Main

There are well known academic2, social3,4, health5,6,7,8,9 and economic10,11 correlates of height, but we know of no studies that have examined its direct implications for evolutionary fitness. To examine the possible evolutionary consequences of stature, we have analysed data from the medical records for 4,419 healthy men aged 25–60 who received compulsory medical examinations between 1983 and 1989 at the Lower Silesian Medical Centre in Wroclaw, Poland. Because the records were not anonymous and many bachelors admitted to having offspring, the risk of false declarations was probably small.

To avoid any confounding pathological effects, we discarded the data from men whose height was more than three standard deviations from the sample mean (172±6.6 cm). Locality of residence had a significant effect on stature (analysis of variance (ANOVA): F3,4409=11.02, P<0.001, with city-dwellers being taller than rural men), so we selected only individuals resident in either rural villages or in cities with more than 100,000 inhabitants (the extreme subpopulations). This yielded a final sample size of 3,201.

Stature is also confounded by a secular trend (linear regression against age: b=−0.00157, F1,3199=184.5, P<0.0001), so we calculated a residual from the common regression line against age for each subject and standardized this against the overall mean for the subject's residence. Stature is also confounded by educational achievement, probably because this correlates with family wealth and status, so we controlled for education.

Figure 1a shows means and variances in stature residuals for men with and without children. When all other variables are held constant, childless men are significantly shorter than those who have at least one child (ANOVA: childedness, F1,3198=25.5, P<0.001; education, F1,3198=93.1, P<0.001). Multiple regression with number of children as the dependent variable, with height and age as independent variables, provides quantitative confirmation of these results (one-tailed tests: city, r2=0.136, n=1,826; height, P<0.001; age, P<0.001; rural, r2=0.208, n=1297; height, P=0.041; age, P<0.0001).

Figure 1: Height (mean±s.d. and 95% range) of men with and without children.
figure 1

a, Residual height for men with different educational experience (removing effects due to both location of residence and a secular trend over time). b, Absolute height for men of different age cohorts, as a function of whether or not they had any children.

Full size image

Comparisons of means for individual age cohorts (Fig. 1b) reveals that men with children are significantly taller than childless men in each case (twenties, t1157=−2.97, P=0.005; thirties, t1115=−3.49, P=0.001; forties, t514=−3.06, P=0.002), except for men in their fifties (t409=0.17, P=0.863). Because these men were born during the 1930s, they entered the marriage market shortly after the Second World War when the population sex ratio was highly skewed in favour of women and sexual selection on males would have been greatly reduced as a result: the sex ratio for adults of working age (18–64 for men, 18–60 for women) in Wroclaw was 114.3 women to 100 men in the post-war decade, but fell to 104–105:100 in subsequent decades12.

These results indicate that the effect of height on reproductive output might be due to shorter men being disadvantaged in the search for a mate. This idea is supported by the fact that bachelors were significantly shorter than married men (ANOVA with residual height as the dependent variable, childedness as the covariate, and marital status as the independent variable: F1,3198=7.82, P=0.005).