Seven third chromosome samples from laboratory populations and one from a wild population were classified by their genotypes at five allozyme loci and by the occurrence of lethals. The data were analysed for independence among classifications by a randomization method, and it was found that when the number of classifications is high the independence hypothesis does not hold. We have split the independence hypothesis among populations and have found that the lack of fit to the hypothesis comes mainly from the wild population. As one of our classifications was lethality, these findings seriously question the interpretation of multilocus fitness estimations. To split the independence hypothesis into interactions, the criterion of Bartlett for non-interaction with exact probability methods was used; interactions at the levels of pairs and triplets were not significant as a whole. It was not possible to carry the interaction analysis any further as the frequencies of chromosomal types were completely determined by their marginal totals at orders of interaction higher than that of triplets. Available parametric estimations of these high order interactions are therefore of dubious meaning. The analyses of chromosome samples and the temporal surveys of population cages described here, as well as data from the literature, suggest that the Est 6-Pgm region is prone to show disequilibrium.