Parent-offspring regression was used to estimate heritabilities (h(2)) of e
ight wing pattern characters plus five principal components of the butterfl
y Bicyclus anynana. The experiment consisted of six metapopulations with fo
ur subpopulations each with an effective subpopulation site Ne = 6 or 12, a
nd additionally, one larger (Ne = 24) panmictic control population. Althoug
h the differences in treatments resulted in substantially different inbreed
ing coefficients at the ninth generation, there were no significantly diffe
rences in estimates of h(2) among populations. The h(2) estimates were mode
rately high (overall mean h(2) = 0.43), even for the smallest most inbred m
etapopulation (mean h(males)(2) = 0.51, mean h(females)(2) = 0.35). We disc
uss why no significant reduction in h(2) was detected in this analysis, whi
le approximately similar sample sizes in molecular and fitness analyses wer
e sufficient to detect severe inbreeding effects.