We have before us then a number of groups of butterflies each with a series of different colour patterns. In each group a portion of the series overlaps a portion of the series belonging to another more or less distantly related group. In the light of recent discoveries connected with heredity and variation the natural interpretation to such a set of phenomena would be somewhat as follows: Each group of Lepidoptera, such as those just discussed, contains, spread out among its various members, a number of hereditary factors for the determination of colour pattern. Within the group differences of pattern depend upon the presence or absence of this or that factor, the variety of pattern being the result of the many possible permutations and
combinations of these colour factors. Within the limits of each group is found a definite number of these factors—more in one group, less in another. But some factors may be common to two or more groups, in which case some of the permutations of the factors would be similar in the groups and would result in identical or nearly identical pattern. To take a simple example in illustration, let us suppose that a given group, (α), contains the eight factors A-H. Since any species in the group may exhibit any combination of one or more of these factors it follows that a considerable number of different forms are possible. Now suppose that another group, (β), distinguished from (α) by definite structural features, also contains eight factors within the group, and that these factors are F-M, F, G, and H being common to both (α) and (β). Any combination therefore in (α) lacking the factors A-E will be paralleled by any combination in (β) lacking the factors I-M. For in both cases we should be dealing only with the factors F, G, and H, which are common to each group. So again a third group might have some factors in common with (α) and some with (β), and so on for other groups. In this way certain of the series of colour patterns found in (β) would overlap certain of those in (α), while others of the groups (β) and (α) might overlap those found in different groups again. The striking resemblances not infrequently found between species belonging to quite distinct groups would on this view depend upon the hereditary factors for pattern and colour being limited
in number, so that the same assortment might not infrequently be brought together even though the group whose members exhibited the resemblance might, owing to structural differences, be placed in different families.
We know from recent experimental work that something of the sort is to be found in the coat colours of different rodents. Agouti, black, chocolate, blue-agouti, blue, and fawn form a series of colours common to the rabbit, the mouse, and the guinea-pig. These colours are related to each other in the same way in these different beasts. In the rat, on the other hand, there occur of this range of colours only the agouti and the black. Each of these species again has certain colour patterns which are peculiar to itself, such as the "English" type in the rabbit, the tricolor pattern in the guinea-pig, or the "hooded" variety in the rat. The total range of colour and pattern is somewhat different for each species, but a few are common to them all. Moreover, there are others which are common to the mouse and the rabbit but are not found in the guinea-pig, and others again which may occur in the rabbit and the guinea-pig but have not been met with in the other two. In certain features the rabbit might be said to "mimic" the mouse, and in other features the guinea-pig. It is not, of course, suggested that the case of the butterflies is so simple as that of the rodents, but so far as we can see at present there would seem to be no reason why the explanation should not be sought along the same lines.
On this view the various colour patterns found among butterflies depend primarily upon definite hereditary factors of which the number is by no means enormous. Many of these factors are common to several or many different groups, and a similar aggregate of colour factors, whether in an Ithomiine, a Pierid, or a Papilio, results in a similar colour scheme. The likeness may be close without being exact because the total effect is dependent in some degree on the size and relative frequency of the scales and other structural features. In so far as pattern goes Hypolimnas dubius and Amauris echeria ([Pl. VIII], figs. 7 and 8) are exceedingly close. But inspection at once reveals a difference in the quality of the scaling, giving to the Hypolimnas, where the black and yellow meet, a softness or even raggedness of outline, which is distinct from the sharper and more clear-cut borders of the Amauris. It is not unreasonable to suppose that these species carry identical factors for colour pattern, and that the differences by which the eye distinguishes them are dependent upon the minuter structural differences such as occur in the scaling. So the eye would distinguish between a pattern printed in identical colours on a piece of cretonne and a piece of glazed calico. Though pattern and colour were the same the difference in material would yield a somewhat different effect.
On the view suggested the occurrence of mimetic resemblances is the expression of the fact that colour pattern is dependent upon definite hereditary factors of which the total number is by no means very great.
As many of the factors are common to various groups of butterflies it is to be expected that certain of the colour patterns exhibited by one group should be paralleled by certain of those found in another group. That cases of resemblance should tend to run in parallel series in different groups is also to be expected, for in some groups the number of factors in common is likely to be greater than in other groups. In consonance with this view is the fact that where polymorphism occurs among the females of a mimicking species the models, though often widely different in appearance, are, as a rule, closely related. Some of the Asiatic Papilios, for instance, resemble Danaines, while others resemble Pharmacophagus Papilios. But although the polymorphism exhibited by the females of a given species may be very marked, we do not find one of them resembling a Danaine and another a Pharmacophagus Swallow-tail. The models of a polymorphic mimic are almost always closely related species[[89]].
In discussing the problems of mimicry more attention is naturally paid to groups which exhibit the phenomenon than to those which either do not do so, or else only do so to a very limited extent. Yet the latter may be of considerable interest. Among the Pieridae of the Old World the phenomenon of mimicry is very rare. Pareronia and Aporia agathon conform
closely to the common Danaid type represented by Danais vulgaris and other species, but apart from these none of the many Pierids in Asia resemble any of the recognised models. Africa is apparently destitute of Pierids which mimic species belonging to other groups. Yet no group of butterflies is more persecuted by birds. Of all the instances of bird attacks collected together by Marshall[[90]] more than one-third are instances of attacks upon this group alone. If birds are the agents by which mimetic likenesses are built up through the cumulative selection of small variations, how can the rarity or absence of mimetic Pierids in the Old World be accounted for? For the species of Pierids, like the species of other families, shew considerable variation, and if this process of selection were really at work one would expect to find many more Pierid mimics in these regions than actually occur. It is true that the white, yellow, and red pigments found in Pierids differ from those of other butterflies in being composed either of uric acid or of some substance closely allied to that body[[91]]. These substances are generally found between the two layers of chitin, of which the scale is composed, whereas the black pigment is intimately associated with the chitin of the scale itself. What is perhaps the principal factor in the formation of a mimetic likeness is the distribution of the black pigment with reference to the lighter pigments; and although the latter are chemically distinct
in the Pierids as compared with other butterflies, there would seem to be no reason why the same factors governing the distribution of black should not be common to members of different groups. A distribution of black pigment similar to that found in a model and its mimic may occur also in a non-mimetic ally of the mimic. Dismorphia astynome, for example, resembles the Ithomiine Mechanitis lysimnia ([Pl. XV], fig. 8) both in the distribution of black as well as of yellow and bright brown pigments. A similar distribution of the black pigment is also found in Dismorphia avonia, but the yellow and bright brown of the other two species is here replaced with white. By a slight though definite alteration in chemical composition this white pigment could be changed into bright brown and yellow with the result that D. avonia would closely resemble D. astynome in its colour scheme and would in this way also become a mimic of Mechanitis lysimnia. Another good instance is that of the females of Perrhybris demophile and P. lorena, the former being black and white, whereas in the latter the white is replaced by yellow and bright brown, giving the insect a typical Ithomiine appearance[[92]]. Here again a definite small change in the composition of the pigment laid down in the scales would result in the establishing of a mimetic likeness where there would otherwise be not even a suggestion of it. It is in accordance with what we know to-day of variation