inconsistent types (e.g. housewife vs.
career woman). For males, there are two types of traditional roles: the professional
roles (especially those that do not require higher education, such as farmer and
labourer), representing the traditional male role of provider, but also the family roles,
such as father,male homemaker and family man. Both of these variants of traditional
male roles are contrasted with the happy-go-lucky, irresponsible, no-commitment types
such as party-crasher,eternal bachelor and wanderer. This category, which appears
to be seen as the male way of being non-traditional has not been reported in previous
studies, as far as we know. We may conclude, then, that the non-traditional males are
lazy and enjoying life, whereas the non-traditional females are industrious working
The masculine–feminine dimension and the traditional–modern dimension have
been obtained consistently in previous studies on gender types. In these other studies,
where male and female types were examined separately, the two were strongly
related, because non-traditional women are seen as more masculine, whereas non-
traditional men are seen as more feminine. In the present study, where male and
female types were considered jointly, this produces two opposing correlations: For
males, traditionality is positively correlated with masculinity and negatively with
femininity, but for females it is the other way round. As a consequence, the
traditional–modern dimension emerged as orthogonal to masculine–feminine. The
masculine–feminine dimension represented the contrast between stereotypical males
versus stereotypical females. In between, the stereotype-inconsistent males and
females were represented. This orthogonality merely reflects a compromise between
the two opposing correlations for male and female types; in fact, the masculine–
feminine dimension is not at all orthogonal to Traditionality, which is correlated
positively with the extent to which the types are prototypic of the overall category.
Thinking about gender types 275
We found two additional dimensions to be useful in describing the structure of
gender types, namely, Settled and Choice. These dimensions, which are correlated
with the three dimensions above, may be seen as specifications or combinations of the
others. The Choice dimension was most strongly related to Age, with the older types
being seen as more likely, and perhaps more capable, to have consciously chosen their
current role. The settled–free distinction was related to Age too, but also to traditional–
modern, with younger and modern types being seen as more free. Because these
dimensions have not been described previously, it is hard to assess their status at this
We examined five other dimensions that have been reported repeatedly in studies
describing the dimensional structure of persons and personality traits. Remarkably,
these dimensions appeared to be of minor importance in the structure of gender types.
First, a conspicuous absentee was Evaluation. A possible explanation is that many of
the types were negative, and very few were positive. For instance, many of the female
sexual types (e.g. slut,dumb blonde) are perceived negatively, but so are the sexless
types (e.g. prissy girl,old maid); for males, the macho types are judged negatively
(e.g. male chauvinist,show-off), but so are the opposite types (e.g. wimp,sissy). On
the positive side of the spectrum, our types collection does not have much to show.
Types such as mother,father, career woman or career man are probably the most
positive ones in our sample, but these types are merely ‘normal’, not extremely good.
Thus, the positive–negative contrast was absent in the list; there was a ‘normal’–
negative contrast at best. The absence of clearly positive types cannot be attributed to
any element in the procedure, because participants in Study 1 were entirely free to list
whatever they wanted, and the questions did not guide them into a negative direction
(see Method section). Instead, it seems more likely that type labels are used to describe
people that differ from normality and are not the average woman or man; and, because
‘normal’ usually means ‘moderately positive’ (e.g. Matlin & Stang, 1978; Parducci,
1968; Peeters & Czapinski, 1990; Sears, 1983), it follows that the types were predomi-
nantly negative. In addition, previous studies suggest that people make fewer distinc-
tions between different types of positive stimuli than between different types of
negative stimuli (see Peeters & Czapinski, 1990; Vonk, 1996). Thus, we should indeed
expect that a representative sample of gender types includes more negative than
A second dimension that did not explain much of the variance in the multi-
dimensional representation was Sexuality, even though there were clearly sexually
related types in the set, as in other studies, such as womanizer and Don Juan for
males, and sex bomb and tramp for females. The reason why Sexuality did not fit into
the representation as a dimension of meaning seems to follow from the fact that male
and female types were combined in our studies. For males, the most sexual types were
also seen as relatively masculine, with high scores for womanizer (mean of 6.78 in
Study 2B) and male chauvinist (6.19). The lower sexuality scores were found for the
soft, feminine types such as Granola/Birckenstocks type (3.87) and wimp (3.19). For
females, the most sexual types were slut (6.34) and femme fatale (6.09). The non-
sexual female types, however, were not the soft, feminine ones but, instead, the
masculine, dominant ones such as bitch (3.84) and virago (3.50). Thus, the non-sexual
types are mostly feminine for males and masculine for females. Given this pattern of
relations, it is hard to fit the sexuality dimension into a configuration that represents
both males and females. In addition, it should be noted that the label ‘sexual’ may in
fact involve multiple dimensions, such as respectability (e.g. bimbo vs. prude), sexual
276 Roos Vonk and Richard D. Ashmore
attractiveness (e.g. beautiful woman vs. slob), and sexual activity or interest (e.g. Don
Juan vs. mother’s boy). If sexuality is itself a multidimensional concept, and if it has
differential associations for males and females, it follows that this dimension cannot be
captured by a multidimensional analysis such as the one we used, which represents the
big picture, rather than local differentiations.
Finally, it is interesting to note that in our representation of the types, the sexually
active female types (e.g. bimbo,prostitute,sex bomb) are quite close to the male side
of the space (and are, in fact, a lot closer than the more independent but ‘sexless’ types
such as feminist and businesswoman). Two explanations for this finding can be
offered. First, because men are generally seen as more sexually active than women, it is
possible that sexual activity is highly correlated with perceived masculinity. Second,
because the sorting task in Study 2A was entirely free, it is possible that some partici-
pants sorted some of the types not by their similarity but by their role relationship. For
instance, there were participants who sorted prostitute and whorehopper into the
same pile—presumably because of the complementary match between the two. Simi-
larly, womanizer and Don Juan may be seen as related to sex bomb or dumb blonde.
Obviously, participants did not predominantly sort the types according to this role
complementary criterion, for otherwise we would not have obtained such a strong
separation between male and female types. But at least for some of the types, they may
have interpreted their task in this way. This possibility illustrates that, when male and
female types are examined jointly, new considerations enter the picture: People have
organized beliefs not only about gender types based on similarities, but also about the
possible relations between men and women, and these beliefs may be part of their
cognitive representation of gender types.
Overall, the present results suggest that it makes sense to examine male and female
types jointly just as much as separately, as in previous work. Although there was, with
a few exceptions, a clear segregation between the male and female types, the differ-
ence between them appears to be dimensional, rather than categorical. This implies
that male and female types can be represented in the same dimensional space.
Although this joint analysis may obscure some features that are more prominent in
separate analyses of male and female types (e.g. the sexuality dimension, which is
different for male and female types), it reveals others, such as the perceived similarities
between male and female types. Thus, our analysis complements the existing literature
on the contents and organization of gender types.
In particular, our results suggest that there are multiple ways to violate gender role
expectancies and to be seen as gender-inconsistent. Males can violate their traditional
role by being lazy or at least not working (which makes them less masculine, but not
more feminine), by taking care of children, family, and household chores (which
makes them less masculine as well as more feminine), or by being homosexual and
‘sissy-like’. Females can be gender-role-inconsistent by being sexually active (which
makes them more masculine), by being feminists, man-haters or lesbians (which makes
them less feminine), or by being liberated, independent and smart. It would be inter-
esting to examine whether these different ways of disconfirming gender stereotypes
have different effects on (a) people’s perceptions of the individual disconfirmer and
(b) on the gender stereotype. For instance, we may speculate that deviations implying
less femininity for a woman and less masculinity for a man evoke (a) relatively negative
judgments of the individual and (b) do not produce any stereotype change, because
the person is not seen as representative of the overall category, so that the impression
of this person does not generalize to the stereotype (see Richards & Hewstone, 2001).
Thinking about gender types 277
However, deviations implying more masculinity for a woman, while maintaining
femininity, and more femininity for a man, while maintaining masculinity (i.e. devia-
tions implying a high level of androgyny, because both masculinity and femininity are
high), (a) may evoke relatively positive judgments of the individual and (b) may
produce more stereotype change, because the person still has characteristics that
are representative of the overall category, so the person is not psychologically
excluded from the category. Experimental studies are needed to address these
Another potential avenue for further research concerns the fact that males and
females were organized more by their gender than by the masculinity or femininity of
their characteristics. This may be the result of the fact that gender is probably the
single most important categorization dimension in our society. It is conceivable, then,
that different results would be obtained for other, less prominent categorizations,
such as race. However, our results might also be indicative of the way in which
people generally organize categorical or dichotomous clusters, such as male–female,
black–white, or categories reflecting different professions or college majors (e.g.
psychology–physics). This would imply that a similar structure is obtained if types of
these other categories were to be examined jointly (e.g. businessman black/white,
streetwise black/white etc.). In this case, our results would be generally informative
about the way in which people mentally represent subtypes of multiple categories,
suggesting that there are limitations in cognitively transcending the borderlines
between these categories.
This study was conducted while the first author was at Leiden University. Thanks to Nathalie van
Kogelenberg, Jord Kooiman, Zwanet Reyne, Beatrice Snel, Joanet van Steenbergen, and Natasja
van Schaik, for interviewing subjects in Study 1; to Laura Sweeney for her help in classifying the
types in Study 1; and to Marjolijn Olde Monnikhof for her help in data collection and data
analysis in Study 2A.
Ashmore, R. D., Del Boca, F. K., & Titus, D. (1984). Types of women and men: Yours, mine,
and ours. Paper presented at the 1984 American Psychological Association Convention,
Ashmore, R. D., Solomon, M. R., & Longo, L. C. (1996). Thinking about beauty: A multi-
dimensional approach to the structure of perceived physical attractiveness. Personality and
Social Psychology Bulletin,22, 1083–1104.
Brannon, R. (1976). The male sex role: Our culture’s blueprint of manhood, and what it’s done
for us lately. In D. David & R. Brannon (Eds.), The forty-nine percent majority. Reading, MA:
Brewer, M. B., Dull, V., & Lui, L. (1981). Perceptions of the elderly: Stereotypes as prototypes.
Journal of Personality and Social Psychology,41, 656–670.
Carpenter, S. (1994). Characteristics of gender subtypes: Group and individual differences. Sex
Chia, R. C., Moore, J. L., Lam, K. N., Chuang, C. J., & Cheng, B. S. (1994). Cultural differences in
gender role attitudes between Chinese and American students. Sex Roles,31, 23–30.
Clifton, A. K., McGrath, D., & Wick, B. (1976). Stereotypes of women: A single category? Sex
278 Roos Vonk and Richard D. Ashmore
Coats, S., & Smith, E. R. (1999). Perceptions of gender subtypes: Sensitivity to recent exemplar
activation and in-group/out-group differences. Personality and Social Psychology Bulletin,
Deaux, K., & Lewis, L. L. (1984). The structure of gender stereotypes: Interrelationships
among components and gender label. Journal of Personality and Social Psychology,46,
Deaux, K., Winton, W., Crowley, M., & Lewis, L. L. (1985). Level of categorization and content of
gender stereotypes. Social Cognition,3, 145–167.
Devine, P. G., & Baker, S. M. (1991). Measurement of racial stereotype subtyping. Personality
and Social Psychology Bulletin,17, 44–50.
Eckes, T. (1994a). Explorations in gender cognition: Content and structure of female and male
subtypes. Social Cognition ,12, 37–60.
Eckes, T. (1994b). Features of men, features of women: Assessing stereotypical beliefs about
gender subtypes. British Journal of Social Psychology,33, 107–123.
Edward, G. H. (1992). The structure and content of the male gender role stereotype: An
exploration of subtypes. Sex Roles,27, 533–551.
England, E. M. (1988). College student stereotypes of female behaviour: Maternal professional
women and assertive housewives. Sex Roles,19, 365–385.
England, E. M. (1992). College student gender stereotypes: Expectations about the behaviour of
male subcategory members. Sex Roles,27, 699–716.
England, E. M., & Hyland, D. T. (1987). The content and structure of stereotypic subcategories of
females and males. Social and Behavioural Sciences Documents,17, no. 2.
England, E. M., & Manko, W. (1991). College student evaluations of stereotypic female
subcategory members. Paper presented at the 62nd Annual Meeting of the Eastern
Psychological Association, New York, April, 1991.
Fiske, S. T. (1992). Thinking is for doing: Portraits of social cognition from daguerreotype to
laserphoto. Journal of Personality and Social Psychology,63, 877–889.
Fiske, S. T., Xu, J., Cuddy, A. J. C., & Glick, P. (1999). Respect versus liking: Status and
interdependence underlie ambivalent stereotypes. Journal of Social Issues,55, 473–489.
Gifi, A. (1990). Nonlinear multivariate analysis. New York: Wiley.
Glick, P., Diebold, J., Bailey-Werner, B., & Zhu, L. (1997). The two faces of Adam: Ambivalent
sexism and polarized attitudes toward women. Personality and Social Psychology Bulletin,
Glick, P., & Fiske, S. T. (1996). The ambivalent sexism inventory: Differentiating hostile and
benevolent sexism. Journal of Personality and Social Psychology,70, 491–512.
Glick, P., & Fiske, S. T. (1998). Sexism and other ‘‘isms’’: Interdependence, status, and the
ambivalent content of stereotypes. In W. B. Swann, Jr., L. A. Gilbert, & J. Langlois (Eds.),
Sexism and stereotypes in modern society: The gender science of Janet Taylor Spence
(pp. 193–251). Washington, DC: American Psychological Association.
Haddock, G., & Zanna, M. P. (1994). Preferring ‘‘housewives’’ to ‘‘feminists’’: Categorization and
the favorability of attitudes towards women. Psychology of Women Quarterly,18, 25–52.
Heiser, W. J., & De Leeuw, J. (1981). Multidimensional mapping of preference data.
Mathe´matique et Sciences Humaines,19, 39–96.
Heiser, W. J., & Meulman, J. J. (1994). Homogeneity analysis: Exploring the distribution
of variables and their nonlinear relationships. In M. Greenacre & J. Blasius (Eds.),
Correspondence analysis in the social sciences: Recent developments and applications
(pp. 179–209). New York: Academic Press.
Holland, D., & Davidson, D. (1983). Labeling the opposite sex: Metaphors and themes in
American folk models of gender. University of North Carolina, Unpublished manuscript.
Holland, D., & Skinner, D. (1987). Prestige and intimacy. In D. Holland & N. Quinn (Eds.),
Cultural models in language and thought (pp. 78–111). Cambridge: Cambridge University
Thinking about gender types 279
Johnston, L., & Hewstone, M. (1992). Cognitive models of stereotypic change. 3. Subtyping and
the perceived typicality of disconfirming group members. Journal of Experimental Social
Kirchler, E. (1992). Adorable women, exert man: Changing gender images of women and men in
management. European Journal of Social Psychology,22, 363–383.
Matlin, M., & Stang, D. (1978). The Polyanna principle. Cambridge, MA: Schenkman.
Maurer, K. L., Park, B., & Rothbart, M. (1995). Subtyping versus subgrouping processes in
stereotype representation. Journal of Personality and Social Psychology,69, 812–824.
Noseworthy, C. M. (1982). A social cognitive analysis of stereotypes. Unpublished doctoral
dissertation, University of Rhode Island.
Noseworthy, C. M., & Lott, A. J. (1984). The cognitive organization of gender-stereotypic
categories. Personality and Social Psychology Bulletin,10, 474–481.
Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Urbana,
IL: University of Illinois Press.
Parducci, A. (1968). The relativism of absolute judgments. Scientific American,219, 84–90.
Park, B., Ryan, C. S., & Judd, C. M. (1992). Role of meaningful subgroups in explaining differ-
ences in perceived variability for in-groups and out-groups. Journal of Personality and
Social Psychology,63, 553–567.
Peeters, G., & Czapinski, J. (1990). Positive–negative asymmetry in evaluations: The distinction
between affective and informational negativity effects. In W. Stroebe & M. Hewstone (Eds.),
European review of social psychology (Vol. 1, pp. 33–60). Chichester, UK: Wiley.
Richards, Z., & Hewstone, M. (2001). Subtyping and subgrouping: Processes for the prevention
and promotion of stereotype change. Personality and Social Psychology Review,5, 52–73.
Rosch, E. H. (1973). Natural categories. Cognitive Psychology,4, 328–350.
Rosenberg, S. (1977). New approaches to the analysis of personal constructs in person
perception. In J. K. Cole & A. W. Landfield (Eds.), Nebraska, symposium on motivation.
Lincoln, NE: University of Nebraska Press.
Rosenberg, S., Nelson, C., & Vivekananthan, P. S. (1968). A multidimensional approach to the
structure of personality impressions. Journal of Personality and Social Psychology,9,
Rosenberg, S., & Sedlak, A. (1972). Structural representations of implicit personality theory. In
L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 6, pp. 235–297).
New York: Academic Press.
Schultz, M. R. (1975). The semantic derogation of women. In B. Thorne & N. Henley (Eds.),
Language and sex: Difference and dominance. Rowley, MA: Newbury House.
Sears, D. O. (1983). The person-positivity bias. Journal of Personality and Social Psychology,
Six, B., & Eckes, T. (1991). A closer look at the complex structure of gender stereotypes. Sex
Slooff, N., & Van der Kloot, W. A. (1985). The perceived structure of 281 personality trait
adjectives. Internet report PRM-85-05. Leiden University.
Vonk, R. (1993). Individual differences and common dimensions in Implicit Personality Theory.
British Journal of Social Psychology,32, 209–226.
Vonk, R. (1996). Negativity and potency effects in impression formation. European Journal of
Social Psychology,26, 851–865.
Vonk, R., & Olde Monnikhof, M. (1998). Gender subgroups: Intergroup bias within the sexes.
European Journal of Social Psychology,28, 37–47.
Weber, R., & Crocker, J. (1983). Cognitive processes in the revision of stereotypic beliefs.
Journal of Personality and Social Psychology,45, 961–977.
Received 3 August 2000; revised version received 10 November 2001
280 Roos Vonk and Richard D. Ashmore