Abstract
The heritability of thorax length in the cactophilic Drosophila buzzatii was determined for flies from each of 10 rotting cactus cladodes. For each rot, emerging flies were used as parents of progeny reared in the laboratory. The methods used were full sib analysis with the parents mated assortatively and also offspring-parent regression. From this, heritabilities were measured for the laboratory environment and for the natural environment of the rotting cladode. For the laboratory environment, h2 = 0.3770 +/- 0.0203 and for the natural environment h2 = 0.0936 +/- 0.0087 within rots and h2 = 0.0595 +/- 0.0123 for a population drawn randomly from different rots. Because of the possibility of genotype-environment interaction between the laboratory and rot environments, the methods of B. Riska, T. Prout and M. Turelli were used to show it is possible that there is no such interaction, but if there is, the above natural heritabilities are approximate lower bounds. These results are related to the general problem of determining heritabilities in nature where it is impractical to measure both parents and progeny in nature. Determining heritability not only in nature but in relation to subdivision into ephemeral patches (cladodes in this case) has an important bearing on natural selection response and to general theories of stabilizing selection proposed to explain the existence of genetic variation. Attempts were made to detect selection by using the size of emerging adults as an indicator of various levels of larval stress. No selection was detected, but the power to do so was very weak. Differences between progeny means from different rots indicated some genetic differences between rots which can be adequately explained by small numbers of founders. This suggests a random fine scale subdivision amounting to FST = 0.1483 +/- 0.0462.