Field cages in southern England enclosing single rows of potatoes and infested with differing initial proportions of S (susceptible), R1 (moderately insecticide resistant) and R2 (very resistant) clones of Myzus persicae (Sulzer) were sprayed three times, at 14-day intervals, with pirimicarb (carbamate), demeton-S-methyl (organophosphorus) or a mixture of deltamethrin and heptenophos (pyrethroid and organophosphorus insecticides). The numbers of aphids on top, middle and basal leaves were counted, pre-treatment and one and eight days after each of the three sprays. The resistance genotype frequencies of the three variants were determined by an immunoplate assay which measures the amount of the carboxylesterase E4, the enzyme conferring resistance, in individual aphids. All three chemicals selected strongly for the very resistant variant. After three sprays, R2 aphid frequencies approached or equalled fixation (1·00) for both starting frequencies. However, the deltamethrin-heptenophos mixture selected for R2 aphids more rapidly than the other chemical treatments. Increasing the initial starting frequency of R2 aphids from 0·02 to 0·20 led to a more rapid increase of their frequencies towards 1·00 for all chemicals. The numbers of aphids on all treated plots were less than on the control. However, the more rapid increase in the proportion of R2's on plots treated with deltamethrin plus heptenophos, coupled with enhanced nymph production, resulted in a smaller reduction in numbers than was achieved by the other chemical treatments. The need for novel control methods is discussed in the light of the strong selection for R2 aphids exerted by all three insecticide classes.