1. Introduction

One characteristic sex difference shared by countless species is that males generally invest more in mating effort and less in parenting effort than do females (Trivers, 1972). Buss and Schmitt (1993) applied this logic to human mating and found, among other things, that men reported preferring more sexual partners over various intervals ranging from 6 months to life.

Pedersen, Miller, Putcha-Bhagavatula, and Yang (2002) questioned the theory and findings of Buss and Schmitt (1993). Pedersen et al. (2002) reanalyzed the data of Buss and Schmitt and also performed their own replication, asking participants to report their desired number of sexual partners over 11 time intervals from the next month to lifetime. Over each time interval, men reported preferring more partners than did women. They note correctly that both their data and that of Buss and Schmitt are highly skewed, calling into question the validity of the parametric tests used by Buss and Schmitt.

Therefore, Pedersen et al. (2002) employed the Maritz–Jarrett procedure to compare the median number of preferred partners (Wilcox, 1996) and found significant sex differences for only 2 of 11 time intervals. After Bonferroni correction, none of the differences were significant. Pedersen et al. conclude that men and women do not differ in the number of desired sexual partners. (It should be noted that Schmitt et al., 2003, argue that the Maritz–Jarrett statistic may be inappropriate for these data.)

Wilcox (2003) reanalyzed a subset of these data using Yuen's method for trimmed means and concluded that the difference between men and women may be significant. He also used S-PLUS to compute the confidence interval for the 75th percentile and rejected the hypothesis that it was the same for men and women.

Schmitt et al. (2003) have since published a massive study of sex differences in sexual desire that sampled 52 nations, six continents, and 13 islands. In that study, they used the appropriate median and Mann–Whitney U tests and found that the differences between men's and women's preferred number of sex partners differed for all time periods.

This paper has four purposes: First, we reanalyze the Pedersen et al. (2002) data and demonstrate that they also clearly show that men report preferring more sexual partners, contrary to their conclusions. Second, we reanalyze the Buss and Schmitt (1993) data and show that they show significant sex differences when the appropriate statistics are used. Third, we demonstrate that considering the entire distribution of preferred number of partners reveals insights about mating preferences that measures of central tendency alone cannot. Finally, we discuss the implication of such data for sexual strategies theory (Buss & Schmitt, 1993) and attachment fertility theory (AFT; Miller & Fishkin, 1997).

2. Method and results

We consider first the Pedersen et al. (2002) data for 30 years because that was the interval that Pedersen et al. chose to graph. Figure 1 of Pedersen et al. shows the preferred number of partners desired by college students over the next 30 years. It is clear that both men and women tend to report desiring one partner and that there is considerable positive skew for both sexes. A bit less obvious, however, is that males tend to predominate in the tail of the distribution.

Our Fig. 1 shows the same data plotted as cumulative percentage of number of preferred partners.

Fig. 1. Cumulative percentage of men and women who desire various numbers of sexual partners over the next 30 years. Data from Figure 1 of Pedersen et al. (2002). The spacing of the data follows Pedersen et al. Note that the spacing of the intervals changes after 20, and the data have been Winsorized (viz. the last bin represents two responses of 150, and 1 of 6000 for males).

Here, a male preference for more partners becomes clear. The Kolmogorov–Smirnov (K-S) test considers cumulative frequencies and, hence, is ideal for these data. The K-S test showed that the sex difference was significant (Z=1.697, P=.003, one tailed). The sex difference was significant for all other time periods as well.

Fig. 2 shows men's desired number of partners as a function of women's for the percentiles 55, 60…95 (cf. Wilcox, 2003). (These percentiles include all participants who desired at least one partner.)

Fig. 2. Number of sexual partners desired by males plotted against the number desired by females at the following percentiles: 55, 60…95. Pedersen et al. 30-year data. Note the different scales on the two axes.

It is obvious that, for all percentiles shown, men desire approximately three times as many partners as women do. The regression equation is Y′=3.0X−1.3, with an r2 of .99. (Although the fit is excellent, inferential statistics are not meaningful because the data are not independent; each data point incorporates the data points to the left. Furthermore, the equation is not valid for fewer than one desired partner.) Reanalysis of the Pedersen et al. (2002) data for the other time periods shows the same pattern of results, as shown in Table 1. To summarize the reanalysis of the Pedersen et al. data, the graphical and statistical methods that they chose obscured the relationships in their data rather than clarifying them.

Table 1. Slope and r2 of the regression line of males' preferred number of sexual partners plotted against number preferred by females over various time periods

	Time period	Pedersen et al., 2002		Buss & Schmitt, 1993
		Slope	r2	Slope	r2
	1 month	2.7	.96	3.5	.92
	6 months	3.1	.95	4.3	.96
	1 year	4.6	.94	6.0	1.0
	2 years	3.0	.99	7.9	.91
	3 years	4.4	.96	8.3	.98
	4 years	9.8	.97	7.2	.96
	5 years	4.8	.86	7.1	.96
	10 years	2.3	.98	6.0	.96
	20 years	3.3	.99	5.9	.96
	30 years	3.1	.99	6.2	.96
	Lifetime	3.4	.91	7.2	.93

Based on the percentiles 45, 50…95 for all participants who desired at least one partner in the time period specified.

We have also reanalyzed the data of Buss and Schmitt (1993) for all time periods and find the sex difference significant by the Kolmogorov–Smirnov test. Plotting them in the manner of Fig. 2 yields a similar pattern as the Pedersen et al. (2002) data does, as also shown in Table 1.

Miller and her colleagues (Miller & Fishkin, 1997; Pedersen et al., 2002) emphasize the importance of the median to represent the typical participant. They make much of the fact that the median number of desired partners is one in their data. There are two problems with this argument. First, the median is only approximately one. By linear interpolation using the cumulative proportions, as is appropriate, the median is .97 for men and .76 for women in their 30-year data. Second, and more important, we can ask how the typical participant relates to the rest of the distribution. Figure 1 of Pedersen et al. strongly implies that the modal participants, that is, those who report preferring one partner, stand in stark contrast with the long tail containing participants who desire more than one. However, let us consider whether the cumulative data are consistent with a normal distribution of desired number of sexual partners. If the z transform of a normally distributed cumulative frequency is plotted against x, the result is a straight line. When we plot the z transform of the Pedersen et al. and the Buss and Schmitt data, we see, surprisingly, that each curve has a linear range between one and about six preferred partners for all but the shortest time periods. A linear portion of such a plot signifies that the data are compatible with a normal distribution over the linear range. For example, the Pedersen et al. 30-year data for men fall on a straight line between one and six preferred partners, as shown in Fig. 3.

Fig. 3. For the Pedersen et al. 30-year data, the z of cumulative percent as a function of number of desired partners. The lines are fitted to the portion of the curve between 1 and 5 (for women) and 1 and 6 partners (for men). Only the data from 0 to 10 partners are shown.

By extrapolation, we find the mean of this line to be 1.5 preferred partners, and the standard deviation to be 7.1. Thus, it is incorrect to conclude that one preferred partner is the norm and that all others represent a different population. (We find evidence of at least one more segment in the Pedersen 30-year data that is normally distributed with a higher mean, but we have not systematically explored the possibility in the rest of the data.)

We determined the range over which the distribution is normal by conservatively judging by eye the portion of the curve that did not deviate systematically from linearity (i.e., that portion of the curve for which deviations from a straight line appear to be random). Each regression line accounted for at least 93% of the variance in the data over that portion. Table 2 shows, for all the Pedersen et al. (2002) and Buss and Schmitt (1993) data, the following estimates by linear interpolation or extrapolation from the normally distributed portion of the curves: (1) the percent who want one or no partners, (2) the mean number of partners desired, (3) the standard deviation, (4) the linear range, and (5) percent of variance accounted for.

Table 2. For all the Pedersen et al. and Buss and Schmitt data, the following estimates by linear interpolation or extrapolation from the normally distributed portion of the curves are given: (1) the percent who want one or no partners, (2) the mean number of partners desired, (3) the standard deviation, (4) the linear range, and (5) percent of variance accounted for

	Time period	Pedersen et al., 2002					Buss & Schmitt, 1993
		Want ≤1	M	σ	Range	r2	Want ≤1	M	σ	Range	r2
	1 month
	Female	NMFa	NMF	NMF	NMF	NMF	NMF	NMF	NMF	NMF	NMF
	Male	79	NMF	3.05	1–4	.99	NMF	NMF	NMF	NMF	NMF
	6 months
	Female	NMF	NMF	NMF	NMF	NMF	86	.94	1.7	1–3	.93
	Male	60	NMF	3.1	1–3	1.0	36	1.2	3.2	1–4	.99
	1 year
	Female	86	NMF	2.6	1–4	.97	NMF	NMF	NMF	NMF	NMF
	Male	50	1.0	2.2	1–3	1.0	40	1.1	3.6	2–6	.98
	2 years
	Female	74	NMF	2.8	1–5	.98	60	.5	2.1	1–4	.96
	Male	46	1.3	2.8	1–4	.98	16	3.4	2.4	1–4	.96
	3 years
	Female	72	NMF	3.9	1–8	.98	40	1.4	1.8	1–4	.96
	Male	43	1.4	3.1	1–5	.98	9	3.6	1.9	1–3	1.0
	4 years
	Female	82	NMF	5.6	2–9	.99	40	1.7	2.8	1–6	.97
	Male	42	1.8	4.4	1–5	1.0	5	4.9	3.3	1–6	.98
	5 years
	Female	62	.1	2.4	1–5	.99	35	2.1	2.8	1–6	.99
	Male	41	2.2	5.0	1–6	.99	9	5.2	3.3	1–5	1.0
	10 years
	Female	59	.2	3.8	1–5	.99	33	2.3	3.3	1–6	.97
	Male	45	1.8	6.3	1–6	.99	10	6.8	4.6	1–6	.99
	20 years
	Female	63	NMF	5.0	1–5	.99	36	2.4	4.0	1–8	.99
	Male	46	1.6	6.7	1–5	.98	11	6.7	4.8	1–6	1.0
	30 years
	Female	62	NMF	5.0	1–5	.98	37	2.4	4.1	1–9	.99
	Male	48	1.5	7.1	1–6	.99	12	7.0	5.0	1–6	1.0
	Lifetime
	Female	53	.6	5.5	1–5	.95	30	3.1	3.9	1–8	.99
	Male	38	3.2	7.2	1–5	.95	12	7.4	5.0	1–6	.99

a Nonmeaningful figure.

Except for the shortest intervals, there is a clear pattern whereby there is a linear range of the distribution that desires from one to about five partners for both men and women. Linear interpolation/extrapolation confirms the analysis shown in Figs. 1 and 2 that men consistently want more partners than women do. Table 2 also shows that the participants of Buss and Schmitt want more partners over all time periods than do those of Pedersen et al.

3. Discussion

Pedersen et al. devote the bulk of their discussion section to the validity of the distinction of Buss and Schmitt (1993) between long- and short-term mating strategies. They state that “most primatologists do not classify humans as having a short term mating system, but instead classify humans as having a predominately long-term mating system, either monogamous or polygynous. Humans fit with monogamous and polygynous primates, and not with short-term maters, on a variety of parameters" (2002, p. 160). Miller and her colleagues propose AFT as an alternative to theorizing that derives from the distinction of Trivers (1972) between mating and parenting effort. According to AFT, men and women both evolved a tendency toward “relatively enduring relationships and few sex differences in mating strategies (Miller & Fishkin, 1997, p. 197).” “We think…the overlap and similarity in [mating preference data] is simply stunning for the most part“ (L. C. Miller, personal communication, 24 June 2003).

However, Hrdy, one of the two primatologists Pedersen et al. (2002) cite, says “there is a dynamic tug of war in [sexual] relationships that is at odds with conventional pipe dreams about humans having an innate tendency to form long-lasting pair bonds, unions in which both sexes have a powerful commitment from within to adhere. Such cases [as those Hrdy describes] make it hard to sustain the illusion that lifelong monogamous families are the natural human condition” (Hrdy, 1999, p. 230).

Gaulin and McBurney (2001) suggest a way of conceptualizing the mixed reproductive strategies that men and women may adopt that differs from both Buss and Schmitt and Miller and her colleagues. Men playing a mixed strategy can invest heavily in one partner but still attempt to attract additional mates. Women playing a mixed strategy can accept investment (e.g., marriage) from one man and seek genes from another. Moreover, in the initial stages of courtship, both men and women may be evaluating the kind of relationship that they may be willing to form. These mixed strategies do not necessarily map onto the long- versus short-term mating distinction. Because neither Buss and Schmitt (1993) nor Pedersen et al. (2002) asked their participants to distinguish between long- and short-term partners in the data at issue here, the relevance of the data to that distinction seems moot. We do not question that both men and women may adopt short-term mating strategies. In particular, differences in criteria for short-term mates between men and women strongly support the existence of short-term mating strategies (e.g., Kenrick, Groth, Trost, & Sadalla, 1993). Nevertheless, consideration of the entire distribution of number of desired mates does not support any interpretation that depends on a distinction between long- and short-term patterns.

Finally, the conclusion that men are more interested in partner novelty hardly rests on these data alone. First, Schmitt et al. (2003), in their massive cross-cultural study, not only replicated the finding on number of desired partners, but also demonstrated that men reported that they require less time knowing a woman before consenting to sex and that more men were currently seeking short-term partners. Men also report higher incidence of having sex the same day they met, having sex after a lot to drink, being unfaithful to a partner, having sex with two people in the same day, having a sexually transmitted disease (Bailey, Kirk, Zhu, Dunne, & Martin, 2000), more sexual fantasies (Ellis & Symons, 1990), and more extramarital affairs (Wiederman, 1997). Evidence not based on self-report includes men's greater use of pornography (Malamuth, 1996) and frequenting of prostitutes (Symons, 1979).

In conclusion, men report preferring more partners than females do, and individuals who prefer one partner come from the same distribution that includes individuals who desire multiple mates. These findings are predictable from the sexual selection theory and are consistent with a host of findings in both animal behavior and human psychology.