How Large Are Gender Differences in Toy Preferences? A Systematic Review and Meta-Analysis of Toy Preference Research. Jac T. M. Davis & Melissa Hines. Archives of Sexual Behavior, January 27 2020. https://link.springer.com/article/10.1007/s10508-019-01624-7
Abstract: It is generally recognized that there are gender-related differences in children’s toy preferences. However, the magnitude of these differences has not been firmly established. Furthermore, not all studies of gender-related toy preferences find significant gender differences. These inconsistent findings could result from using different toys or methods to measure toy preferences or from studying children of different ages. Our systematic review and meta-analysis combined 113 effect sizes from 75 studies to estimate the magnitude of gender-related differences in toy preferences. We also assessed the impact of using different toys or methods to assess these differences, as well as the effect of age on gender-related toy preferences. Boys preferred boy-related toys more than girls did, and girls preferred girl-related toys more than boys did. These differences were large (d ≥ 1.60). Girls also preferred toys that researchers classified as neutral more than boys did (d = 0.29). Preferences for gender-typical over gender-atypical toys were also large and significant (d ≥ 1.20), and girls and boys showed gender-related differences of similar magnitude. When only dolls and vehicles were considered, within-sex differences were even larger and of comparable size for boys and girls. Researchers sometimes misclassified toys, perhaps contributing to an apparent gender difference in preference for neutral toys. Forced choice methods produced larger gender-related differences than other methods, and gender-related differences increased with age.
Discussion
We
found a broad consistency of results across the large body of research
on children’s gender-related toy preferences: children showed large and
reliable preferences for toys that were related to their own gender.
Thus, according to our review, gender-related toy preferences may be
considered a well-established finding. Our results, with 75 studies and a
range of toy preference measurements, complement and extend a previous
meta-analysis of 16 studies focused on free play (Todd et al.,
2018).
However,
our meta-analyses also revealed some gaps that could prevent confident
inferences about the drivers and consequences of children’s
gender-related toy preferences. These gaps could form priority targets
for future research. Our analyses also revealed some emergent patterns
in the data, especially in how gender-related preferences for broad
groups of toys differed in some respects from those for dolls and
vehicles, how study results varied according to study method, and how
gender-related differences in toy preferences related to child age.
Toy Selection and Gender Categorization
The
way that toys are selected, and categorized, as boy-related or
girl-related, is not standardized in the present research. Studies in
our review appeared to treat the gender categorization of toys as
uncontroversial, even though, according to our review, it was not
uncommon for toys to be assigned to different gender categories in
different studies. For example, in some studies, blocks were classified
as boy-related toys (e.g., Alexander & Saenz,
2012; Benenson et al.,
1997; Fagot & Patterson,
1969), and in other studies they were classified as neutral toys (e.g., Cherney et al.,
2003; Guinn,
1984; Wood, Desmarais, & Gugula,
2002). Similarly, drawing toys were sometimes categorized as girl-related toys (e.g., Berenbaum & Hines,
1992; Martin et al.,
2013), and sometimes as neutral toys (e.g., Berenbaum & Snyder,
1995; Pasterski et al.,
2005); and stuffed toys were equally likely to be classified as girl-related toys (e.g., DeLucia,
1963; Jacklin et al.,
1973) as neutral toys (e.g., Alexander & Saenz,
2012; Idle et al.,
1993; Moller & Serbin,
1996), but were also sometimes classified as boy-related toys (e.g., Stagnitti, Rodger, & Clarke,
1997). This pattern suggests that researchers sometimes disagree on what toys are boy-related, girl-related, or neutral.
In
addition to finding that researchers sometimes disagreed on toy
classifications, we also found that researchers typically did not report
how they had selected toys for study or how they had assigned the toys
to gender categories. We suspect that, in most cases, researchers used a
simple heuristic method based on perceived cultural stereotypes. There
are two problems with this type of approach. First, as noted above, toys
categorized using this approach do not always fall into the same gender
category in different studies. If one study includes a stuffed toy in
the category “girls’ toys” and another study includes a stuffed toy in
the category “neutral toys,” they may well report different results,
even if the true underlying effect they are measuring is the same.
Second, at its extreme, this problem may manifest as criterion
contamination, in which gender-typed toys are defined by the results of
the study. That is, the researchers may use many toys and select as
“gender-related” toys the ones that they find to be differentially
preferred by gender. At best, this tautology limits the generalizability
of study results to other samples. At worst, it could invalidate the
study.
Using methods that avoid confusion about toy categorization
could be a priority for future research on children’s gender-related
toy preferences. As also suggested by Fine (
2015),
this field could benefit from researchers specifying more clearly the
ways in which they selected and categorized toys. Depending on the goal
of the study, this selection and categorization might be based on
different criteria. For example, a study examining whether stereotypes
about children’s toy preferences relate to children’s actual
preferences, might select toys based on adults’ independent ratings of
the gender stereotyping of toys. In contrast, a study of the effect of a
particular mechanism, such as social, cognitive, or hormonal
influences, on toy preferences might select toys based on prior studies’
findings that certain toys are on average preferred by girls or boys.
Overall, the important point is that researchers report more clearly how
they selected toys and assigned toys to gender categories.
Researchers
also have begun to investigate specific hypotheses about what
characteristics of different toys might make them appeal more to boys or
to girls. For instance, it has been suggested that color or shape might
influence children’s gender-related preferences (e.g., Jadva et al.,
2010; Weisgram et al.,
2014; Wong & Hines,
2015).
Similarly, it has been suggested that affordance of activity, motion,
or propulsion might influence these preferences (Alexander & Hines,
2002; Benenson et al.,
1997; Hassett et al.,
2008; for a review, see Zosuls & Ruble,
2018).
To evaluate these suggestions, it would be useful if researchers could
provide color images, or full descriptions, of the toys used in the
research they report. Similarly, it would be useful for this purpose, as
well as for future reviews, if researchers could provide descriptive
statistics, including means and SD or similar, by sex, for individual
toys used, and not just for toy groupings.
To test whether the
meta-analysis results were affected by researchers’ definitions of toy
gender, we analyzed the subset of effect sizes that related to a very
narrow definition of boy-related toys and girl-related toys:
specifically, vehicles and dolls. These toys were the only ones for
which sufficient data had been reported to allow reliable meta-analyses.
The gender effects observed in the overall meta-analyses were broadly
replicated with this more narrowly defined subset of toys, giving us
confidence that our overall meta-analytic results were not entirely
dependent on how researchers had chosen to categorize toys in regard to
gender.
Furthermore, we found that girls’ gender-specific
preference for dolls over vehicles was larger than their preference for
broadly defined groups of girl-related toys. However, despite the large
effect size, girls’ gender-specific preference for dolls over vehicles
was not statistically significant, as this effect also showed large
meta-analytic statistical variance. The large meta-analytic statistical
variance is due to a combination of large variances in girls’ preference
for dolls within the studies, variation between studies introducing
additional statistical variance, and a smaller total number of studies
that reported separate statistics for dolls as compared to broadly
defined toy groups. In addition, the broadly defined toy groups included
toys that, as mentioned above, were classified as neutral in some
studies but girl-related in others. If toys are classified consistently,
girls may show gender-related preferences at least as large as those of
boys.
Culture and Gender-Related Toy Preferences
Cultural
perceptions of play, including play with toys, may differ in different
cultural, ethnic, or socioeconomic groups. For example, play is viewed
as central to children’s cognitive and social development in many
Western, technologically developed societies, but as less important in
more traditional societies (Roopnarine,
2010).
Children in different cultures may also have different referential
concepts for appropriate gender-related behavior, due to cultural
variation in gender norms (Pfeiffer & Butz,
2005; Wood & Eagly,
2002).
This possibility is particularly relevant to toy preferences, because
there may be cultural variations in the toys that are available,
culturally relevant, and gender-related.
Nevertheless, little
empirical research is presently available on cultural variation in
gender-related toy preferences. Our review revealed that most toy
preference studies focus on the U.S., Canada, the UK, and Australia. Of
those studies conducted outside English-speaking industrialized nations,
one was conducted in France (Le Maner-Idrissi,
1996), one in Finland (Lamminmäki et al.,
2012), four in Sweden (Nelson,
2005; Nordenström, Servin, Bohlin, Larsson, & Wedell,
2002; Serbin et al.,
2001; Servin, Bohlin, & Berlin,
1999), and one in the Netherlands (van de Beek et al.,
2009). An additional study included some participants from Hungary, along with participants from the UK (Turner & Gervai,
1995).
These studies did not report different results to the studies from the
English-speaking countries, even when researchers had specifically
hypothesized that they would (e.g., Nelson,
2005).
In global perspective, however, these countries are very similar in
terms of industrialization, wealth, education, media access, democracy,
and gender equality. Consequently, children in these countries probably
have very similar toys available to them and similar access to
information about dominant social stereotypes around these toys. It
remains an open question, then, whether children in cultures with
radically different stereotype referents and social norms would show the
same gender-related toy preferences to those found in the current
meta-analysis.
We did not formally investigate other aspects of
cultural diversity, such as ethnicity and socioeconomic status, because
these also have not received much attention in empirical studies of
gender-related toy preferences. Participants in most toy preference
studies are not very ethnically diverse, and so it may not be practical
to report results by ethnicity. We found three studies (out of our total
75) that reported toy preferences by ethnicity. Two of these studies
were conducted in the USA and reported no significant differences in
gender-related toy preferences between children of Hispanic and
non-Hispanic background (Goble, 2012), or Native American and non-Native
American background (Guinn,
1984).
In contrast, another study based in the U.S. found that ethnicity might
affect children’s preferences for gender-related activities, including
play with toys, via children’s social networks (Martin et al.,
2013).
Furthermore, in recent years, the wider field of gender development
research has paid increasing attention to the intersectionality of
gender, ethnicity, and other identities (e.g., Shields,
2008).
This trend in the wider field may translate in future to more studies
investigating gender-related toy preferences in diverse social groups.
Methods of Measuring Toy Preference Are Important
Studies
may find different gender effects on children’s toy preferences,
depending on the method they use to measure toy preferences. We
evaluated four categories of study methods: free play methods, where
children were given access to a set of toys and observed playing,
however, they liked; visual preference measures, where children were
asked to look at pictures of toys; forced choice methods, where children
were asked to choose toys or pictures of toys, typically in front of an
experimenter; and naturalistic methods, where children’s toy options
were not predefined by the researchers or other adults. We found that
forced choice methods consistently showed larger gender differences than
other methods.
There are two possible explanations for this
pattern. One is the potential demand characteristics of forced choice
paradigms. A request to publicly choose an option may be interpreted as
evaluative by children, who then feel obliged to give the answer that
they feel is “correct,” rather than indicate their actual preference.
Children’s propensity to misunderstand requests for information as tests
has been noted in other contexts (e.g., Lamb et al.,
2003).
Another possibility is that the paradigm creates a false dichotomy. In
forced choice methods, the child is usually presented with one
boy-related option and one girl-related option and asked to choose
between them. There is usually not a neutral option, and, generally, the
child must choose only one option and reject the other. In contrast, in
a free play paradigm, children typically have more response options
available, such as several toys associated with each gender, or neutral
toys as well as gender-related toys. Even if only two toys are
available, the child has more options than in most forced choice
paradigms. For example, if a doll and a car are available, a child may
choose to play with the doll, play with the car, play with both the doll
and the car, or play with neither. In most forced choice methods,
however, children must choose one and only one of two options.
Forced
choice methods, in their current form, do not give comparable results
to other methods of measuring gender-related toy preferences.
Nevertheless, forced choice methods can be an efficient and easily
administered measurement tool and therefore may be appropriate for
studies where, for example, data need to be collected across a very
large group or under difficult conditions. Future investigators wishing
to measure gender-related toy preferences with an easily administered
tool might do so, however, with the aim of minimizing artificial
inflation in effect sizes. For instance, a procedure in which the
experimenter cannot see which option the child selects, and the child
knows that their response is not seen, might be useful. It also might be
useful to include neutral options, as well as gender-related options,
and allow the range of possible choices to include “both” or “neither.”
These modifications of forced choice methods could provide results that
are more comparable to other methods of measuring toy preference and
perhaps are more reflective of children’s actual gender-related
preferences.
Child Age and Gender-Related Toy Preferences
We
found that gender differences in preferences for gender-related toys
increased linearly with child age. Our results further suggested that
this pattern could be explained by boys’ showing increased preference
for gender-typical over gender-atypical toys with age, while girls’
preferences for gender-typical over gender-atypical toys did not
increase significantly with age. Similarly, the previous meta-analysis
of free play studies (Todd et al.,
2018)
found an increase in gender-related play with age in boys, although
they did not find increasing gender differences. This may reflect a
difference in the power of the two meta-analyses; the previous
meta-analysis included 16 studies, whereas the current meta-analysis
included 75 studies. We did not find significant curvilinear effects of
age on children’s gender-related toy preferences.
Our findings of
linear effects contrast with those of some prior investigations of age
effects on children’s gender-related toy preferences. For example,
Campbell et al. (
2000)
measured infants’ gender-related visual preferences longitudinally at
ages 3, 9, and 18 months. They found that preferences did not change
with age, but the infants were all very young compared to the age range
in the wider literature and in the current meta-analysis.
In
contrast, our meta-analytic findings suggest that boys’ and girls’
gender-related toy preferences increase with age in a linear fashion.
These findings resemble findings for a broader measure of children’s
gender-typical behavior, the Pre-School Activities Inventory (PSAI). The
PSAI is a 24-item parent report inventory that asks about children’s
gender-typed toy preferences and about children’s gender-related
activity and playmate preferences. A longitudinal, population study in
which the PSAI was completed by a parent to describe their child at ages
2, 3, and 5 years also found that both boys and girls became
increasingly gender-typed with age (Golombok et al.,
2008).
Our
results suggest that children’s toy preferences might become more
gender-related with age, as predicted by several theories of gender
development. Children might be encouraged, through socialization
pressures such as modeling and reinforcement, to prefer same
gender-related toys, and the effects of this socialization may
accumulate as they get older (Fagot, Rodgers, & Leinbach,
2000).
Additionally, based on their early gender-related toy interests,
children might gravitate to different social environments, enhancing
their early preferences (Liben & Bigler,
2002; Martin et al.,
2013).
Finally, differences in children’s prenatal and early postnatal hormone
exposure may dynamically interact with social environments and
cognitive processes to increase children’s gender-related preferences
over time (Hines,
2012).
Together, these social and cognitive effects, and their interactions
with early hormonal influences, may explain the linear increase in
gender-related differences with age.
The findings of our
meta-analysis, however, are not a substitute for a large, longitudinal
study of children’s gender-related toy preferences. We used
meta-analytic techniques to compare gender-related preferences in
children from different age groups, reported in different studies. Our
analysis, therefore, was cross-sectional and does not have the
inferential power of a well-controlled longitudinal study. Our results
would be best confirmed by a future longitudinal study of children’s
gender-related toy preferences from infancy to pre-pubertal age. The
longitudinal parent report study using the PSAI (Golombok et al.,
2008) is the closest existing example and found similar results to our meta-analysis.
Gender-Related Toy Preferences Over Time
We
found no change in the magnitude of gender-related differences in toy
preferences across year of publication. The results of the moderator
analyses suggested that gender effects on children’s toy preferences
have remained generally constant in magnitude across the past five
decades. This finding might seem surprising. Since the earliest studies
on gender-related toy preferences, gender-atypical behavior and
preferences have become increasingly socially acceptable. Perhaps the
lack of any discernible pattern of change results from different social
pressures influencing gender-related toy preferences in different
directions. For example, growing acceptance of gender-atypical behavior
may be countered by increasing gender segregation of the toy market.
Contrary to our results, a previous meta-analysis of children’s toy preferences (Todd et al.,
2018)
found that boys and girls played more with gender-related toys in
earlier studies than in more recent studies. Todd et al. suggested that
increasing gender equality in Western societies could influence children
to play with neutral toys, due to increased advertising to children
about gender-neutral toys. A recent analysis of online toy marketing,
however, found that more toys were marketed for “boys only” or for
“girls only” than for both (Auster & Mansbach,
2012),
and an analysis of department store catalogs concluded that gender
differentiation in toy advertising had increased since the 1980s as
marketers employed gender stereotypes to encourage sales (Sweet,
2013).
Taken together, these analyses challenge the view that gender-related
toy advertising is decreasing with time. Alternatively, the previous
finding could be partly explained by the smaller time frame considered
in the prior meta-analytic review; the prior review covered about
35 years of research, while the present review covered 50 years.
It
may be that children’s preferences are robust to social influences at
this macrolevel; or that, despite social change, the underlying cultural
environment regarding gendered toys has not changed. A similar result
was found in a systematic review of gender stereotypes from the 1970s to
the present. Rudman and Glick (
2008)
hypothesized that women’s changing social roles would be reflected in
changing stereotypes of women. Although they found a change in women’s
self-concept over time, they also found that more general stereotypes of
women’s personalities had not changed. They suggested that the lack of
change might be due to people viewing personality as part of the
fundamental essence of gender, and therefore being reluctant to modify
their stereotypic beliefs about personality. A similar explanation may
also apply to toy preferences: if people view toy preferences as an
essential part of a child’s gender, they may be unlikely to change their
gender-related beliefs about toy preferences. Children may then adapt
their actual toy preferences to reflect broader societal beliefs.
Limitations
The
meta-analysis could only include data that were reported in the
individual toy preference studies. Therefore, we could not analyze
variables such as color or shape, or individual toys other than dolls
and vehicles. In future research, if investigators report more
information about toy characteristics and about individual toys, it may
be possible to discover more about what characteristics of different
toys make them more likely to be preferred by one gender or another.
Our
literature search covers papers published to March 2014 and does not
include papers published outside of this time frame. More recent papers
may therefore be missing from the current meta-analysis. The current
meta-analysis, however, synthesizes 50 years of research on toy
preferences and finds that toy preference effect sizes have not changed
significantly over time. Thus, results from a new review including more
recent papers would be unlikely to differ from what we report.
We
focused on gender-related preferences in typically developing children.
Some studies selected participants specifically because they were not
typically developing (for example, clinical samples of children with
genetic variants causing atypical early hormone environments, or
children who showed gender-related behavior that was noticeably
different from their peers). To include these atypical populations in
our study might have skewed the results, so we did not include them. Our
results, therefore, may not apply to clinical populations.
Additionally,
we meta-analyzed only direct measures of children’s toy preferences. We
did not, for example, include parent report measures. Similarly, we did
not include broader aspects of children’s gender-related behavior, such
as activity preferences, playmate preferences, or sex role
identification (e.g., Brown,
1956).
Additionally, we did not search for these broader terms, so we may have
missed papers that included toy preferences in a broader measure of sex
role identification or androgyny (e.g., Zucker & Torkos,
1989).
It would be interesting to know whether meta-analyses from these other
sources of data and types of gender-related behavior would show similar
outcomes. We hope that the current systematic review and meta-analysis
will encourage such studies.
Posts
