In the butterflies, therefore, a perfect congruence of form-relationship does not exist, inasmuch as the imagines constitute one large group of higher order whilst the larvæ can only be formed into families. If it be admitted that the common characters of butterflies depend on their derivation from a common ancestor, the imagines must have retained certain common characters which enable them to be recognized as allies, whilst the larvæ have preserved no such characters from the period at which the families diverged.
Without going at present into the causes of these phenomena I will pass on to the consideration of further facts, and will now proceed to investigate both the form-relationships within the families. Here there can be no doubt that in an overwhelmingly large majority of cases the phyletic development has proceeded with very close parallelism in both stages; larval and imaginal families agree almost completely.
Thus, under the group Rhopalocera there is a series of families which equally well permit of their being founded on the structure of the larva or on that of the imago, and in which the larvæ and imagines therefore deviate from one another to the same extent. This is the case, for instance, with the families of the Pieridæ, Papilionidæ, Danaidæ, and Lycænidæ.
But there are also families of which the limits would be very different if the larvæ were made the basis of the classification instead of the butterflies as heretofore. To this category belongs the sub-family Nymphalinæ. Here also a very characteristic form of caterpillar indeed prevails, but it does not occur in all the genera, being replaced in some by a quite different form of larva.
In the latest catalogue of Diurnal Lepidoptera, that of Kirby (1871), 112 genera are comprised under this family. Of these most of the larvæ possess one or several rows of spines on most or on all the segments, a character which, as thus disposed, is not met with in any other family.
This character is noticeable in genera 1 to 90, if, from those genera of which the larvæ are known, we may draw a conclusion with reference to their allies. I am acquainted with larvæ of genus 2, Agraulis, Boisd. (Dione, Hübn.); of genus 3, Cethosia, Fabr.; 10, Atella, Doubl.; 12, Argynnis, Fabr.; 13, Melitæa,9 Fabr.; 19, Araschnia, Hübn.; 22, Vanessa, Fabr.; 23, Pyrameis, Hübn.; 24, Junonia, Hübn.; 31, Ergolis, Boisd.; 65, Hypolimnas, Hübn. (Diadema, Boisd.); 77, Limenitis, Fabr.; 81, Neptis, Fabr.; 82, Athyma, Westw.; and finally with those of genus 90, Euthalia, Hübn. – which, according to Horsfield’s figures, possess only two rows of spines, these being remarkably long and curved, and fringing both sides. It may be safely assumed that the intermediate genera would agree in possessing this important character of the Nymphalideous larvæ, viz., spines.
After the genus 90 there are 22 more genera, and these are spineless, at least in the case of the two chief genera, 93, Apatura, and 104, Nymphalis. Of the remainder I know neither figures nor descriptions.10 In the two genera named the larvæ are provided with two or more spine-like tentacles on the head, and the last segment ends in a fork-like process directed backwards. The body is otherwise smooth, and differs also in form from that of the larvæ of the other Nymphalinæ, being thickest in the middle, and tapering anteriorly and posteriorly; neither is the form cylindrical, but somewhat flattened and slug-shaped. If therefore we were to arrange these butterflies by the larvæ instead of by the imagines, these two genera and their allies would form a distinct family, and could not remain associated with the 90 other Nymphalideous genera.
We have here a case of incongruence; the imagines of the genera 1–90 and 91–112 are more closely allied than their larvæ.
From still another side there arises a similar disagreement. The larvæ of the genera Apatura and Nymphalis agree very closely in their bodily form and in their forked caudal appendage with the caterpillars of another sub-family of butterflies, the Satyrinæ, whilst their imagines differ chiefly from those of the latter sub-family in the absence of an enlargement of certain veins of the fore-wings, an essential character of the Satyrinæ.
This double disagreement has also been noticed by those systematists who have taken the form of the caterpillar into consideration. Thus, Morris11 attempted to incorporate the genera Apatura and Nymphalis into the family Libytheidæ, placing the latter as transitional from the Nymphalidæ to the Satyridæ. But although the imagines of the genera Apatura, Nymphalis, and Libythea may be most closely related – as I believe they actually are – the larvæ are widely different, being at least as different as are those of Apatura and Nymphalis from the remaining Nymphalinæ.
Now if we could safely raise Apatura and Nymphalis into a distinct family – an arrangement which in the estimation of Staudinger12 is correct – and if this were interpolated between the Satyridæ and Nymphalidæ, such an arrangement could only be based on the larval structure, and that of the imagines would thus remain unconsidered, since no other common characters can be found for these two genera than those which they possess in common with the other Nymphalideous genera.
The emperor-butterflies (Apatura), by the ocelli of their fore-wings certainly put us somewhat in mind of the Satyrinæ, in which such spots are always present; but this character does not occur in the genus Nymphalis, and is likewise absent in most of the other genera of this group. The genus Apatura shows in addition a most striking similarity in the markings of the wings to the purely Nymphalideous genus Limenitis, and it is therefore placed, by those systematists who leave this genus in the same family, in the closest proximity to Limenitis. This resemblance cannot depend upon mimicry, since not only one or another but all the species of the two genera possess a similar marking; and further, because similarity of marking alone does not constitute mimicry, but a resemblance in colour must also be added. The genus Limenitis actually contains a case of imitation, but in quite another direction; this will be treated of subsequently.
It cannot therefore be well denied that in this case the larvæ show different relationships to the imagines.
If the “natural” system is the expression of the genetic relationship of living forms, the question arises in this and in similar cases as to whether the more credence is to be attached to the larvæ or to the imagines – or, in more scientific phraseology, which of the two inherited classes of characters have been the most distinctly and completely preserved, and which of these, through its form-relationship, admits of the most distinct recognition of the blood-relationship, or, inversely, which has diverged the most widely from the ancestral form? The decision in single instances cannot but be difficult, and appears indeed at first sight impossible; nevertheless this will be arrived at in most cases as soon as the ontogeny of the larvæ, and therewith a portion of the phylogeny of this stage, can be accurately ascertained.
As in the Rhopalocera most of the families show a complete congruence in the form-relationship of the caterpillars and perfect insects, so a similar congruence is also found in the majority of the families belonging to other groups. Thus, the two allied families of the group Sphingina can also be very well characterized by their larvæ;13 both the Sphingidæ and the Sesiidæ possess throughout a characteristic form of larva.
Of the group Bombycina