1.1. Testing the hypotheses

Beginning in the early 1960s, dramatic changes in the roles and opportunities available to girls and women have occurred in many Western societies, including the United States. Changes include increased participation and acceptance in the labor force, higher education, political office holding, and organized athletics (Browne, 2002, Shulman & Bowen, 2001, Twenge, 2001). These changes offer an opportunity to test the explanatory power of the sociocultural conditions hypothesis and the evolved predispositions hypothesis. The former predicts convergence in social psychological attributes, while the latter predicts that sex differences will remain stable.

Determining whether there has been convergence in social psychological attributes has proven difficult, however. Meta-analyses of gender stereotypes, sex typing, and personality have produced evidence of convergence (Twenge, 1997, Twenge, 2001; cf. Diekman & Eagly, 2000) but have also indicated stability or even increased differentiation (Feingold, 1994, Lueptow et al., 2001). Furthermore, the characterization of psychological attributes in these meta-analyses has relied almost exclusively on written assessments rather than behavior under real-world conditions. Thus, it is unclear whether these kinds of studies could, in principle, provide strong evidence of changes in social psychological attributes, as opposed to changes in self-presentation (Johnson, 1981, Twenge, 1997).

Here I investigate male and female performances in distance running, an activity in which competitiveness can be assessed at the population level and that permits straightforward historical comparisons. In Study 1, I argue that the occurrence of relatively fast distance runners in matched populations should estimate a sex difference in competitiveness. I then demonstrate that in matched populations of U.S. runners, there are presently many more relatively fast males. In Study 2, I investigate whether this sex difference in the occurrence of relatively fast runners is in the process of diminishing due to the growth in opportunities and incentives for female athletes. I show that the number of fast female runners is not increasing, either absolutely or relative to males, implying that the sex difference in competitiveness is not diminishing.

2.1. Methods

Open data were taken from Track & Field News (2004); D1 collegiate data were taken from an unpublished national championships qualifiers list; national-class high school data were taken from Shephard (2003). High school cross-country data were gathered from state athletic association internet sites, most of which were linked from a single source (Dyestat, 2004). In most states, races were run in separate divisions (e.g., AA, AAA); times were pooled across all divisions, and the fastest 40 performances were used for analysis. The number of cross-country participants for each state was taken from the National Federation of State High School Associations (2004). To investigate the confound of race duration (see below), data from state championship high school outdoor track meets were taken from state athletic association internet sites (Dyestat, 2004). Comprehensive data sets for both studies are available upon request.

World records were taken from IAAF (n.d.). The 10 fastest world performers of 2003 were taken from Track & Field News (2004). High school distance races are 1600 and 3200m, but these are not contested at the Open level; world records and 10-Fastest standards for these distances were estimated from world-class performances at 1500 and 3000m, following conversion guidelines from Track & Field News (2000).

To control for the effects of race duration on relative speed (see Section 2.2), least squares linear regression was first used to assess the relation between (log-transformed) seconds and speed relative to sex-specific world records or 10-Fastest standards. Regressions were calculated for each sex at each of the three competitive levels (national-class high school, Division 1 collegiate, and Open), using, for each point, the mean race duration and relative speed of the first 40 males and females represented in each event performance list. Regressions were similarly calculated for the 11 states that conduct “all-state” outdoor high school championship track meets; each point represented the mean race duration and relative speed of the first five finishers in each event in 2003. Although intercepts correlated with competitive levels ([competitive levels were defined as Open, D1 collegiate, national-class high school, and states, in descending order of number of participants] males: n=14, rs=.82, p<.001; females: n=14, rs=.78, p<.001), regression coefficients did not (males: n=14, rs=.40, p=.15; females: n=14, rs=.09, p=.92). The uniform decrease in ratio proximity to world-class standards as races lengthen indicates that a constant can be used to establish duration-corrected male performances at all distances and competitive levels. The constant used here was 0.043, which is the upper 95% confidence limit of the regression coefficient across all competitive levels; calculations of this constant using world records and 10-Fastest standards were identical to three decimal places. Therefore, to produce a duration-corrected male performance measure, one multiplies the individual male time by the ratio of mean female to mean male duration for that event and context (e.g., typically 1.10–1.17), log-transforms this value, and then multiplies it by the constant of 0.043. The individual's original time is also log-transformed and multiplied by the constant; then, the latter number is then subtracted from the former. This remainder is then added to the originally computed relative performance measure (e.g., 10-Fastest ratio), and the duration-corrected relative speed measure is obtained.

Prior to calculating the variance for a sample of performances, durations were normalized by first calculating the mean duration of the 40 performances in that particular event list or race, dividing each performance by this mean and then multiplying by 100. Durations were normalized because, all else being equal, larger numbers produce larger variances. Normalized variance was computed for performances ranked 6th–40th because preliminary analyses indicated that the fastest five performances were disproportionately influential. However, the general result of male samples tending to be less variable was obtained no matter which performances were included.

Comparisons of male and female relative speed measures among the 40 performances in each event list or race were made using t tests with separate variance estimates. To test for sex differences in normalized variance, I calculated the ratio of male to female normalized variance (Sokal & Rolf, 1995, p. 189). All statistical modeling in both studies was conducted with Statistica 6.0 (Statsoft, Tulsa, OK). All statistical tests were two tailed and α was set at .05.

2.2. Results

2.2.1. National class runners

For each event commonly contested by national-class high school, D1 collegiate, and Open runners, I calculated the best performance of the 40 fastest males and females as a ratio of the relevant sex-specific world record (current January 1, 2004). In 20 of 21 cases, the performance of the 40 fastest males was significantly closer to the sex-specific world record than was the performance of the 40 fastest females. However, in 17 of 21 cases, the performance of the first ranked male competitor was closer to the world record than was the performance of first ranked female competitor (binomial probability <.01), and, in a paired t test, the first ranked males were significantly closer to world records overall [t(20)=3.70, p<.01]. These findings are consistent with the suggestion that some female world records could be anomalously fast (Seiler & Sailer, 1997).

Because world records could provide a biased standard for comparison, I repeated calculations of relative speed using an alternative world-class standard, the mean best time of the 10 fastest performers in the world in 2003 (hereafter “10-Fastest”). In 20 of 21 cases, the performances of the 40 fastest males were significantly closer to the 10-Fastest standard than were the times of the 40 fastest females (Fig. 1; Table 1). Despite this consistent sex difference across the fastest 40 competitors, there was no sex difference in the proximity of the first competitor's performance to the 10-Fastest standard [11 events females closer; paired t test, t(20)=0.41, p=.68]. More crucially, in seven of the eight Open events, the first ranked female was closer to this standard. Thus, the 10-Fastest standard is not biased against females, and I therefore report results based on it for the rest of this study. Nevertheless, all analyses in Study 1 were also conducted with world records. Results were highly similar, except that the sex difference was generally far larger with world records.

Fig. 1. Plots of performance rank and relative speed for a representative sprint and distance event, for three competitive levels. Open and collegiate runners commonly contest 1500m races, whereas high school runners generally contest 1600m races. Females are represented by filled boxes (▪) and males by unfilled boxes (□). Relative speed is defined as the ratio of the performance to the event-specific, sex-specific 10-Fastest standard (see text).

Table 1. Comparison of the top 40 male and female high school, collegiate, and Open runners for commonly contested distances

	Context	Distance (m)	Mean male 10-Fastest ratio	Mean female 10-Fastest ratio	pa	Mean male duration corrected 10-Fastest ratio	pb	Variance ratioc	pd	Duration-corrected bias ratio
	Open	100	1.02	1.03	<.001	1.02	<.01	1.78	<.05	2.1
	Open	200	1.02	1.03	<.01	1.02	<.05	1.81	<.05	1.8
	Open	400	1.02	1.04	<.001	1.02	<.001	0.95	NS	3.1
	Open	800	1.04	1.05	<.001	1.04	<.01	3.89	<.001	2.7
	Open	1500	1.05	1.06	NS	1.05	NS	8.32	<.001	1.7
	Open	5000	1.06	1.07	<.05	1.07	NS	4.74	<.001	1.7
	Open	10 000	1.06	1.09	<.001	1.07	<.001	1.71	NS	2.0
	Open	42 195	1.09	1.13	<.001	1.10	<.001	3.59	<.001	3.6
	Collegiate	100	1.03	1.04	<.01	1.03	NS	1.96	<.05	1.8
	Collegiate	200	1.03	1.04	<.001	1.04	<.05	2.43	<.05	1.7
	Collegiate	400	1.03	1.05	<.001	1.03	<.001	1.74	NS	2.9
	Collegiate	800	1.05	1.07	<.001	1.05	<.001	3.38	<.001	2.7
	Collegiate	1500	1.06	1.08	<.001	1.06	<.001	6.02	<.001	3.6
	Collegiate	5000	1.08	1.11	<.001	1.09	<.001	6.87	<.001	3.6
	Collegiate	10 000	1.08	1.13	<.001	1.08	<.001	2.42	<.01	4.0
	High School	100	1.06	1.07	<.001	1.06	<.01	1.40	NS	2.5
	High School	200	1.06	1.07	<.01	1.06	<.05	1.04	NS	2.1
	High School	400	1.05	1.07	<.001	1.06	<.001	1.93	<.05	3.6
	High School	800	1.08	1.10	<.001	1.08	<.001	2.28	<.01	5.0
	High School	1600	1.10	1.14	<.001	1.10	<.001	0.64	NS	13.3
	High School	3200	1.11	1.15	<.001	1.11	<.001	0.94	NS	8.0

a Probability of sex difference between 10-Fastest ratios. b Probability of sex difference between duration-corrected male and female 10-Fastest ratios. c Normalized variance of males divided by normalized variance of females. d Probability of a significant difference in variance ratio.

Races that are longer in distance and duration are associated with relatively slower performances, suggesting that these results might reflect that, for any given distance, female races are longer in duration. However, even after male running performances were duration corrected, they were significantly closer to the 10-Fastest standard in 18 of 21 cases (Table 1).

The prediction of less variance among top male performances was also supported. Specifically, the ratio of female to male normalized variance was greater than 1 in 18 of 21 cases, and in all 14 cases where the variance ratio reached significance, males were less variable (Fig. 1; Table 1). Thus, not only are the performances of elite males closer to world-class standards than are the performances of corresponding females, but these males generally contend with more competitors of similar ability.

A straightforward explanation for these findings is that more males participate in competitive running. In fact, there were 9% more female teams and 2% more female participants in D1 collegiate track and field in 2003 (NCAA Research, 2004). In high school, there were 1% more male teams and 21% more male participants (National Federation of State High School Associations, 2004). This male bias in high school participation is longstanding (National Federation of State High School Associations, 1980) and might be argued to ultimately produce small but consistent sex differences at all levels.

One method of estimating the magnitude of the sex difference is to tabulate how many females meet or beat the relative performance of the 40th ranked male and to then calculate a bias ratio, the ratio of 40 over the number of females meeting this performance. For instance, in the Open 100m, 19 females met or exceeded the duration-corrected 10-Fastest ratio of the 40th ranked male (Fig. 1, upper left panel), resulting in a bias ratio of 2.1; this means that twice as many males as females ran relatively fast in this event. For the 21 events detailed in Table 1, the mean bias ratio was 3.5 (median=2.7). These bias ratios are far too large to be explained by slight differences in high school participation.

2.2.2. State championship cross-country runners

In the 26 states where males and females competed on the same 5000m course on the same day, the duration-corrected time of the fastest 40 male finishers was significantly closer to the 10-Fastest standard in every case. The ratio of female to male normalized variance was greater than 1 in 21 of 26 cases, and in all nine instances where normalized variance differed significantly, males were less variable. The mean bias ratio across the 26 states was 5.1 (median=3.6). Although more males than females participated in cross-country in these states (mean 25% male bias; National Federation of State High School Associations, 2004), these participation differences are much too small to account for the more than threefold difference in the number of relatively fast runners.

Another approach to estimating how many more males than females ran relatively fast in cross-country races is to examine the relationship between participation and performance across states. Although courses vary in difficulty, for both males and females, states with more participants exhibited faster relative times (Fig. 2A) and less variation (Fig. 2B) among their fastest 40 finishers. For relative speed, the regression slopes did not differ significantly between the sexes [homogeneity of slopes model, interaction Sex×Log-Participants: F(1,48)=0.1, p=.90], but the intercepts did [ANCOVA, sex: F(1,49)=21.1, p<.001]. Normalized variance exhibited different slopes for males and females [F(1,48)=6.7, p<.05], but sex remained a significant main effect in a separate slopes model [F(1,48)=8.9, p<.01]. The crucial point is that the regression equations (Fig. 2) indicate that, to match the characteristics of a hypothetical average-sized male population, a female population would have to be 2.3 (normalized variance) or 3.2 (10-Fastest) times as large.

Fig. 2. Regressions of state cross-country population size on (A) mean relative speed of the fastest 40 finishers in the championship race and (B) variance among finishers 6–40. Regression equations: Female 10-Fastest=1.678−0.099×log-participants; male duration-corrected 10-Fastest=1.613−0.094×log-participants; female log-normalized variance=1.753−0.396×log-participants; male log-normalized variance=1.199−0.274×log-participants.

2.3. Discussion

The fastest 40 male and female performances for 47 event lists or races were examined relative to a sex-specific world-class standard, the mean best time of the 10 fastest performers in the world in 2003 for that event. In all 47 cases, the mean performance of the males was closer to the 10-Fastest standard, and in 44 cases, this difference was statistically significant.

This sex difference cannot be ascribed to the 10-Fastest standard being somehow biased against females because the top-ranked U.S. female in each performance list was generally closer to this standard than was their male counterpart. Further evidence that the sex difference is not an artifact of world-class standards is that the variance among performers was less among males than females in 39 of 47 cases, and, in all 23 cases where variance differed significantly, males were less variable. Because variance estimates are calculated completely independent of world-class standards, this sex difference cannot be attributed to them.

Bias ratios indicated that two to four times as many males as females ran relatively fast in each event list or race, a difference that is far too large to be accounted for by slight (or nonexistent) differences in the number of male and female runners. Furthermore, using a technique that fully controlled for population size, a sex difference of the same magnitude was estimated in cross-country runners.

The proportional sex difference in the number of relatively fast runners demonstrated here is both large and widespread, occurring across a range of competitive levels, events, and geographical areas. In addition, analyses of large U.S. road races indicate the same sex difference (Deaner, in preparation). To my knowledge, this sex difference has never been previously documented. Nevertheless, evidence of the phenomenon can be readily found. For instance, the qualifying time for the 2004 U.S. Olympic marathon trials of 2:48:00 for women (USATF, n.d.a) was 124% of the world record (2:15:25) and 119% of the 10-Fastest standard (2:20:55). The corresponding time for men of 2:22:00 (USATF, n.d.b) was 114% of the world record (2:04:55) and 113% of the 10-Fastest standard (2:06:12). The bias ratio shown in Table 1 indicates that if females had a qualifying standard proportionally similar to the 10-Fastest standard for males, about one third as many females as males would qualify for the Olympic trials. If females had a qualifying standard proportionally similar to the world record for males, about one fifth as many females would qualify.

3.1. Methods

Historical information on Open and collegiate performances was obtained from “Track & Field News.” Open performance lists were presented in the December issue of each year or in the January or February issue of the following year. Collegiate performances were generally presented in July or August issues. Historical information on high school performances was taken from large newspapers (e.g., L.A. Times, Star Ledger, and Boston Globe). High school championship meets occurred in early June, and results were generally printed on the following day. Some performances from state championship high school track meets required conversions for hand timing or the use of English rather than metric distances; conversions followed “Track & Field News” (2000).

3.2. Results

3.2.1. Open runners

I first tested whether the time of the 10th best female Open performer has improved since 1976, the first year for which such data are available. In each of the eight commonly contested events, females improved significantly, both in absolute terms and relative to the corresponding male performance (all distances: least-squares linear regression, p<.05; Fig. 3). Visual inspection indicated, however, that most improvement occurred in the 1970s and early 1980s. I therefore repeated these analyses using only data from the last 20 years for which they were available, 1984–2003. Performance improved in the 5000m (absolutely and relatively) but worsened in the 800m (absolutely and relatively), marathon (absolutely) and 100m (relatively; Table 2). Because the 5000m marathon was rarely contested by females prior to the 1990s and was not an Olympic distance until 1996 (Runner's World, 1996, Track & Field News, 1988), improvement in this event may not reflect an increased number of fast female distance runners.

Fig. 3. Plots of year and race duration for a representative sprint and distance event, for three competitive levels. Open data represent the 10th ranked U.S. performer; collegiate data represent the fifth place finisher at the D1 Outdoor National Championship meet; high school-CA data represent the fifth place finisher at the California All-State Outdoor State Championship meet. See Table 2 for regression equations.

Table 2. Absolute and relative changes in performance for high school, collegiate, and Open runners for commonly contested distances

	Context	Distance (m)	Sample	Female			Male			Female–Male
				β	t	p	β	t	p	β	t	p
	Open	100	1984–2003	−.15	−0.63	NS	−.79	−5.41	<.001	.53	2.64	<.05
	Open	200	1984–2003	−.14	−0.59	NS	−.09	−0.38	NS	−.10	−0.43	NS
	Open	400	1984–2003	−.27	−1.18	NS	.04	0.16	NS	−.35	−1.59	NS
	Open	800	1984–2003	.64	3.53	<.01	.11	0.46	NS	.61	3.29	<.01
	Open	1500	1984–2003	.36	1.63	NS	.33	1.49	NS	.18	0.77	NS
	Open	5000	1984–2003	−.52	−2.58	<.05	.33	1.48	NS	−.61	−3.28	<.01
	Open	10 000	1984–2003	−.18	−0.77	NS	.36	1.66	NS	−.39	−1.82	NS
	Open	42 195	1984–2003	.64	3.58	<.01	.77	5.11	<.001	.16	0.68	NS
	Collegiate	100	1982–2004	.29	1.37	NS	.23	1.10	NS	.08	0.36	NS
	Collegiate	200	1982–2004	.03	0.13	NS	.60	3.42	<.01	−.44	−2.27	<.05
	Collegiate	400	1982–2004	−.23	−1.07	NS	−.33	−1.62	NS	−.04	−0.19	NS
	Collegiate	800	1982–2004	−.07	−0.31	NS	−.30	−1.45	NS	.12	0.56	NS
	Collegiate	1500	1982–2004	−.10	−0.48	NS	.10	0.48	NS	−.19	−0.87	NS
	Collegiate	5000	1982–2004	.10	0.44	NS	−.05	−0.23	NS	.28	1.30	NS
	Collegiate	10 000	1982–2004	.27	1.29	NS	.17	0.81	NS	.01	0.06	NS
	High school-CA	100	1980–2004	−.32	−1.61	NS	−.30	−1.51	NS	−.16	−0.77	NS
	High school-CA	200	1980–2004	−.75	−5.40	<.001	−.31	−1.55	NS	−.49	−2.72	<.05
	High school-CA	400	1980–2004	−.02	−0.11	NS	.13	0.63	NS	−.10	−0.46	NS
	High school-CA	800	1980–2004	.41	2.18	<.05	.46	2.52	<.05	−.10	−0.46	NS
	High school-CA	1600	1980–2004	.47	2.53	<.05	.47	2.56	<.05	−.02	−0.10	NS
	High school-CA	3200	1980–2004	.24	1.20	NS	.38	1.98	NS	.15	0.75	NS
	High school-NJ	100	1980–2004	−.82	−6.85	<.001	−.72	−4.96	<.001	−.52	−2.91	<.01
	High school-NJ	200	1980–2004	−.70	−4.75	<.001	−.50	−2.76	<.05	−.34	−1.76	NS
	High school-NJ	400	1980–2004	−.49	−2.67	<.05	−.17	−0.82	NS	−.36	−1.82	NS
	High school-NJ	800	1980–2004	−.04	−0.20	NS	.26	1.28	NS	−.22	−1.07	NS
	High school-NJ	1600	1980–2004	.46	2.49	<.05	.19	0.94	NS	.38	1.97	NS
	High school-NJ	3200	1980–2004	.40	2.11	<.05	.38	1.99	NS	.19	0.91	NS
	High school-MA	100	1980–2004	.00	0.02	NS	−.10	−0.49	NS	.08	0.40	NS
	High school-MA	200	1980–2004	−.42	−2.20	<.05	−.21	−1.00	NS	−.38	−1.92	NS
	High school-MA	400	1980–2004	−.49	−2.62	<.05	−.44	−2.29	<.05	−.03	−0.16	NS
	High school-MA	800	1980–2004	−.10	−0.48	NS	−.30	−1.46	NS	.08	0.36	NS
	High school-MA	1600	1980–2004	−.28	−1.34	NS	−.07	−0.34	NS	−.30	−1.48	NS
	High school-MA	3200	1980–2004	−.28	−1.34	NS	−.01	−0.05	NS	−.32	−1.60	NS

3.3. Discussion

With the exception of high school sprints, there is no indication of improvement among elite or sub-elite female runners since the early to mid-1980s, indicating that the number of fast female runners has not increased over this time. It might be argued that statistical power was insufficient to detect such improvement. However, the approach used here is reasonably sensitive, as is shown by the fact that improvements were detected among high school sprinters and among all Open runners when data from the late 1970s and early 1980s were included. Moreover, if the hypothesis of female improvement were true but the data too variable to reveal significant effects, one would predict that the regression coefficients would be predominantly negative. However, in terms of absolute performance, only 7 of the 18 female distance tests (≥800m) had negative regression coefficients (Table 2). Even more crucially, of the six significant regression coefficients, five were positive, indicating a decrease in the number of fast runners. Relative female performance also showed no trend towards a predominance of negative regression coefficients (only 8 of 18 coefficients negative, only 1 of 2 significant coefficients negative; Table 2). Therefore, these analyses collectively provide strong evidence that the number of elite and sub-elite female distance runners has remained stable since the early to mid-1980s.

This stability is surprising, given the well-documented gains in participation and incentives for female athletes and the celebrated accomplishments attributed to these gains (e.g., Gavora, 2002, Lopiano, 2000, Shulman & Bowen, 2001). However, distance running differs from most other sports in that widespread opportunities for female participation occurred relatively early. Specifically, in U.S. high schools in 1980, 46% of high school outdoor track and field teams were female, and the number of female participants was only 10% less than in 2003 (National Federation of State High School Associations, 2004). At the collegiate level in 1982, 38% of teams were female, and the number of female participants was only 40% less than in 2003 (NCAA Research, 2004). These data indicate that most talented female runners could have found opportunities to excel by the early 1980s. Of course, compared with their male contemporaries and females of today, these athletes almost certainly would have faced many disadvantages, including fewer scholarship opportunities, smaller travel and equipment allowances, and poorer professional prospects. Apparently, these disadvantages did not undermine the motivation of these runners.

4.1. Alternative explanations for a failure to increase competitiveness

One possibility is that, although progress in leveling the playing field has not yet increased the competitiveness of female distance runners, eventually it will. Because the growth in opportunities and incentives for female athletes first began on a large scale in the 1970s and may be still occurring today, there has been less than a generation for these altered sociocultural conditions to become effective. Thus, if the historical analyses in Study 2 were conducted in another generation, there might be evidence of an increasing number of fast female distance runners. For this reason, and because of the competition for athletes hypothesis (Section 3.3.1), historical analyses must be repeated in the future.

Another possibility is that the sex difference in competitiveness could be eliminated if the society were altered appropriately but that increased opportunities and incentives for female athletes is not an appropriate modification. Specifically, one might argue that competitiveness in athletics is dependent on receiving childhood encouragement from parents and peers (Eccles & Harold, 1991) and that the increased female athletic opportunities at high schools and colleges has not actually altered the general level of encouragement. To my knowledge, there is no cross-temporal data on childhood athletic encouragement, so this issue requires study.

Eagly's social role theory Eagly, 1987, Eagly & Wood, 1999, arguably the most comprehensive version of the sociocultural conditions hypothesis, specifically holds that increased female participation in traditionally male-occupied social roles, including in athletics (Diekman & Eagly, 2000), will ultimately lead to females developing more instrumental or agentic psychological attributes, including competitiveness. In addition, in addressing females' lesser interest in competitive athletics, several U.S. courts have adopted a stance of “If you build it, they will come,” essentially claiming that if policies would ensure that males and females participate in intercollegiate athletics at the same rate, sex differences in competitive athletics interest would disappear (Gavora, 2002). Therefore, although alternative versions of the sociocultural conditions hypothesis can be formulated, the failure of increased opportunities and incentives for female distance runners to increase their competitiveness must be considered a genuine failure for the sociocultural conditions hypothesis.

Finally, in assessing claims that the sociocultural changes occurred too recently or were not appropriate, it is notable that several social scientists have posited that sex differences in physical strength may be decreasing due to these changes (Diekman & Eagly, 2000, Dowling, 2000, Messner, 2002). Although there is no empirical support for these suggestions (Shephard, 2000), their occurrence shows that the expectation of enhanced female competitiveness is realistic.

4.2. The generality of the sex difference in sport competitiveness

The hypothesis that males have an evolved predisposition to directly compete for status is based on a wealth of empirical observations (Browne, 2002, Buss, 2004, Campbell, 2002, Maccoby, 1998). For instance, compared with their female counterparts, young boys more frequently engage in activities with clear winners and losers (Lever, 1976), older boys compete more often in sports, particularly in informal contexts (Kirshnit, Ham, & Richards, 1989), and men typically show greater commitment to attaining high-status jobs (Browne, 2002). In this context, it is not surprising that proportionally more male distance runners are motivated to seek competition (Callen, 1983, Johnsgard, 1985, Leedy, 2000, Ogles & Masters, 2003, Walter et al., 1989) and, in fact, run relatively fast.

These observations imply that proportionally more males will be highly competitive in sports besides distance running. Study 1 provides some support for this prediction among sprinters, although Study 2 indicates that this sex difference could be shrinking. In addition, there are reasons to expect that there will be some sports with no sex difference in competitiveness or where competitive females will outnumber competitive males. For instance, Sargent, Zillman, and Weaver (1998) found that males preferred viewing sports characterized by direct competition and elements of combat, whereas female spectators were partial to stylistic sports such as gymnastics, diving, and figure skating. Thus, the generality of the sex difference in sport competitiveness requires empirical investigation.

4.3. Policy implications

These studies provide substantial evidence that there is a large sex difference in the number of competitive distance runners and that this difference is not diminishing. This stability, despite substantial growth in opportunities and incentives for female athletes, strongly suggests that a sex difference in competitiveness is not merely due to sociocultural conditions but instead partly reflects evolved psychological predispositions. These data thus provide a powerful challenge to the assumption that all sex differences in sports interest and competitiveness should be ascribed to past or present discrimination. As this assumption lies at the heart of intercollegiate athletic policies in the United States, this policy may require revision (Gavora, 2002, Rhoads, 2004).

Recognizing that males and females may generally have different levels of interest in (some) competitive sports should not deter attempts to equalize the opportunities. As reviewed above, in the United States, opportunities and external incentives in distance running are not biased towards males. Nevertheless, biases in opportunities and incentives almost certainly remain in other sports and, when demonstrated, should be rectified.