METHODS

Subjects

The participants were 468 undergraduate students from Brock University, 334 females and 134 males. This research was approved for use with human subjects by the Brock University Research Ethics Board (file number 00-072). Participants did not receive compensation for participating in the research. The questionnaire and raw data are available on request (from M.H.R.).

Procedures

The participants in the present study were requested to complete a questionnaire in which they indicated the minimum and maximum acceptable scores for particular traits in a potential partner, given a particular level of relationship. The possible scores ranged from one to 10 and were considered to be a measure of choosiness. The personal traits assessed were chosen to replicate as closely as possible those used in the early studies by Kenrick et al. (1990, 1993), with a few changes to remove gender-biased terms such as “good housekeeper”; these changes were made in response to suggestions received during preliminary tests of the questionnaire. The characteristics used were kind, understanding, religious, exciting personality, creative, artistic, intelligent, good earning potential, hospitable, wants children, easy-going, free of genetic disease, ethnicity, university graduate, physically attractive, healthy, aggressive, ambitious, emotionally stable, friendly, popular, powerful, sexy, wealthy, good sense of humor, high social status, likes to be in charge, grooming, and hygiene. The participants were also asked to rate themselves on the same criteria. An overall choosiness score was constructed by averaging all of the criteria for a given level of relationship.

The levels of relationship examined were single date, one-night stand, casual dating, steady dating, and marriage. The levels of involvement were derived from Kenrick et al. (1993) but were slightly modified in name. The following definitions were included in the questionnaire: single date, a single date that would not include any sexual activity; one-night stand, this type of situation involves sexual intercourse with a person you did not know and would not see again; casual dating, which refers to partners involved in a relationship including sexual intercourse but not necessarily exclusively with one partner; steady dating, which refers to partners involved in a committed relationship, and would include sexual intercourse; and marriage, which refers to formal marriage.

The instructions on the questionnaire emphasized that all relationships were to be treated hypothetically and that by answering the questions, no subjects implied that they had ever engaged in any of these relationships. We did not ask questions about subjects' sexual orientation, attitudes to sex before marriage, or use of birth control. Two orders of presentation of the levels of relationship (ascending or descending) within the questionnaire were used in order to control for any order effects. To minimize the biasing effect of self-rating on mate score (Kenrick et al., 1993), self scores were assessed at the end of the questionnaire, after the assessments of hypothetical mates.

Statistical methods

Principal components analysis using choosiness scores for the 29 characteristics as variables was undertaken to assess the feasibility of aggregating the characteristics into a smaller number of composite trait groups to simplify the analyses. Kenrick et al. (1990, 1993) aggregated their set of characteristics into eight composites such as dominance and family orientation. In this study, we found that the total variance explained by the first four principal components was only 57%, and that only four of 29 traits (ambitious, good earning potential, friendly, and intelligent) had even moderate factor loadings (more than 0.70), indicating no overall and easily interpretable pattern of trait preferences. Thus, for our data set, principal components analysis was not revealing, and we elected to analyze the data using general linear models (as Kenrick et al. also did).

For choosiness scores, we used ANCOVA to determine the influences of gender (a nominal variable), level of relationship (a ranked, interval variable), and self-rating on choosiness (minimum score for a partner), according to the modelWe analyzed both overall choosiness scores, and the individual choosiness scores for each of the 29 individual characteristics. One-way ANOVA was used to compare gender differences in choosiness or self-scores within particular levels of relationship. All analyses were performed by using SPSS.

RESULTS

For all types of relationships, the overall mean choosiness scores of females were higher than those of males (Figure 1 and Table 1). In both males and females, choosiness tended to increase as risk of parental investment increased, with a decrease in choosiness at the level of one-night stand. The difference between males and females was especially marked at the level of one-night stand (ANOVA: F_1,490 = 16.707, p < .05), which probably accounts for a significant interaction between gender and level of relationship (Table 1).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 1

Overall choosiness scores of male and female subjects ordered by real risk of parental investment as defined in the questionnaire.

Separate analyses of choosiness scores for each characteristic revealed that for some characteristics (Table 2), females were choosier than were males, whereas for others males were choosier than were females, or there was no difference between the sexes. Because self-rating and choosiness scores tend to be correlated (Table 3), we looked for evidence that either gender is choosier for those traits in which that gender has higher self-ratings. As is evident from Table 2, neither males nor females tend to be choosier for traits on which they are more highly self-rated.

DISCUSSION

The reliability of self-report data on questionnaires for assessing human behavioral ecology is somewhat controversial (Daly and Wilson, 1999), because self-assessment may be biased, subjects may lie, or they may not have perfect memory of events. We believe that to assess mate choosiness levels as they are associated with expected parental investment is only possible by using this method. First, lying and mischief-making would simply have increased the variance in responses, making it less likely that we would detect differences between males and females or a correlation between choosiness and parental investment risk. Second, subjects were stating their responses to hypothetical situations and did not need to rely on memory. Third, an observational study would require that individuals be observed in a variety of mate choice situations, and it is difficult to see how this could be achieved without resorting to memory-based accounts of past romantic adventures. Fourth, observational studies of mate choice have one serious drawback, which is that subjects (human or animal) must choose from the available pool of mates (Pawlowski and Dunbar, 1999), rather than from the completely free choice offered by a hypothetical situation. Thus, it is unlikely that in observing actual human mate choices, we would be able to reliably associate them with risk of parental investment. In short, Kenrick's studies and ours actually are at an advantage to be studying “the garrulous animal” (Daly and Wilson, 1999).

The present study and those of Kenrick et al. (1990, 1993) support the general prediction, based on parental investment theory (Trivers, 1972), that mate choice criteria become more stringent as risk of parental investment increases. Moreover, the prediction that females should be choosier than are males was also generally supported, although it was not true for all criteria nor was the difference between males and females statistically significant within every level of relationship (Table 1). In fact, the choosiness scores of males and females become more and more similar as the risk of parental investment increases, so that at the level of marriage, they were almost identical. This suggests that in long-term reproductive relationships, both males and females expect to invest heavily, and perhaps equivalently, in their offspring.

An important difference between the present study and the earlier work of Kenrick et al. (1990, 1993) is that we explicitly defined levels of sexual activity in the relationships. In the present study, this meant that the study subjects did not need to interpret the level of sexual activity involved in the hypothesized relationship, and allowed us to distinguish post hoc between real and perceived risk of parental investment as it relates to choosiness. Moreover, Kenrick et al. (1990, 1993) found that male choosiness declined at the level of one-night stand, whereas we found that the decline actually occurs in both sexes (Figure 1). This is evidence that study subjects do not judge risk of parental investment based solely on the level of sexual activity. This is not surprising given that widespread use of contraceptives may mean that the perceived risk of pregnancy in a one-night stand is zero, even though the risk is actually greater than zero. However, we suspect that contraceptive use is not the only explanation for the decline. As we were informed by the undergraduate testers of our questionnaire, a date requires planning and financial investment, and may be regarded as a potential prelude to a longer-term relationship that will eventually include sex—whereas a one-night stand often requires neither planning nor financial investment. As a result, the perceived risk of parental investment in a date is greater than zero, even though the actual risk is zero, whereas the perceived risk for a one-night stand is zero, even though the real risk is greater than zero. If we re-ordered the relationship levels so that one-night stand preceded single date, we would have perfect support for Trivers' (1972) hypothesis (Figure 1).

We found that male subjects seemed to be choosier for characteristics (e.g., attractive, grooming, and hygiene) related to physical appearance or health, whereas female subjects tended to be choosier for characteristics related to the resources of prospective mates (e.g., earning potential, high social status, university graduate, and wealthy) (Table 2). Although we did not directly address the issue of age in the present study, it appears to have been an indirect criterion, because physical attractiveness in females and resource acquisition in males are both associated with age (Bereczkei et al., 1997; Buss et al., 2000; Buunk et al., 2001; Kenrick and Keefe, 1992; Townsend and Wasserman, 1997). These patterns have been interpreted as evidence that human males choose mates primarily based on female health and child-bearing ability, whereas females choose mates based on male resource acquisition and holding ability (Pawlowski and Dunbar, 1999). Such interpretations must be made cautiously, especially when based on forced choices such as in our questionnaire, as missing sex-specific criteria might be as important or more important than those actually assessed (Herz and Inzlicht, 2002).

Alternative models of human mate choice are the social exchange and biological markets models, in which prospective partners attempt to exchange their assets for those of a partner, in a manner analogous to an economic exchange system. Social exchange theory predicts that individuals will assess their own qualities, and attempt to find partners of the same quality for relationships (Kenrick et al., 1993), and so predicts a correlation between self-scores and minimum mate choice criteria (Table 1). Mate similarity in quality is also predicted by biological markets models in which two classes of traders (males and females) exchange commodities (sex; Noe and Hammerstein, 1995; Pawlowski and Dunbar, 1999). In fact, the parental investment model also predicts mate similarity in quality, especially at higher levels of investment and in monogamous mating systems (as implied by marriage in the present study). So all three models make the same prediction. However, social exchange models do not predict the existence of gender differences in choosiness, except insofar as would be predicted by gender differences in the self scores (Thibaut and Kelley, 1959). In other words, females could be choosier than are males, but this would be because they rate themselves more highly. However, we found that females were choosier, despite having lower self-scores on average than do males. Furthermore, the characteristics for which each gender's self-scores were higher were not necessarily those for which they were choosier. Thus, the social exchange models are contradicted by the patterns revealed in the present study, as well as Kenrick's. Ultimately, our data fit the parental investment model but not the socioeconomic alternatives.