In this connection a few words on another source of variation may not be out of place. The patterns of butterflies are often very sensitive to changes in the conditions to which they are exposed during later larval and pupal life. Many moths and butterflies in temperate climates are double brooded. The eggs laid by the late summer brood hatch out, hibernate in the larval or pupal state, and emerge in the following spring. This spring brood produces the summer brood during the same year. In these cases it often happens that the two broods differ in appearance from one another, a phenomenon to which the term "Seasonal Dimorphism" has been applied. A well-marked instance is that of the little European Vanessid, Araschnia levana. The so-called levana form which emerges in the spring is a small black and orange-brown butterfly ([Pl. VI], fig. 10). From the eggs laid by this brood is produced another brood which emerges later on in the summer, and is, from its very different appearance, distinguished as the prorsa form ([Pl. VI], fig. 9). It is very much darker than the spring form and is characterised by white bands across the wings. The eggs laid by the prorsa form give rise to the levana form which emerges in the following spring. It has been shewn by various workers, and more especially by the extensive experiments of Merrifield[[84]], that the appearance of the levana or the prorsa form from any batch of eggs, whether laid by prorsa or levana, is dependent upon the conditions of temperature under
which the later larval and early pupal stages are passed. By cooling appropriately at the right stage levana can be made to produce levana instead of the prorsa which it normally produces under summer conditions. So also by appropriate warming prorsa will give rise to prorsa. Moreover, if the conditions are properly adjusted an intermediate form porima can be produced, a form which occurs occasionally under natural conditions. The pattern is, in short, a function of the temperature to which certain earlier sensitive stages in this species are submitted. What is true of A. levana is true also of a number of other species. In some cases temperature is the factor that induces the variation. In other countries where the year is marked by wet and dry seasons instead of warm and cold ones moisture is the agent that brings about the change. In some of the South African butterflies of the genus Precis the seasonal change may be even more conspicuous than in A. levana. In Precis octavia, for example, the ground colour of the wet season form is predominantly red, while in the dry season form of the same species the pattern is different, blue being the predominating colour (cf. [Pl. VI], figs. 11 and 12). Such examples as these are sufficient to shew how sensitive many butterflies are to changes in the conditions of later larval and earlier pupal life. The variations brought about in this way are as a rule smaller than in the examples chosen, but in no case are they known to be inherited, and in no case consequently could variation of this nature play any part in
evolutionary change. Before any given variation can be claimed as a possible stage in the development of a mimetic likeness satisfactory evidence must be forthcoming that it is not of this nature, but that it is transmissible and independent of climatic and other conditions.
Many species of butterflies, especially such as are found over a wide range, exhibit minor varieties which are characteristic of given localities. These minor varieties may be quite small. In Danais chrysippus, for example, African and Asiatic specimens can generally be distinguished. On examples from India a small spot is seen just below the bar on the fore wing and on the inner side of it. Eastwards towards China this spot tends to become larger and confluent with the white bar, giving rise to an L-shaped marking; westwards in Africa the spot tends to disappear altogether. The existence of such local races has been used as an argument for the hereditary transmission of very small variations—in the present instance the size of a small white spot[[85]]. For if it can be supposed that small differences of this nature are always transmitted, it becomes less difficult to imagine that a mimetic resemblance has been brought about by a long series of very small steps. But before this can be admitted it is necessary to shew by experiment that the size of this spot is independent of environmental conditions, both climatic and other. Apart from temperature and moisture it is not improbable that the formation of
pigment in the wings may depend in some degree upon the nature of the food. The larvae of D. chrysippus feed upon various Asclepiads, and it is at any rate conceivable that the pigment formation, and consequently the details of pattern, may be in slight measure affected by the plant species upon which they have fed. The species of food plants are more likely to be different at the extremities of the range of a widely distributed form like D. chrysippus, and if they are really a factor in the pattern it is at the extremities that we should expect to find the most distinct forms[[86]]. Actually we do find this in D. chrysippus, though it does not, of course, follow that the cause suggested is the true one, or, if true, the only one. Of the nature of local races too little at present is known to enable us to lay down any generalization. We must first learn by experiment how far they remain constant when transported from their own environment and bred in the environment under which another distinct local race is living. The behaviour of the transported race under the altered conditions would help us in deciding whether any variation by which it is characterised had a definite hereditary basis or was merely a fluctuation dependent upon something in the conditions under which it had grown up. The decision as to whether it is hereditary or not must depend upon the
test of breeding, through which alone we can hope to arrive at a satisfactory verdict upon any given case.
The particular geographical variation which has just been considered happens to be a small one. But it may happen that a geographical variety is much more distinct. Indeed it is not impossible that butterflies which are at present ranked as distinct species may prove eventually to be different forms of the same species. Especially is this likely to be true of many forms in South America, of which Bates long ago remarked "that the suspicion of many of the species being nothing more than local modifications of other forms has proved to be well founded." Since Bates' day more material has been forthcoming[[87]] and it has been shewn that certain colour schemes are characteristic of distinct geographical regions in South America where they may occur in species belonging to very different genera and families. In Central America, for example, the pattern common to many species is determined by horizontal and oblique black bands on a bright fulvous brown ground, with two broken yellow bars towards the tip of the fore wing. The general type is well shewn by Mechanitis saturata and the female of Dismorphia praxinoe ([Pl. X], figs. 7 and 3). Belonging to this pattern group are a number of different species belonging to various families, including several Heliconines and Ithomiines, Pierids such as Dismorphia and Perrhybris, Nymphalines of the genera Eresia and
Protogonius, and other forms. In Eastern Brazil the predominant pattern is one characterised by a yellow band across the hind wing and a white or yellow apical fore wing marking (cf. [Pl. XV], figs. 3 and 8). Here also, with the exception of the Perrhybris, all the various genera which figured in the last group are again represented. It is true that the members of this second group are regarded as belonging to different species from those of the first group, but as species here are made by the systematist chiefly, if not entirely, on the colour pattern this fact may not mean much. Passing now to Ega on the Upper Amazons the general ground colour is a deep chestnut purple and the apical area of the fore wings presents a much mottled appearance (cf. [Pl. XV], figs. 4 and 9). In this group again we find represented the different genera found in the other groups, the only notable absentees being Eresia and Perrhybris. Lastly in Ecuador, Peru, and Bolivia the general pattern scheme consists of orange-tawny markings on a black ground (cf. [Pl. XV], figs. 5 and 10). This group differs somewhat in composition from the preceding in that it contains no Pierid and no Danaid. On the other hand its numbers have been strengthened by the accession of a Papilio, an Acraea, and two species of the Satyrid genus Pedaliodes. Certain writers have seen in the theory of mimicry the only explanation of these peculiar geographical pattern groups. The fashion is in each case set by the most abundant form, generally an Ithomiine of the genus Melinaea. The rest are mimics of this dominant species, either in the
Batesian or Müllerian sense. Batesian mimics are such genera as Dismorphia and Protogonius, to which there are no reasons for attributing disagreeable properties. Of the nature of Müllerian mimics on the other hand are the various Heliconines and Ithomiines which enter into the combination. In each case the whole assemblage is a great "mimicry ring," of which the pattern is dictated by the Ithomiine that predominates in point of numbers. It is, however, very doubtful whether this can be accepted as a satisfactory explanation. The four groups which we have considered are all characterised by a peculiar and distinctive coloration, and in each case the pattern must on the theory of mimicry be regarded as a highly efficient warning pattern. One or other of these patterns must doubtless be looked upon as the most primitive. If so the question at once arises as to why a distasteful genus should change from one efficient warning pattern to another quite distinct one. If the newer pattern affords better protection we should expect it to have spread and eventually to have ousted the older one. That it has not done so must probably be attributed to the old pattern being as efficient as the new one. But if this is so we are left without grounds for assuming the change to have been brought about by natural selection through the agency of enemies to whom warning colours appeal. For natural selection can only bring about a change that is beneficial to the species. Hence we must suppose the change on the part of the dominant model to have been independent of natural selection by