1. Introduction

Recent studies of female mate preferences suggest that women display strategic pluralism; that is, they possess context-specific psychological adaptations that lead them to favour different mate characteristics in different mating contexts. Much of this research has involved assessing how female preferences differ across hypothetical short-term or long-term relationship scenarios, and this has proved a successful research strategy. The short-term vs. long-term dichotomy is intuitively plausible and has generated data indicating that women's preferences for male partners do indeed change in different mating situations. Moreover, these apparent facultative shifts in female preferences provide support for adaptationist hypotheses.

Men provide indirect genetic benefits and, often, direct phenotypic benefits to their offspring in the form of paternal investment. It is likely that women have faced selection pressures that favour accurate detection of cues to both these benefits (as is the case in many nonhuman species; Jennions & Petrie, 2000). As men with cues to high genetic quality seem likely to provide low levels of paternal investment, and vice versa, women may derive fitness benefits from adopting conditional mate choice strategies (Gangestad & Simpson, 2000).

Contextual variation in female preferences for male faces also supports the theory that trade-offs between genetic and phenotypic benefits are occurring in female mate choice. Trade-off theories would predict, and preference data demonstrate, that prosociality should be more important in a long-term than in a short-term relationship, and that putative good genes cues should be prioritized in short-term situations (Little, Jones, Penton-Voak, Burt, & Perrett, 2002).

Women's preferences for faces seem consistent with strategic pluralism: depending on contextual factors, women seem to either prioritise cues to good genes (e.g., facial masculinity) or cues to future paternal investment (e.g., attributions of prosocial personality) when choosing a mate (e.g., Penton-Voak et al., 1999). However, one potential criticism of this body of work is that there is a lack of good evidence that the attributions that women make, particularly to faces, are accurate enough to be useful. Although there is increasing evidence that certain aspects of facial appearance are associated with genetic and/or endocrine factors that are likely to be fitness related (e.g., Roberts et al., 2005, Penton-Voak & Chen, 2004), the links between the personality attributions made to faces and future parental investment (the key trait in the strategic pluralism argument) are poorly understood. There is some evidence that personality attributions to facial characteristics have a ‘kernel of truth’ (see Zebrowitz & Collins, 1997, for review) but accuracy tends to be low (e.g., Penton-Voak, Pound, Little, & Perrett, 2006).

One study (Roney, Hanson, Durante, & Maestripieri, 2006) has directly addressed the question of whether women can accurately ‘track’ men's genetic and paternal quality from facial images. In this study, men were photographed, provided a saliva sample to assay circulating testosterone levels, and completed an ‘interest in infants task’ (in which they had to choose either an adult or infant face as most preferred in a series of forced choice trials). The rationale of expecting women to be able to ‘track’ variation in these measures is that testosterone levels are a putative correlate of genetic quality, while the ‘infant interest’ task is a plausible measure of willingness to provide paternal investment. Roney et al. (2006) found that men's scores on the ‘infant interest’ task predicted women's ratings of how much the men liked children. Both the men's scores on the ‘infant interest’ task and women's ratings of likely paternal quality predicted the attractiveness of men in a long-term attractiveness judgment, whereas the testosterone levels and masculinity ratings (which were significantly related) predicted women's attractiveness ratings in the context of a short-term relationship. This paper, then, provides the best evidence to date that women can discriminate between cues to genetic and parental quality, and make strategic mate preference decisions based on these traits.

The present study aimed to expand upon the findings of Roney et al. (2006). In addition to collecting images of men for rating by women on a variety of traits and attractiveness, we collected two measures of the men's affinity for children. The first was a brief questionnaire measure, analogous to, although less sophisticated than, the task of Roney et al. The second measure was derived from a behavioral task in which the men were asked to produce speech as though talking to a child. This is a novel and potentially useful way to assess a man's potential interest in, and likely success in interaction with, their offspring. When parents talk to their children, they adopt a speech register that is characterized by elevated pitch, exaggerated prosody, slower speech rate, shorter utterances, and simplified morphosyntax (Fernald et al., 1989). This speech register, referred to as child-directed (CD) speech, serves to regulate the child's arousal and emotion, and to direct the child's attention. There is also increasing evidence that CD speech seems to be specifically suited to the task of language learning, as the various speech adjustments provide multiple cues that can facilitate the child's language acquisition (Kuhl et al., 1997, Kempe et al., 2003).

Most studies of CD speech production by adults have examined the speech of mothers to their small children, and fewer studies have included fathers, grandparents, and older siblings. These studies show, however, that CD speech produced by other relatives tends to be similar in terms of its acoustic properties to mothers' CD speech (Golinkoff & Ames, 1979, Keller et al., 1982, Lipscomb & Coon, 1983). The apparent similarities in the way adults talk to children have led to the suggestion that CD speech is a form of ‘species-specific didactic support’ (Papousek, Papousek, & Haekel, 1987). The ability to perform CD speech may be a useful behavioral indicator of a man's affinity for children. Given that CD speech has been shown to be beneficial for child development, the degree to which a man displays characteristic features of CD speech may be related to factors that affect his willingness to invest in his own progeny.

We used two measures of men's CD speech to assess two aspects of vocal accommodation to communication with children. The first was average pitch increase. Studies of vocal expression have demonstrated that variation in pitch is one of the prime indicators of emotional state (Scherer, 2003). Pitch variation is related to modifications at the vocal folds and may be influenced by physiological changes accompanying various emotional states. Pitch increase, therefore, might be related to the participants' positive affect towards children. The second measure of CD speech accommodation was increase in vowel length. Lengthening of vowels is a major factor responsible for the slower speech rate of CD speech and, in conjunction with the resulting hyper-articulation (Bradlow, Toretta, & Pisoni, 1996), may be viewed as a reflection of an individual's deliberate attempt to facilitate communication with a child (Liu, Kuhl, & Tsao, 2003). Interestingly, in pet-directed speech, which is characterized by a large degree of vocal affect as indicated by increased pitch variation compared to AD speech, hyper-articulation does not occur (Burnham, Kitamura, & Vollmer-Conna, 2002). This suggests that vowel lengthening does not necessarily have to be affectively modulated, but may represent a didactic effort on the part of the speaker to aid the child's understanding and acquisition of language.

Based on theories of strategic pluralism in mate preferences, and the findings of Roney et al. (2006), we aimed to further investigate two research questions in the current study.

2. Methods

3. Results

3.1. Female responses to the male faces

Women's ratings of the male faces for each of the seven traits were reliable (all Cronbach's alphas >0.75). Consequently, women's ratings were averaged across raters to give a score for each target face on each of the five traits and two attractiveness ratings.

There were a number of moderate to very strong correlations between ratings of the male faces on some of the five traits (Table 1). To simplify further analyses, factor analysis (PCA, varimax rotation) was used to reduce the data. This analysis identified two factors with eigenvalues over 1, which accounted for 89% of the variance in trait attributions. The first factor accounted for 68% of the variance and loaded on the following trait ratings: good father, likes children, trustworthy, healthy. We labelled this factor ‘prosociality’. The second factor accounted for 21% of the variance and loaded solely on the ‘masculinity’ trait rating.

Table 1. Correlations between trait ratings of male faces by female participants

		Masculine	Dislikes children	Healthy	Trustworthy
	Good father	0.11	−0.90⁎⁎	0.73⁎⁎	0.94⁎⁎
	Masculine		0.12	−0.02	−0.11
	Dislikes children			−0.64⁎⁎	−0.94⁎⁎
	Healthy				0.65⁎⁎

⁎⁎ p`<.01.

In order to establish whether patterns of association between trait attributions and facial attractiveness ratings differed for long- and short-term mating contexts correlation coefficients were calculated. The prosociality factor was strongly correlated with attractiveness ratings in both the long-term (r=0.82, n=63, p<.01) and short-term contexts (r=0.69, n=63, p<.01). The masculinity factor predicted neither long- nor short-term attractiveness (r=0.19 and 0.17, respectively; n=63; NS in both analyses). The correlation between mean attractiveness ratings for each face as a short-term partner and as a long-term partner was very strong (r=0.96; n=63, p<.001) suggesting that ratings of attractiveness for a given face in one context predict ratings in the other nearly perfectly.

3.2. Do women's attributions predict men's self-reported affinity for children or men's CD speech characteristics?

Roney et al. (2006) found that those men that women thought would make good fathers (on the basis of their facial appearance) exhibited higher self-reported affinity for children. In the present study, our two self-report measures of affinity for children derived from the questionnaire items given to the male participants (‘wants children’ and ‘likes children’) correlated with each other significantly (r=0.51, n=63, p<.01). We tested whether these two self-report measures were associated with the rated prosociality and masculinity factors. Consistent with the findings of Roney et al., we found that women's ratings of prosociality were significantly associated with men's self-reported liking of children (r=0.25, n=62, p<.05). All other relationships were nonsignificant (see Table 2). Controlling for number of younger siblings had no effect on the pattern or significance of these findings, or on any subsequent analyses, and is not reported.

Table 2. Correlations between perceived personality and self-reported liking and wanting of children for the male targets, and men's accommodation in CD speech

	Pearson's r (n=63)	‘Wants children’ (n=63)	‘Likes children’ (n=63)	Mean vowel duration CD–AD difference (n=52)	Mean vowel pitch CD–AD difference (n=45)
	Prosociality factor	0.00	0.25⁎	0.17	−0.40⁎⁎
	Masculinity factor	0.18	0.06	0.37⁎	0.00

⁎ p<.05. ⁎⁎ p<.01.

Performance of men on the CD speech task provides an alternative measure of their affinity for children. For our rated sample of 63 men, 52 provided complete and analysable speech samples for the vowel duration variable, and 45 participants had analysable data for the fundamental frequency variable. We examined whether women's assessments of prosociality and masculinity were associated with the amount of accommodation exhibited by men in their CD speech. These analyses are illustrated in Table 2. Two results are noteworthy—first, women's judgements of men's prosociality are negatively associated with pitch increase in CD speech (i.e., men who look like they posses positive personality characteristics that would make them good fathers do not accommodate children in pitch as well as those men who look less prosocial). Second, men's rated masculinity positively predicts the degree to which they lengthen vowel sounds in the CD speech task. As with the self-report data, controlling for number of younger siblings had no effect on the pattern or significance of these findings.

3.3. Do women's attractiveness judgements predict men's self-reported affinity for children or CD speech characteristics?

Women's preferences when choosing mates may be functional irrespective of whether they are associated with conscious attribution of particular personality traits—i.e., women may perceive the faces of men who like children to be more attractive as long-term partners but this aspect of mate choice may not necessarily be mediated by conscious attributions of prosocial personality traits to their faces. To test this possibility, we examined the relationships between short- and long-term attractiveness, men's self-reported affinity for children, and the two CD speech variables (Table 3). There were significant positive associations between men's self-reported liking for children and both their short-term (r=0.42, n=63, p<.01) and long-term (r=0.38, n=63, p<.01) attractiveness. So men who self-report liking children are judged as more attractive in both relationship contexts. However, there was a significant negative association between a measure of performance on the CD speech task (mean vowel pitch CD–AD difference) and male attractiveness as a long-term partner (r=−0.29, n=45, p<.05) but no attractiveness as a short-term partner (r=−0.26, n=45, p<.1). In summary, men who were poor at accommodation in CD speech were more attractive in the context of a long-term relationship than men who were better at accommodation in CD speech.

Table 3. Correlations between women's attractiveness judgements, self-reported affinity for children, and the CD speech variables

	Pearson's r	‘Wants children’ (n=63)	‘Likes children’ (n=63)	Mean vowel duration CD–AD difference (n=52)	Mean vowel pitch CD–AD difference (n=45)
	Long-term attractiveness	0.13	0.38⁎⁎	0.16	−0.29⁎
	Short-term attractiveness	0.21	0.42⁎⁎	0.13	−0.26

⁎ p<.05. ⁎⁎ p<.01.

4. Discussion

We aimed to replicate and extend the finding of Roney et al. (2006) that when women judge the attractiveness of male faces, their judgements appear to ‘track’ men's interest in infants. Our results provide some support for this earlier work—when women judged men's faces, attributions of prosociality and long- and short-term attractiveness were significantly related to the men's self-reported liking of children in our data. We also examined whether women's attributions of parenting skill accurately predict performance on a behavioral task in which men gave instructions to an imaginary child. Women's attributions of prosociality and long-term attractiveness were significantly related to men's performance on this task, but in the opposite direction to our hypothesis: men whose faces elicit ratings of high prosociality actually perform worse on this task than men who look less prosocial. These findings raise several issues for discussion.

We found a number of strong correlations between attributions made to the male faces. For example, attractiveness ratings strongly predict various ratings of prosociality (which are themselves closely related to one another). In the evolutionary psychology literature, correlations between trait attributions and attractiveness are often assumed to have functional significance. An alternative explanation, well documented in the social psychology literature, is that attractive faces are attributed desirable personalities simply because they are attractive—in other words, pervasive ‘halo effects’ of attractiveness mean that personality attributions may lack specificity. Although halo effects fail to explain much of the data in the literature investigating human facial attractiveness (e.g., menstrual cycle shifts in preferences), they may perhaps be underestimated in studies conducted from an evolutionary perspective.

The identified dissociation between self-report and behavioral affinity for children requires consideration. It was hypothesized that the two variables would be positively related to women's attributions of prosociality and attractiveness (especially when women are asked to indicate their preferences for a long-term relationship). We found, however, that our two measures of affinity for children were related to women's attributions, but in opposite directions. Men who are perceived by women to be attractive and whose faces elicit attributions of prosociality do self-report a liking for children but in fact are poor at accommodating children in their speech through pitch increase. This dissociation must be considered seriously: it seems that one, or perhaps both, of the measures of affinity for offspring does not reflect the assumed underlying trait (i.e., men's willingness to invest in their offspring) in our sample of men.

Both of the tasks are subject to some potential criticisms. Our short self-report measure generates responses similar to the affinity to infants measure employed by Roney et al. (2006), yet is clearly transparent and may be biased by a strong social desirability effect: who, after all, would be willing to admit that they ‘dislike children’ in the face of societal norms in 21st century Britain? Despite this hypothesised taboo, however, we have found noteworthy variance in this measure, and this variance leads to a correlation between a man's self-reported liking for children and their perceived prosociality and attractiveness. If this measure does not reflect true affinity for children, we must ask why is it that more attractive men self-report a liking for children when less attractive men do not. Even if responses were affected by a self-presentation bias, why do some men present themselves as having greater affinity for children than do others? As attractiveness predicts social competence (see Feingold, 1992, for review), attractive men may be more aware of the importance of presenting themselves positively. Alternatively, since self-reports on this sort of measure must be made relative to what is perceived as normal for the population at large, it may be that more attractive men are simply more aware of the desirability of affinity for children in women's partner choice (see Daly & Wilson, 1999, for a discussion of the potential problems of interpreting self-report data from the ‘garrulous animal’).

Our behavioral measure of affinity for children has the advantage of being less transparent than simple questionnaire measures, but its ecological validity is limited by the fact that no actual child was present during the recording of the speech. Furthermore, all of the men in the study were childless and therefore had limited opportunities to have developed parental skills such as a fine-tuned means of adjusting one's speech to the limited cognitive and linguistic capacity of small children (although controlling for experience of younger siblings had no influence on the outcomes of our analyses). There is, however, evidence that the ability to produce CD speech is not limited to parents addressing their progeny. In fact, when asked to address an imaginary child, speakers with no experience of childcare display typical CD prosodic features, albeit in a somewhat attenuated manner (Jacobson, Boersma, Fields, & Olson, 1983). Nevertheless, it remains possible that some individuals may react differently to the ‘imaginary child’ situation, and that some of these differences may co-vary with the perceived prosociality of the men's faces. For example, attractive, socially competent men may find the experimental situation less motivating or potentially more embarrassing than less socially competent men and therefore may underperform on the task. There is also the possibility that some men, despite a genuine affinity towards children, may choose not to use CD speech to avoid ‘talking down’ to infants.

The findings of the present study can be considered in relation to the behavior of males in animal species with similar mating systems (i.e., pairbonding with biparental care). Research in various species has examined how male parental behavior varies with attractiveness. Empirical studies (mainly with passerine bird species) indicate that male phenotypic condition is one of several factors that influence the trade-off between male parental effort and mating effort: highly attractive males invest less in offspring, and less attractive males invest more (see Magrath & Komdeur, 2003, for review). In humans, a similar pattern of results has been found—men with low levels of fluctuating asymmetry (a trait that correlates with attractiveness in several modalities) invest less in relationships than more asymmetric men (see Gangestad & Simpson, 2000, for review). These comparative and human findings suggest that physically attractive men provide relatively less parental effort than less physically attractive men. If this is the case, then our findings indicate that the behavioral speech task is more likely to be an accurate measure of future parental effort than self-report.

This study failed to identify the expected relationships between masculinity and the other variables. We predicted negative relationships between ratings of personality and masculinity, and that masculine faces would be relatively more attractive in the short-term rather than long-term attractiveness condition, but neither relationship was found. The association between masculinity and attractiveness remains unresolved—when composite faces are used as stimuli, overall preferences favour femininity, but shift towards masculinity when judgements are made in the context of a short-term relationship (or as a result of other factors outlined in the introduction above, e.g., Penton-Voak et al., 1999, Little et al., 2002). When individual faces are presented, the general pattern is that more masculine faces are preferred, but this effect is not found exclusively (Rhodes, 2006). Our data add little to this debate, other than to confuse matters still further. Although ratings of masculinity based on facial appearance predict little in this study, they positively predict vowel length increase in the CD speech task. This unexpected result is hard to interpret in the absence of other expected associations with masculinity, but it is worth noting again that some CD speech researchers have speculated that while pitch increase is associated with the expression of positive affect towards offspring, vowel length increase is associated with a conscious effort to accommodate—whether affective or motivational factors are most important in determining paternal effort over the long-term is unknown.

In conclusion, the major contribution of this research has been first to replicate the finding of Roney et al. (2006) that women can predict men's self-reported affinity for children and then additionally to demonstrate that these same women's ratings predict men's performance in an alternative measure of affinity for children, namely, a CD speech task, but in the opposite direction to our hypothesis. It is unclear which measure of affinity for offspring is more valid, but findings from human studies of other characteristics (such as symmetry) and other species with similar mating systems indicate that high-quality (attractive) males invest less in offspring than low-quality males. Moreover, despite the absence of a real child the CD speech task has arguably more ecological validity and is perhaps less subject to social desirability effects than self-report data. Consequently, our findings suggest that women's attributions of parental quality may represent halo effects of attractiveness rather than accurate social perception of male behavioral traits.

This study does not suggest that women, in real social interactions, are unable to accurately predict men's behavior (and vice versa). Extensive social psychological research demonstrates people's ability to make accurate personality judgements after surprisingly brief interactions (Zebrowitz & Collins, 1997) Indeed, our remarkable aptitude in this regard may lead to overconfidence in our abilities to judge personality accurately from impoverished stimuli such as photographs. The rapidity of accurate social judgment in real social interactions may consequently convince observers that they have made judgements on the basis of facial appearance alone when, in fact, they have been made on the basis of behavioral cues (e.g., facial movements) that are available very early in the interaction. This raises an interesting question of validity for facial attractiveness studies that suggest personality attribution to static faces plays a key role in judgements of male attractiveness (e.g., Penton-Voak et al., 1999), as clearly, such attributions may not necessarily be accurate.