With regard to the former of these two difficulties we have seen that the supporters of the theory regard birds as the main selective agent. At the outset we are met with the fact that relatively few birds have been observed to prey habitually on butterflies, while some at any rate of those that do so shew no discrimination between what should be theoretically pleasant to eat and what should not be pleasant. Even if birds are the postulated enemies it must be further shewn that they exercise the postulated discrimination. It is required of them that they should do two things. {140} In the first place they must confuse an incipient or "rough" mimic with a model sufficiently often to give it an advantage over those which have not varied in the direction of the model. In other words, they must be easily taken in. Secondly, they are expected to bring about those marvellously close resemblances that sometimes occur by confusing the exact mimicking pattern with the model, while at the same time eliminating those which vary ever so little from it. In other words, they must be endowed with most remarkably acute powers of discrimination. Clearly one cannot ask the same enemy to play both parts. If, therefore, birds help to bring about the resemblance we must suppose that it is done by different species--that there are some which do the rough work, others which do the smoothing, and others again which put on the final polish and keep it up to the mark. This is, of course, a possibility, but before it can be accepted as a probability some evidence must be forthcoming in its favour.