1. Study 1: Pupil size preference and sexual selection

In a classic study of pupil size and attraction, Hess (1965) showed that men found drawings of women's faces to be more attractive when the pupils were rendered larger, a finding that has been replicated using various methods Bull & Shead, 1979, Hess, 1975, Stass & Willis, 1967, Tomlinson et al., 1978. Studies that additionally or independently assessed the corresponding responses of women to pupil sizes of men, however, have shown inconsistent results (Bull & Shead, 1979, Stass & Willis, 1967, Tomlinson et al., 1978).

The present study sought to resolve this discrepancy within the concept of sexual selection. Sexual selection seemed an apt framework for several reasons. For one, pupillary dilation has been shown to be foremost an expression and signal of sexual arousal Aboyoun & Dabbs, 1998, Bernick et al., 1971, Hamel, 1974. For another, effects of pupil size on attraction begin at about puberty, which suggests that it has a role in reproductive strategies Bull & Shead, 1979, Flade & Lindner, 1979, Hess, 1975, Tarrahian & Hicks, 1979. Foremost, however, the sexual selection concept provided a precise hypothesis about the total pattern of sex differences.

The hypothesis was based on a fundamental divergence in reproductive strategies between sexes, which has been well established in humans and other species displaying at least some degree of polygyny (Daly & Wilson, 1983, pp. 113–136). In such species, male reproductive success is mainly limited by access to willing, fertile females; thus, it may be assumed that males have been programmed by selection to respond without much equivocation to female sexual arousal and attention. Females, on the other hand, do not require intense male interest to obtain mating privileges. Furthermore, the female reproductive strategy is based on selective mating, and her interests are probably better served by moderate levels of male arousal. Overzealous sexual attention on the part of the male may presage forced copulation, overpossessiveness, excessive sexual jealousy and/or promiscuity, all of which can operate to the detriment of the female's fitness. Thus, it is plausible to assume that females have been programmed by selection to prefer moderation in the sexual attentions and arousal of prospective suitors.

This phenomenon is not unique to humans. Trivers (1985, p. 349), for example, gives an engaging account of the courtship strategy of the male black grouse, noting that subtlety and restraint are critical to success. In fact, the successful male will often turn his back to the female and begin to walk away during courtship, leaving her to take the initiative.

The hypothesis of the present study, then, was that the relationship between pupil size and attraction for men viewing women will be positive and linear, whereas the relationship for women viewing men will be a negative quadratic (inverted U) function. The latter aspect of this prediction may explain inconsistencies in prior studies, in that most used two, rather than three, pupil sizes, and females' preferences may have depended on whether the smaller or larger pupil more closely approached the moderate range. In fact, one study that employed three pupil sizes (Tomlinson et al., 1978) did find the inverted U–shaped function predicted here, though the authors interpreted this as an artefact, occurring with relatively unattractive male models.

1.2. Results and conclusions

Fig. 1 shows mean attractiveness ratings by pupil size for males viewing female photos and females viewing male photos. Using a 2×3 ANOVA, the interaction effect was significant [F(2,57)=4.78, p<.01] but not entirely in the directions predicted. Males viewing females showed the expected linear relationship, with a Tukey test of simple effects showing significant differences at p<.05 between small and medium pupils and between medium and large pupils. Females viewing males, however, showed significantly higher ratings for both medium and large compared to small pupils, but no differences at or approaching significance between them. Analyses of photos for individual models all showed the same general trends as the grouped data.

Fig. 1. Mean attractiveness ratings of males rating females and females rating males by small (S), moderate (M), and large (L) pupil sizes.

Though not directly relevant to the present hypothesis, patterns of same-sex ratings were also investigated. Female raters showed the same positive and linear trend in their ratings of other females as did male raters. Male raters, on the other hand, did not show the same nonlinear trend in their ratings of other males as did their female cohorts. Instead, the function was linear, and the interaction of pupil size and groups (males rating males vs. females rating males) reached borderline significance at F(2,57)=2.99, p=. 05. This may suggest that males project their own mating priorities when assessing the attractiveness of other males to females.

Thus, the hypothesis was partially supported, the exception being that the female sample showed preferences for large pupils about as frequently as middle sized. In an attempt to account for this unanticipated result, we turned to the question of whether large pupil preferences were randomly distributed across female participants or whether they reflected, to some extent, individual differences. Pearson rs for ratings of the large pupil stimulus, between pairs of models, were .20, .26 and .31 (Ns=30). None reached significance, although the Ns may have been inadequate for correlational analyses. A coefficient of reliability was applied to these data for an estimate of the overall magnitude of the intercorrelation across the three models, yielding a moderate but acceptable .51. Thus, it appeared that attraction to large pupils by women may be an individual characteristic.

2. Study 2: pupil size and dating partner preferences in women

A second study followed up on the suggestion above. Its first purpose was to replicate the findings for females of the first study with an adequate N for correlational analyses, in order to better establish whether the positive attraction value of large pupils was idiosyncratic to a subsample of women. Second, given that females preferences for large pupils were subject to individual differences, we wished to explore whether they related to particular predilections for personality attributes of dating partners.

2.2. Results and conclusions

The distribution of attractiveness ratings by pupil size was similar to the prior study, with mean ratings for small, medium and large pupils at 5.07 (S.D.=1.69), 6.57 (S.D.=1.15), and 6.36 (S.D.=1.72), respectively. As in Study 1, ratings for both medium and large pupils were significantly higher than for small [t(59)=5.41, p<.01 and 3.35, p<.05, respectively], but there was no notable difference between large and medium pupil ratings. Similar also to the first study, ratings of the large pupil showed a coefficient of reliability across models of .56. Correlations of ratings of individual photos with each other were .42, .37 and .16 (Ns=60), with the first two significant at p<.05.

Analysis of the DPPT yielded a three-factor result, using a criterion of eigenvalue greater than 3. Based on items with loadings of .4 or greater, factors were labelled, in order, nice guy, mover shaker, and bad boy(Table 1). Scores for individual participants on individual factors were calculated using means across items loading at criterion, and correlations of factor scores with attraction scores for large pupils were assessed. The correlation with bad boy was positive and significant (r=.32, N=60, p<.05), whereas the other two were of zero-order magnitudes.

Table 1. Eigenvalues (EVs) and traits loading at minimum .40 for factors of the Dating Partner Preference Questionnaire

	Nice guy (EV=6.70)	Mover shaker (EV=5.37)	Bad boy (EV=3.04)
	Sensitive (.77)	Foresighted (.71)	Fickle (.71)
	Sentimental (.69)	Inventive (.65)	Frivolous (.62)
	Sympathetic (.63)	Ingenious (.64)	Opportunistic (.58)
	Jolly (.64)	Shrewd (.62)	Hardheaded (.49)
	Helpful (.63)	Industrious (.60)	Handsome (.45)
	Appreciative (.61)	Enterprising (.55)	Confident (.40)
	Warm (.61)	Strong (.53)	Conceited (.40)
	Considerate (.59)	Arrogant (.51)
	Cooperative (.57)	Assertive (.50)
	Friendly (.58)	Forceful (.47)
	Talkative (.53)	Timid (−.44)
	Forgiving (.48)	Dependent (−.43)
	Emotional (.46)
	Aggressive (−41)

Specific trait loadings are in brackets.

Thus, the data suggested that individual women tended to differ in their preferences for both pupil size and dating partner characteristics, and these preferences were related such that women preferring maximally enlarged male pupils also preferred personality attributes ascribed to proverbial “bad boys”.

3. General discussion

Although the predicted distributions of attraction ratings of pupil size by sex of rater and model were only partially obtained in both studies, the broader underlying concept of the studies was supported. As hypothesized on the basis of sexually dimorphic mating strategies, women, overall, did not show the same relative degree of attraction as did men to intense sexual attention signalled by maximally dilated pupils. A subsample of women, however, did find large pupils in male photos most attractive, and this predilection was positively related to preferences for proverbial “bad boys” as dating partners.

The results of the factor analysis of dating partner preferences in Study 2 were largely consistent with Simpson and Gangestad's (1992) analysis of preferences for “romantic partner attributes.” The nice guy factor of the present study appears similar to their personal/parenting qualities factor, and their attractiveness/social visibility factor seemed to be a mix of our mover/shaker and bad boy factors. The emergence of the bad boy factor in the present study but not in Simpson and Gangestad's is likely due to differences in the scales used, in that the subject samples were quite similar.

Simpson and Gangestad's (1992) findings seemed congruent to ours in another aspect, as well. In the present study, females who exhibited preferences for male photos showing heightened sexual arousal also revealed preferences for bad boys as dating partners. Simpson and Gangestad's female subjects who manifested an unrestricted sociosexual orientation, characterized by frequent sex with varied partners, also expressed preferences for males high on their attractiveness/social visibility factor, which is presumed to be somewhat of a counterpart to our bad boy factor. Thus, attraction to bad boys appears to be associated with variables related to the female sexual response. This is consistent with Kruger, Fisher and Jobling's (2003) finding that women were most likely to prefer “cads” to “dads” (their counterpart labels for bad boys and nice guys) for transient sexual affairs, but not for long- or even short-term relationships.

But why would the attraction for bad boys have evolved at all? A frequently quoted explanation is Fisher's (1930)sexy sons theory, which asserts that under the condition of female choice, genes influencing the expression of sexually attractive male traits will persist because they are proliferated by the sons who have inherited them. This explanation, however, begs the question of the adaptive value of bad boy traits for women, the question of why such traits would have become sexually attractive at the outset.

In this vein, several writers Gangestad & Simpson, 2000, Gangestad & Thornhill, 2003, Penton-Voak et al., 2003 have proposed that bad boy characteristics may signal “good genes,” a predisposition for health, vigor, dominance, sexuality, and other attributes favoring male fitness. This line of theory presumes that males of this type are more likely to adopt short-term mating strategies, deterring them from being reliable provisioners and generating the nice guy/bad boy dichotomy. It is also presumed by some writers that dual preferences correspondingly evolved in females, which often lead them to seek nice guys as long-term mates and bad boys for transitory, extra-pair copulations (Scheib, 2001).

The latter notion, however, seems antithetical to the evidence in the present studies of a dual typology among women in terms of nice guy versus bad boy preferences. It may be that the observed duality was not a function of different types of women, but of differences among our sample in terms of life stages or other transient circumstances affecting mating priorities. This appears consistent with Gangestad and Simpson's (2000) concept of “strategic pluralism” in female mate choices.

Finally, an alternative theory maintains that bad boy preferences evolved because attributes in a mate such as toughness and dominance bestowed advantages in resource competition and afforded greater protection to the female and her offspring from predators or other marauding males (cf. Ellis, 1992).