The studies of sexual attractiveness are notoriously univariate. Quantitative genetics has repeatedly emphasized the need for considering multivariate phenotypes, especially if their undelrying traits are involved in genetic correlations of varying sign and magnitude. This bias arises mainly in situations when female preference (which defines the selection gradient acting on attractive male traits) is not aligned with the direction of the dominant axis of genetic variance in the complex of attractive traits.
Here we reiterate the important conclusion about considering full multivariate phenotypes instead of univariate isolated traits. We demosntrate how not accounting properly for a multivariate character of attractiveness traits may significantly bias conclusions that are drawn from studies of sexual selection. We discuss the influence of the number of traits, the imbalance of their attractiveness (i.e. difference between female selection gradients acting on specific traits) and their genetic correlations on the outcome of sexual selection for indirect genetic benefits. We provide also a short review of the existing literature that demonstrates how the published estimates of heritabilities in male attractiveness reflect this bias. Among the published accounts of genetic variance in male attractiveness those obtained using specific, isolated traits have substantially higher heritabilities than genuine attractiveness traits (measured as the female response to male phenotypes). We also review existing studies that publish G-matrices of multiple male attractiveness traits and demonstrate that the heritability of the true multivariate male attractiveness in those trait complexes can be an order of magnitude lower than the observed haritability of isolated characters.
Our results have important implications for the evolutionary genetics of sexual selection. They suggest that in many cases the indirect genetic benefits from mating may be greatly overestimated if one uses only isolated univariate male phenotypes. They also emphasize the need for using truly multivariate assays of male attractiveness, ideally by subjecting males to direct assessment by females, and scoring female responses as the male attractiveness index.