Girls know how to choose: Fathers lived in larger cities, had higher education, were heavier and taller , more attractive & masculine, had lighter eyes, darker hair, & were more agreeable, conscientious, & emotionally stable than non-fathers

She Always Steps in the Same River: Similarity Among Long-Term Partners in Their Demographic, Physical, and Personality Characteristics. Zuzana Štěrbová, Petr Tureček and Karel Kleisner. Front. Psychol., February 5 2019. https://doi.org/10.3389/fpsyg.2019.00052

Abstract: In mate choice, individuals consider a wide pool of potential partners. It has been found that people have certain preferences, but intraindividual stability of mate choice over time remains little explored. We tested individual consistency of mate choice with respect to a number of demographic, physical, and personality characteristics. Only mothers were recruited for this study, because we wanted to find out not only whether women choose long-term partners with certain characteristics but also whether the father of their child(ren) differs from their other long-term (ex-)partners. Women (N = 537) of 19–45 years of age indicated the demographic, physical (by using image stimuli), and personality characteristics of all of their long-term partners (partners per respondent: mean = 2.98, SD = 1.32). Then we compared the average difference between an individual’s long-term partners with the expected average difference using a permutation test. We also evaluated differences between partners who had children with the participants (fathers) and other long-term partners (non-fathers) using permutation tests and mixed-effect models. Our results revealed that women choose long-term partners consistently with respect to all types of characteristics. Although effect sizes for the individual characteristics were rather weak, maximal cumulative effect size for all characteristics together was high, which suggests that relatively low effect sizes were caused by high variability with low correlations between characteristics, and not by inconsistent mate choice. Furthermore, we found that despite some differences between partners, fathers of participants’ child(ren) do fit their ‘type’. These results suggest that mate choice may be guided by relatively stable but to some degree flexible preferences, which makes mate choice cognitively less demanding and less time-consuming. Further longitudinal studies are needed to confirm this conclusion.

Results

Mate choice consistency was higher than expected in all assessed qualities except for facial masculinity and beardedness. Difference between observed and expected consistency was statistically significant in most qualities, but effect sizes differed substantially. While consistency of mate choice in residence or weight was substantial, it was only medium-sized or small with respect to hair or eye color. Complete results are summarized in Table 1 and Figure 1.

Table 1. Mate choice consistency: complete results.

Figure 1. Visualization of permutation tests of mate choice consistency centered around observed image and normalized along the SD of expected image distribution. Difference between the observed and expected value is expressed in standard deviations from the expected value distribution. The higher the bell curve above the Observed image value, the higher the actual mate choice consistency. Bell curve below Observed image value indicates a trait where the observed mate choice was less consistent than expected.

The average effect size was highest in demographic variables, but none of the pairwise comparisons between groups of variables (demographic, physical, and psychological) was statistically significant (p > 0.1). Permutation test results are visualized in Figure 1. All sample sizes and descriptive statistics of all variables are listed in the Appendix. The different estimates of effect size were highly correlated. The proportion of males who had to be relocated between respondents correlated with the variance accounted for by the respondent at 0.93, whereby a linear model of relationship between these two measures supports the idea that the latter is approximately double of the former. The slope in the model where respondent-attributable variance regressed on the proportion of partners to relocate was 2.08 (95% CI = 1.72–2.45) with minimal (not significantly different from 0) intercept of -0.18 (95% CI = -3.19–2.83). Results yielded by the simple Pearson correlation correlated at 0.91 with the percentage of partners to relocate and at 0.98 with respondent-attributable variance. All of these measures can be thus treated as functionally equivalent.

Links between pairs of partners’ qualities are summarized in Table 2. In total, 103 out of 210 correlations were significant even after Benjamini–Hochberg correction for multiple comparisons. Maximal cumulative effect size was 50.95% (expressed in the proportion of partners to switch between individuals). The first 10 variables ordered according to their unique contribution starting with the highest (residence, weight, relative height, age difference, attractiveness, hair color, openness, BMI, height, agreeableness, in this order) explained 48.30% of partner assignment. The other 11 variables contributed little (their unique contributions were less than 1%) or not at all (after the inclusion of all other variables, facial masculinity and beardedness failed to show any positive numbers). Full results are visualized in Figure 2.

Table 2. Relations between investigated qualities of romantic partners expressed in shared effect sizes and Pearson correlations.

Figure 2. Visualization of maximal cumulative effect size. Variables are added in order given by maximal unique contribution to overall consistency.

Reaching maximal possible effect size suggests that adding yet other variables to a similar model of cumulative consistency would add little to our current sum. On the other hand, it is conceivable that one might select precisely those variables which are not intercorrelated and explain a majority of mate choice consistency in just a handful independent dimensions. In theory, complex interaction patterns may lead to an even higher cumulative effect size since 50% of partners to relocate as an effect size limit applies to a single variable with two levels and represents the difference between maximal and minimal consistency (i.e., not maximal and expected). The high proportion of significantly correlated pairs of variables (49%), does, however, fit well within the impression of a substantial redundancy in our model.

Permutation test of changes in mate choice consistency revealed that fathers are significantly exceptional amongst participants’ long-term partners in beardedness, muscularity, hirsuteness, extraversion, and openness. The average image without these individuals was lower than the image in permutation runs where an equivalent proportion of random partners (i.e., fathers and non-fathers) was excluded. Fathers were not significantly typical long-term partners in any of the assessed qualities. Complete results of these tests are summarized in Table 3 and visualization is provided in Figure 3.

Table 3. Permutation test of father exceptionality, complete results.

Figure 3. Visualization of permutation tests of father exceptionality centered around the observed image when fathers were excluded from the sample of partners and normalized along the SD of expected image distribution in such a situation. Difference between observed and expected values is expressed in standard deviations of expected value distribution. The higher the bell curve above the observed image value, the more exceptional were the fathers among the long-term partners of an individual. Bell curve below the observed image value indicates a trait where fathers were more typical representatives of an individual’s long-term partners.

In qualities where fathers were indicated as exceptional individuals (except for extraversion), mean trait values differed between fathers and non-fathers, while variances differed in beardedness, muscularity, and hirsuteness. Fathers were more bearded, hairier, more muscular, and showed a higher openness to experience. These differences might explain the overall exceptionality of fathers except for extraversion. It seems that fathers are outliers within partner sets even where the group means and variances of father and non-father sets do not differ. Moreover, fathers lived in larger cities, had higher education, were heavier and taller (although relatively, their height was closer to the height of respondents), more attractive and masculine, had lighter eyes, darker hair, more masculine faces, and were more agreeable, conscientious, and emotionally stable than non-fathers.

Group variances differed in several qualities. Fathers were significantly more variable than non-fathers with respect to age difference from the respondent and less variable in attractiveness, masculinity (general and facial), BMI, conscientiousness, and agreeableness. It seems that along these variables, either or both of the extremes are not the right for the ‘father material’. A graphic overview which compares densities that indicate differences between group means and variances is presented in Figure 4. Complete results in a textual form are listed in Table 4.

Figure 4. Visualization of differences between fathers and non-fathers. Significance of difference between group means and variances is estimated from mixed effect models with respondent ID treated as a random factor. Significance levels are indicated as follows: ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Table 4. Results of Mixed effect models comparing father/non-father means and variances.