A correlation (grey, double-sided arrow) between the phenotypic expression of a trait and a measure of fitness is often assumed to be a causal relationship, i.e. the phenotype is thought to affect fitness. Therefore, the genetic and environmental correlations of the trait with fitness are expected to be the same as the phenotypic correlations with fitness (rA = rE = rP). However, correlation does not imply causation, and the causal route (black single headed arrows) may go via one or more unmeasured phenotypic traits that may be correlated with the trait of interest. In this scenario, phenotypic selection patterns need not be the same as genetic selection patterns, and a genetic perspective is necessary to elucidate patterns of response to selection.
Bolund et al. BMC Evolutionary Biology 2011 11:327 doi:10.1186/1471-2148-11-327