Transverse striping occurs as the sole mode of marking in species which live on bushes and trees whose leaves have strong lateral veins, such as willows, poplars, oaks, privet, syringa, and so on, and these markings associated with the leaf-green of their colouring protect them most effectively from discovery.

The third scheme of marking, namely by spots, occurs in various forms in species of the genera Deilephila and Chærocampa, and it varies in its biological significance; in many species the spots serve as a warning colour, by making the caterpillar conspicuous and easily seen from a distance (Deilephila galii, Fig. 117); in others they imitate the eyes of a larger animal, and have a 'terrifying' effect, as we have already said (Fig. 4); in still other and rarer cases they heighten the resemblance of the caterpillar to its food-plant by mimicking parts of it, as, for instance, the red berries of the buckthorn (Deilephila hippophaës, Fig. 8, r).

Thus all three modes of marking possess a biological value, and protect the soft and easily wounded animal in some way, and, in the case of at least two of them, it is clear that they must have arisen at the very end of the caterpillar's development, since they can only be effective as the animal is approaching full size, and would be valueless in the very young caterpillar. The transverse striping only makes the caterpillar like a leaf when the stripes bear about the same relation to each other as those on the leaf, and eye-spots can only scare away lizards and birds when they are of a certain size. Only longitudinal striping is effective as a protection in the case of young caterpillars, supposing, that is, that they live in or on the grass (Fig. 116).

Let us consider the ontogeny of these different forms of markings, beginning with the eye-spots. It appears that these develop from a sub-dorsal stripe, which appears in the young caterpillar in the second stage of its life, and from it, in the course of the further development, two pairs of large eye-spots are formed. Even in the young caterpillar, scarcely one centimetre in length (Fig. 116), it can be observed that the fine, white sub-dorsal line takes a slight curve upwards on the fourth and fifth segments (C), and on the lower edge of these curves a black line is laid down (D). This is then continued to the upper side (E), and encloses the piece of the sub-dorsal stripe (F and G), and thus there arises a white-centred, black-framed spot which only requires to grow and to differentiate a blackish shadow-centre, the pupil (G), to give the impression of a large eye. This occurs as the caterpillar goes on growing, and after the fourth moult or ecdysis the eyes have already some effect, as the animal is six centimetres in length, but they become even more perfect in the fifth and last stage. During this development of the eye-spot the sub-dorsal stripe disappears completely from the greater part of the caterpillar, persisting only on the first three segments (Fig. 116, B-F).

When we consider that this stripe in the little caterpillar a centimetre long, which lives on the large leaves of the vine, or on the obliquely ribbed willow-herb (Epilobium hirsutum), is quite without protective value, its occurrence at that stage can only be regarded as a phyletic reminiscence due to the fact that the ancestors of these species of Chærocampa possessed longitudinal stripes in the adult state, probably because at that time they lived on plants among the grass, and that, later, when the species changed their habitat to plants with broad leaves which had arisen in the meantime, eye-spots were developed in addition to the green or brown protective colouring which they retained. Thus the modern development of these spots mirrors their phyletic evolution very faithfully; on the two segments there were formed, from pieces of the sub-dorsal line, first white spots ringed round with black, then unmistakeable eyes with pupils (C, D, G). This transformation can only have begun in the fairly well-grown caterpillar, because it was only of any use to it; but later on it was shunted further back in the ontogeny, from the sixth and fifth to the fourth and third caterpillar stage, not in its complete development, but in more and more incipient form; and nowadays the first traces of eyes, as we have already seen, are visible in the course of the second stage. The marking of the more remote ancestors, the longitudinal striping, is now lost in proportion as the eye-spots develop, perhaps because the former would take away from the full effect of the latter. The longitudinal stripes are still quite plainly visible on the first three segments, but these segments are drawn in and are scarcely noticeable when the caterpillar assumes a defiant attitude (Fig. 4).

In the case of marking with ring-spots, which is found especially in species of the genus Deilephila, the ontogeny discloses that it has developed phyletically from the sub-dorsal stripe; in the young stage of this caterpillar also, the sole marking is longitudinal striping; in Deilephila zygophylli, from the steppes of Southern Russia, this persists apparently through all the stages, but in the others it disappears almost completely in the later stages, but only on the segments on which the spot-marking has developed from it. This happens in a manner similar to that in which the eye-spot in Chærocampa arises, a piece of the white sub-dorsal stripe is enclosed above and below by a semicircle of black, and later these semicircles unite, and cut off the portion of the sub-dorsal line, and form a black spot with a light centre within which a red spot frequently appears (Fig. 117, A).

In most species these ring-spots occur on many segments (10-12) (Fig. 117, B), and in cases where they are of importance in making the caterpillar conspicuous and easily seen they sometimes form a double row. But we know one species, Deilephila hippophaës, in which only a single ring-spot exists, and it is a large brick-red spot on the second last segment, mimicking the red berry of the buckthorn (Fig. 8, A and B, r). But individuals also occur in which there are, on the five or six segments in front, smaller ring-spots which become less distinct the further forward they are, and in most caterpillars it is possible, on careful examination, to recognize little red dots on the faded sub-dorsal stripes of these segments (Fig. 8, B). We might be disposed to think, on this account, that the ancestors of D. hippophaës bore rings on all the segments, and that these had gradually become vestigial on the majority of them, because they had lost their earlier biological importance, and now, by adaptation to the buckthorn, could only be of use on the second last. But when we take the ontogeny also into account we find in the young caterpillar only a simple sub-dorsal line, upon which, in the third stage, the red spot of the tail-horn segment appears (Fig. 8, A).

No spots ever occur on the other segments at this stage; they only appear in the last stage, but as they may be entirely wanting, they must have arisen as the result of internal laws of correlation, that is, they must be recapitulations of the hindmost spots which arose in the phylogeny through natural selection. We may conclude this, at least, if we believe in the truth of the fundamental proposition of the biogenetic law, and admit that there is in the ontogeny some more or less distinct recapitulation of the phylogeny.

This proposition may be recognized as true in the case of Deilephila also, if we compare the different species with one another as regards their ontogeny. We find here too that not only the sub-dorsal, that is, the phyletically oldest marking of the Sphingid caterpillars, occurs everywhere in the young stages, but also that it is being shunted back to younger and younger stages, in proportion to the degree of the development of the spot-marking reached in the full-grown caterpillar. Thus, for instance, in the caterpillar of Deilephila euphorbiæ the highest form of spot-marking is reached, and in this species the sub-dorsal line is no longer the sole marking element at any stage. Leaving out of the question the absolutely unmarked little caterpillar which emerges from the egg (Fig. 118, A), there appears at once in the second stage a series of ring-spots connected by a fine white sub-dorsal line (Fig. 118, B). In the following stage, the third, this sub-dorsal line disappears without leaving a trace, and there remains only the spot-marking, which is subsequently duplicated.

Let us compare with this the ontogeny of the bed-straw hawk-moth, Deilephila galii (Fig. 117). The full-grown caterpillar possesses only a single row of ring-spots (B), and accordingly the young stages of the caterpillar up to the fourth show a distinct sub-dorsal line (A), although spots are seen upon it. A still earlier phyletic stage of development is illustrated by Deilephila livornica, in which the ring-spots are all connected by the sub-dorsal line.

It can thus hardly be doubted that the biogenetic law is guiding us aright when we conclude from a comparison of the ontogeny of the different species of Deilephila, that the oldest ancestors of the genus possessed only the longitudinal stripes, and that from these small pieces were cut off as ring-spots, and that these were gradually perfected and ultimately duplicated, while at the same time the original marking, the longitudinal stripe, was shunted back further and further in the young stages, until it finally disappeared altogether.

Let us now refer for a moment to the third form of marking in the caterpillars of the Sphingidæ—transverse striping. This has not arisen out of the sub-dorsal line, but quite independently and at a later date. This is proved with great certainty by the ontogeny of species of the genus Smerinthus. The full-grown, and usually also the young caterpillars, of these species have quite regularly the seven broad oblique stripes which run in the direction of the tail-horn at equal intervals on the lateral surfaces of the body (Fig. 3). They are absent only from the three anterior segments, and upon these a part of the older marking, the sub-dorsal stripe, has persisted. But we find this fully developed in the youngest stages of other species. In Smerinthus populi, the little caterpillar, which has no markings at all when it leaves the egg, very soon shows the white sub-dorsal line, and simultaneously with it the seven transverse stripes, which cut obliquely through it; in the older caterpillars the sub-dorsal then disappears (Fig. 119).

When I was investigating these matters at the beginning of the seventies I did not succeed in procuring eggs of the species of the genus Sphinx, which likewise almost all exhibit the oblique striping in their full-grown stages. But from what I knew of the ontogeny of Smerinthus species I was able to predict that, among the young stages of Sphinx, there must be some with sub-dorsal lines. This was confirmed later, for Poulton found in Sphinx convolvuli that in the first stage there are no oblique stripes, but only the sub-dorsal stripe, while in Sphinx ligustri both kinds of marking were present at the same time.

From all these facts, which I have summarized as briefly as possible, we see that the older phyletic characters are gradually crowded by the newer into ever-younger stages in the ontogeny, until ultimately they disappear altogether. We have now to ask to what this phenomenon is due; is it a simple crowding out of the old and less advantageous by the new and better characters as a result of natural selection, or is there some other factor at work? It is clear in regard to these forms of marking that they can have been developed at first only in the almost full-grown larva by natural selection, because they are of use only there, and that, at the same time, the old marking must have been set aside through the influence of the same factor, in as far as it prejudiced the effect of the new adaptation. This seems to be indicated by the persistence of the sub-dorsal line on those segments which are drawn in when Chærocampa assumes a terrifying attitude, or which do not bear oblique stripes in the leaf-like caterpillars, e.g. the three anterior segments in the species of Sphinx and Smerinthus. When newly acquired schemes of marking like the eye-spots of Chærocampa are transmitted from the last stage to the stage before, this can be explained by following the same train of thought, for the caterpillar is already of sufficient size to be able to inspire terror with its eyes; but in still younger stages the spots would not be likely to have that effect, and yet they occur in quite small animals (20 mm.). More obvious still is the uselessness of the oblique striping in the young stages of the Sphinx and Smerinthus caterpillars, for in the earliest stages of life the caterpillars are much too small to look like a leaf, and the oblique stripes stand much closer together than the lateral ribs of any leaf. Moreover, the little green caterpillars require no further protection when they sit on the under side of a leaf; they might then very easily be mistaken in toto for a leaf-rib. Thus it is certainly not natural selection which effects the shunting back of the new characters. Nor can this be caused by the fact that the new character can only be developed gradually and in several stages, for the oblique striping at any rate arises in the ontogeny all at once. There must therefore be some mechanical factor in development to which is due the fact that characters acquired in the later stages are gradually transferred to the younger stages. But this shifting backwards can be checked by the agency of natural selection as soon as it becomes disadvantageous for the stage concerned.

It is in this way that I explain the fact that the majority of the caterpillars of the Sphingidæ are absolutely without markings when they emerge from the egg. Thus, for instance, the caterpillars of Chærocampa (Fig. 116, A), of Macroglossa (Fig. 115), and of Deilephila (Fig. 118, A), as well as those of the Smerinthus species, are at first without stripe or mark of any kind; they are of a pale green colour, almost transparent, and very difficult to recognize when they sit upon a leaf. How very greatly the different stages can be independently adapted to the different conditions of their life, when that is necessary for the preservation of the species, is shown in the most striking manner by many species. Thus the little green caterpillar of Aglia tau, when it leaves the egg, bears five remarkable reddish rod-like thorns, which in form and colour resemble the bud-scales of the young beech-buds among which they live, and which disappear later on; the full-grown caterpillar shows nothing of these, but is leaf-green, marked with oblique stripes. Even if the use of these reddish thorns be other than I have indicated, we have in any case to deal with a special adaptation of one, and that the first caterpillar-stage, and what can happen at this stage is possible also at every other. Nor is it only animals which undergo metamorphosis that can exhibit independent phyletic variation at every stage, but those also with direct development, and indeed, in the case of these, we may assume adaptation of this kind at almost every stage in the history of the organs, as we have already seen, because the great abridgement of the phylogeny into the ontogeny necessitates a very precise mutual adaptation of the organ-rudiments and of the diverse rates of development.

We have thus been led by the facts discussed—and numerous others from other groups in the animal kingdom might be ranked along with them—to two main propositions, which express the relation of phylogeny to ontogeny. The first and fundamental proposition is the one already formulated. The ontogeny arises from the phylogeny by a condensation of its stages, which may be varied, shortened, thrown out, or compressed by the interpolation of new stages. The second proposition refers to individual parts, and may run as follows: As each stage can undergo new adaptations by itself, so can every part, every organ; such new adaptations very often show a tendency to be transferred to the immediately antecedent stage in ontogeny.

It is not my intention to formulate the laws of ontogeny just now, otherwise many others might be added to these, such as that of the regular transference of characters acquired at one end of a segmented animal to the other segments: I must confine myself here to bringing the two main propositions into harmony with the principles of our theory of heredity.

How phylogeny is condensed in ontogeny can be understood readily enough in a general way, although we cannot profess to have any insight into the detailed processes. The continuity of the germ-plasm brings about inheritance, in that it is continually handing over to the germ-plasm of the next generation the determinant-complex of the preceding one. Every new adaptation at any stage whatever depends on the variation of particular determinants within the germ-plasm, and this in its turn depends on germinal selection, that is, on the struggle of the different determinant-variants among themselves, and on the variation in a definite direction which arises from this, as we have already shown. A new kind of determinant can never arise of itself, but always only from already existing determinants, and through variation of these. But as spontaneous variation never causes all the homologous determinants of a germ-plasm to vary in quite the same way, but only a majority of them, there always remains a minority of the old determinants, which may, under certain circumstances, predominate again, as is proved by the aberrations in Vanessa species due to cold, and by many other kinds of reversion.

But it is not this variation which leads to the prolongation of ontogeny, and the repetition of the phyletic stages within it. In this case it is rather that a new character takes the place of an old one, not that it is added to it. A black spot may arise instead of a red one, but not first a black spot and then a red one. Of course we still know far too little in regard to the intimate succession of events in the stages of ontogeny to be able to say definitely that, in such apparently simple transformations, the older stage does not, in every ontogeny, precede the more recent one as a preparation for it, though it may be only for a brief and transient period.

It is certain, however, that variations such as the addition of a new stage in ontogeny are undergone, and that this implies the occurrence of something really quite new. Therefore such a new stage can arise only from the germ-plasm, by the duplication, and in part variation, of the determinants of the preceding stage. If, for instance, the body of a Crustacean be lengthened by a segment, this must be due to a process of this kind, and in such a case it is intelligible enough that the new segment can be formed in the ontogeny only after the development of the older preceding one, for its determinants come from that, and are from the beginning so arranged that they are only liberated to activity by the formation of the preceding segment.

Now, if in the course of the phylogeny numerous new segments were added to the body of the Crustacean, the ontogeny would be materially prolonged, and condensation would become necessary in the interests of species-preservation. To bring this condensation about, whole series of segments which were added successively in the phylogeny succeeded each other with gradually increasing rapidity in the ontogeny, until finally they appeared simultaneously: the determinants of the segments n, n + 1, n + 2, ... n + x varied in regard to their liberating stimuli, and were roused to activity no longer successively, but simultaneously, in the cell complexes controlled by them. We have thus recapitulation, but with abridgement and compression, of the phyletic stages in the ontogeny. Thus in the nauplius of Leptodora we see the rudiments of five of the pairs of legs of the subsequent thorax (Fig. 111, IV-VIII), and in the Zoæa larva the rudiments of six thoracic legs may be seen behind the already developed swimming-leg (Fig. 114, VI-XIII).

But in the course of the phylogeny a segment may also become superfluous, and we know that it then degenerates and is ultimately eliminated altogether. Thus in a parasitic Isopod, which lives within other Crustaceans, a segment of the thorax is wanting in the relatively well-developed larva, and in the Caprellidæ among the Amphipod Crustaceans the whole abdomen of from six to seven segments has degenerated to a narrow, rudimentary structure. In such cases the gradual degeneration of the relative determinants has preceded step for step the degeneration of the part itself, and when this is complete the ontogeny shows nothing of what was previously present, and so we may speak of a 'falsification' of the phylogeny. But that the complete disappearance of the determinants only comes about with extreme slowness, so that whole geological periods are sometimes not enough for its accomplishment, we have already learnt from our study of rudimentary organs, instances of which can be demonstrated in every higher animal, bearing witness to the presence of the relevant organs or structures in the ancestors of the species.

We can infer with certainty, from the observational data at our disposal, that the disappearance of useless parts is regulated by definite laws; but it is too soon to attempt to formulate these laws, or even to trace them back to their mechanical causes. As we have already said, a much more comprehensive collection of facts, and above all one which has been made on a definite plan, is a necessary preliminary condition to this. But so much at least we may gather from the facts before us, that the degeneration of an organ begins at the final stage, and is transferred gradually backwards into the embryogenesis. Thus the two fingers of birds which have disappeared since Cretaceous times are still indicated in every bird-embryo, though they subsequently degenerate. In various mammals 'pre-lacteal tooth-germs' have been demonstrated in the jaws of embryos, which show us that not only did ancestors exist whose dentition was the modern 'milk-teeth,' but that still more remote ancestors possessed another set of teeth, which was crowded out by the 'milk-teeth'; thus the teeth of the ancestors of the modern right whale (Balæna mysticetus) are only represented in the embryo of to-day in the form of dental pits. And, as we saw already, the Os centrale so characteristic of the wrist of lower vertebrates only appears in Man at a very early embryonic stage, and disappears again as such in the further course of the embryogenesis.

We may perhaps give a preliminary statement of this law as follows: It is impossible that any part or organ should be removed suddenly from the ontogeny without bringing the whole into disorder, and the least serious disturbance of the course of development will undoubtedly be caused if the final stage of the part in question become rudimentary first. Only after this has happened, and the neighbouring parts have adapted themselves to the disappearance, can this extend to the stages immediately preceding it, so that these too degenerate, and allow the surrounding parts to adapt themselves. The further back into the ontogeny the disappearance extends the greater will be the number of other structures affected in some way or other by the degeneration, and these must not all be brought suddenly into new conditions, else the whole course of development would suffer. Thus at first only those determinants may disappear—and can disappear according to the laws of germinal selection—which control the final form of the useless organ, then those just preceding them, which controlled, let us say, its size, and thus more and more of the previously active determinants disappear, and hand in hand with this disappearance there is variation of all the parts correlated with the dwindling condition of the organ, so that their own development and that of the animal as a whole suffers no injury. If it were otherwise, if when a part became useless its collective determinants were all to disappear at the same time, the whole ontogeny would totter, in fact it would be much as if a man who wished to remove the breadth of a window from a house standing on pillars were to begin by taking away the foundation pillar.

It is, of course, to be understood that these processes go on so exceedingly slowly that personal selection takes a share in them, at least at the beginning. Later on, the further degeneration of a useless organ or rudiment has no effect on the individual's power of life, and therefore depends solely upon the struggle of the parts within the germ-plasm (germinal selection).

If we could see the determinants, and recognize directly their arrangement in the germ-plasm and their importance in ontogeny, we should doubtless understand many of the phenomena of ontogeny and their relation to phylogeny which must otherwise remain a riddle, or demand accessory hypotheses for their interpretation. Several years ago Emery rightly pointed out that the phenomena of the variation of homologous parts might be inferred by reasoning from the germ-plasm theory. If one hand has six fingers instead of five, it not infrequently happens that the other also exhibits a superfluity of fingers, and sometimes the foot does so too. The phyletic modification of the limbs in the Ungulates has taken place with striking uniformity in the fore and hind extremities; no animal has ever been one-hoofed in front and two-hoofed behind. Although I might suggest that this primarily depends on adaptation to different conditions of the ground, and that the Artiodactyls were evolved in relation to the soft marshy soil of the forest, and the Perissodactyls for the steppes, it cannot be denied that germinal conditions may have co-operated in bringing about this uniformity of the direction of variation, especially as the whole structure of the fore- and hind-limbs exhibits such marked similarity. Emery is inclined to refer this to 'germ-plasmic correlations,' and we have assumed from the very first that the different determinants and groups of determinants do indeed stand in definite and close relations to one another. But it seems to me premature to say anything more precise and definite than that in the meantime. I should like, however, to say that determinants or groups of determinants which had in old ancestral germ-plasms to give rise to a series of quite similar structures by multiplication during the ontogeny, and therefore only needed to be present singly in the germ-plasm, would, in later descendants, have to shift their multiplication back into the germ-plasm itself, if necessity required that the homologous parts which they controlled should become different from each other. Then the previously single group of determinants in the germ-plasm would have to become multiple. But as new determinants can only arise from those which already exist, these new ones must have had their place beside the old, and would therefore probably be exposed to any intra-germinal causes of variation in common with them—that is to say, they will tend to vary even later in a similar manner. For instance, we might think of the segments of primitive Annelids, which in form and contents are for the most part alike, as arising from one germ-rudiment, from which, when, in the higher Annelids, the various regions of the body had to take a different form, several primary constituents of the germ-plasm separated themselves off; and in a similar way the much higher and more complex differentiation of the somatic segments in the Crustaceans must have been brought about. Thus we understand how the determinant groups of the germ-plasm multiplied according to the need for increasing differentiation, but remained in intimate relation, which exposed them in some measure to a common fate, that is, to common modifying influences, and in many cases determined them to similar variation.

But we cannot see directly into the germ-plasm, and are therefore thrown back on the inductions we can make from the facts presented to us by the phenomena of visible living organisms. As yet the material for such inductions is scanty, because it has been got together haphazard, and not collected on a definite plan. I therefore refrain for the present from attempting any further elaboration of my germ-plasm theory. It is only when an abundance of observation material, collected according to a definite plan, lies at our disposal that anything more in regard to the intimate structure of the germ-plasm, or the mutual influences and relations of its determinants and its modification in the course of phylogeny can be deduced with any certainty. Meanwhile, we must content ourselves with having, through the hypothesis of determinants, made intelligible at least the one fundamental fact, how it is possible that in the course of the phylogeny single parts and single stages can be thrown out or interpolated, or even only caused to vary, without giving rise to variation in all the rest of the parts and stages of the animal. A theory of epigenesis cannot do this, for, if no representative particles were contained in the germ-plasm, then every variation of it would affect the whole course of development and every part of the organism, and variations of individual parts arising from the germ would be impossible.