ABSTRACT.1. The continuous spectrum of palatability can be divided without difficulty into two halves: the distasteful forms in one half tend to become models and muellerian mimics, the palatable forms in the other half to become batesian mimics.2. Most of the traditionally recognized differences between batesian and muellerian mimicry are valid and useful: negative versus positive frequency‐dependence, detriment versus benefit to the model, convergent evolution versus advergent evolution. None the less in both kinds, mimicry usually evolves by means of a major mutation whose effects are later modified. ‘Supergenes’ tend to be discovered more in batesian mimics than in muellerian mimics. This results from a complex interaction between natural selection on the one hand, and the origin of clusters of genes of related function through tandem duplication on the other.3. Adaptation is not simply the creation of the best of all possible worlds: what natural selection can do is limited by what is possible for the