Correlated-response-to-selection experiments designed to test for a genetic correlation between female preferences and male traits yield biased results.

Correspondence: D. A. Gray, Biological Sciences, Brock University, St Catharines, Ontario L2S 3A1, Canada (email: [email protected]). The genetic correlation between mating preferences and sexually selected (epigamic) traits is a central issue in both Fisherian and good genes models of mate choice (Fisher 1930; O’Donald 1980; Lande 1981; Pomiankowski 1988). One method of demonstrating the existence of a genetic correlation is to select on one trait, and observe if a second trait shows a correlated response to selection (Lande & Arnold 1983; Falconer 1989). This has become the most commonly adopted method of testing for a genetic correlation between mating preferences and epigamic traits: eight of 12 studies reviewed by Bakker & Pomiankowski (1995) used this method. In this commentary, we extend and clarify methodological issues first raised by Butlin (1993), and argue that correlated-response (CR) methods will overestimate the genetic correlation, and may even incorrectly identify a genetic correlation when none in fact exists. The logic behind applying the CR method is fairly straightforward: if a genetic correlation between traits X and Y exists, then individuals selected for X should harbour genes for Y different from the population mean of Y. Thus selection on X produces a response not only in trait X (given that X is heritable), but also a response in trait Y (also given heritability of Y). One can then determine the extent to which an XY phenotypic correlation reflects an XY genetic correlation. In the special case of mate choice, individuals do not express both X (preference) and Y (epigamic trait), thus no phenotypic correlation exists. If a genetic correlation exists, then selection on the male trait should produce a CR in the female preference (and vice versa). The genetic correlation is specifically due to the selected males (or females) harbouring unexpressed preference (trait) genes. For example, high trait-selected males possess highpreference genes, and low trait-selected males possess low-preference genes. These unexpressed genes (in whichever sex) are the proximate cause of the genetic correlation.