We may now turn to one of the most ingenious developments of the theory of mimicry. Not long after Bates' original memoir appeared attention was directed to a group of cases which could not be explained on the simple hypothesis there put forward. Many striking cases of resemblance had been adduced in which both species obviously belonged to the presumably unpalatable groups. Instances of the sort had been recorded by Bates himself and are perhaps most plentiful in South America between species belonging respectively to the Ithomiinae and Heliconinae. On the theory of mimicry all the members of both of these groups must be regarded as specially protected owing to their conspicuous coloration and distasteful properties. What advantage then can an Ithomiine be supposed to gain by mimicking a Heliconine, or _vice versâ_? Why should a species exchange its own bright and conspicuous warning pattern for one which is neither brighter nor more conspicuous? To Fritz Müller, the well-known correspondent of Darwin, belongs the credit of having suggested a way out of the difficulty. Müller's explanation turns upon the education of birds. Every year there hatch into the world fresh generations of young birds, and each {15} generation has to learn afresh from experience what is pleasant to eat and what is not. They will try all things and hold fast to that which is good. They will learn to associate the gay colours of the Heliconine and the Ithomiine with an evil taste[6] and they will thenceforward avoid butterflies which advertise themselves by means of these particular colour combinations. But in a locality where there are many models, each with a different pattern and colour complex, each will have to be tested separately before the unpalatableness of each is realised. If for example a thousand young birds started their education on a population of butterflies in which there were five disagreeable species, each with a distinct warning pattern, it is clear that one thousand of each would devote their lives to the education of these birds, or five thousand butterflies in all[7]. But if these five species, instead of shewing five distinct warning patterns, all displayed the same one it is evident that the education of the birds would be accomplished at the price of but one thousand butterfly existences instead of five. Even if one of the five species were far more abundant than the others it would yet be to its advantage that the other four should exhibit the same warning pattern. Even though the losses were distributed _pro rata_ the more abundant species would profit to some extent. For {16} the less abundant species the gain would of course be relatively greater. Theoretically therefore, all of the five species would profit if in place of five distinct warning patterns they exhibited but a single one in common. And since it is profitable to all concerned what more natural than that it should be brought about by natural selection?