Introduction

Women undertake dramatic changes in reproductive effort in the process of conceiving, carrying a pregnancy to term, and in postpartum care. This amounts to life history shifts from mating to parenting effort (Ellison 2003; Trivers 1972; Stearns 1992). The postpartum phase is characterized by intensive parenting effort, typically at expense to mating effort. Indeed, intensive breastfeeding interacts with other energetic and health factors in contributing to lactational amenorrhea (Ellison 2003; Ellison 2009), where the maternal benefits of lactation operate at expense to ovarian cycling. These shifts in reproductive effort are reflected in steroid hormone changes. For example, low testosterone and estradiol levels, both sex steroids which are lower after birth compared to pre-pregnancy levels (see Barrett et al. 2013; Nelson 2011), may decrease women’s sexual behaviors (Meston and Frohlich 2001; Simon et al. 2005; Sarrel et al. 1998). An increase in oxytocin, particularly during breastfeeding, after giving birth may indicate a shift in care towards an infant (Brizendine 2006; Lancaster and Kaplan 2009).

A downregulation of mating effort may be seen through a decline in sexual function (Weeks-Shackelford et al. 2007; Escasa-Dorne 2015), including selectivity for mating-related traits. Postpartum mothers tend to exhibit lower rates of sexual activity compared to pre-pregnancy levels (Hicks et al. 2004; Hipp et al. 2012; Leeman and Rogers 2012; Sayasneh and Pandeva 2010; Woranitat and Taneepanichskul 2007), and rates of dyspareunia during the year after giving birth are as high as 92 % (Andrews et al. 2008).

A preference for sexually dimorphic traits provides us with information on the evolution of human mating psychology (Puts 2010, 2016). Sexually dimorphic traits, including masculine facial features, may be one cue indicating genetic potential or phenotypic quality of a possible mate (Andersson 1994; Johnstone 1995; Roberts et al. 2004; Zahavi 1991). Women’s preferences for masculinity, particularly for traits of masculinity in faces, in a potential partner appear to be contextual (see Gildersleeve et al. 2014). For example, Little et al. (2002) found preferences for feminine facial features appear higher among women who are not in a relationship and for a long-term (rather than a short-term) context among women who were not using oral contraceptives; however, face preferences differed among women who were currently using oral contraceptives. Additionally, while masculine faces may be indicative of certain beneficial traits when selecting a partner, many studies find only slight preferences for masculine features (Rhodes et al. 2003; Puts 2005; DeBruine et al. 2006; Little et al. 2008) or preferences for feminine features (Perrett et al. 1998; Penton-Voak et al. 1999a, b; Little et al. 2001; Welling et al. 2007). In one study among the Hadza of Tanzania, breastfeeding women preferred feminized male voices played over an audio recording compared to women who were not breastfeeding (Apicella and Feinberg 2009). Among 28 women in the UK, recruited during pregnancy and followed through to 12 weeks postpartum, participants had a stronger preference for masculine faces during pregnancy compared to postpartum (Cobey et al. 2015). Further, compared to a group of nulliparous women, women using hormonal contraceptives had a higher preference for masculine male faces (Cobey et al. 2015).

Preferences for masculine traits appear to be highest around the time of ovulation, when conception risk is highest (Gildersleeve et al. 2014; Gangestad et al. 2015; Gangestad and Thornhill 1998). The ovulatory-shift hypothesis (Gangestad et al. 2005) argues that the preference for cues indicative of high genetic quality in a potential mate are highest when risk of conception is highest, compared to times of low risk of conception. For example, Penton-Voak and Perrett (2000) analyzed facial masculinity preferences of 139 women who were not using oral contraceptives. Based on self-reports of menstrual cycle status, women who were placed in a high-conception risk group (i.e., around ovulation) had higher preferences for masculinized male faces compared to women in a low-conception risk group. Other studies using similar self-report methods find similar effects of higher masculinity preference around ovulation (Johnston et al. 2001; Jones et al. 2005a, b, 2008; Penton-Voak et al. 1999a, b; Penton-Voak and Perrett 2000). Shifting preferences over the menstrual cycle may be due to an adaptive benefit of extra selectiveness of traits indicative of high genetic quality in a partner during a time of high risk of conception, an adaptive benefit of selectivity of high partnership quality during most non-conceptive times of a cycle.

Shifting preferences may also be a result of fluctuations in hormones, particularly the sex steroid hormones, across the menstrual cycle. Estradiol and progesterone shift, with the highest levels of estradiol near ovulation, whereas progesterone levels peak during the luteal phase. These shifts in sex steroid hormone levels may alter women’s facial preferences over the menstrual cycle. Within-participant changes in estradiol level predicted women’s preferences for facial masculinity (Roney and Simmons 2008; Roney et al. 2011). Among 70 women, Welling et al. (2007) found women’s preferences for masculine faces were higher on days on which testosterone levels were higher. In another study, Marcinkowska et al. (2016) analyzed risk of conception using estradiol as a marker of between-subjects cycling status of 115 Polish women and found that estradiol was not associated with masculinity preferences of faces in a short-term or long-term context, nor did the study find differences in preferences when the women were divided into groups of high or low probability of conception.

Preferences for masculine male faces appear to provide cues of a high quality partner; however, masculine features may be contextually less desirable for numerous reasons. Males with higher testosterone levels, for example, may also be more aggressive, more likely to have extra-pair relationships, and be less invested in spouses or parenting (Gray et al. 2002; Mazur and Booth 1998). Males with higher facial masculinity were rated as more likely to cheat on a partner, to get into more physical fights, to be a lower quality spouse and father, and to be less emotionally supportive in a relationship compared to males with higher facial femininity (Boothroyd et al. 2007; Kruger 2006). Moreover, cross-cultural variation appears to impact the degree of preference for masculine or feminine facial features, with arguments that the preference for masculinity may be impacted by the pathogen load in the environment (DeBruine et al. 2010a; Moore et al. 2013; Scott et al. 2014).

Despite the availability of research on women’s face preferences for males, little research focuses on post-birth shifts in female preferences for male traits. Because of the change in sex steroid hormones, and the behavioral shifts from mating to parenting effort, during the postpartum phase, an understanding of facultative face preferences among women after giving birth is warranted. Women’s post-birth preferences for male traits may be altered, also complementing research on facultative female preferences across the menstrual cycle. Women’s sexual and romantic preferences in a partner after giving birth have received little previous research focus, leaving an empirical gap to be filled. Additionally, the preferences of women who are in lactational amenorrhea may provide insight into the evolution of our behavioral reproductive milieu, as women in natural fertility groups spend a majority of their reproductive years in this phase, and unlike industrialized groups (Escasa-Dorne et al. 2013; Eaton et al. 1994; Strassmann 1997, 1999).

Here, we test preferences for facial masculinity or femininity among mothers caring for a young baby. Based on tradeoffs between mating and parenting effort, we predicted mothers from our sample would have a preference for feminine faces, and that this would be stronger for breastfeeding women. To explore potential variation in female preferences of male faces during the time after giving birth, we asked participants to complete a face preference task. Participants were recruited from two populations: an online, convenience sample of women who had given birth in the past year and a sample of women from the Philippines who had 1) given birth in the past year, or 2) had never before given birth. The Philippines sample of nulliparous women provided a comparative sample of women who were not breastfeeding nor had given birth, and would more closely resemble previous study populations where facial preferences have been studied. As the Philippines sample also provided saliva samples for steroid hormone analysis, this population was chosen to provide a pool of participants who were more likely to breastfeed extensively and less likely to use hormonal contraceptives (NSO 2009).

Stimuli

All participants were asked to view multiple sets of photographs of male faces. A total of five sets of male faces were presented, each set presented a male of a different ethnic background (composite images were obtained from male participants in South Asia, East Asia, South America, Europe, and African-Caribbean). The photograph was formed from a composite of images from each ethnicity. Each set of photographs consisted of one photo that was an unaltered photograph of a male, one photograph that was digitally altered to increase masculine features by 60 %, and one photograph that was digitally altered to increase feminine features by 60 % (see Scott et al. 2014 for more information on the methods used to digitally alter the photographs). The stimuli have been used in previous studies on multiple populations ranging from foraging groups to market economy groups (Scott et al. 2014). In random order, each set of three photographs was presented and participants were asked to choose which male was the most attractive in a short-term context and in a long-term context. Participants were also asked to select the male photo which appeared to be the “nicest”, “nastiest or meanest”, and “healthiest.” Each participant rated five sets of photographs.

Data were analyzed using SPSS v23.0 for Mac. All analyses assumed an alpha of .05 and were two-tailed. Once participants selected a photograph for each trait, the preference was entered into a database. The masculinized photograph was entered as a 60, the feminized photograph was entered as a −60, and the neutral photograph was entered as a 0 to reflect the percentage of manipulation (in either a feminized or masculinized direction) to each photograph. Each participants’ ratings were averaged across all 5 ratings (for the 5 photographs). Independent t-tests and ANOVAs were used for group comparisons, and bivariate correlation analyses were used for correlation effects. The methods for study 1 and 2 were reviewed and approved by the Institutional Review Board at the University of Nevada, Las Vegas.

Study 1

Methods

All questionnaires were completed online using SurveyMonkey.com. A link to the study was posted on various outlets (e.g., parenting blogs), and provided a convenience sample pool based primarily in the US. Women between the ages of 18–45 who had given birth in the past year were asked to complete the study. A total of 287 women started the survey. After excluding duplicate IP addresses and incomplete questionnaires, 211 participants’ data were used for analysis. After providing informed consent online, participants were directed to a questionnaire including demographic information (see Table 1). After completing the survey questions, participants were directed to the photographs to complete the stimuli task.

Table 1 Demographic data of the participants from Study 1 (online sample)
Full size table

Study 1 Results

Of the 211 participants ratings, there was a significant preference for feminine faces (t(165) = 2.98, p = .003, d = 0.64) in rating long-term attractiveness, and a non-significant masculinity preference for a short-term context (t(165) = −.99, p = .324, d = −0.154). There was no significant difference for either long-term preferences between the women who were cycling (35 % preference for feminized faces) and women who were non-cycling (34.2 % preference for feminized faces) (t(151) = −.21, p = .836, d = −0.03), nor short-term preferences between cycling (30.2 %) and non-cycling women (27.7 %) (t(151) = −.27, p = .785, d = −0.04). There was no significant difference for either long-term preferences between the women who were breastfeeding (35 % preference for feminized faces) and women who were not breastfeeding (35.5 % preference for feminized faces) (t(164) = .15, p = .882, d = 0.02), nor short-term preferences between breastfeeding (30.9 %) and women who were not breastfeeding (24.2 %) (t(164) = 1.07, p = .287, d = 0.17).

There was no correlation between the age of the infant (M = 6.6, SD = 3.8 months) and long-term preference for faces (r(151) = −.10, p = .243) nor short-term preference for faces (r(151) = −.05, p = .539), nor age of the participant (M = 29.01, SD = 5.5 years) and long-term preference for faces (r(139) = .04, p = .625), nor short-term preference for faces (r(139) = −.13, p = .113). There was no significant difference between women who were or were not using a hormonal contraceptive and long-term preference for faces (t(164) = −.32, p = .753, d = −0.05) nor short-term preference for faces (t(164) = −.56, p = .575, d = −0.09).

Study 2

Methods

In the second study, 260 participants were recruited from neighborhood health centers in Manila, Philippines (see Table 2 for demographic data). Participants were placed into one of two groups: women who had never before given birth (n = 105) and women who were breastfeeding an infant (n = 155). For additional analyses, the breastfeeding women were grouped into ‘cycling’ or ‘non-cycling’ based on information from the questionnaire. Cycling status was determined based on the woman’s last reported menstrual cycle and time since birth. Eighty-six women had reported having at least two menstrual cycles since giving birth and were at least 12 weeks postpartum, while 48 women reported not having resumed any menstrual cycles since becoming pregnant. We were unable to determine the cycling status of 21 women, thus these women were excluded from additional analyses.

Table 2 Demographic data of the participants from Study 2 (Philippines sample)
Full size table

After the researchers explained the purpose of the study in Tagalog, and obtained informed consent, each participant provided a saliva sample via passive drool to be analyzed for testosterone and estradiol levels. Saliva samples were stored in a cooler with ice immediately after collection and placed in a freezer at the end of each collection day. Samples were shipped to the UNLV Evolution and Behavior Lab via overnight delivery on dry ice, and were stored in a −20 °C freezer until analysis. Saliva samples were assayed in batches, and run in duplicate, for testosterone and estradiol at the UNLV Evolution and Human Behavior Lab using Salimetrics kits. Inter-assay coefficients of variation were 3.3 % and 3.0 % for testosterone and 7.8 % and 4.3 % for estradiol. The intra-assay coefficients of variation were 4.2 % and 7.3 %, for testosterone and estradiol respectively.

The same demographic questionnaires and photo stimuli as Study 1 were used in Study 2. All questionnaires were translated into Tagalog, and participants were given the option to complete the study in English or Tagalog. Although Tagalog is spoken by most individuals as either a first or second language in the Philippines, Tagalog (or Filipino) and English are the national languages of the Philippines. Participants completed all methods with the researcher in a semi-private location of the health centers. After the participants completed the questionnaire, the researchers placed the photographs in front of the participant and asked the participant to choose the photo which represented her choice for the same traits listed in Study 1 (most attractive in a short-term context, most attractive in a long-term context, nicest, nastiest, and healthiest).

Study 2 Results

Of the 260 participant’s ratings, there was a statistically significant preference for feminine faces in a long-term (t(238) = 6.58, p < .001, d = 0.85) and short-term (t(238) = 4.02, p < .001, d = 0.52) context. There was no significant difference for either long-term preferences between the women who had never given birth (56.2 % preference for feminized faces) and women who were breastfeeding (51.6 % preference for feminized faces) (t(237) = .58, p = .561, d = 0.08), nor short-term preferences between women who had never given birth (42.9 %) and women who were breastfeeding (48.4 %) (t(237) = −0.97, p = .335, d = −0.13) (see Table 3). There was no significant difference between cycling status and long-term preferences (F(2, 219) = .25, p = .779, ηp2 = 0.002) nor short-term preferences (F(2, 219) = .35, p = .703, ηp2 = 0.003).

Table 3 Study 1 and 2 differences in means in rating attractiveness of male faces in a long-term or short-term context
Full size table

There was no correlation between the age of the infant (M = 8.5, SD = 5.8 months) and long-term preference for faces (r(140) = −.14, p = .089) nor short-term preference for faces (r(140) = .01, p = .918). There was a significant correlation between age and long-term preference for faces (r(203) = −.16, p = .026), but not for short-term preference for faces (r(203) = −.03, p = .636). After entering age as a covariate, overall preference for feminized faces was still significant (F(1, 204) = 5.06, p = .026). There was no significant difference in face preferences between relationship status and long-term preference for faces (t(208) = −.85, p = .399, d = −0.12) nor short-term preference for faces (t(208) = −1.48, p = .141, d = −0.21).

There was no correlation between estradiol levels (M = 1.46 pg/mL, SD = .97) and long-term preference (r(176) = −.01, p = .934) nor short-term preference (r(176) = −.03, p = .671), nor between testosterone levels (M = 58.79 pg/mL, SD = 24.25) and long-term preferences (r(173) = .02, p = .833) nor short-term preferences (r(173) = −.01, p = .671).

Discussion

This study was designed to determine male facial masculinity preferences among women caring for a young baby. Data from both samples indicate no masculinity facial preferences among postpartum women. Women in both studies had a preference for feminine faces; however, there was no difference between the breastfeeding and nulliparous women in the Philippines sample. Previous studies find support for masculine face preferences, albeit with small to modest effects (Gildersleeve et al. 2014; Rhodes et al. 2003; DeBruine et al. 2006), while other studies suggest a feminine face preference or no preference (see DeBruine et al. 2006). The variation in face preferences may be due to the variation in methods utilized by each study, low perception of variation in morphology of stimuli, or could suggest the wide variation in preferences across ecological contexts (Germine et al. 2015).

The preference for feminine faces has been interpreted as a potential mating strategy, in which females select traits which may be indicative of pair-bonding quality (Little et al. 2002), or avoidance of characteristics detrimental to pair-bonding (Boothroyd et al. 2007; Gray et al. 2002; Kruger 2006; Mazur and Booth 1998). While the interpretations of mate preference have occurred primarily among women who are regularly cycling, the preference of male faces may be of particular importance during the postpartum phase, where women are negotiating trade-offs between mating and parenting effort, and undergoing birth spacing strategies (Escasa-Dorne et al. 2013). Also consistent with this view, one study reported lower preferences for masculine faces during pregnancy, compared to women who were not pregnant, with high preferences for feminine faces during the third trimester (Limoncin et al. 2015). Moreover, a longitudinal study entailing two assessments of women’s preferences for faces showed decreases in preferred facial masculinity but no differences in female facial masculinity postpartum compared to pregnancy (Cobey et al. 2015). As also noted in the study by Cobey et al. (2015), changes in preferences for masculinity or femininity of male faces may be a feature of physiological trade-offs in parenting status, rather than primarily attributed to changes in the physiology of the post-partum phase.

Previous data have suggested a preference for masculine traits or features as they may indicate a tolerance to disease in an environment of higher pathogen load (risk) (DeBruine et al. 2010a, b, c) However, the preference for feminine faces overall in the Philippines study may be consistent with cross-cultural data which find a higher preference for feminine faces among women in higher pathogen load populations (Scott et al. 2014). The lack of differences in Filipina women’s preferences depending on motherhood status (mothers vs. childless) for male faces suggests some broader cultural or contextual factor (rather than reproductive condition) may underlie face preferences. Yet the similar patterns in women’s face preferences, perhaps with potential differences in pathogen loads, between the U.S.-based Internet sample and Philippines samples also argues against any clear or direct connection to pathogen loads.

There were no differences in facial masculinity preference between breastfeeding status or cycling status. The authors analyzed cycling status based on self-reported data and did not measure any surges in luteinizing hormone (LH) to determine cycling status. Testosterone and estradiol were both lower among the non-cycling group compared to the cycling and nulliparous groups (F(2, 172) = 12.3, p < .001, ηp2 = 0.13 and F(2, 175) = 23.1, p < .001, ηp2 = 0.21 respectively), consistent with previous studies (see Nelson 2011) which find lower sex steroid hormone production among women in lactational amenorrhea. Thus, these data confirm expected differences in hormone levels according to cycle status, even as cycling status was unrelated to women’s preferences for male faces. Future studies may benefit from employing a longitudinal design among women, following them from pregnancy through lactational amenorrhea to the resumption of ovulatory cycles based on LH surges to determine cycling status.

To summarize, the current two studies tested preferences for facial masculinity among mothers caring for a young baby. Both populations of mothers preferred feminine faces; however, there were no significant differences between breastfeeding status or cycling status. The preferences of women during the time after giving birth may help provide contextual information for the reproductive strategies of women, and may be useful to a broader understanding of tradeoffs between mating and parenting effort, birth spacing strategies, differential investment, and parenting-mating interactions over successive offspring. Research on women’s mate preferences after birth calls for further analysis, and the current study provides a glimpse into mate preferences throughout life history transitions.