PSYCHOLOGICAL SCIENCE Research Article VOL. 13, NO. 2, MARCH 2002 Copyright © 2002 American Psychological Society 157 EVOLVED SEX DIFFERENCES IN THE NUMBER OF PARTNERS DESIRED? The Long and the Short of It William C. Pedersen, Lynn Carol Miller, Anila D. Putcha-Bhagavatula, and Yijing Yang University of Southern California Abstract—Do men seek more short-term mates than women? Buss and Schmitt (1993) showed a pattern of mean difference in the ideal number of sexual partners men and women desired over various time frames. We replicated these mean sex differences (e.g., ideal number over the next 30 years: Ms = 7.69 and 2.78 for men and women, re- spectively), but in both data sets the sampling distributions were highly skewed. In Study 1, we found few sex differences in medians across time frames (e.g., ideal number over the next 30 years: Mdn = 1 for both men and women). In Study 2, most college men (98.9%) and women (99.2%) said they wanted to settle down with one mutually ex- clusive sexual partner at some point in their life, ideally within the next 5 years. Neither medians in number of partners desired overall before settling down (replicating Study 1) nor medians in short-term partners desired before settling down (Mdn = 0) differed significantly by gender. Rather, men and women concurred: Short-term mating is not what humans typically seek. Do men and women differ in the number of short-term partners that they seek? Sexual strategies theory (Buss & Schmitt, 1993) argues that the answer to this question is yes. Buss and Schmitt (1993) argued that, although men and women should seek both short-term and long- term relationships, “because of a fundamental asymmetry between the sexes in minimum levels of parental investment, men devote a larger proportion of their total mating effort to short-term mating than do women” (p. 205). According to Buss and Schmitt, one prediction that stems from this assumption is the following: “For any given period of time (e.g., a month, a year, a decade, or a lifetime), men will desire a larger number of mates than will women” (p. 210). This desire for a large number of (presumably short-term) mates helps solve a problem that, according to sexual strategies theory, men confront in pursuing short-term, but not long-term relationships (viz. “the problem of part- ner number,” p. 207). In fact, Buss and Schmitt reported a mean differ- ence in the number of partners men (M = 16) versus women (M = 4) desire in the next 30 years. Do these findings support the claim that men seek more short-term mates than do women? Because the question Buss and Schmitt (1993) posed to their participants did not specify the nature of the relation- ships men and women seek, this question cannot be answered directly. Still, the mean number of mates desired is large for men compared with women (e.g., over the next 30 years: Ms = 16 vs. 4). Therefore, this pattern might suggest, at least indirectly, that men typically desire more short-term mates than women. There are many problems, however, with this inference. Later, we return to the problem of directly assessing the number of relationships of each type (e.g., short term, intermediate term, long term) that men and women desire. First, however, we consider the problem with indi- rectly inferring differences for men and women in preferences for short-term mates from mean values of the number of mates desired. Buss and Schmitt’s (1993) indirect inferences rely heavily on t-test comparisons of means. This matters both conceptually and statisti- cally. Buss and Schmitt did not merely argue for sex differences in ex- pressed behavior, they argued that these are due to fundamental evolved sex differences in minimal levels of parental investment (e.g., pregnancy, labor, and nourishing young vs. mere insemination). The logic of sexual strategies theory, so tied to nonoverlapping gender dif- ferences in biological propensities (e.g., to produce sperm or not; to bear offspring or not), seems consistent with the expectation that cer- tainly, the typical man and the typical woman would differ in their sexual strategies. Moreover, open-ended sexual-preference and behavior data are apt to be heavily skewed (Rothspan & Read, 1996). As distributions be- come progressively more skewed, means become increasingly poor measures of the typical response. In such cases, it “has been argued that a comparison of medians is more appropriate . . . because medians generally lie closer to the ‘bulk’ of the data” (Wilcox & Charlin, 1986, p. 264). Not surprisingly, therefore, researchers who report numbers of sexual partners routinely use medians (e.g., Davies et al., 1992; King et al., 1994). Still, the conventional wisdom has been that even when distribu- tions are badly skewed or replete with outliers, resulting in means that are not representative of the “typical” response (Wilcox & Charlin, 1986), the inferences derived from t tests are still accurate (e.g., Bo- neau, 1960; Box, 1954; Glass, Peckham, & Sanders, 1972). Even if psychologists once thought this, today’s statisticians certainly dis- agree, as t tests are sensitive to even moderate violations of the as- sumptions (e.g., normality and homogeneity of variance) of these tests (Cliff, 1993; Wilcox, 1992, 1994, 1996). Violations of these assump- tions can affect both Type I and Type II errors. When assumptions of the t test are violated, “the actual probability of a Type I error can be substantially higher or lower than the nominal  level” (Wilcox, 1997, p. 70). Similarly, Tabachnick and Fidell (1996) noted, “Especially worrisome is that an outlier can produce either a Type I or a Type II er- ror, with no clue in the analysis as to which is occurring” (p. 381). Furthermore, nonrobustness and statistical inferences are apt to be- come more problematic as “skewness or kurtosis of the sampled popu- lation departs increasingly from its normal-distribution value” (Bradley, 1982, p. 87), rendering t tests inferior to other two-sample (e.g., non- parametric) tests (Cressie & Whitford, 1986; Neave & Granger, 1968; Wilcox, 1990). Therefore, medians and inferential tests associated with them may—for a variety of conceptual and statistical reasons— provide a better vehicle to examine a hypothesis about evolved, bio- Address correspondence to Lynn Carol Miller, Annenberg School for Com- munication, University of Southern California, Los Angeles, CA 90089-0281; e-mail: Lmiller@rcf.usc.edu.