Sex Differences

‘‘Sex differences’’ is the label used in describing variations between women and men. The term ‘‘sex’’ reflects the division of women and men into two groups on the basis of their unique biological features; the term ‘‘differences’’ derives from the tradition of differential psychology, in which distinct groups of people (defined by natural categories such as sex or constructed categories such as socioeconomic class) are compared in terms of an outcome.

Both terms have been criticized. To many people, references to sex differences imply a biological determinism that ignores the role of socialization and context. The more contemporary term ‘‘gender’’ directs attention to the social meanings assigned to the categories of male and female. With either term, however, it is not necessary to assume a particular causal factor for an observed pattern of variation. A problem with the term ‘‘differences’’ is that it suggests that dissimilarities are the norm and similarities are the exception. More appropriately, one makes comparisons between groups to see whether there are similarities or differences.

Regardless of the rubric used to characterize the investigation, comparisons between women and men pervade social science literature. Thousands of studies have analyzed sex-related patterns in physical performance, cognitive abilities, personality traits, moral reasoning, social interaction, occupational choice, sexual behavior, attitudes, aggression, depression, self-esteem, leadership, nonverbal behavior, self-disclosure, intelligence, life satisfaction, workplace achievement, and almost every other domain of human activity. To learn how these behaviors emerge, developmental psychologists explore the ways in which specific socialization practices contribute to observed differences between girls and boys and, by extension, women and men. Sociologists study structural features that shape the roles of women and men in organizational settings, family units, and labor markets. All these analyses contribute to the understanding of gender.

Not all of these topics can be discussed in this entry, and this review will not deal with sexual behavior or the biological basis of sex-related differences. Instead, this article will examine some of the main theoretical accounts of sex differences and then turn to

1) sex differences in four domains that were highlighted in early reviews (verbal, quantitative, and spatial ability and aggression),

2) differences in mathematics and science achievement,

3) other analyses of sex differences, and

4) contextual influences on sex-related patterns of behavior.

Theoretical accounts of sex differences

Two predominant and contrasting theoretical accounts of sex differences are evolutionary approaches and a variety of social roles, expectations, and/or status perspectives (Eagly and Wood 1999). Evolutionary theory predicts that ‘‘the sexes will differ in precisely those domains in which women and men have faced different sorts of adaptive problems’’ in evolutionary history (Buss 1995, p. 164). These domains primarily are those relevant to reproduction, such as mate selection and sexual behavior (Feingold 1992), but also may include physical skills, the commission of violent crimes (Daly and Wilson 1988), and thresholds for physical risk taking (Wilson and Daly 1985). The evolutionary perspective suggests that in domains irrelevant to sexual selection, the sexes should and do tend to be psychologically similar.

The social roles and/or expectations perspective focuses on the relationship between male gender and high status in American and other societies. Sex differences in behavior may emerge because men are expected to be more competent and authoritative than women, and many social situations are structured to support this outcome (Foschi 1998; Geis 1993; Ridgeway and Diekema 1992). Eagly’s (1987) social role theory further suggests that the unequal distribution of men and women in social roles (e.g., homemaker versus paid employee) contributes to both biased perceptions of the sexes and the development of skills and behavior that fit those roles.

The literature on self-fulfilling prophecy (Geis 1993; Rosenthal and Jacobson 1968) is consistent with this perspective. Expectations lead people to behave in ways that make the expected outcome more likely to occur. Gender-stereotyped beliefs of parents, for example, predict students’ performance in mathematics courses: Parents who believe that girls are inferior in mathematical ability are more likely to have daughters who do poorly in mathematics (Eccles 1985). Deaux and Major’s (1987) interactionist perspective further details how contextual factors determine whether and how gender-related expectations are translated into gendered behavior.

Other accounts of sex differences exist, including biological approaches (emphasizing hormones, brain structure, and genetics), developmental and/or learning approaches (e.g., social learning theory and gender schema theory) ( Jacklin and Reynolds 1993), and constructionist perspectives (which describe ‘‘doing gender’’ rather than ‘‘having gender’’) (West and Zimmerman 1987). The basic theme of ‘‘biology or environment’’ pervades these perspectives. As is discussed below, any theoretical account faces the challenge of accounting for the size and pattern of sex differences across domains and the susceptibility of those differences to contextual moderation.

Well-established sex differences

In discussing sex differences, one can begin with Maccoby and Jacklin’s (1974) landmark book The Psychology of Sex Differences. In this volume, the authors review the literature on sex differences and highlight several that are ‘‘fairly well established’’ (p. 351): Females tend to have greater verbal ability than males, and males tend to surpass females in quantitative ability, spatial ability, and aggression.

After the appearance of this narrative review, the statistical technique of ‘‘meta-analysis’’ was developed, enabling researchers to quantitatively combine results across many studies. Rather than rely on subjective impressions of the literature or a ‘‘count’’ of significant and nonsignificant effects, meta-analysis is based on calculation of an effect size (d) that reflects the mean sex difference in pooled standard deviation units in a given study. D’s are combined across studies to compute an average effect size. As a rough guide, Cohen (1977) suggests that d’s of .20 are small, those of .50 are medium, and those of .80 are large in magnitude. (To take a familiar example of a physical sex difference, the effect size for U.S. male–female differences in height is very large at d = 1.93.)

Meta-analyses of the four ‘‘established’’ sex differences have reached somewhat different conclusions than did Maccoby and Jacklin’s (1974) narrative review. For example, females and males appear to be comparable in verbal ability in general and in most specific types of verbal ability, such as vocabulary, verbal analogies, and reading comprehension (overall d = .11) (Hyde and Linn 1988); positive d’s indicate a relative female advantage. This effect was fairly stable across age despite Maccoby and Jacklin’s (1974) conclusion that the onset of sex differences in verbal ability occurs at around age 11. A possible exception to the smallness of sex differences is verbal fluency (e.g., speed and accuracy of speech production and the production of sentences meeting meaning requirements), where females more clearly outperform males (d = .33) (Hyde and Linn 1988). It is also the case that males are more prevalent at the low end of the verbal abilities distribution. For example, there are three to four times as many male stutterers as female stutterers (Skinner and Shelton 1985), and severe dyslexia is about 10 times more common in males than in females (Sutaria 1985; see also Halpern 1992).

In regard to quantitative ability, some metaanalyses have reported modest differences between males and females in performance on tests of mathematical skill (d = −.36 and −.43) (Feingold 1988; Hyde 1981). A more recent study, however, found a very small overall effect in the direction of female superiority (d = .05), although differences favoring men emerged in high school (d = −.29) and college samples (d = −.32) (Hyde et al. 1990). Sex differences also tend to be greater among selected samples of mathematically talented youth tested on the mathematics section of the Scholastic Aptitude Test (Benbow 1988). Overall, the average scores of 12- and 13-year-old boys are higher than those of girls of the same age in these samples; the variability in the boys’ scores is also greater. As a result, in the upper 3 percent of the distribution defined as mathematically precocious, boys outnumber girls in ratios that sometimes are dramatic (Benbow and Lubinski 1993).

Spatial ability has a possible (though still unclear) link to mathematical aptitude and scientific and engineering achievement. Meta-analyses of sex differences in visual-spatial ability have shown a variety of effect sizes, depending on the specific type of task or skill assessed. For example, on tasks of mental rotation, in which one is asked to visualize the rotation of a three-dimensional object, large sex differences favoring males have been found on one specific type of test (the Shepard- Metzler test; d = −.94) but not on others (d = −.26) (Linn and Peterson 1986). Tasks involving spatial perception (e.g., determining the true vertical plane when one is seated in a tilted chair) also indicate relatively smaller sex effects (d = −.44), and those involving spatial visualization (e.g., finding a simple shape in a complex pattern of shapes) show virtually no sex difference (d = −.13) (Linn and Peterson 1986; see also Masters and Sanders 1993; Voyer et al. 1995). Thus, sex differences in this domain are quite task-specific.

The only social behavior highlighted in Maccoby and Jacklin (1974) was aggression. Two metaanalyses of sex differences in aggression that were published a decade apart produced comparable, small effect sizes that indicated less evidence of aggression among females than among males (d = −.24 and d = −.29 in Eagly and Steffen 1986 and Bettencourt and Miller 1996, respectively). Bettencourt and Miller (1996) also identified an important moderating condition of this sex effect: Unprovoked men tend to be more aggressive than are unprovoked women (d = −.33), but under conditions of provocation, this sex effect is much smaller (d = −.17). A sense of the relatively small size of these effects can be gleaned by comparing them to the effect of provocation on aggression (d = .76). Although sex differences in aggression emerge, they do not seem to be as strong or straightforward as previously was thought.

Mathematics and Science achievement

Although meta-analyses indicate relatively small sex differences in quantitative ability, an important related question is whether females and males differ in real-world achievements in quantitative domains such as mathematics, science, and engineering.

A report on women’s representation in science and engineering by the National Science Foundation (1999) describes several significant findings. First, with regard to secondary education, male and female students are similar in their completion of high school mathematics and science courses (about 17 percent of both sexes have taken trigonometry, and about 25 percent have taken physics). Furthermore, average mathematics scores for females and males in the eighth and twelveth grades are not significantly different, although science scores among twelveth-graders are slightly higher for males than for females (based on the 1996 National Assessment of Educational Progress mathematics and science assessment).

On the mathematics portion of the Scholastic Assessment Test (SAT), men tend to score about 35 points higher than do women, a gap that remains even for students who reported having taken calculus and physics courses in high school. However, this difference appears to be due in part to the fact that a larger number of women than men who take the SAT are from lower-income families. Because parental income is related to SAT scores, the greater proportion of low-income women may reduce the overall female average (National Science Foundation 1999). It is unclear whether this income factor also accounts for findings regarding men’s and women’s performance on the 1996 College Board Advanced Placement (AP) tests. These data indicated a male advantage over females on AP tests in physics, computer science, chemistry, and calculus (d from −.26 to −.52) as well as an overrepresentation of males in the upper tail and an underrepresentation in the lower tail of the distribution of test scores (Stumpf and Stanley 1998). Interestingly, across twenty-nine different AP tests, the size of the effect favoring males was highly correlated with the percentage of males taking a given test (r = .71). That is, the greater proportion of male to female high school students who choose to take a given AP test, the greater the male advantage in performance. Students taking AP tests are college-bound and selfselect test subjects on the basis of their levels of preparation and expertise. Thus, this finding demonstrates that males feel more competent than females in the domains where they outperform females and that sex differences are apparent even among this highly self-selected group of ‘‘prepared’’ male and female students.

Women are more  focused  on their goals than men today

Figure 1: Women are more focused on their goals than men today

The National Science Foundation report (1999) details larger differences in female versus male achievement in regard to the attainment of advanced degrees and employment. Although women received 46 percent of science and engineering bachelor’s degrees in 1995, they earned only 31 percent of the total science and engineering doctoral degrees in that year (up from 26 percent in 1985). The only science field in which women received the majority of doctoral degrees (64 percent) was psychology.

With regard to employment, women represent 46 percent of the total U.S. labor force but only 22 percent of scientists and engineers. Representation differs dramatically across fields: More than half of psychologists and 47 percent of sociologists are women, but women account for only 12 percent of physicists and 9 percent of engineers (National Science Foundation 1999). Among those with academic employment, women with science training are more likely than are men to work in elementary or secondary schools and two-year colleges. In four-year colleges, women are less likely than men to be tenured (35 percent versus 59 percent) or full professors (24 percent versus 49 percent). Men also are more likely than women to be managers, and across fields, full-time male scientists and engineers earn more than do females (overall median salary of $52,000 versus $42,000). Some, but not all, of these differences in employment placement may be due to differences in age, and women in science and engineering tend to be younger than men. Similarly, salary differences are due in part to age and field differences, since men are more likely than women to enter the high paying fields of computer science and engineering. Nonetheless, differences in academic rank
remain after one controls for age, and with increasing age, the gap in salaries between female and male scientists and engineers widens (National Science Foundation 1999)

Other sex differences and similarities

It is impossible to provide a thorough overview of other research on sex differences in this brief article. In addition to the thousands of original empirical studies, over 180 meta-analyses have been published since Hall’s (1978) initial metaanalytic study of sex differences in the decoding of nonverbal cues (where d = .40).

In a review of the ‘‘science and politics of comparing women and men,’’ Eagly (1995) summarizes meta-analytic research by suggesting that the sizes of sex differences vary considerably across domains. The largest differences (d > .80) tend to be found for

(1) some physical abilities (e.g., throwing distance, speed, and accuracy),

(2) the Shepard- Metzler mental rotation task described above,

(3) social behaviors such as facial expressiveness,

(4) sexual behaviors such as the frequency of masturbation and attitudes toward casual sex, and

(5) nurturant personality traits (tender-mindedness) (Feingold 1994).

Most other examined attributes, however, tend to support sex differences that are small to moderate (or negligible) in size. For example, the mean effect size is −.14 for studies of sex differences in self-esteem (Major et al. 1999), .21 for fine eye-motor coordination (Thomas and French 1985), .02 for perceived leadership effectiveness (Eagly et al. 1995), .26 for negative attitudes toward homosexuality (Whitley and Kite 1995), and .07 for competitiveness in negotiation (Walters et al. 1998). Ashmore (1990) summarizes his review of the meta-analytic sex differences literature by noting that ‘‘relatively large sex differences are demonstrated for physical abilities and for body use and positioning; more modest differences are shown in abilities and social behaviors; and many negligible sex differences are sprinkled across all domains’’ (p. 500).

Whether one views observed sex differences as ‘‘meaningful’’ depends on one’s opinion and perspective. On the one hand, as Eagly (1995) points out, even a large effect size translates into substantial overlap between the sexes: A d of around .80 indicates about a 53 percent overlap in female– male distributions. In this manner, findings of difference may mask substantial similarity between the sexes. Furthermore, many sex differences are quite small in magnitude compared with other important psychological phenomena (e.g., the provocation–aggression link of .76 described above and the effect of group pressure on conformity of d = 1.06) (Bond and Smith 1996).

On the other hand, a number of other sex differences are comparable in magnitude to psychological effects that typically are interpreted as both theoretically and socially meaningful. For example, the effect of exposure to media violence on aggression is d = .27 (Wood et al. 1991), and the effect of having a type A personality style on systolic blood pressure is d = .33 (Lyness 1993). It is also the case that sex differences based on general population samples may be smaller than the differences that appear when one looks at selected populations or those in the ‘‘tails’’ of frequency distributions. The greater representation of males among the mathematically precocious, the verbally challenged, and the most violently criminal, for example, suggests that for some highly salient important outcomes, sex differences may be more striking than one might assume based on observations of ‘‘average’’ people encountered in everyday experience.

Some have questioned the wisdom of overreliance on meta-analysis, as this quantitative approach cannot overcome the shortcomings or biases in the original set of studies on which the data summary is based. For example, in Eagly and Crowley’s (1986) meta-analysis of sex differences in helping behavior, the authors note that their findings are based primarily on studies that involve short-term helping of strangers (‘‘heroic helping’’), a type of helping that is more consistent with the male role than the female role. Any finding of sex differences must be taken in light of this fact and not generalized across all helping situations.

Others have illustrated the limitations of meta- analysis for making causal inferences (Knight et al. 1996). For example, a number of meta-analyses of sex differences have indicated a ‘‘year of publication’’ effect: Studies published relatively recently suggest smaller sex differences in domains such as aggression; verbal, cognitive, and mathematical abilities; helping behavior; influenceability; leadership; and sexuality. Although some have interpreted these findings to mean that sex differences are disappearing (an interpretation more supportive of an environmental than a biological account), changes in research methodology over time may provide an alternative explanation. Indeed, in revisiting a meta-analysis of sex differences in aggression, Knight et al. (1996) found that statistically controlling for method characteristics (e.g., experimental or not, observational or selfreport measures, physical or verbal aggression) removed the year of publication effect; small sex differences in aggression appear to have remained stable over time.

The entire enterprise of studying sex differences has its critics, and a number of relevant issues have been questioned and debated in several recent venues (American Psychologist 1995; Feminism and Psychology 1994; Journal of Social Issues 1997). Criticisms center on the tendency in this work to treat females and males as belonging to polar, global categories that ignore the other social characteristics of individuals (race, class, age, sexual orientation), invoke unsophisticated nature versus nurture explanations, ignore social context, and imply that a difference means a deficit. Despite these criticisms, the research literature continues to grow, and there appears to be a trend ‘‘toward a more contextualized version of genderrelated behavior’’ (Deaux and LaFrance 1998, p. 816).

Contextual influence on sex-related patterns

As is indicated in many of the patterns of sex difference described above, researchers often find important moderators of the sex differences that emerge. For example, experimental work has demonstrated that the way in which a task is framed can affect performance outcomes. Thus, when the ‘‘spatial nature’’ of a spatial reasoning task was emphasized, men outperformed women, but when this aspect of the task was ignored, no sex difference emerged (Sharps et al. 1993). Similar findings have appeared with regard to mathematics performance, as predicted by Steele’s (1997) theory of ‘‘stereotype threat.’’ This perspective suggests that women’s (and minorities) weaker mathematics performance may be based in part on the threat of being judged consistently with negative group stereotypes. When that threat is lifted, performance differences should disappear. In one study, when a mathematics test was described as producing sex differences, women performed worse than did similarly qualified men; when the test was described as not producing sex differences, women and men did not differ in performance (Spencer et al. 1999). Other testing characteristics also may affect the size of sex differences. For example, structured tests that use a free-response format tend to produce smaller sex differences than do those which use a multiple-choice format (Kimball 1989).

The contextual setting also can play a role in the size of sex differences. A case in point is the analysis of leadership effectiveness, where military settings promote sex differences of moderate size (d = −.42) but other organizational contexts produce effect sizes of a much smaller magnitude or reversed direction, ranging from −.07 to .15 (Eagly et al. 1995). An analysis of gender and self-esteem indicated the importance of considering age and ethnicity in discussions of sex differences (Major et al. 1999). No sex difference in self-esteem was apparent before adolescence (d = −.01), but very modest differences began to emerge around ages 11 through 13 (d = −.12). Similarly, sex differences in self-esteem are apparent among whites (d = −.20) but virtually nonexistent among North American minority groups, especially African-Americans (d = .03). Finally, many abilities addressed in studies of sex differences are easily modified by experience. For example, spatial skills improve with exposure, and specific training programs designed to improve spatial skills are equally effective for women and men (Baenninger and Newcombe 1989).

Thus, task characteristics, familiarity, and social context can create sex differences or make them disappear. As Deaux and Major (1987) suggest, the basic behavioral repertoires of women and men are quite similar (e.g., both women and men know how to be aggressive, how to be helpful, how to smile). What men and women actually do is determined less by differential abilities than by the context in which they act. Norms, expectations, the actions of others, and the actor’s goals and objectives may all combine to produce sex differences in behavior in some circumstances.

No doubt people will continue to ask how or why women and men differ, but the answers will never be simple and the explanations will not fall squarely on the side of biology or that of environment. Evolutionary and social role models ultimately may complement each other in their relative emphases on distal (distant) versus proximal (immediate) causal factors. The causal direction of some reasoning may need to be examined further. For example, observed differences between the sexes cannot be used as a simple explanation for broader gender roles. Instead, accepted roles may channel men and women into different patterns of behavior. Whatever the patterns observed, most sex differences will continue to reflect a gendered environment and be subject to change as social factors shift over place and time.

Related Links

This Aricle was Written by

This Article was Published in
Second Edition
A Book by

University of Washington, Seattle


Managing Editor
University of Kansas, Lawrence Copyright 2010 - 2012 © All Rights Reserved
  Home | About | Contact | Links