Anholocyclic Myzus persicae (Sulz.) from the glasshouse and field can be classified into six variants, each with a characteristic carboxylesterase activity and consequent resistance to carbamate and organophosphorus insecticides. Overlapping activity distributions prevent unequivocal identification of an individual from measurements of its enzyme activity, but this may be done by cloning and measuring the frequency distribution of the enzyme activity in its progeny. Although each of the four lowest esterase variants has a frequency distribution with a single peak, only the susceptible (USLL) is best described by a single Normal distribution. Those of the other three clones (MSIG, French R and T1V) correspond better with two Normally‐distributed components. However, these are not well separated and the lower component accounts for only 10‐25% of the individuals. The very resistant clones from glasshouses have higher mean activities but with very broad distributions covering the whole range observed in this aphid. These broad distributions have peaks corresponding to those of the less resistant variants and arise from the appearance at each generation of a small proportion of individuals with much less active enzyme than their parents. Careful monitoring of the changes in activity between parents and offspring of the most resistant variant (clone G6) showed that complete loss of resistance can occur either in a single step or over several generations, and that below a threshold value, reversion to higher levels is very rare. Such reversion could only be detected by selection, with insecticide, of large clonal populations from an individual that had lost activity. High esterase activity was not stabilised by breeding for 16 generations only from individuals with high esterase activity, and is probably maintained in glasshouses by continuous exposure to insecticides. Spontaneous loss of esterase activity and resistance was observed only in glasshouse populations.