Owen and Whiteley (1986) have recently drawn attention to the enormous amounts of colour-pattern variation within populations of the African noctuid moth, Achaea lienardi. They state that 'just about every individual is unique in terms of the monochrome pattern of the forewing' (the only part visible when at rest) and recall that Moment (1962a) commented on similar 'massive' colour polymorphisms in the North Atlantic brittlestar (Ophiopholis aculeata) and the butterfly clam (Donax variabilis). Moment (1962a) had speculated that massive polymorphism is 'protective' in the sense that it might restrict the ability of predators to learn that every colourpattern is palatable, on the grounds that any predator which discovered and ate one individual would be unlikely to encounter other prey with the same colour pattern. This idea has obvious parallels with that of 'protean behaviour' (Chance and Russell 1959, Humphries and Driver 1967, 1970) which proposes the existence of patterns of erratic behaviour in prey whose function is to confuse predators. Moment (1962a) suggested that massive colour polymorphism is generated by 'reflexive selection', 'because it is the variation per se which is adaptive, and the frequency of any one type is determined by a feedback relationship with all the other types'. Coincidentally, Clarke (1962) proposed that frequency-dependent ('apostatic') selection by predators could be responsible for the maintenance of non-mimetic colour polymorphisms. If predators preferentially kill common morphs then rare morphs (or 'apostates') would be at a selective advantage and thus would be actively maintained within the population. Nowadays the term 'reflexive selection' is much less commonly used than 'apostatic selection' and Owen and Whiteley (1986) suggest that it should be resurrected to describe the process that maintains massive polymorphism. In this paper I argue that reflexive selection is no different from apostatic selection, an interpretation that was implied both by Li (1962) and Moment himself (Moment 1962b) in addenda to the original report.