A test of how well the repeatability of courtship predicts its heritability

We compared repeatability and heritability estimates for courtship traits in the housefly Musca domestica L., to test the expectation that the repeatability of a behaviour is an upper-bound estimate of its heritability. The repeatabilities across and within mating partners for two courtship traits, male wing buzzing (Buzz) and female wing extension (Wing-out), were compared to heritabilities calculated from the divergence among subpopulations due to founder-flush events or artificial selection. The male's Buzz showed highly significant repeatabilities within and across female partners, but the values ranged widely (0.3-0.5). The intensity of Buzz decreased significantly with successive courtship attempts to the same female, apparently as a male strategy to optimize his efforts to gain copulation by modulating behaviour for specific females. Repeatabilities for female Wing-out within and across male partners were not significant, but this trait yielded significant heritabilities in two assessments (based on successful courtships). We attribute this contradiction to biased sampling because only the subset of reluctant females met the necessary restriction to refuse copulation throughout four successive courtships. These results demonstrate that repeatabilities are labile within individuals and may actually underestimate heritabilities when there is a genetic predisposition not to repeat performance. We concur with related studies that repeatabilities can identify evolutionary potential, but we advise caution in interpreting low values or comparing their absolute values. (C) 1998 The Association for the Study of Animal Behaviour.