Composite Image Calibration

The maie face prototypes images were rated on both attractiveness and masculinity and the female faces were rated for attractiveness and femininity. The raters were recruited through an introductory psychology class at the University of Colorado at Colorado Springs for course credit. For our analyses, we included only female raters under 25 years of age (N = 38, Mean Age = 18.7 ± 1.2, range 17-23 years) and not taking hormonal contraceptives or pregnant. Male raters were under 25 years of age (N = 39, Mean Age = 19.6, ± 1.4, 17-23 years). Analyses are based on opposite sex ratings. Images were presented individually and in random order amongst filler items, and rated on a 7-point scale.

Paired t-test analyses revealed that the high masculine male face prototype (M = 3.97 ± 1.03) was judged more masculine than the low masculine male face

prototype (M = 2.45 ± 1.01), t(37) = 7.3,p < .001, T|^= .59. When judged on

attractiveness, the high masculinity (M-3.12 ±1.18) and low masculinity (M=3.15 ± 1.31) face prototypes were found not to be significantly different, t (33) = -. 114, p = .91.

For the high and low attractive male images rated on attractiveness we found that the high attractive male face (M = 3.85 ± 1.18) was rated significantly more attractive than the low attractive face (M = 3.15 ± 1.11), t(33) = 4.112, < .001, = .34. When rated on masculinity the two were not significantly different t(37) = 1.02, p = .31 (high attractiveness: M = 3.26 ± 1.13; low attractiveness M = 3.08 ± .82). These data show that the intended manipulation of male face prototypes along one dimension while controlling a second dimension was successful.

For the female images, however, the calibration results indicate that segregating attractiveness and femininity in female faces was not successful. The high and low feminine face pairs were rated differently on femminity; high feminine (M = 2.34 ± 1.74), low feminine (M = 1.62 ± 1.545), t(28) = 3.92, p = .001, t|^ = .354, but the same images were also judged differently on attractiveness: high feminine (M = 3.33 ± 1.43), low feminine faces (M = 2.62 ± 1.12), t(20) = 3.25, p = .004, r\ =

.346. The high (M = 3.86 ± 1.389) and low (M = 2.90 ± 1.136) attractiveness female faces were rated differently on attractiveness t(20) = 3.301, p = .004, i f = .353; but also differed on rated femininity: high (M = 2.86 ± 1.98), low (M = 2.07 ± 1.65); t(28) = 3.111, p = .004, Tj^ = .257.

5.3.3 Experimental Images

Three composite ‘base’ male faces were made by averaging 8 randomly

chosen Caucasian male face images, aged between 18 and 24. Three ‘base’ female faces were similarly created. This created base faces differing in apparent identity which were then transfonned by +/-115% of the difference in face shape, colour, and texture between the high and low sexual-dimorphic and high/low attractive

prototypes (see Psychomorph methods). Finally a sequence of 25 images was created interpolating between the +115% and -115% end point images. This effectively created 3 face continua (of 25 images) for each sex that differed along one dimension but were matched in other respects (i.e. different apparent masculinity but matched in identity and attractiveness). For illustration see Figures 5.1 and 5.2.

These continua were used to create three interactive sequences, with 25 individual images in each sequence. Participants were asked to choose the image that

they considered to be the most attractive from the range available (Little et al. 2001a; Perrett et al. 1998).

As sexually-dimorphic characteristics may be associated with age (masculine faces look older, while feminine faces look younger), we included an age transform sequence. These images were made from a different cohort of male and female images, and each image was judged for age by 10 raters. Male composite aging- prototype images were made by creating 3 individual faces fi'om blends of 6 young

male faces (perceived age between 15-20 years) and 3 individual blends of 6 older

male faces (perceived age between 50-60 years). All images were made symmetrical. The faces were validated via online tests by students at the University of Colorado at Colorado Springs. Twenty-one male and 6 6 female participants (males; mean age =

20. 65, SD= 5.21; female: mean age = 20.03, SD = 4.67) were asked to enter the estimated age of each face. Interrater reliability for perceived age was high (Cronbach’s alpha = .996).

The interactive tests were made by pairing one of the young-aged male composite images with one of the older-aged male composite images. To create the continua between the young and older faces, 35 morphs were made by beginning with the young image and incrementally increasing the facial characteristics of the older image and decreasing those of the young image. The last image of the continua then would be the older composite image. This approximated an increase of 1 year of age to each of the 35 morphs. This process was repeated for the remaining paired male young-old images. Female aging images and continua were made in the same manner.

5.4 Methods

5.4.1 Participants

Heterosexual undergraduate students were recruited from the University of St. Andrews: 46 women not taking hormonal contraceptives or reporting pregnancy (age range 18-23, Mean 19.50 ± 1.36), and 52 men (age range 18-24, Mean 20.62 ± 1.60). 5.4.2 Materials

To assess preferences for facial masculinity and attractiveness, interactive face-sequence trials were used, consisting of three male and three female Caucasian faces.

Participants were also asked to complete a questionnaire, which included life- history questions relating to age of menarche/puberty, and Age of First Sex. For individuals reporting not having had sex or sexual paifners: 9 males and 16 females) current age was used as Age of First Sex.

Additionally, family relationship questions included warmth toward father and mother, quality of parent’s relationship with one another, and current age of parents, which used a 9-point Likert-type scale. Also relevant to this study was self-rated attractiveness, which used a 7-point Likert-type scale. Father absence was assessed with questions relating to the participant’s age at the time of parents’ separation. 5.4.3 Procedures

After reading and signing a consent form, participants were asked to complete an on-line questionnaire. After completion of the questionnaire, they were presented with three conditions of interactive face-sequence trials, sexual dimorphism,

attractiveness, and age for both opposite-sex and same-sex. Both the conditions and the faces within each condition were randomised. For each face, the participants were

asked to select the face they found most attractive by moving the cursor over the image to change through the continua of sequenced faces. By clicking on the mouse, the participant chose the face he or she found most attractive as well as moving them on to the next trial.

5.5 Results 5.5.1 Men

We hypothesised that timing of developmental markers would influence adult mate choice preferences. Correlations indicate that both Age of Puberty (Spearman T49 = -.331, /j = .020) and Age of First Sex (rsi = -.286, p = .042) related to

preferences for facial femininity. Thus early sexual development was associated with increased preference for feminised facial characteristics.

Age of First Sex correlated with apparent age of face preference (rsi = .380, = .006), other correlations between maturation and face preferences for facial age and attractiveness were non-significant (all p > .25). Wliile the attractiveness stimuli did not reveal male preferences influenced by sexual development, it is interesting to note that there was a stronger preference overall for the attractive female face (M = 6.50, SEM .043) compared to the feminine female face (M = 2.945, SEM .033; t4g = 5.847,

p < .001), and we found a negative trend in the zero-order conelation for preferences for the two stimuli (r49 = -273, p = .058).

Despite findings in previous research linking age of puberty with age of first sex, we only found a positive but non-significant correlation (rso = .204, p = .156), perhaps due to the small sample size.