Abstract
Allelic frequency distributions are insufficient to characterize the genetic behavior of a population subject to selection. Two methods for estimating adaptive values which differ in the use of genotypic versus allelic frequency distributions yield disparate descriptions of selection when applied to any particular data set. Adaptive values estimated from genotypic frequency distributions provide a more complete description of gene dynamics and are therefore to be preferred. Experimental populations of Drosophila melanogaster which differ in genotypic frequency distributions but maintain a constant allelic distribution differ both in observed allelic frequency change per generation and also in fitness of the homozygotes. Adaptive values of a genotype whose frequency is constant also depend on the frequency distribution of the other genotypes. Minimal information necessary to predict population change is at least the genotypic frequency distribution.