The Morphology of the Tunicata, and its bearings on the Phylogeny of the Chordata

As many special points have required detailed examination in the course of this essay, it may be convenient to the reader to be provided with a summary of the main argument.

The ancestry of Chordata must be consistent with the systematic sequence: Echinoderm—Hemichordate—Protochordate—Vertebrate, which, at bottom, implies an evolutional progress of plankton-feeding organisms from a fixed condition with external ciliated tentacles and food-grooves to an eventually free and motile state with endopharyngeal apparatus of gill-slits and endostyle.

This implication is corroborated by the secondary character of the coelomic (locomotive) metamerism of Amphioxus, by the association of an external lophophore with the simplest known condition of gill-slits (Cephalodiscus), and, negatively, by the absence of any proof that pelagic larvae necessitate pelagic ancestors.

The alternative view that Protochordates and Hemichordates are degenerate Vertebrates is precluded by the existence in Vertebrate embryos of vestiges of the complete Protochordate organization of endostyle (=thyroid), epipharyngeal groove (=subnotochordal rod), and tongue-bars (=thymus).

The metamorphosis of Ascidians conflicts at various points with the idea that Tunicata have been derived from Amphioxus-like ancestors, especially in the absence of metamerism, the development of lateral atria before the gill-slits, and the independence of larval and adult nervous systems. The neuromuscular relations in Ascidian larvae and Appendicularians are much more consistent with a theory of incipient than of vestigial metamerism. The development of atria before the gill-slits is in accordance with the phyletic history of the Protochordate type of gill-slit, while both the form of the atrium and its mode of development in Amphioxus are probably secondary modifications. The discontinuity between larval and adult nervous systems is unintelligible on the Amphioxus theory, and points to an early divergence between larva and adult in pre-Chordate ancestors, i. e. to a derivation of Tunicates from ancestors with a metamorphic life-history, before the typical Chordate nerve-tube had come into existence.

On the other hand, the form of the gill-slits in Ascidians has undoubtedly undergone great elaboration from the original type, which, as in Amphioxus, was U-shaped, tonguebarred, and synapticulate. In the retention of this type Amphioxus is more primitive than any Tunicate, and in showing the origin of this type from a simple pharyngeal pouch, Balanoglossus is more primitive than either. As the common ancestor could not have possessed more than two to three pairs (the limit in Ascidians), the multiplication of gill-slits in Balanoglossus and in Amphioxus must be regarded as independently acquired since the separation of these forms from the common stock, presumably in correlation with a secondary elongation of the body.

These conclusions with regard to the original form and ancestral number of gill-slits, together with the points previously mentioned with regard to metamerism, atria, and nervous system, become coherent and intelligible on a theory of the derivation of the Chordata from a line of fixed metamorphic ancestors leading from Pterobranchia to Tunicata, Balanoglossus and Amphioxus being derivable from this sequence at different evolutional levels by loss of fixation, retention of the larval symmetry, and independent adaptation to a burrowing life.

The theory just outlined imposes the necessity of explaining the Ascidian tadpole as an interpolation in the lifehistory. The author’s Auricularia theory of 1894 is therefore drawn upon to suggest a way in which the muscular Chordate larva may have been evolved from an original ciliated larva of the Dipleurula type, by the substitution of muscular for ciliary means of locomotion. A parallel to this is furnished by certain starfish larvae in which muscular flappings of one or more body-processes have replaced the original ciliary locomotion. A primitive Chordate tadpole may well have arisen from an elongated Auricularia-like larva which acquired the power of undulating from side to side. The organization of such a larva would provide the rudiments of characteristic Chordate features, the circumoral band and its underlying nervous system being regarded as predecessors of the medullary folds, and the adoral band with its ventral loop as the predecessor of the peripharyngeal band and endostyle. By postponement of fixation, the rudiments of the gill-slits, which were primarily organs of the adult stage, may have come into working relations with the predecessor of the endostyle, leading to a synthesis of the full Chordate combination of organs. The loss of the larval tail and nervous system at the metamorphosis of an Ascidian would thus correspond to the loss of muscular appendages and larval nervous system at the metamorphosis of a starfish larva. On the other hand, the persistence of these structures in other Chordata would be attributable to the loss of fixation and metamorphosis and consequent survival of the larval type of organization (paedomorphosis).

This chapter closes with a reinterpretation of the development of the endostyle in Amphioxus, which is confirmatory of its origin from the ventral loop of the Echinoderm adoral band.

The prevalent view of the primitive nature of Appendicularians is here examined in detail. The horizontal tail is a common feature of Synascidian larvae, but the dextral twist of the intestinal loop, in opposition to the sinistral twist in fixed Ascidians, militates against an origin of Appendicularians directly from that source. On the other hand, the passage of the nerve-cord across the right side of the oesophagus is found again in the larva of Doliolum, together with a straightening out of the intestinal loop and even a dextral twist in many species. This indication of affinity is corroborated by agreement between the two types in what appears at first sight to be a primitive feature, the position of the endostyle in front of the gill-slits; but this, in Doliolids, is shown to be a secondary modification, caused by the backward rotation of the cloaca and the withdrawal of the peribranchial cavities from the front part of the pharynx. From the mode of working of the pharyngeal apparatus, and from the structure of the endostyle, it is shown that the Appendicularian pharynx has been derived from that of Doliolidae, which is itself derivable from that of normal Ascidians. The oozooid of Doliolum possesses the equivalent of only two pairs of gill-slits, so that the single pair of Appendicularians represents no great reduction. The sequence is not reversible.

In addition to these points the differentiation of the Appendicularian ectoderm into oikoplastic and non-oikoplastic areas is a unique and extremely modified character. As the nonoikoplastic area lies posteriorly, and receives the apertures of gill-slits, intestine, and gonad’s, it is claimed as the equivalent of a shallow atrio-cloacal cavity, like that of Doliolum, from which it has probably been derived by suppression of the cavity and eversion of the epithelial lining (cf. gastrozooids of Doliolum). The dorsal hood of various Appendicularian genera is regarded as a modification of the dorsal outgrowth of the cloacal rim which carries the creeping stolon in A n c h i n i a and the free buds in Doliolum.

On this theory, the ‘Haus’ of Appendicularians, formed from the oikoplastic area, becomes the exact equivalent of a complete Doliolid test, which has been shown by Uljanin to be subject to exuviation and renewal like the Appendicularian ‘Haus The primitive form of house is shown to correspond with the requirements of this theory by possessing two opposite apertures, one oral and the other posterior.

The mixture of larval and adult characters which is so characteristic of the young Doliolum, and is a result of the abandonment of fixation, provides the key to Appendicularian structure, which combines the simplified body of an adult Doliolum with retention and specialization of the larval tail. The alleged metamerism of the Appendicularian muscle-cells is examined, and Martini’s theory of ‘Eutely’, as a substitute for ‘partial Neoteny’, is criticized.

This chapter deals primarily with a number of anomalous and puzzling endodermal glands, derived from the pharynx immediately behind the endostyle, and either retaining that connexion or acquiring a secondary communication with the exterior by fusion with the ectoderm. These are the ‘pharyngeal pockets’ of Fritillaria, the ‘oral glands’of Oikopleura, and the ‘club-shaped gland’ of Amphioxus. All are claimed to be vestiges of the Ascidian organ of budding known as the ‘epicardium’, with a reservation in the case of the Oikopleura glands, owing to uncertainties with regard to their development.

In the case of the Polyclinid Ascidian Euherdmania, which possesses two separate epicardial tubes, it is known that one undergoes a glandular modification, apparently in consequence of the loss of its original function. In Molgulid Ascidians both epicardial tubes have been arrested in development, and, after fusing with one another in the usual way, have been transformed into a closed renal organ; and the endocrine modifications of endostyle, tongue-bars, gill-pouches, &c., in Vertebrates furnish additional illustrations of the tendency of vestigial organs of the Protochordate pharynx to be transformed into glands of various kinds.

It is suggested that the endodermal ‘cement-glands’ of larval Teleostomes may be further derivatives from the same source, their original position having apparently been ventral and postoral.

The asymmetry of the larval Amphioxus is a remarkable phenomenon which affects the interpretation of many transitory features as well as the general relationships of Amphioxus. It has therefore been restudied and is here explained as the consequence of a secondary reduction of yolk in the egg, entailing premature hatching and the improvisation of a larval feeding mechanism. A great enlargement of the mouth, and special ciliation of its entrance, seem to form the basis of this mechanism, which, under the circumstances of the case, involves a temporary dislocation of all the adjacent parts, and is held to have entailed changes which have left a mark on the permanent organization of the adult. The club-shaped gland probably functions as a source of mucilage in place of the imperfectly developed endostyle. The precocious multiplication and distension of gill-slits, as part of the feeding mechanism, are held to be responsible for the peculiar form and mode of development of the belated atrium. These considerations, added to the presence of a vestigial epicardium, lead to the conclusion that the ancestors of Amphioxus were essentially primitive Ascidians, and that Amphioxus itself, while retaining some primitive characteristics, is another example of paedomorphosis. The ‘caudal appendage’ or ‘urostyle’ of Asymmetron, which carries the larval fin, is claimed to be the homologue of the Ascidian larval tail, of which the larva of Amphioxus retains a vestige in its provisional caudal fin.

This chapter surveys the modes of budding in Tunicata, since any evidence that the epicardium was of relatively late origin within the group would be inconsistent with the conclusion that Amphioxus possessed a representative of it. Apart from two cases, in which the evidence is doubtful, it is shown that there is a sharp cleavage of the fixed Ascidians into two groups, here called Endoblastica and Periblastica, in one of which the inner vesicle of the buds is derived from pharyngeal outgrowths (epicardia), and in the other from peribranchial outgrowths, and that these differences correspond with equally clear distinctions between the two groups in regard to adult structure and larval characters. These two groups must be derivable from some third stock more primitive than either, but such a stock among fixed Ascidians is unknown and may possibly be quite extinct.

On the other hand, the mode of budding in Thaliacea combines both types found in Ascidiacea, owing to the fact that the internal organs of the buds are built up from paired rudiments derived both from the pharynx and the cloaca of the parent. There is every reason to believe therefore that the primitive Tunicata, from which both the Thaliacea and Ascidiacea were derived, resembled the Thaliacea in possessing both pharyngeal and peribranchial outgrowths in the stolo prolifer with which they were doubtless provided. Thus a pair of epicardia probably existed in the earliest stock of Tunicata, together with a pair of peribranchial diverticula.

If Amphioxus is correctly regarded as a paedomorphic representative of that primitive stock, its possession of vestigial epicardia is accordingly consistent. Lankester’s ‘atrio-coelomic funnels’ may conceivably represent the corresponding pair of peribranchial diverticula, the dislocation between epicardial and peribranchial outgrowths being explained by the precocious specialization of the former for larval purposes and the retarded development of the atrium.

It is suggested that these two pairs of elements in the early Tunicate stolon may have originated in the fixed ancestors of the Protochordata as regenerative discs or pockets at the base of the pharynx and atrium respectively, to serve for the rejuvenation of the thoracic region after periodic atrophy (cf. Diazona, Distaplia, &c.). When the ectoderm of the primitive stolon became too specialized to play an important part in budding, these regenerative pockets may then have extended into the stolon and taken over its function, collectively at first (Thaliacea), alternatively in the two sections of Ascidiacea, the inner (or epicardial) elements persisting in those which retained the original median stolon (e.g. Clavelinidae), the outer (or peribranchial) in those which widened the original base of fixation and developed their buds laterally (e.g. Botryllidae).

In a subsequent communication the author will deal with the origin of the Chordate nervous system and with the various cephalic organs associated with it as hypophysis, wheel-organ, subneural gland, &c.

The development of Ascidians has hitherto been interpreted as an example of adult degeneration consequent upon the assumption of sessile habits by a free-swimming ancestor. The organization of this ancestor is generally assumed to have been more or less like that of an Amphioxus or an Ammocoete, sharing an open endostyle and ciliated gill-slits with the former, and cerebral sense-organs and absence of an atrium with the latter, together with the neuromuscular metamerism of both. The tailed larva is supposed to ¹ represent’ such an ancestor, which, having fixed itself to a rock, was eventually superseded as an adult by the modern sessile creature. As larvae are usually smaller than adults, appeal is made to reduction of size as a factor capable of accounting to some extent for the manifest imperfection with which the actual larvae record the features of their former sires.

The most striking imperfections of the biogenetic record which have to be explained on this hypothesis are the following: (1) absence of coelomic and neuromuscular metamerism; (2) development of lateral atria before the gill-slits; and (3) independence of larval and adult nervous systems.

These, possibly, are small matters compared with the possession by Tunicate larvae of the typical Vertebrate combination of notochord, gill-slits, and neural canal. At the same time it is well to realize that since Kowalevsky’s demonstration of the Vertebrate affinities of the Tunicata sixty years ago our knowledge of various types of organization connecting Vertebrates and Invertebrates has greatly increased, and the idea that the ancestor of Vertebrates was a kind of free-swimming pelagic Annelid has long been discarded as untenable. Below the Protochordata (i. e. A m p h i o x u s and the Tunicata) come the Hemichordata, with gill-slits, but little or no notochord and neural canal; and below Hemichordata come the sedentary Echinoderms with neither notochord, neural canal, nor gillslits, yet all linked in a sequence of relationships by common features of structure and development which render their consideration essential in a discussion of Tunicate origins. Even if the Enteropneusta were unknown, the anal blastopore, the azygos coelomic water-pore, and the mesodermal skeleton would remain as remarkable bonds between Echinoderms and the Chordata.

If therefore we can disregard for a moment all theoretical prepossessions, we must admit that, on the face of things, systematic morphology points to conclusions very different from the ideas of sixty years ago. Firstly, the myomerism of Amphioxus and the Vertebrates is probably a secondary phenomenon, derivable from the simple, tripartite organization of the Enteropneusta (Morgan, 1894; MacBride, 1898, 1909)—a conclusion which, physiologically expressed, implies a change from relatively feeble to relatively vigorous means of locomotive activity. Secondly, the gill-slit apparatus of Vertebrates is admittedly in its origin an apparatus for filtering plankton as a food-supply, and in Cephalodiscus, as Harmer pointed out in his famous ‘Challenger’ appendix (1887), we have a Hemichordate that bridges the gap between the earlier, tentaculate, type of plankton-collecting and the later endopharyngeal type with ciliated gill-slits or stigmata. If this bridge be regarded as fairly safe—and without it we can see no way across—the conclusion again implies an evolutional origin of Vertebrates from ancestors of essentially sessile habits, for there is hardly an exception to the overwhelming rule that lophophores are the feeding organs of sedentary or tubicolous types, of which Echinoderms were amongst the earliest examples.

The views advanced by Dohrn and Gaskell that Protochordates and Hemichordates are the result of a retrograde evolution from typical free-swimming Vertebrates are inconsistent with the fact that the Vertebrate pharynx from Cyclostomes upwards gives substantial ontogenetic evidence of descent from a previous Protochordate condition, the extent of which is hardly yet appreciated as fully or as widely as it deserves. That the thyroid gland is the modified remnant of an endostyle is generally admitted, but the equally cogent evidence that the metameric rudiments of the thymus betray an origin from vestigial tongue-bars, as suggested by Willey (1894, p. 29), has been neglected. The developmental evidence in such a form as 8 pin ax (Fritsche, 1910) is unequivocal (cf. Brachet, 1921). Less certain, but not negligible, is the evidence which points to the epipharyngeal groove as the functional predecessor of the subnotochordal rod, the chief difficulty being the greater extent of the latter—a discrepancy which may possibly be explained by the embryonic vestige having become secondarily utilized for the formation of a sub-vertebral ligament and thus extended beyond its original limits.

There is of course nothing novel in regarding the ancestry of Vertebrates as intimately connected with that of Echinoderms and Hemichordata, but hitherto the relations which I have just emphasized between these types of organization and a sedentary, or relatively sedentary, mode of life have been thrust aside by a purely speculative prepossession in favour of ‘free-swimming pelagic ancestors’. So in 1894 we had my Auricularia ancestor, in 1897 Masterman’s Actinotrocha ancestor, and in 1909 MacBride’s, which, though claimed to have been constructed synthetically by ‘adding the arms (of Cepha1odiscus) to the collar of a short-bodied Balanoglossus’, can be seen from its portrait (MacBride, 1909, p. 336, fig. 10) to be identical with an Actinotrocha, except in lacking those details of structure which alone render possible a pelagic career for that anomalous larva (locomotive telotroch, hoodlike pre-oral lobe to protect and guide the food-currents, &c.). Yet in each of these theories we neglected two vital points, firstly that by magnifying tiny larvae into effective adults we exceeded the critical limits possible for simple ciliary methods of locomotion and feeding; and secondly that, had these putative adults ever become mature and laden with gonads, it would have been physically impossible for them to keep afloat, unless the original type of ciliary locomotion were reinforced or replaced by muscular. MacBride’s ancestor, indeed, might well have resorted to wriggling, and so initiated that metamerization of its trunk which was required to make a Vertebrate of it; but it is obvious that violent locomotion, after the manner of a pelagic Annelid, would have been utterly inconsistent with the rest of its organization (ciliated arms for catching plankton).

There is a famous sentence in the preface to Thomas Huxley’s ‘Manual’ which the late Professor Weldon was fond of quoting: ‘The growing tendency to mix up aetiological speculations with morphological generalizations will, if unchecked, throw Biology into confusion.’ I think it might be said with equal truth that the tendency, now happily past, to restrict Zoology to ‘morphological generalization’ has itself caused a good deal of illfounded evolutional speculation.

That the morphological theory of adult recapitulation need no longer limit us in the study of phylogeny, I have already attempted to show (1922); but in view of Professor MacBride’s reply (1926) I may perhaps add the following remarks.

In his text-book of Embryology and elsewhere (to which detailed references are given in my paper on the subject) Professor MacBride had selected a number of cases which seemed to him to be free from ambiguity as tests of the truth of the theory as he regarded it. In addition to other criticism, I took the whole series of these examples and their ancestors from Professor MacBride without exception or qualification, and showed that, contrary to his theory, the test of special resemblance between the larval stage of the * descendants’ and the adult stage of their ‘ancestors’ broke down in every case (1922, § 11). In such a case as that of Port union the absence of legs on the last thoracic segment is indeed so unequivocally a larval character throughout the Isopoda, and not a normal adult character, that argumentation on the point is impossible.

Yet in an article devoted to this subject and to my criticism Professor MacBride ignores the results of this crucial test altogether, and devotes his ‘limited space’ (actually some thirteen pages) to repetition of the general arguments to be found in his earlier writings, with additional illustrations, always readable and full of interest, but in no case directed towards the point at issue.

In view of these facts I am entitled to repeat that the theory of adult recapitulation is dead, and need no longer limit and warp us in the study of phylogeny. Instead of assuming that the phases of a given ontogeny represent a compressed succession of adult ancestors, I shall simply assume (1) that under similar conditions egg, larva, and adult of one ontogeny inherit—or tend to preserve—the characters of egg, larva, and adult of a previous ontogeny; (2) that, instead of new characters tending to arise only towards the end of the ontogeny, they may arise at any stage of the ontogenetic sequence; and (3) that, instead of new characters always tending to push their way backwards in the ontogeny, they may extend into adjoining stages in either direction, either backwards from the adult towards the larva and the embryo (tachygenesis) or forwards to the adult from the embryo and the larva (paedomorphosis).

We can now proceed to deal with the problem immediately before us, the current views on which are so compactly and authoritatively expressed in the article to which I have referred: ‘It is really impossible for a competent marine biologist, and especially for a specialist in Tunicata … to deny the fact that the Tunicate tadpole indicates a former free-swimming ancestor of the Tunicata, which was provided with a notochord’ (MacBride, 1926, p. 470).

To deny a fact is, of course, impossible; but I will try to show that certain facts, which are demonstrable and undeniable, admit of, and in my j udgement demand, a very different interpretation.

Before comparing the life-history of Ascidians as a whole with that of lower groups, it will be convenient to examine the points already mentioned (p. 52), since an intimate relationship between the Tunicata and Amphioxus cannot be doubted, and the points raised are essentially differences between the modes of development of these two types.

Mere reduction of size from an assumed ancestral condition seems to be quite inadequate as an explanation in this case, since the larvae of Amphioxus are just as small as those of Ascidians, if not smaller, yet in them successive coelomic sacs are clearly defined, the gill-slits, or many of them, are perforated before the atrium develops, and there is no distinction to be drawn between larval and adult nervous systems.

An important condition undoubtedly is the extreme specialization of the Ascidian larva for larval purposes, which in this case are two only, viz. locomotion and fixation. The larva does not feed, its mouth being covered by a coating of test-substance (Seeliger in Bronn, p. 812); locomotion is effected, not by ciliary action, or by serpentine movements of the whole body, but by a tail which is sharply defined from the visceral region; and the ventral sucker, by which fixation is accomplished, is given its terminal position in front by a temporary suppression of pharyngeal development, rotation forwards of the endostylar region, dislocation of the mouth to the dorsal side, and complete obliteration of the pre-oral lobe.

To these structural peculiarities of the Ascidian larva must be added the differentiation of the embryonic neural tube into two successive and independent nervous systems, one subservient to the larval, and the other to the adult organization, a remarkable and almost unique phenomenon. Possibly this profound specialization of the larva is causally connected with the direct and ‘determinative’ character of the early development, which contrasts so markedly with that of Amphioxus.

When the egg itself is already differentiated into zones of predetermined fate, it is also useless to look for vestiges of ancestral structures like pre-oral and collar enteroceles, which are wanted functionally neither by larva nor adult.

The absence of regular neuromuscular metamerism in the tail of Ascidian larvae (and in the Appendicularians as well, cf. § 4, end) is, however, a different matter, for metamerism is undoubtedly a means of securing locomotive efficiency, and it is hardly likely that it would be discarded, when once it had been acquired, so long as the larvae at any rate required to swim. It would seem, therefore, that the imperfect segmentation of the Tunicate tail (for details see Seeliger, 1900, and Ihle, 1913) is a case of metamerism in the making, and not of a secondary degeneration, since, as already remarked, the argument based on small size cannot with consistency be invoked in support of the latter theory.

The second difference between the development of Tunicata and Amphioxus concerns the atrium and gill-slits. The atrium of Tunicata is essentially and primarily a pair of lateral canals or chambers. The incorporation of the mid-dorsal integument between them, including the anal aperture, to form a cloaca, is admittedly a secondary phenomenon. The whole peribranchial complex is now known to be of purely ectodermal origin in the embryos of every type of Tunicate.

The earliest—and still, perhaps, the most generally accepted—theory of the larval atrial canals was based on the simple arrangements prevalent in Appendicularians, in which the pair of spiracles, on the analogy of some Vertebrate gill-slits, could be regarded as a pair of similar structures, formed by the fusion of endodermal pouches with ectodermal pockets, the spiracular ring of cilia marking the boundary between the two elements. Although the basis for this interpretation has since been undermined by Delsman’s observation (1910,1912) of the endodermal unity of the whole structure, it remains true that the lateral atria of Ascidians on this theory have been regarded as secondary expansions of the outer (‘ectodermal’) pockets, and the pharyngeal stigmata as secondary multiplications of the originally simple perforations (Fol, 1872, pp. 5-6; Seeliger, 1893, pp. 391, 398; Julin, 1904, p. 607). On this theory Tunicata are primitive Vertebrates with a single pair of ‘true’ gill-clefts.

This view was accepted by Willey in his first paper on Ascidian development (1892, p. 517), but abandoned in his larger paper of the following year (1893, i) owing to further reflection on his discovery of the U-shaped form assumed by the primary protostigmata in C i o n a, with its striking resemblance to the tongue-bar formation in Amphioxus. He concluded that ‘the stigmata in the branchial sac of the Ascidians are derived from and represent three pairs of primary gillclefts’ (1. c., pp. 335, 336, 341). With this conclusion Willey associated an attempt to homologize the atrium of Tunicata with that of Amphioxus, basing his argument on the groovelike rudiments of the larval canals in Clavelina. Two years later, however, he again changed his mind as to the number of primary gill-clefts, which he once more restricted to a single pair (1894, pp. 232-4) and emphasized the tubular character of the atrial ‘inpushings’ as against an origin from branchial grooves (1. c., pp. 211-12), the case of Clavelina being relegated to a footnote (p. 241). This appears to be Willey’s final interpretation, since he reaffirmed it five years later: ‘There are strong grounds for the interpretation of the numerous branchial stigmata as having originated by the subdivision of a single pair of gill-slits which persist in their undivided condition in Appendicularia’ (1899, p. 238).

I confess I have never been able to appreciate the reasons given by Willey for this reversion to his earlier view, for the differences which he showed to exist between Molgula and Ci on a, as regards the order of development of the protostigmata, seem to be much less significant than the agreement between these Ascidians in showing three pairs of primary U-shaped gill-slits. This phenomenon has since been claimed to occur in all types of ‘simple’ Ascidians (see Julin, 1904), and a renewed survey of the facts only strengthens the illuminating comparison which Willey originally drew between this mode of origin of the Ascidian rows of stigmata and the development of the tongue-barred gill-slit of Amphioxus (and B a 1 a n o glossus) with its potentialities of subdivision by synapticula (see below, p. 68). Willey’s later restriction of the number of similar slits in Ascidians to one ‘primitive’ pair was based on minor differences of behaviour and sequence in M o 1 g u 1 a and Ciona; but his selection of the first pair in Amphioxus as alone represented in Ascidians is invalidated at once by the fact that the first pair in Amphioxus (one member of which is claimed to be the club-shaped gland) arises and atrophies without ever undergoing the tongue-barred modification displayed by the rest of the series (Willey, 1894, p. 148).

It is true that Julin’s scheme of Ascidian evolution ignores this Amphioxus-trait altogether, and in its broad outlines, which are consistently based on other data, agrees essentially with Willey’s final interpretation, so that the two views may be thought to be mutually corroborative. Julin’s scheme, however, is purely formal and deductive, assuming as selfevident propositions that a condition with one pair of gillclefts is more primitive than a condition with two pairs, and a condition with two pairs more primitive than one with three pairs. So he gives us in the first line the Appendicularian condition, with a single pair of tubular spiracles, and the other pelagic Tunicates as very debatable derivatives from it; in the second the Archiascidia-Clavelina-Distaplia series, with stigmata derived from two pairs of primary slits; and lastly the Ciona-Molgula-Styela series with three pairs of primary slits, in which the U-shaped form attains its most complete expression. An orderly arrangement of numerical facts is always arresting, but when it conflicts with the order revealed by more substantial facts of structure, it is not doubtful which of the two must give way: priority in Tunicate classification must be given to those characters which most intimately link the group to its primitive relatives. Willey’s discovery of the tongue-barred, synapticulate gill-slit of Amphioxus in the pharynx of certain simple Ascidians has given us such a character, not in its primitive simplicity, but disguised by a number of developmental and evolutional peculiarities which demonstrate its transmission to the group from without, rather than its origination from within. If this be admitted, then the Appendicularian, with its mathematical simplicity, is morphologically degraded, and the ancestral Tunicate was polytrematic, like Balanoglossus and Amphioxus, its pharynx possessing a series of at least two, possibly three, pairs of tongue-barred, synapticulate gillslits.

It is curious that Damas, in his concluding paper on the Ascidian pharynx (published, like Julin’s, in 1904), should have come very nearly to the position here maintained, and then, like Willey, have dropped back into the current Appendicularian theory, with the formula ‘Les Tuniciers ne possèdent qu’une seule paire de fentes branchiales’ (1904, pp. 820, 825). For he first confirmed and extended my early observations (1892) on the apparently independent origin of the protostigmata in Styelopsis (=Thylacium) and Botryllus, and their homology with the indefinite number of perforations in Pyrosoma; and then, after dealing with the various modifications of the type of development discovered by Willey in Ciona, remarked upon the close similarity between the series of three independent crescentic protostigmata of Molgula and the tongue-barred slits of Amphioxus (p. 822). He even admitted that ‘logiquement nous devrons admettre que le type le plus primitif est représenté par la Molgule, dont le développement de la branchie présente, parmi les protostigmates, la régularité la plus grande.

Damas gave two principal reasons for not adopting the one conclusion which he admitted to be logical: (1) the difficulty of accepting the development of Mo1gu1aas more primitive than that of Ciona, when its adult organization is more anomalous; and (2) the occasional connexion between protostigmata I and IV in Ciona before they give origin to II and III.

The first of these reasons is clearly invalid. The adult organization of all modern Ascidians known to us is modified in one way or another (not even excepting Julin’s remarkable Archiascidia), and there are several features of Molgulid structure, notably its tendency towards bilaterality of the gonads, in which it is quite conceivably more ‘primitive’ than Ciona. Moreover, the argument is at variance with established facts: it would prevent us, for example, from accepting Auricularia as a more primitive type of larva than Bipinnaria (although the latter goes through an Auricularia stage), since the adult Holothurian structure is admittedly more modified than that of Asteroids.

The second reason confessedly had great influence with Willey himself (1893, p. 322), as it has had with two subsequent workers—Damas (1904, p. 824) and Huntsman (1913, p. 441). It demands therefore a closer examination, although the connexion is known and admitted to be a mere abnormality, and should no more influence morphological interpretation than any other casual abnormality. This view, though with reservation, has indeed already been foreshadowed by De Selys Longchamps, who has seen but one example of the abnormality in Ciona and two in Ascidiella (1900, p. 678). Taking the normal development of Ciona and Molgula as our basis, we see the pharynx perforated by a succession of imperfectly U-shaped slits, which recall the tongue-barred synapticulate slits of Amphioxus, and ultimately attain complete resemblance to them. Let us assume that in their completed form the two structures are fully homologous. Then we must recognize that the development of the slit in Ascidians has been modified in a quite specific manner. Each gill-slit is represented by a genetically connected pair of protostigmata which are characteristically curved and orientated towards one another like the two halves of a tongue-barred slit, which has already been divided by fusion of the tip of the tongue-bar with the ventral wall of the slit. If the two halves appeared simultaneously, it would be a simple case of precocious or direct development (cf. Doliolum, p. 104), but normally this is not the case. Probably by adaptation to the conditions of interstitial growth in the narrow atrium, the gill-slit develops on the instalment system, first one-half (the parent protostigma), and then the second half (the daughter protostigma), by curvature and upgrowth from the ventral end of the first. To avoid further confusion, let us name the first or parental protostigma a ‘hemitreme’¹ and note that each hemitreme possesses a special growing point at its ventral end, by which its growth into a complete tongue-barred slit is completed. In Molgula the hemitremes arise successively, like the gill-slits of Amphioxus, or rather of Balanoglossus, but in Ciona and its relatives the first two hemitremes arise simultaneously. In Molgula also and its allies the successive hemitremes are all orientated alike, with their growing points behind, while in Ciona and its allies (except Perophora)a mutual symmetry is exhibited by the first two hemitremes (in obvious relation to their simultaneous development), so that their growing points are directed towards one another. Occasionally these curving, growing ends accidentally meet and fuse (is it surprising ?), and produce the illusion of a large tongue-barred slit, which has sent Willey and his followers so far astray in their interpretations (cf. Huntsman, 1. c., p. 441: ‘Willey concluded from this that the first four protostigmata were derived from a single primary gill-slit. This view seems to be justified by the facts’). But that this putative slit is an illusion, and not a morphological unit, is clearly shown by the normal independence of its two halves (cf. Julin, 1904, figs. 23-5), by the irregular

and variable forms which the temporary fusion takes (Willey, 1893, figs. 7, 8, 16, 20), and by the absence of any observation that it ever arises from a single perforation or rudiment. Moreover, in Perophora (Damas, 1904) the two primary perforations become hemitremes of a slightly modified type, symmetrically orientated back to back, so that their principal growing points curve outwards instead of inwards, exactly the reverse of their direction in C i o n a !

It follows that the number of gill-slits represented in the Tunicate pharynx should be determined without regard to this abnormality in C i o n a, and with reference solely to the number of actual hemitremes observable, whenever the development of the pharynx takes place through the intermediation of these structures. This number in Ciona and Molgula alike is three pairs (two in Perophora).

If now we turn to Huntsman’s ingenious diagram (1913, fig. 2) where all the known types of development of stigmata are synoptically assembled and classified, we see that in the series of * Ptychobranchia’ (=Stolidobranchia of Lahille) a gradual evolutional change has taken place in the mode of development, the hemitremes (represented by crooks) of the Molgulids (=‘Caesiridae’) being replaced by couples of long and short protostigmata in Cynthiids (=‘Tethyidae’), in which the genetic connexion between the two is broken at a very early stage, and in the hinder part of the pharynx discontinued altogether. Finally, in Styelopsis (Dendrodoa) and the Botryllidae the protostigmata are represented as independent from the beginning.

The completion of this series, largely by Huntsman’s own observations on Boltenia and Sty el a, makes it clear that the metameric regularity of the protostigmatic units in Styelopsis and Botryllus (as well as Pyrosoma), to which I drew attention in 1892, is a secondary phenomenon. Damas (1. c., p. 820), in his comments on my work, inadvertently says that I considered ‘chacun des stigmates du Pyro so ma comme une fente branchiale’, a remark which, if true, would imply that I also held the same view of the protostigmata of Ascidians generally. Reference, however, to the introductory paragraph of my paper (1. c., p. 505) will show that this was never my opinion. On the contrary, what I did say, when introducing the new term ‘protostigmata’, was simply: ‘It cannot be denied that these structures present striking analogies with the true gill-clefts of Amphioxus and the lower Vertebrata’ (1. c., p. 510). At the time in question zoologists were working under Van Beneden and Julin’s conclusion that true gill-clefts in Ascidians had disappeared. It is not irrelevant to remark that Willey’s paper on Ciona and mine on Styelopsis and Botry 11 us were read, as it happened, at the same meeting in 1892, and neither of us had any previous knowledge of the other’s work. From the moment of hearing Willey’s results, I have never ceased to regard his discovery of the crescentic type of slit as furnishing the real key to Ascidian morphology, and my own as being in some way merely a special modification of it. If I have now succeeded in putting Willey’s discovery on its true basis, the controversy as to the number of gill-slits in Ascidians must automatically cease. For whether Damas and Huntsman are right in supporting, or Julin is right in qualifying, my account of the apparent independence of the protostigmata in Styelopsis and Botryllus, it is now perfectly clear that these structures represent, not gill-slits, but gill-slits bisected. The primitive unit in Tunicate ancestors was a tongue-barred slit, as in Amphioxus, which formed its stigmata by synapticulae. It then became a hemitreme, which, curving dorsally, divided into two protostigmata before it had assumed its final form, and each protostigma produced a row of stigmata by subdivision. Then the protostigmata became independent developmental units, each producing stigmata at first by septation, and in later stages of growth by a kind of budding process both dorsally and ventrally. Finally, the climax of this transformation is reached in Clavelina and its allies, in which even the half gill-slit (or protostigma) fails to complete itself, the unit of formation becomes a self-reproducing stigma, and the whole process of development is so modified that interpretation in terms of gill-slits is largely arbitrary (see Julin, 1. c., figs. 9-20).

The two first pairs of larval perforations in Clavelina are generally admitted to be homologous with the two initial pairs in C i o n a; but whereas Ciona develops a third pair, and the three pairs divide into six pairs of protostigmata, which by septation, longitudinal and transverse, give rise to the entire fenestration of the pharynx, in Clavelina each primal perforation proliferates a transverse series of stigmata, without first going through the protostigmatic elongation, and then each row of stigmata divides into two rows. At tins stage the C1ave1ina pharynx possesses four rows of stigmata derived from two initial perforations. Are these rows homologous with the four rows in Ciona which are produced from the homologous pairs of perforations ? I think the answer must undoubtedly be in the negative, for the rows of stigmata in C i o n a also possess the power of dividing into two, yet it is not by this process that the rows in question are produced. It follows that each primary perforation in Clavelina, while genetically homologous with the corresponding one in C i o n a, is actually equivalent only to a protostigma. Potentially it is a gill-slit, but de facto it is only a half gill-slit; and it never becomes more, because the physiological precocity which characterizes it has rendered morphological completion unnecessary, if not impossible. In Distaplia this precocity is still greater, for each of the primary perforations divides into two, one behind the other, as soon as produced, so that the four rows of stigmata are proliferated from the beginning (Julin, 1. c., figs. 30-2). This is not Julin’s interpretation, it is true, for he regards the precocious division of the larval perforations as equivalent to the splitting of a hemitreme into two protostigmata; but this theory is at variance with the morphological fact that division into protostigmata takes place at the ventral end of the rows of stigmata (i. e. at the point where tongue-bar and ventral wall fuse in Amphioxus), not at the dorsal end as in Distaplia; and if all the circumstances, especially the extra yolk in the Distaplia egg, are taken into account, the interpretation I suggest will probably commend itself as correct. It agrees essentially with the opinions expressed by Damas (1904) after a careful study of these problems from a different standpoint. It is in any case a good example of the arbitrary nature of morphological interpretation in this section, where the morphological units have lost their original boundaries and acquired new physiological powers.

By means of this precocity of proliferation the pharynx of Clavelina is able to keep pace with the growth of the atrium without ever completing the form of its two potential gill-slits. So when Julin contends that the Clavelina condition with two pairs of gill-slits is more primitive than the M o1gu1a or Ciona condition with three pairs, I cannot assent to the proposition on these grounds. It is more probable that the primitive Ascidian pharynx possessed three pairs of tongue-barred synapticulate gill-slits, one of which has been lost in the ‘Krikobranchiate’ series in consequence of an increase of yolk in the egg and precocious subdivision of the series of stigmata formed from the first two perforations. The only ground I can admit for the primitive nature of two pairs of gill-slits is the fact that in Perophora only two pairs of hemitremes are developed; but these hemitremes divide so precociously that true protostigmata are never developed from them, and it must for the present remain a doubtful point whether Perophora really retains the primitive number or has undergone a secondary reduction. The internal armature of the pharynx in Perophora is very variable, a feature which is more characteristic of degenerate than of progressive types.

Thus the Ascidians show clear signs that their gill-slits have been derived from an Amphioxus-like condition, and not from one resembling that of Appendicularians. But the resemblance to Amphioxus ceases with the attainment of three pairs (=six pairs of protostigmata), and all further fenestration is accomplished by subdivision of the rows of stigmata into which these protostigmata become converted. Clearly the tendency to indefinite multiplication of gill-slits is a peculiarity of Amphioxus which Ascidians never inherited: the divergence of one stock from the other must have taken place when the gill-slits were limited to three pairs, or even less, and had not yet acquired the power of proliferation by division.

These considerations fix an important detail in the organization of the primitive Protochordate: its pharynx was provided with not less than two pairs, nor more than three pairs, of tonguebarred synapticulate gill-slits. This is just the organization that can be correlated physiologically with a change from lophophoral to endopharyngeal methods of food collection in a series of small sessile animals. In Cephalodiscus the lophophoral method is combined with a single pair of exhalant water-canals, but from such a condition the change could not be accomplished simply by dropping a tongue-bar into the canals and cutting off the tentacles. An animal of the same size, collecting the same amount of food by endopharyngeal means, would require to pass at least ten times as much water through its pharynx as does Cephalodiscus, so that an increase in the number of canals would seem essential before the lophophore could be entirely replaced. The current would become increasingly dependent on a powerful ciliation of the internal orifices, and this power would be secured without waste of food by down-growth of tongue-bars as the orifices themselves stretched across the pharyngeal wall. In its individual development of course the early Protochordate probably began its life with one pair, and added the second and third pairs as it grew in size.

We can now return to the question of the atrium. As the original Protochordates were polytrematic, and their gill-slits elongated and synapticulate, the perforations themselves could not have been freely exposed, or the pharynx would soon have been torn to shreds. Since in the embryonic development of Ascidians the first rudiments of the gill-slits appear as pharyngeal pouches, or as proliferations recalling pouches which fuse with the atrial epithelium directly, and without the formation of corresponding ectodermal ‘pockets’ (Julin, 1904, fig. 7; cf. Delsman, 1912, on Oikopleura), it would seem probable that the original mode of development was similar to that seen in Balanoglossus to-day, and that the tongue-bar grew downwards at the pharyngeal end of the trematic pouch, the ciliated slit being thus protected within the endodermal pouch itself (cf. MacBride, 1894, p. 411). But in various burrowing Enteropneusts the outer pores of the gill-pouches may come to lie in a pair of dorso-lateral ‘branchial grooves’ (cf. Willey’s Spengelia, 1898, Pl. 47, fig. 1), and in others, e.g. Ptychod e r a, lateral flaps of the body-wall beneath the gill-pores may arch over the back of the body and thus define a functional peribranchial cavity (Spengel, 1893; Willey, 1897, p. 179, Pl. 5). Under the latter circumstances the protection afforded by the distal chambers of the trematic pouches is no longer required, and the U-shaped slits open directly to the surface under cover of the peribranchial flaps (=‘genital pleurae’), almost exactly as the shallow gill-slits of Amphioxus and Ascidians open into their respective ‘atria

Thus, independently of the evidence (whatever its value) afforded by the atrial grooves of the larval Clavelina, a study of the gill-slits themselves points strongly to the conclusion that Willey was right when he derived the lateral atria of Ascidians from primitive ‘branchial grooves’ into which the original series of gill-slits opened; and that he was wrong in deserting this idea and adopting the orthodox Appendicularian theory of ancestry. Shallow gill-slits, like those of Amphioxus and Ascidians, are not primitive structures, even when of the simplest circular form. They possess an inherent tendency to become tongue-barred and synapticulate, a condition which cannot have arisen except under cover. They are therefore to be derived from the complex endodermal funnels exemplified in Balanoglossus, and acquired their shallow form subsequently under cover of protective atrial chambers.

From this point of view, therefore, Ascidians in their development preserve the phyletic order more faithfully than Amphioxus, which flagrantly displays a whole series of large shallow slits without an atrial cover. It is significant, however, that while these primary slits of Amphioxus do not become tongue-barred until the end of the larval growth-period, the secondary slits, which arise under complete cover, are tonguebarred almost from the start (Willey, 1894, figs. 77, 80), thus closely conforming to the tendency in Ascidians.

That Amphioxus larvae should be able to do with impunity what we have assumed was impossible for their ancestors is accounted for by the changed conditions of their development, as discussed more fully below (p. 147). Here it is sufficient to note that the single series of gill-slits, which is abnormal in its development and in its relation to the atrium, is also highly specialized. The slits are arrested in an embryonic condition, enormously dilated, and provided with special sphincters (Goldschmidt, 1905, Taf. IV, V; Van Wijhe, 1914, Pl. II), which doubtless regulate the flow of water through them. These peculiarities are emphatically not primitive: they disappear at the metamorphosis and do not recur, even as transitory features, in the formation of the subsequent right-sided series (Van Wijhe, 1. c., pp. 11, 16, 36, 53). They are adaptive to the exigencies of a precocious larval feeding mechanism, the significance of which can best be appreciated when the abnormalities of the larval stage are considered as a whole (p. 147). In the meantime I need only recall that in a relatively primitive larva, Tornaria, the U-shaped gill-slit is developed almost as rapidly as in Ascidians (Agassiz, 1873, Pl. II; Morgan, 1891, 1894); so that it is not the precocious and specialized left gillslits of Amphioxus, but the retarded and unspecialized right slits, which adhere to the primitive mode of development, and this takes place under cover of the atrium.

Thus, on our second point, we reach the conclusion that in Ascidians the development of the atrium before the gill-slits is much more closely in accord with the probable course of phylogeny than is the developmental sequence of Amphioxus, since the common ancestor of these types must already have possessed an atrium before its tongue-barred, synapticulate gillslits could have assumed their shallow form. As.the gill-slits of Protochordata normally assume this form from the beginning of ontogeny, the atrium should already be in existence.

The original atrium must have been paired with lateral apertures, as in Ascidian tadpoles, though no existing animal is known to retain this type unmodified in the adult condition. In Ascidians the two apertures are united during the metamorphosis by the depression of the dorsal surface to form a cloaca; in Amphioxus, owing to precocious development and dislocation of the gill-slits, the original paired formation is precluded, and a single atrium is produced from a median ventral groove (Lankester and Willey, 1890, especially p. 456; cf. MacBride on ‘atrial folds’, 1898, 1900; Lankester, 1898; Smith and Newth, 1917). The causes of this divergent procedure are not far to seek, if we assume a common starting-point in an ancestor provided with lateral atria which developed ontogenetically from a pair of ‘atrial grooves’ (cf. Willey, 1893, figs. 5, 6). With the great development of the dorsal musculature along the Amphioxus line, the embryonic grooves would be driven more and more ventrally until, on the development of asymmetry, if not before, the dislocation of the left series of gill-slits to the mid-ventral line would render a ventral fusion of the two grooves inevitable (cf. development of atria in B o t r y 11 u s from a mid-dorsal involution, without any ontogenetic trace of the phyletic migration).

In Balanoglossus, moreover, it is a point of great interest that in the development of certain species the first two gill-slits, together with the collar-pore, open on each side into a definite ectodermal depression or chamber, which is closely comparable with the larval atrium of an Ascidian. This was described by Morgan in his account of the metamorphosis of the large West Indian Tornaria (1894, pp. 51, 66, figs. 67-9), which has been referred by Stiasny (1927) to a species of Ptychodera. It is therefore highly probable that these larval chambers become produced and modified, as development proceeds, to give rise to the cavities, already mentioned, which in the adult are bounded by the ‘genital wings’.

The third point in which Ascidian development differs from that of Amphioxus and the Vertebrates, viz. the independence of the larval and adult nervous systems, raises deep issues. The conservatism of the nervous system is almost axiomatic in morphology, and the test of innervation one of the surest guides in the determination of homologies. If it can be shown that in Ascidians the larval nervous system contributes nothing to the adult organization, we shall have come upon a crucial fact of great significance for the interpretation of Ascidian development and evolution.

In other cases of profound metamorphosis (e.g. barnacles, Diptera, &c.), however great the destruction of larval tissues, the central nervous system remains intact, i. e. it is laid down in the embryo, and survives all the changes of form which go on in the body, subject only to minor changes in the degree of concentration. On the other hand, in the metamorphosis of an Ascidian, apart from the absorption of the purely locomotive tail, with its muscles and notochord, there are no radical changes in the visceral organization. The atrial cavities enlarge, the stigmata multiply, the digestive organs differentiate, but all the changes are of a simple and progressive character on the basis of essential features already established in the larva, and are far slighter in character and extent than those involved in the metamorphosis of one of the higher Insects. Yet in spite of this continuity of the visceral organization, the continuity of the nervous system is completely snapped. The whole of the larval nervous system—brain, sense-organs, and nerve-cord—is destroyed; a new brain is developed ab initio; and new nerves grow out from the new brain to supply the various parts of the body. The process is not one of simple obliteration of the neural canal, of atrophy of those parts connected with locomotion, and survival of those which control the viscera. This is an erroneous statement commonly repeated in the text-books, and the error is perpetuated by the unfortunate name ‘visceral ganglion’ which has gained currency in lieu of * brain’ for the principal nerve-centre of the larval Ascidian. Yet, as shown by Salensky (1893), and confirmed by Grave and Woodbridge (1924), this so-called visceral ganglion gives off no nerves to the viscera at all, and is simply and solely the co-ordinating centre by which the cerebral sense-organs of the larva are connected with the adhesive papillae in front- and the caudal muscles behind. Our own observations on Botrylloides are in entire agreement with this account, and we have seen no need to supplement Grave and Woodbridge’s detailed figures of the similar conditions in Botryllus.

This larval nervous system is entirely destroyed and removed by phagocytes at the metamorphosis, and the whole of the adult nervous system is a new development proliferated from a minute remnant of the embryonic neural canal. The region in which this proliferation takes place is usually included in the neurohypophysial canal, immediately behind the sensory vesicle (Botrylloides, Pl. II, figs. 6, 7; Ciona, Willey, 1893, fig. 12; Distaplia, Salensky, 1893, Tab. V, fig. 8), and, conformably with this, was assigned to their cul-de-sac cérébral in Clavelina by Van Beneden and Julin (1887). In this form, however, Willey figures the rudiment as arising from the wall of the sensory vesicle after separation of the neurohypophysial canal (1. c., fig. 44), a discrepancy which is not met by his remark that brain and canal both ‘proceed from the same epithelial tract’ (p. 314). It would seem probable that the real point of origin of the brain rudiment lay a section or two farther back in Willey’s preparations, in that ‘hinder region of the cerebral vesicle’ where the ‘boundary line between the hypophysis and the developing ganglion was by no means so distinct as in fig. 44’ (1. c., p. 313). From the existence of an area of continuity (‘fusion’ in the language of Metcalfe, 1900, p. 507) between the adult brain and neural gland in Clavelina, it is practically certain that in this type, as in all others investigated, the rudiment of the adult brain is bound up with the hypophysial tube, and not with the larval sense-vesicle.

This segregation in space and time between the larval and adult brains applies equally to the posterior extension of the nervous system described in various adult types of Ascidian—the ‘cordon ganglionnaire viscéral’ of Van Beneden and Julin, the ‘Ganglionzellstrang’ of Seeliger. According to Metcalfe (1900) this dorsal or ‘visceral’ nerve-cord, which is absent altogether in Botryllus and many simple Ascidians, and has such varied and confused relations with posterior extensions of the neural-gland system, is anatomically an extension of the adult brain in Cynthiids and Molgulids.¹ In Clavelina, the only form in which its development has been investigated, Van Beneden and Julin established its independence of the actual nervous system of the larva, and its development at the time of metamorphosis in continuity with the adult brain. It does not represent a persistence of any part of the larval visceral ganglion, as stated by Sedgwick (1. c., p. 23) and others, but develops separately, exactly like the adult brain, by transformation of the epithelial lining of a portion of the neural canal. This lining is continuous with that of the cul-de-sac cérébral (=Willey’s neurohypophysial canal) from which the adult brain develops, and the two form together a single unit of structure which can only be regarded, as all investigators since Van Beneden’s time have in fact regarded it, as a persistent part of the neural canal of the embryo, from which the sensory vesicle and * visceral ganglion’ of the larva have been, in a sense, precociously segregated. At the metamorphosis, the epithelial cells of this section of the canal lose, for the first time, their embryonic, undifferentiated character, and become transformed into the solid brain and ganglionated cord of the adult, while the larval brain, sensory vesicle, and caudal nerve disintegrate and disappear.

This independence of the two nervous systems was clearly recognized long ago by Van Beneden and Julin, who drew attention to it in the following words: ‘C’est un fait bien remarquable que l’atrophie complète … de toutes les parties différenciées du système nerveux de la larve, alors que le système nerveux de l’adulte se développe aux dépens de parties restées jusque-là à l’état embryonnaire ou épithélial’ (1. c., pp. 353-4).

So also Seeliger (Bronn, p. 836): ‘Die Zellen des Rumpfganglions haben sich allmählich abgelöst und degenerirten, wie es scheint, sämmtlich, ohne sich in Elemente des Ganglionzellstrangs verwandeln zu können.’

Thus, although larval and adult nervous systems are alike traceable to an origin from the neural canal, they are completely independent one of the other, and it is an error to assert that any part of the actual nervous system of the adult has formed a part of the larval nervous system.¹ The neural tube of an Ascidian has no more, and no less, to do with the nervous system, larval or adult, than has the epineural canal of a Holothurian. Equally erroneous are the widespread text-book accounts which represent the larval nervous system as something much more elaborate than that of the ‘degenerate’ adult, e. g. in this account of the metamorphosis: ‘The nervous system dwindles away to a mere ganglion from which a few nerves come off’ (Dendy, ‘Evolutionary Biology 1912, p. 401). For in a larva like that of Distaplia, in which, owing to abundance of yolk, the adult characters arise precociously, the two brains can be seen side by side, and from the larval brain not a nerve can be traced except to the sucker and the tail, while nerves to pharynx and viscera, as well as to oral and cloacal siphons, can be seen growing out from the adult brain (Salensky, 1. c.).

These facts of development seem quite inexplicable on the current theory ‘that the tailed larva represents the primitive or ancestral form from which the adult Ascidian has been evolved by degeneration’ (Herdman, 1904, p. 62). They are, on the other hand, like many other cases of metamorphosis, perfectly consistent with the theory that in the past history of the group larva and adult have undergone divergent adaptive modifications, with the proviso in this case, that the divergence must be dated back to an epoch before the typical Chordate nervous system had been established.

As the adult solid brain is not a remnant of the larval nervous system, it may be the direct representative of the brain of sedentary metamorphic pre-Tunicate ancestors, which has been preserved, generation by generation, to function during the adult stage, by a kind of ‘imaginal disc’, exactly as the legs of a fly are preserved in a dormant condition for adult development during the legless early stages of the insect’s life.

Thus on each of the three points which were selected for preliminary examination we have reached the conclusion that the current view of Tunicate ancestry breaks down. The absence of coelomic segmentation cannot be explained as a consequence of reduction of size, and the detailed studies made by Seeliger and Ihle of the neuromuscular relations in the tail of Ascidian larvae and Appendicularians suggest incipient, but not vestigial, metamerism (cf. § 4, p. 93). The peculiarities of the Ascidian pharynx are admittedly not primitive, and are traceable to modifications of a limited number of tongue-barred gill-slits (two to three pairs) similar to those of Amphioxus; but those were plainly derived from gill-pouches of the Balanog 1 o s s u s type, and were converted into shallow slits after the development of a pair of lateral atrial chambers. Hence the precocious development and multiplication of the slits in Amphioxus, together with the form and mode of development of its atrium, are secondary modifications, and Ascidians preserve more primitive relations in these respects. Finally, the discontinuity between the larval and adult, nervous systems in Ascidians is too profound to be compatible with derivation from an Amphioxus-like ancestry, and is intelligible only by descent from ancestors with a long history of metamorphosis behind them.

On the other hand, all these points are consistent with the alternative view that the Chordata are derivable from a line of fixed ancestors leading directly from primitive Pterobranchia to Tunicata. A small pharynx of two to three pairs of tongue-barred gill-pouches may well have succeeded the Cephalo-discus type, and so have initiated the various lines which have culminated in Balanoglossus, Amphioxus, and the Ascidians. The primitive stock which first discarded the lophophore is now, so far as we know, extinct; but the Ascidians, with their habits of fixation and budding, may be regarded as the most direct descendants of the Pterobranchia, since their pharynx, highly specialized as it now is, contains the equivalents of not more than three pairs of gill-clefts. Balanoglossus and Amphioxus are explicable on this theory as derived from the sequence at different evolutional levels by loss of fixation, retention of larval symmetry, and independent adaptation to a burrowing life, the flexible, elongated body and indefinite multiplication of gill-slits being secondary, but natural developments under such conditions.

The principal difficulty of such a theory is the consequence it imposes of explaining the Ascidian tadpole as an interpolation in the life-history. Some preliminary remarks will therefore be offered on this point before attempting to corroborate the hypothesis in detail.

When Annelids became changed into Molluscs, the Trochosphere was succeeded by the Veliger, which is simply a Trochosphere with the Molluscan characters of shell and foot superadded, and with the circular prototroch produced laterally into a powerful bilobed velum. This velum is plainly no relic of adult ancestry, but a larval development to sustain the weight of the additional structures carried during the pelagic phase. The bigger the velum, the more fully can the adult organization be developed in the larva without jeopardizing the larval function of distribution. Sooner or later, however, the limit of equilibrium is exceeded; the larva sinks to the bottom; the velum is absorbed, and the snail, already equipped with its essential organs, is ready at once to pursue its particular career.

In the case of a starfish the divergence between larva and adult is greater, and the metamorphosis more profound. In most cases, however, the larval means of locomotion is a slender convoluted ciliated band, which is of little use in the later stages of development, so that some species still anchor themselves when undergoing metamorphosis. But in a species of Bipinnaria, whose habits were described by me many years ago (1893), the pelagic period is prolonged by a substitution of muscular flappings of the pre-oral lobe for the primitive ciliary mode, and the developing starfish is easily carried about by the larval movements. Here the analogy with Ascidian metamorphosis is close, for not only does the larva swim by muscular means, but at the metamorphosis a large part of the larval body, including the muscular appendage, ciliated band, and larval nervous system, is thrown off or absorbed, like the muscular tail and entire nervous system of the Ascidian tadpole. The nervous system of the adult starfish, like that of adult Ascidians, is quite independent of the larval nervous system which it succeeds. This latter system in starfish larvae underlies the ciliated band (Semon, 1888). Whether it acquires any control over the locomotive contractions of the pre-oral lobe is not known, but, as the muscles of this appendage are mesenchymatous and lie directly beneath the ectoderm, the possibility of such a connexion is not excluded.

By analysis and analogy we are now in a position to attempt a theory of Ascidian metamorphosis more consistent with the indications of Comparative Anatomy and Bionomics, that the immediate ancestors of the Chordata were benthic animals, collecting microplankton by means of external ciliated tentacles, and of sessile habit. We assume, from the evidence of Comparative Embryology, that the pelagic larvae of these ancestors were simple Dipleurulae, of the Echinoderm-Enteropneust type, provided with apical eye-spots and a circumoral ciliated band.

It is already clear from our present knowledge of Echinoderm and Enteropneust life-histories that the circumoral band is only capable of discharging locomotive functions in the simplest of these larvae (Auricularia) or in their youngest stages (most Holothurians). In other cases it is drawn out into long arms (Plutei), or broken up and rearranged in a series of girdles (Holothurian pupae, possibly Crinoids) and ‘epaulettes’ (Echinoids), or functionally superseded altogether by an additional girdle or girdles (Tornaria) or by muscular contractions of all the body-processes (Bipinnaria asterigera),or of a special pre-oral appendage (Bipinnaria, sp.).

Our new theory invokes yet another method of prolonging the pelagic phase, but on the lines of the last example. Instead of adding new and more powerful ciliary mechanisms, the larval body elongated, became increasingly muscular, with an increasing tendency towards a segmental arrangement, in consequence of its resorting to lateral undulations as a means of locomotion. This had the effect of bringing the lateral halves of the circumoral band, and its underlying nerve-tract, into closer parallelism on either side of the mid-dorsal line. Subsequently, as a result of additional yolk in the egg, the early ciliated Dipleurula phase was relegated to the embryonic period, and the larvae were hatched as muscular tadpoles, the circumoral band, with its underlying nervous system and its apical sense-organs, having been rolled up meanwhile beneath the surface as a neural canal.

The result of this process would be to bring together for the first time within a single dorsal epithelial tube the rudiments of two nervous systems—a mid-dorsal adult nerve-centre, and a pair of lateral larval nerve-tracts derived from the circumoral band, and connected with the apical sense-organs. If to these features we add the persistence of the larval adoral band and ventral loop as the beginnings of a peripharyngeal band and endostyle, this muscular Dipleurula, which may be distinguished as a ‘Notoneurula’, would lack only gill-slits and a notochord to transform it into a regular Chordate tadpole.

At the same time, however, that these changes were taking place in the larval form, the organization of the sessile adult was presumably changing from external tentaculate to internal trematic modes of collecting plankton. If we imagine the pelagic larval life prolonged until the rudiments of the adult gill-slits made their appearance (as happens, e.g. in Tornaria), the various organs which had hitherto had a purely larval value would come into possible working relations with the gill-slits, which hitherto had been organs only of the adult phase. This would be a real moment of creative evolution, and it is in this way, as I shall hope to show, that the typical Chordate combination of characters has come into being. Certain features, especially the gill-slits and the beginnings of a notochord, have had their origin in the adult phase of sessile Pterobranch ancestors, while the endostyle and neural canal are the results of a further development and transformation of early larval characters. The neuromuscular metamerism, which is merely incipient in the Tunicate larva, did not attain its complete expression until, by the abandonment of fixation, the larval type of organization was enabled to persist to maturity (paedomorphosis). That this, in part, has been the history of Appendicularians, and of Amphioxus more fully, will be shown in the sequel.

This outline, while not claiming at this stage to be demonstrative, will show, I think, that there is no inherent impossibility in the thesis I have to sustain, that the Tunicate tadpole may have been an interpolation within the life-history of a line of sedentary organisms, and that it does not necessitate ‘a former free-swimming ancestor’ to account for it. The homologies between the Dipleurula larva and the Chordate embryo, which were raised in my Auricularia note of 1894, acquire a fuller significance under this new interpretation, and will prove, I believe, more fruitful of results. In that note it was pointed out for the first time how remarkably similar are the medullary (or neural) folds of Vertebrate embryos and the circumoral band of Echinoderm larvae both in their relation to nervous system and sense-organs as well as to mouth and blastopore, and also how strikingly the combined peripharyngeal bands and endostyle of the Protochordata recall the adoral band of Echinoderm larvae, with its significant loop along the floor of the endodermal oesophagus of the larva (Semon, 1888).

The significance of the first point is discussed in a later section of this essay. On the second point the constitution of the endostyle has immediate bearings upon our problem, and it will be convenient to deal with certain aspects of it at once.

The endostyles of Amphioxus and Ascidians are practically identical in structure and relations except in the following points: (1) In Amphioxus the endostyle is a nearly flat, or slightly concave pad, while in Ascidians its glandular sides are steeply arched and enclose a deep groove; (2) the median tract of flagellated cells is adapted to these conditions; it carries relatively short flagella in Amphioxus, but very long ones, extending above the top of the groove, in Ascidians; (3) there are only two tracts of glandular cells on each side in Amphioxus, but usually three in Ascidians, the upper one in Ascidians usually presenting the largest surface towards the endostyle cavity, the lower two being more compact and wedgelike in transverse section.

In both Ascidians and Amphioxus there is a marginal ciliated band surrounding the whole structure except anteriorly where its right and left limbs diverge into the peripharyngeal bands with which they are continuous. But the relations of the marginal band to the endostyle are not alike in the two cases. In Amphioxus the band is part of the thickened pad itself, and is directly contiguous with the outer glandular tract along its whole length (Willey, 1894, fig. 13, after Lankester). It is also, on its outer side directly continuous with the ciliated epithelium of the branchial bars. But in Ascidians the marginal band, although crowning the right and left ridges of the endostyle, is quite isolated by unciliated tracts on either side, i. e. both from the glandular part of the endostyle itself and from the perforated region of the pharynx (Seeliger in Bronn: Clavelina, fig. 71; Cynthia, fig. 72; Botryllus and Ciona, Taf. XVIII).

Alternative views are possible as to the bearings of these differences on the homology of the two organs in Amphioxus and Ascidians. In the development of Ascidians the first sign of differentiation is a longitudinal constriction of the thickened wall of the groove on each side, which defines the outer glandular zone above and separates it from the ventral half as a whole, so that at this stage the Ascidian larva has a certain resemblance to Amphioxus by possessing only two glandular tracts on either side, separated by a narrow band of ciliated cells (Ihle, 1913, p. 500). But the ventral half is not purely glandular, and invariably proceeds to divide into two more or less compact homogeneous glandular tracts, separated by another band of ciliated cells (cf. Seeliger, 1. c., pp. 342, 344), and this ventral half of the endostyle, when so completed, bears by itself an even closer resemblance to the conditions in Amphioxus.

Since the greatest differences between the two types are recognizable in the marginal zone of the endostyle, we are on safer ground, I think, if we homologize the two pairs of glandular tracts in Amphioxus with the two lowest pairs in adult Ascidians, and regard the broad uppermost pair in Ascidians, together with the peculiarities in the marginal bands, as special modifications peculiar to Tunicata. In support of this interpretation may be cited the fact that in the few Ascidians in which only two glandular zones are preserved on either side, it is the upper one which has disappeared (e.g. Distaplia, Lahille, 1890, p. 167, fig. 84), and this condition leads directly to that exhibited by Doliolum (Fol, 1876, p. 232, Taf. VII; Uljanin, 1883, Taf. VI). This identification simplifies the phyletic problem, for the elevation of the sides of the endostyle in Ascidians is clearly a secondary feature to be associated with the development of their accessory endopharyngeal apparatus, on which the transport of food and mucilage to the dorsal groove is largely dependent (Orton, 1913; Hecht, 1918; i. e. ciliated papillae, internal longitudinal bars, transverse membranes, &c.). The additional (upper) tract of gland-cells, as well as the peculiarities of the marginal band, then fall into line as correlated with these new developments.

It follows from these considerations that Amphioxus, in its endostyle as well as in its gill-slits, retains the fundamental or ancestral structure with fewer modifications than do the Ascidians, and the remarkable changes which it shows in the development of its endostyle may well be expected to throw light on the original evolution of the organ.

As Willey’s well-known investigations showed (1891), the endostyle of Amphioxus arises at a very early stage in front of the first gill-slit as a transverse patch of thickened epithelium (Hatschek’s ‘Flimmerstreif’). This becomes V-shaped and, after a considerable pause during the process of ‘symmetrization’, grows backwards in the mid-ventral line between the two rows of gill-slits, with its right and left limbs now closely approximated. The front end of the endostyle, however, retains its bifid or double character up to the stage with eight pairs of gill-slits, when the larva has become practically symmetrical, and the endostyle has extended backwards half-way down the series (1. c., fig. 15). The anterior gap in the endostyle then closes up. From a very early stage the peripharyngeal bands are seen to be ‘continuous’ with the front ends of the two limbs.

In view of the differences to which I have drawn attention between the endostyles of Amphioxus and Ascidians, a complete histological treatment of the endostyle of Amphioxus during its successive phases of development would be of great value, but Willey’s account does not go into details of structure. Two points in particular need further elucidation. What is the precise mode of ‘continuity’ between the peripharyngeal bands and the limbs of the endostyle ? What is the origin of the median tract of flagellate cells ? From the examination of several series of sections kindly lent me by Professor Goodrich, I believe that the median tract of flagellate cells is not derived by the fusion of lateral tracts on the inner side of the limbs of the V, but by the incorporation of a narrow median tract of the general pharyngeal epithelium during the process of approximation. A little group of cells arising in this way was readily traceable for some distance in the median line of the ¹ fused’ portion of the endostyle, but I was unable to follow their later history. One of two things must happen: either the pharyngeal cells so enclosed are not really incorporated, but eventually absorbed, in which case the median cells must be differentiated de novo from the fused limbs of the V; or, if incorporated, they have still to undergo modification to transform them into the definitive flagellate cells. Each of these alternatives raises subsidiary questions, but provisionally I assume that the latter process ensues. This assumption leaves only the distinctly paired elements of the endostyle to be derived from the limbs of the V, which are in continuity with the peripharyngeal bands, and the process involved is a simple differentiation into parallel tracts of glandular and shortly ciliated cells.

There are thus three ontogenetic stages in the development of the endostyle of Amphioxus, which, if summed up with due allowance for the more obvious consequences of the temporary asymmetry, are as follows: (1) a peripharyngeal ciliated ring, the ventral portion of which is thickened by the differentiation of parallel tracts of ciliated and glandular cells. This thickened portion runs transversely across the floor of the pharynx inside the mouth, and is the first part to appear; (2) a postero-ventral glandular loop. This—the V-shaped stage—is produced by the bending backwards along the floor of the pharynx of the glandular half of the previous ring; (3) the endostyle proper, which is produced from the previous stage by closure of the anterior gap and differentiation of the enclosed tract of pharyngeal epithelium into the median row of flagellated cells. This integrated organ then grows backwards along the floor of the pharynx (apparently along a predetermined path, see Gibson, 1910, p. 247).

If we interpret this series phyletically, only two stages are really represented, because the glandular tract in Stage I probably owes its transverse position simply to the general suppression of the right-sided organs at this period of development. The transverse patch of cilio-glandular epithelium, which first arises in the floor of the pharynx, consists at the outset entirely of the left limb of the ultimate V-shaped loop, and, as soon as the right limb begins to develop, it immediately pushes the pre-existent transverse limb backwards (see p. 92).

If this interpretation is correct, the glandular tract did not arise phyletically as a differentiation of the lower half of a simple transverse circlet, but by modification of a posteroventral loop already in existence. The corresponding phyletic stages then become:

1. A peripharyngeal ciliated band with ventral loop.

2. A cilio-glandular elaboration of each limb of the loop.

3. Differentiation of the median area between the two limbs of the loop as a central tract of flagellated cells, and closure of the anterior gap.

It is hardly necessary to point out how closely the first of these stages corresponds with the actual adoral band and loop of Echinoderm larvae, a type of structure which occurs in no other form of animal, adult or larval. In view of the admitted relationship between the two groups, there can be only two theoretical explanations of this correspondence: either Echinoderms are degenerate Chordata, or the endostyle of Protochordata is a further elaboration of the Echinoderm larval loop. The former hypothesis may not be untenable, but I see no way of sustaining it. It is at any rate upon the basis of the second alternative that I proceed.

From the standpoint now defined, it is first of all necessary to justify the view already taken that Appendicularians are not primitive Tunicates. We have given one reason for rejecting their claims in this direction, viz. the simplicity of their gillslits; since, if this is regarded as an ancestral Tunicate feature, the interpretation of the pharynx in terms of tongue-barred slits, like those of Balanoglossus and Amphioxus, becomes impossible. From the high degree to which synapticulation has been carried in Tunicata, it is clearly probable that the Balanoglossus-Amphioxus condition came first, and, with the single exception of the Appendicularians, the various Tunicate conditions are readily derivable from it by processes of further elaboration. How, then, has the Appendicularian condition come about ?

The latest summary of the Appendicularian problem is that of Ihle (1913), and it will be convenient to take this as our starting-point. According to Ihle (1. c., p. 518) there are two views to be considered: (1) that Appendicularians are the most primitive of living Tunicates, and (2) that they are neotenic Ascidian larvae. The first view appears to him to be the most correct, although he admits that Appendicularians have been, on the one hand, so specialized for pelagic life and, on the other, so much reduced and simplified ‘that in many respects the larvae of Ascidians exhibit more primitive features’. He rejects the neotenic theory because of important differences between Appendicularians and Ascidian larvae, such as the single pair of gill-slits, ventral anus, absence of atrium, &c., and because of the primitive nature of the differentiating characters. He adds from Seeliger (Bronn, pp. 915-16) that there is no group of Ascidian larvae to which the Appendicularians show any special relations.

To these considerations I would reply as follows: (1) the gillslits of Appendicularians may be primitive only in the sense that they retain embryonic characters (cf. Euherdmania, Ritter, 1903, p. 255, iii); (2) a mid-ventral position of the anus characterizes only the most advanced family of Appendicularians; the intestinal loop of Appendicularians is essentially asymmetrical, with a dextral flexure (the very few exceptions are admittedly not primitive. See Lohmann and Bückmann, 1926, pp. 100, 129); and the anus is definitely right-sided in Kowalevskia and the Fritillariidae, under conditions which are obviously more primitive than in Oikopleuridae (viz. anus behind the gill-slits); and (3) the absence of an atrium may also be a secondary character, as it is in the case of another possible paedomorph, Amphioxides (Goldschmidt, 1905 and 1906). Only the second of these points therefore would be crucial, if true, but it breaks down upon examination. Oddly enough, Ihle himself inadvertently admits it: ‘Die Fritillariidae haben die primitive Lage des Enddarms beibehalten. Der After liegt rechts und unmittelbar vor der Schwanzwurzel’ (1. c., p. 505). The full significance of this unusual position of the anus will be discussed below.

We come then to Seeliger’s argument that Appendicularians cannot be regarded as neotenic since there is no group of existing Ascidian larvae to which they can be related. This is a shrewd point, for, as Ascidian larvae never feed, an Appendicularian that fully resembled an Ascidian larva would be unable to live at all! There are, however, two ways out of this difficulty. While it is a fact that the pharynx is never functional during the larval period of any Ascidian, this is due to the liberal amount of yolk with which the egg is provided. But the amount varies between wide limits in different types, so that the larva, on hatching, may be in an early or late stage of visceral development. In a Simple Ascidian of the Ciona type the larva hatches with a pair of open gill-slits; in M o 1 g u 1 a or Botryllus the cloaca is developed simultaneously with the atria; in various Synascidians the larva is already a diminutive adult, with larval brain and tail, but with cloaca, fenestrated pharynx, and even a chain of buds all complete; and these differences are mainly dependent on variations in the amount of yolk-provision. Consequently it is possible that Appendicularians may be the modified larvae of some early type of Ascidian the eggs of which were less yolky than those of modern Ascidians, or that they are derivable from one or other of the familiar stocks in which a secondary reduction of yolk has taken place, and the larva has thus become dependent on its own resources for feeding. The first hypothesis would meet the views of those who regard the Appendicularians as primitive, since the larva of a primitive Ascidian would be likely to possess a simpler organization than one of a later type. It would be likely, for example, to possess only a single pair of gill-slits (cf. E u h e r d mania, Ritter, 1. c.), instead of the two pairs with which most modern larvae are provided; and one of Ihle’s difficulties would thereby be removed (‘Wären die Appendicularien tatsächlich neotenische Ascidienlarven, dann wäre es nicht einzusehen, warum bei ihnen die Zahl der Kiemenspalten auf ein Paar beschränkt bleibt’, 1. c., p. 521). This hypothesis, however, cannot easily be tested. The second can be, if we can discover any structural features which plainly point to a relationship with one of the existing types of larva.

The absence of a cerebral eye in Appendicularians may be such a character, since it is also lacking in the larvae of M o 1 g u 1 a and B o t r y 11 u s; but there seems to be nothing else in Appendicularian structure to support a relationship with this group except possibly the indications in Chun’s Megalocercus of a rudimentary epineural gland (1887, Taf. V, fig. 4). The significance of this character, however, even in the fixed Ascidians, is still obscure, and for the present no special importance can be attached to it.

The horizontal tail, on the other hand, has bearings on our problem which are fundamental, though hitherto very inadequately appreciated, except by Damas (1904). Even Ihle makes no comments on its significance except as an indication of Appendicularian specialization. Lohmann has so frequently—and rightly—stressed its relations to the working of the ‘Haus’ that it has apparently come to be regarded as an Appendicularian peculiarity (e.g. 1896, p. 6). Nevertheless, the sinistral twist of the tail into the horizontal plane has long been known as a larval character in Synascidians, temporarily in Clavelina (Seeliger, l. c., p. 774, fig. 162, and Taf. XXVIII), permanently in Distaplia (Damas, 1904) and the Polyclinidae (Grave, 1921). It is only the ventral position of the tail, not its lateral torsion, that is peculiar to Appendicularians, so that a phyletic connexion between Appendicularians and Synascidians is at once suggested. From the nature of the case it is obvious that the condition in Synascidians is more primitive than that in Appendicularians, since the tail in Synascidian larvae, in spite of the twist, retains its primitive position in continuation of the body-axis, while in Appendicularians, both Oikopleura and Fri tillaria, it begins in this condition and then becomes further modified. The tail of the young Appendicularian on hatching is extended behind the body, but is already fully twisted to the left, in both respects resembling the tail of a Distaplia or an Aniaroecium larva. Later on, when the genital hump of the Appendicularian develops, the tail is carried to the ventral side and becomes directed obliquely forwards. (For references see Ihle, pp. 517, 518.)

Unless this remarkable twist of the tail has taken place twice independently in the same group, and in the same direction (for in each case the nerve-cord has been carried round to the left side of the tail), it is obvious that Appendicularians have inherited this condition from Ascidian ancestors; and, as simple Ascidians never show this larval twist, we reach the provisional conclusion that the remote ancestors of Appendicularians were not only fixed Ascidians, but compound Ascidians, which multiplied by budding.

Zoologists interested in the inheritance of acquired characters should note, however, that the horizontal tail provides an illustration of this phenomenon exactly comparable with Packard’s famous case in Ammonites. When the tail is horizontal in the larva it is derived from an egg in which the tail is coiled vertically (meridionally) around the embryo, while the vertical larval tail is usually coiled horizontally round the equator of the embryo, and thus undergoes no torsion. Clavelina is exceptionally interesting because it connects the two conditions and proves that the horizontal position of the larval tail is an ‘impressed character’, due to the mode of embryonic coifing. In it the embryonic tail is coiled vertically, and, when hatched, it retains the horizontally twisted tail for some time. Eventually, however, the tail rectifies itself and becomes vertical as in Ciona. On the other hand, Distaplia and the Polyclinids retain the horizontal position throughout their larval life. Here, then, is a clear example of the ‘inheritance’ of an acquired character. I freely admit, under these circumstances, that the value of the horizontal tail, as a mark of affinity, is debatable.

There are, moreover, several difficulties which militate against the idea that the immediate ancestors of Appendicularians were fixed Ascidians, one of which is the fact that in the latter the intestinal loop is almost invariably twisted to the left, so that the intestine lies on the left side of the oesophagus and stomach, whereas in Appendicularians, even when the anus is median, the general tendency is just the reverse, and in Fritillariidae, as already mentioned, the anus itself is definitely on the right side. The few exceptions to this rule are limited to cases in which the intestinal loop has been secondarily opened out as part of the flotational adaptation of certain types (Lohmann u. Biickmann, 1926, pp. 101, 104). The effect of this dextral torsion of the intestinal loop is seen in the course of the nerve-cord as it passes down the body to the base of the tail. Owing to the deviation of oesophagus to left and intestine to right, the nerve-cord usually runs between the two, i. e. to the right of the oesophagus and to the left of the intestine (cf. Fol, 1872, Pl. iv, fig. 1; Chun, Taf. V, figs. 1 and 4). Only when the loop is very compact, as in Kowalevskia and Fritillaria formica, is this course short-circuited and the nerve-cord reaches the base of the tail by a direct path on the left side of the entire intestinal loop (Fol, 1. c., Pl. vii, figs. 1, 2; xi, fig. 1). These various relations are diagrammatically represented in Text-fig. 4.

In Ascidian larvae with a vertical tail, owing to the tail retaining its postero-dorsal position, the nerve-cord is not involved in the intestinal loop; but in Synascidians with a horizontal tail, the nerve-cord, in spite of its rotation to the left, crosses the rectum on the right side of the latter (e. g. Distaplia, Damas, 1. c., Pl. xxiii, figs. 12, 13). This, as stated, is the exact opposite of the Appendicularian arrangement. (Cf. Text-figures 4 and 5.)

Examining this matter more closely, we find that, in addition to the Appendicularians, there is only one significant exception to the rule that throughout the Tunicata the intestinal loop, when not secondarily straightened out, displays a sinistral twist. The various types are diagrammatically shown in Text-fig. 6. In all Ascidiacea (apart from C o r e 11 a and its relatives, which are characterized by a secondary reversal of the normal symmetry, Longchamps, 1900), in Pyrosoma, and in Salp a the intestine is invariably left-sided; but in the oozooids of D o 1 i o 1 u m the loop has been straightened out, with the anus posterior, and in the gonozooids either this condition is retained (Doliolum mulleri, Anchinia) or the intestine is retwisted, but now to the right side as in Appendicularians (Doliolum denticulatum, &c.). The condition in the gonozooids is important, because the sexual sterility of the oozooid is admittedly a secondary phenomenon, and in any comparison with Appendicularians we have to take account of the possible effect on the form of the intestinal loop which the absence of gonads may have occasioned in the oozooid.

Here then in Doliolum we have another hint of Appendi cularian affinities; and one which is not inconsistent with the former, since it is commonly admitted that the Thaliacea are derivable from Ascidian ancestors (cf. Neumann, ‘Die Pyrosomen der Deutschen Tiefsee-Expedition, 1913, p. 362; Korschelt u. Heider, 1. c., p. 1420).

Unfortunately the tail of Doliolum is not functional as a locomotive organ, and its nerve-cord appears to atrophy at a very early stage (Neumann, 1. c., p. 106). We can neither assert nor deny that its tail was originally horizontal, as in Synascidians, but a number of striking peculiarities in the tail and other organs corroborate the view of a close relationship with Appendicularians:

(1) The tail, although continuing the axis of the body, is forced downwards by the backward rotation of the cloaca, so that it presents a condition remarkably similar to that of a young Appendicularian (cf. Delsman’s figure of the ‘larval’ Oikopleura in Ihle, 1. c., fig. 27, p. 517, with Neumann’s figures of Doliolum denticulatum, Taf. I, fig. 5; III, figs. 1, 2, 4).

(2) The nerve-cord, as it passes backwards between the peribranchial involutions, bends down the right side of the future oesophageal region as in Oikopleura (Neumann, Taf. II, figs. 21, 22, copied in Korschelt u. Heider, 1910, fig. 610, A—this being the posterior section, and c the anterior, of the three figured). Neumann, strangely enough, makes no remark on this interesting feature shown by his sections. I infer the bend to be on the right side from his remarks on the peribranchial involutions, p. 109, where the deeper one is described as ‘links’.

(3) The single otocyst, developed outside the neural tube, lies on the left side of the brain, as in Appendicularians. In Ascidian larvae the otolith lies on the right side. The homology of these organs may be disputed, but the coincidence is worth noting (see ‘The Development of Botrylloides’, this journal, vol. 72, p. 34).

(4) The curious ventral proboscis of the larval Doliolum may be represented by the prominent ventral lip of Oikopleurids (Neumann has shown that Uljanin misinterpreted his early larvae, and drew them upside down).

In Appendicularians the shortness of the endostyle, its anterior position, and the presence of a single pair of gill-slits behind it, undoubtedly suggest the retention of very primitive features. This presumption is supported by two considerations which to me at any rate made a strong appeal in the first instance: firstly, the general arrangement is remarkably similar to that presented by the early larva of Amphioxus, if allowance be made for the temporary asymmetry of that animal; and secondly, it accords with the theory which this essay is concerned to defend, viz. that the endostyle was primitively a differentiation of the short ventral loop of the Echinoderm adoral band. We cannot deal more fairly with this idea than to assume it as a working.hypothesis, and trace its consequences. It will be admitted, I think, that if the Appendicularian arrangement is truly primitive, it must furnish a possible step in the development of the normal Ascidian arrangement, in which the endostyle is coextensive with the pharynx and the gill-slits lie not behind it, but on either side of it. A comparison of the actual feeding process in an Appendicularian and an Ascidian will best reveal the meaning of these differences and throw light on the evolutional problem.

The Appendicularian, by means of its short anterior endostyle, encircles the mouth with a ring of mucilage. The cilia of the endostyle work forwards, like those of the ventral loop of the adoral band of Echinoderm larvae according to MacBride (1914, p. 463). Those of the peripharyngeal bands work upwards on either side of the mouth, thus maintaining a constant stream of mucus round the entrance to the pharynx. The single pair of ciliated slits in the hinder part of the pharynx gape widely and draw through the mouth a stream of water, which trails the mucilage into tangled strings; other cilia set the whole mass rotating and transform it into a conical sieve which entraps food particles; the twisted apex of mingled slime and food is carried backwards as a continuous cord into the stomach (Fol, 1872, fig. 5).

On the other hand, the adult fixed Ascidian does not make and suspend a sieve within the entrance to its pharynx. It has multiplied and elaborated the pair of simple exhalant tubes of Cephalodiscus into at least four to six rows of minute stigmata, derived by the synapticulation of two to three pairs of U-shaped slits. The wall of the pharynx itself is the filter and requires an endostyle as long as itself that will direct the mucilage upwards and sideways over its fenestrated surface, instead of forwards towards the mouth (Orton, 1913, fig. 7).

Thus although morphologically the simple arrangement of organs in the Appendicularian pharynx appeared to be a possible stage in the development of the Ascidian pharynx, the theory breaks down as soon as the parts are seen at work, and we can understand why no Ascidian goes through an Appendicularian phase in the development of its pharynx. The two mechanisms differ radically in principle; and of the two it is the Ascidian arrangement which is relatively primitive and simple, the Appendicularian that is elaborate and rare. The Ascidian arrangement, by which food particles are entangled and swept together along the walls of the pharynx is merely a variant of the general methods employed by other plankton-feeding organisms (e. g. Pterobranchia, Lamellibranchia, &c.); but the free suspension of a rotating sieve of mucus within the pharyngeal cavity is one of the most delicate and elaborate contrivances for food-prehension in the animal kingdom. The Appendicularian arrangement could no more grow into the Ascidian arrangement while maintaining a functional continuity than can the mouth-parts of a caterpillar into those of the imago. The topographical differences could readily be adjusted by an extension of the endostyle backwards between the gill-slits, and by subsequent multiplication and subdivision of the latter. But the Appendicularian endostyle is specialized to drive its mucus forwards, and the extension backwards of such an organ would be useless. By a multiplication of gill-clefts the inflowing current could be diverted from the axial cavity of the pharynx to its walls; but without a shmy coating on the walls themselves, and a mechanism for renewing it and sweeping it upwards, the net result would simply be to destroy the rotating sieve, and allow the food particles to be carried untrammelled through the gill-slits.

On the other hand, the differences between the Appendicularian and Ascidian types of pharynx reveal all the more clearly the affinities of Appendicularians with Doliolum. Here again we meet with a pharynx in which the perforations are limited to the hinder part of the cavity, behind the endostyle (cf. e. g. Sedgwick, fig. 46; Korschelt and Heider, figs. 767, 835), and the method of feeding is essentially similar to that in Appendicularians, with minor variations associated with differences in the number and obliquity of the row of perforations in different types of zooid (Fol, 1872, fig. 4; 1876, Taf. VII). In each case the endostyle propels its mucus forwards, and as it is swept dorsally along the peripharyngeal bands, ‘fringes’ of slime are trailed off them by the current, are entangled together on the dorsal side by rotation, and driven backwards as a free cord diagonally through the pharyngeal cavity to the oesophageal aperture. Again, the delicacy of the adjustment of parts to the whole is manifest, and one of Fol’s comments may be quoted with advantage: ‘Beobachtet man das Thier im Augenblicke wo es eben anfängt Nahrung aufzunehmen [feeding being intermittent], so sieht man … dass dieser, am unteren [i. e. morphologically posterior] Ende freie Faden gerade auf den Schlundeingang lossteuert; es muss somit die Richtung der Wasserströmungen dermassen combi n i r t sein, dass der Faden mit Nothwendigkeit genau den Schlundeingang trifft. Niemals sah ich den neugebildeten Faden etwa durch eine Kiemenspalte austreten’ (1876, p. 236).

These remarks, though applicable generally, were made with special reference to the gastrozooid of Doliolum, in which Fol observed that three or four of the anterior gill-slits lashed inwards instead of outwards—a fact of considerable physiological interest and worth noting, even if unique.

Now in the bud-generations of Doliolidae the pharynx is peculiar in possessing a single row of small perforations on each side, set obliquely or transversely to the longitudinal axis. The oblique condition is plainly intermediate between the normal or longitudinal series of perforations shown by Pyrosoma and the transverse series of A n c h i n i a; and the perforations in these types have been generally regarded as having the same morphological value. The view put forward by Lahille (1890, pp. 53-4, 63-6), and subsequently adopted by Julin (1904), that the series in Pyrosoma represents one of the transverse rows of stigmata of an ordinary Ascidian rotated at right angles to its normal position, was shown by me long ago to be completely at variance with the evidence adduced in its support (1892, p. 506; 1893, p. 245), and this view has been corroborated on other grounds by Seeliger (1895) and Damas (1904, p. 780). It follows then, if the perforations in Dolio-1 um are homologous with those of Pyrosoma, that it is the Doliolid series which has undergone rotation. The obliquity of the series in, for example, Doliolum Ehrenbergi marks, as it were, the beginning of an angular rotation backwards through 90° of the whole series of protostigmata, pivoted on a postero-ventral centre near the back of the endostyle; and the rotation may be regarded as completed with the transverse position of the slits in A n c h i n i a.

This theory, unlike Lahille’s, seems to be in complete agreement with the facts of the case. It is not the mere shift backwards of the cloacal aperture that has entailed a change in the structure of the pharynx. So long as the lateral peribranchial cavities remained coextensive with the pharynx, the rotation of the cloacal aperture had little effect, as shown by Pyrosoma itself. But in the Doliolids the peribranchial cavities have been withdrawn altogether from the antero-ventral regions of the pharynx, thus preventing the formation of any perforations except in the postero-dorsal region. This withdrawal, in conjunction with the backward shift of the cloaca, is an essential part of their adaptation for muscular propulsion. The peribranchial and cloacal chambers have been merged into one posterior space, and the partition between pharynx and atriocloacal chamber has been reduced practically to a transverse or oblique septum across a barrel-shaped cavity.

Thus the position of the gill-slits in Doliolum behind the endostyle is not a primitive character, but emphatically a secondary modification, and we have to face the fact that this is the very feature which we assumed to be primitive in Appendicularians. There can, I think, be no other conclusion than that our assumption was fallacious so far as Tunicata are concerned. The shortness of the endostyle and its position in front of the gill-slits seemed to be primitive features indicative of original larval characters: in reality they are secondary features due to the Appendicularian’s retention of the larval form and locomotive tail in combination with an adult Doliolid pharynx, secondarily reduced and simplified.

This interpretation, once reached, is readily confirmed. For, as we have already seen in dealing with the adoral band, the primitive endostyle, as shown by the larva of Amphioxus, was an open loop of the peripharyngeal band, lacking a median tract of flagellated cells, and functioning in all probability simply as a right and left pair of glands to supply the peripharyngeal bands with mucus. It is this organ, and not the definite endostyle, which lies in front of the gill-slits in Amphioxus, and from it the Appendicularian endostyle differs even more deeply than does the normal endostyle of adult Ascidians and Amphioxus. For this organ, as we have seen, is but a further elaboration of the larval peripharyngeal loop; but the Appendicularian endostyle is based not on a simple larval loop, but on the fully integrated adult organ. Its anterior part, which in the larval Amphioxus is an open gap, is here the seat of its greatest complexity. For details see Ihle’s special paper on the subject (1907) and his later summary (1913, pp. 496-500).

In Oikopleura, in which type alone its mode of working is reasonably established, the mucus is undoubtedly directed forwards and upwards, as we have seen, but not by the primitive mechanism of the larval Amphioxus. The front end carries a highly specialized tuft of large baling cilia or setae, which are pivoted in front and directed backwards along the floor of the groove. These are plainly the hypertrophied representatives of the median flagella of an ordinary Ascidian endostyle, the rest of the row having been obliterated pari passu with a shortening of the whole organ. A similar remark, mutatis mutandis, can be made with regard to the lateral rows of gland-cells, which in our common Oikopleura dioica are represented by a single cell on each side, as seen in transverse section, in place of the two groups of gland-cells on each side in Amphioxus, or the three of normal Ascidians. Such a condition as this is obviously less primitive than that of Ascidian larvae, and more specialized than that of Ascidian adults. It cannot possibly represent an ancestral stage in the evolution of the organ. As Ihle has remarked: ‘Von dem zwar einfachen, aber doch stark spezialisierten und zellenarmen Endostyl der Appendicularien können wir also keinenfalls den Endostyl der Ascidien ableiten’ (1913, p. 500). From the functional standpoint accordingly the Appendicularian endostyle discharges a simple and primitive function (i. e. supplying mucus to the peripharyngeal band), but does so by the secondary reduction and adaptation of an organ evolved for use under polytrematic conditions (viz. supplying mucus to the pharyngeal walls generally). The larval loop of Amphioxus has only lateral tracts of cilia by which to propel its mucus, in this respect resembling the Echinoderm adoral loop; the median tuft of cilia of the Appendicularian endostyle represents a later evolutional product altogether, and is additional to the lateral (marginal) bands which still persist. It is neither larval nor primitive, but essentially an adult Ascidian endostyle peculiarly modified and reduced.

The reduction of the endostyle in other species of Oikopleura is not quite so great as in Oikopleura dioica, since in various species examined by Ihle there are two rows of gland-cells on each side separated by a narrow intervening tract of minute ciliated cells. These plainly correspond to the two multicellular tracts of gland-cells on each side of the endostyle of Amphioxus; but the ordinary Ascidian, as we have seen, possesses three pairs of such glandular tracts, and the same is true of Pyrosoma and most species of Salp a. Here again D o 1 i o 1 u m furnishes the connecting link with only two groups of gland-cells on each side (Fol, 1876, Taf. VII, 7; Uljanin, 1884, pp. 17, 18, Taf. VI), so that we are able to trace every step in the retrograde evolution of the Appendicularian endostyle from that of the fixed Ascidians by way of Dist a p 1 i a (see p. 89) and D o 1 i o 1 u m.

Moreover, while I have shown how difficult would be a change from the feeding process of an Appendicularian to that of a fixed Ascidian, in consequence of the specialization of the endostyle in the former case, this difficulty does not arise when the evolution from one type to the other is assumed to have taken place in the opposite direction, and D o 1 i o 1 u m shows how the change was accomplished.

I must first point out that Orton’s admirable study of the feeding mechanism of Amphioxus and the Ascidians (1913, 1914), while making perfectly clear the essential similarity of the process in the two cases, leaves a good deal of uncertainty with regard to the functioning of the various ciliated tracts, especially the action of the marginal bands. It is no easy matter, when a current of water is pouring through a complicated pharynx, to determine the precise influence of different factors in producing the general result. The principal motive-power in these cases is undoubtedly provided by the cilia of the gill-clefts, but the direction taken by the stream within the pharynx is influenced by many other factors, e. g. position of the atrial aperture, ciliation or non-ciliation of the pharyngeal (interserial) bars, &c. Thus in Amphioxus, as shown by Orton himself (1. c., fig. 3), the gill-bars are ciliated on their pharyngeal surfaces, and these cilia lash along the length of the bars from below upwards (in morphological orientation), yet the strings of food particles do not move dorsally in strict parallelism with the direction of the bars, but slide obliquely backwards across successive gill-bars and gill-slits. In this case the main current of water pouring in through the mouth is distributed through the evenly perforated walls and sets from all quarters towards the atriopore on the ventral side. The postero-dorsal direction of the food-strings on each side thus makes an angle with the postero-ventral axis of the water-current, and this resultant deflexion is doubtless due to the action of the frontal cilia described by Orton on the gill-bars, the lashing of which (if truly along the length of the bars) is necessarily an terodorsal owing to the obliquity of the bars, i. e. the movement of particles on the wall of the pharynx is approximately at right angles to the direction of motion of the frontal cilia.

Now both in Amphioxus and the Ascidians Orton describes the lateral (marginal and intermediate) cilia of the endostyle as lashing mucus transversely from the endostyle on to the sides of the pharynx (pp. 24, 25, 31, 35), and the question arises whether this is strictly correct. Mucus certainly passes sideways from the endostyle to the pharyngeal walls, but this result may be due simply to the indiscriminate centrifugal waving of the long median flagella combined with the general direction of the pharyngeal currents. The median flagella can be imagined as lifting shreds of mucus above the level of the endostyle, and the bipartition of the main current right and left would necessarily carry these shreds on to the pharyngeal surface. But in Amphioxus, without a local current antagonistic to the general backward flow, mucus could hardly be transported at right angles to the long axis of the endostyle, as actually happens. Even if the marginal cilia lash transversely, as Orton claims, the direction taken by particles would be obliquely backwards, as we have seen is the case with particles over the frontal cilia, so that (apart from any contributions of mucus from the peripharyngeal bands) the front part, and especially the antero-dorsal region, of the pharynx would thus receive no mucus and food-collection would there be impossible. On mechanical grounds, therefore, it is probable that the lateral cilia (both marginal and intermediate) actually lash from behind forwards, thus ensuring an even distribution of mucus over the whole side-walls.

If this criticism should prove to be correct, morphological inference and actual observation would not conflict, as they do at present if Orton’s account is final, since theoretically the motion of the lateral cilia on the endostyle of Amphioxus should correspond with that along the peripharyngeal bands, of which the lateral tracts, as we have seen, are simply a ventral continuation.

In Ascidians the probability is even greater, for whereas in Amphioxus the frontal cilia of the gill-bars come close up to the marginal bands of the endostyle, and provide a possible basis for Orton’s theory of the direct conveyance of mucus outwards by ciliary action, in Ascidians the gill-bars themselves are not ciliated, and the marginal band is cut off by unciliated pavement epithelium both from the gland-cells of the endostyle and the accessory pharyngeal apparatus which moves the slimeropes dorsally (see § 3, p. 84). Under these circumstances it is difficult to accept Orton’s account of the details of the feeding process. It seems probable that mucus is simply flicked out of the groove by the long median flagella, and carried sideways, not by the marginal bands (which probably work forwards), but by the general set of the water-currents towards the gillslits and dorsal cloaca. The main function of the lateral bands (both intermediate and marginal) is probably to keep up a pressure of mucus towards the anterior end, and thus compensate for the prevalent backward tendency of the currents in that region.

This digression may be summarized in the remark that while the broad features of the feeding process in Amphioxus and the Ascidians have been made clear by Orton’s studies, the precise role of the median flagella and the lateral ciliated tracts of the endostyle remains uncertain. Provisionally it seems probable—in spite of Orton’s observations to the contrary—that the marginal bands are not directly concerned in conveying mucus outwards, but forwards; and that the distribution of mucus on to the side-walls of the pharynx is especially the task of the long median flagella in conjunction with the general set of the water-currents.

We can now return to Doliolum as throwing light on the steps by which the Appendicularian feeding mechanism may have been derived from that of fixed Ascidians. Although the gill-slits in the oozooid are reduced to four pairs of small perforations in a pharyngo-cloacal diaphragm which forms the posterior wall of the pharynx, the endostyle, with its marginal bands, runs along the whole length of the pharynx. If Orton’s theory were correct, we should expect mucus to be thrown over the side-walls of the pharynx as in other cases, but we have seen from Fol’s careful and circumstantial accounts that this is not the case. It is practically all driven forwards on to the peripharyngeal circlet and trailed off from the latter in strings and fringes, not by local ciliary action, but by the force of the inflowing current which sets through the pharyngeal cavity towards the orifices in the posterior diaphragm, where the exhalant cilia are at work. This change is partly due to the concentration of the exhalant apertures posteriorly, and partly to the fadt, which Fol repeatedly emphasizes, that the lips of the endostyle are here maintained in close contact all along their length except in front, at the base of the peripharyngeal bands, where the mucus is set free and driven dorsally on each side around the pharynx. Structurally the endostyle of Doliolum is a long open groove as in Ascidians, but, in adaptation to the new conditions, it transforms itself functionally into a closed tube¹ opening only in front—a condition which, it is interesting to recall, led Huxley, during his observations on the ‘Rattle-snake’, to describe and name the organ as an axial rod. Plainly the endostyle of Doliolum does not function in the original manner displayed in Amphioxus and the fixed Ascidians, and is already partly degenerate by disappearance of one of the typical pairs of glandular tracts. If the body were to be still further diminished in size, and the protostigmata reduced to one pair instead of four, the endostyle would undergo the further abbreviation which is exhibited by Appendicularians. Thus the rotating sieve of mucus is the end-result of structural changes which can be traced seriatim: (1) backward rotation of the cloaca, withdrawal of the peribranchial cavities, and reduction of gill-slits; together with (2) functional closure of the endostyle behind, reduction of glandular tracts, abbreviation of the whole organ to a small anterior gland, and suppression of the median flagella except at the anterior extremity.

During the transitional period when the peribranchial cavities were being withdrawn and reduced (probably, of course, through a series of generations), the endostyle was able, by the simple device of opening or closing its groove, to check or increase the lateral or anterior discharge of mucus, as circumstance required. Physiologically, as well as morphologically, therefore, Doliolum furnishes an evolutional bridge from Ascidians to Appendicularians, so far as the essential features of the pharynx and its working are concerned. But there is no passage in the reverse direction.

The agreement between Appendicularians and Doliolids in regard to their pharyngeal mechanism becomes still closer when we distinguish, as we must, between oozooid and blastozooid, and compare the Appendicularian with the former in the earlier stages of its development before budding has begun. In the oozooid of Doliolum (Korschelt and Heider, fig. 832, A) the gill-slits are only four in number on each side, and they undergo no increase either by subdivision or by the addition of new perforations. According to the latest account (Neumann, 1913, pp. 113, 114) these four slits arise in Doliolum denticulatumin rapid succession as independent perforations, the first dorsally, the last ventrally, an order which is in accordance with our theory of the rotation, the dorsal end of the series corresponding with the anterior end in Pyrosoma. These slits accordingly, like those of Pyrosoma, must be interpreted as protostigmata, each corresponding to a transverse row of stigmata in a fixed Ascidian. In most of Neumann’s figures (Taf. XIII, fig. 4; XIV, 13; XV, 2), which were drawn from preserved material, these protostigmata appear as a transverse row of equidistant, independent, transversely elongated slits, but in figures drawn from living specimens of the two commonest Mediterranean species, Uljanin (1. c., VII, 11, and XII, 8) represents the slits as arranged in two pairs on each side, one dorsal and one ventral. The two slits of each pair lie parallel with one another, but the axis of the dorsal pair is at right angles with that of the ventral pair. The same arrangement recurs in one of Neumann’s own figures (1. c., XV, fig. 1), and also in Grobben’s (cf. Korschelt and Heider, fig. 832), so that its constancy cannot be without significance. Uljanin’s brief account (1. c., p. 58), while agreeing essentially with that of Neumann as regards the independence of the perforations and the order of development, distinguishes only between the pairs of slits, the dorsal pair arising before the ventral. Each pair, therefore, probably represents a single horseshoe-shaped gill-slit, divided into two equal halves, which arise by independent perforation—a mode of development not known in fixed Ascidians, but intimately related to the second type previously described (Text-fig. 2, b).

It follows that the oozooid of Doliolum is a Tunicate provided with only two pairs of primitive gill-slits. Reduce these to one pair, then suppress the tongue-bar formation altogether, and the pharynx of Doliolum would attain the simplicity of an Appendicularian’s (cf. Text-fig. 7).

If this condition be imagined in a young Doliolum, in which the persistent tail was still functional (cf. Korschelt and Heider, fig. 767, or Sedgwick, iii, fig. 48, with Text-fig. 5, c), and if we further imagine the cloacal invagination, with its shallow atrial wings, to be suppressed altogether (the muscular barrel and a functional tail being inconsistencies), a condition NO. 285 I would be produced which would hardly be distinguishable from that of a primitive Appendicularian before the ventral rotation of the tail. As already pointed out (p. 94) the anus would not then occupy a mid-ventral position between and in front of the gill-slits, as in Oikopleura, but a position above the base of the tail, probably on the right side as in the gonozooids of various species of Doliolum, in Fritillaria and Kowalevskia, and the nerve-cord would cross the visceral loop into the tail to the right of the oesophagus (see p. 101) and to the left of the intestine, the tail being in all probability horizontal (Text-fig. 5, c).

Is there any evidence that such a process has actually taken place ? I believe there is, and that we have only to consider the problem of the origin of the Appendicularian ‘Haus’ to arrive at the same conclusion.

The advocates of the primitive nature of Appendicularians have always kept in the background the fact of the extreme and peculiar differentiation of the Appendicularian integument. Whereas in every other Tunicate the entire surface ectoderm participates in the formation of the superficial tunic or test, in Appendicularians the ectoderm is sharply divisible into an antero-dorsal oikoplastic zone and a large posteroventral area incapable of test-production. The boundary between the two is a line drawn obliquely across the body from the front of the genital hump dorsally to the region of the endostyle ventrally (see Text-fig. 7, d). The contrast between the thick secretory epithelium in front of this line and the thin membrane behind it is exceedingly sharp, and the edge of the oikoplastic zone is often raised into a low ridge encircling the body along this line, from beneath which the genito-spiracular area appears to protrude as a kind of hernia (see especially Uebel’s figure of Oikopleura najadis, 1913, p. 626, and various figures of Lohmann’s, 1896, Taf. XIII, XV, XVI, XVIII). Into this posterior area devoid of test substance open the spiracles, anus, and gonads. It is in fact morphologically equivalent to a shallow atrio-cloacal chamber, like that of Doliolum, the cavity of which has been suppressed and the epithelial lining everted. As is well known, the epithelium of the atrial and cloacal involutions of Tunicates generally, though derived from the ectoderm, never participates in test-production. Consequently its suppression, in the way I have suggested, would bring to the surface a large posterior area of unsecretory epithelium, together with the apertures of gill-slits, rectum, and gonads, as openings within it (cf. Text-fig. 9).

The boundary line between oikoplastic and non-oikoplastic areas is very commonly the site of some very peculiar outgrowths. In Oikopleurids the ridge already noticed in various species (p. 114) is raised in certain cases (Oiko pleura longicauda, Oikopleura cornutogastr a) into a thin dorsal fold which extends forwards as a ‘veil’ over the secretory region (see Text-fig. 9). In Fritillaria the homologue of this ‘veil’ is present as the well-known dorsal hood (see Text-fig. 4), and functions as a receptacle for the gelatinous nose-bag, which in this genus replaces the ‘Haus’ of other Appendicularians. Lohmann has shown (1896, Taf. I) that a similar hood also occurs in Kowalevskia, and apparently in Appendicularia (1899, Taf. Ill, fig. 7), but is so delicate and transparent that it escaped the keen eyes of Fol and previous observers. An ingenious suggestion has been put forward by Lohmann and Bückmann in their recent monograph (1926, pp. 124,125, figs. 35, 36) to account for the retention and size of this veil in Oikopleura longicauda (=Oikopleura spissa of Fol, and ve lifer a of Langerhans), but it is enough to notice that the presence of this hood or veil in each of the four chief types of Appendicularians, although with three different functions, is clear proof that it is based on a structure which must have been present in the earliest Appendicularians, although originally, in all probability, serving some altogether different purpose. This hood or veil almost certainly represents the dorsal lip of the original cloacal aperture, which in Anchinia is produced into a long tentacle-like outgrowth. This process in the oozooid probably supports the creeping stolon and buds, and is doubtless homologous with the similar outgrowth in Doliolum, although the latter has been shifted forwards from the cloacal margin. It also strikingly recalls the’ atrial languet’ of many Synascidians.

The mid-ventral margin of the original cloacal aperture has apparently left no remnant, and can scarcely be identified with the ventral prominence behind the oikoplast boundary shown in Text-fig. 9, owing to the position of the latter in front of the tail. This prominence, originally discovered by Fol, is only known in the species figured (Oikopleura longicauda), and is of unknown significance (see below, p. 117). In Fritil1 a r i a the sides of the hood curve downwards and gradually die away ventrally without definitely uniting with one another. The explanation of this feature becomes clear when the form of the original atrio-cloacal chamber is borne in mind, and the process of obliteration followed in the analogous case afforded by the gastrozooids of Doliolum. In these, as is well known, the atrio-cloacal cavity has opened out almost as completely as I claim it to have done in Appendicularians (cf. diagram in Sedgwick, fig. 52, G; Korschelt and Heider, fig. 831, after Grobben), and the ontogenetic process can be followed in Neumann’s admirable account (1913, pp. 201-6, Pl. xxii). The process begins in the cloacal (ventral or anal) region, from which it extends laterally and dorsally into the atrial or branchial region, which is the last to be exposed, as it is the first to be formed. It begins with an elevation of the cloacal floor, and is continued by a gradual widening of the cloacal aperture until the opening is almost coextensive with the area originally enclosed. The median ventral region is completely flattened out, bringing the anus to the surface, and leaving beneath it no trace of a cloacal rim at a stage when there is still a distinct rim around the dorsal and lateral margins (Neumann, Taf. XXII, fig. 7). The base of the Fritillarian hood, and its incompleteness ventrally, thus corresponds closely with the last boundaries of a vestigial cloacal cavity. In the lateral buds of Doliolum there is no tail to complicate the process, and the enlarged aperture retains a simple vertical oval outline; but in a tailed Doliolum a similar process would expose a saddle-shaped area, consisting of a posterior cloacal region above the tail, and a pair of lateral wings extending forwards from it (cf. my diagram, Text-fig. 5 c, which does not, however, display the full dorso-ventral dimensions attainable by the atrio-cloacal cavity). Theoretically a mid-ventral strip of test-surface should persist between the spiracles of an Appendicularian to connect the general oikoplastic area in front of the hood with the base of the tail, but the maintenance of an isolated strip of test in such a position would be useless in itself, and would obviously destroy the possibility of an architectural continuity between test and ‘Haus’, which will next be dealt with. It is in fact along this median tract that the anus must have migrated forwards from the right side of the tail to attain the position which it occupies in O i k o p 1 e u r a. It is just possible that the peculiar lyrate ventral prominence in Oikopleura longicauda already referred to (Text-fig. 9) may be an incidental result of this migration, the anus, metaphorically speaking, having pushed before it a remnant of the ventro-lateral cloacal rim (Fol, Pl. ii, fig. 8).

Having now interpreted the differentiation of the Appendicularian epidermis into oikoplastic and non-oikoplastic areas as the consequence of the opening out of a former atrio-cloacal cavity, I can pass to the subject of the ‘Haus’ itself, the elucidation of which as an elaborate mechanism for the filtration and capture of ‘Nanno-plankton’ must always be associated with Lohmann’s brilliant and careful researches. With the details and development of this mechanism, however, I only propose to deal in so far as they throw light on the phylogenetic problem. An excellent and well-illustrated summary of Lohmann’s work is given by Ihle (1. c., 1913).

Lohmann has shown that, in spite of the complexity of its final structure, the ¹ Haus’ begins as a continuous sheet of cuticular deposit, differing, for example, from that of an Arthropod merely in carrying the differentiation of parts to a further degree. The ventro-lateral wall is almost uniformly gelatinous, but the dorso-lateral wall at the outset is distinguishable into zones of gelatinous material alternating with others which are thin and membranous or deep and laminated. The conversion of the cuticle into an accessory organ by means of this differentiation and by subsequent foldings and readjustment of parts to one another does not qualify the fact that at first it forms a continuous mosaic of cuticular products resting on a continuous matrix of diversified oikoplastic epithelium. ‘Hiernach kann kein Zweifel sein, dass die gesammte Gehäusesubstanz als cuticulare Ausscheidung anzusehen ist’ (1899, p. 382; cf. my Text-fig. 9).

The peculiarities of the Appendicularian cuticle begin with its exuviation, for it is then seen that instead of being simply peeled off and cast aside, it is being converted to something else with a new and different function. The primitive cuticle was a simple exoskeleton, moulded on the body and subservient to it. In the Appendicularian the roles are reversed: the differentiations of the ectoderm are subservient to the formation of the cuticular house, and ecdysis is accompanied by so many modifications that it is merely preliminary to a process of functional deployment. ‘Die Trennung der Anlage von den Oikoplasten ist nichts anderes als eine Hautung. …. Die Entfaltung aber hat, so weit mir bekannt, sonst im ganzen Tierreich kein Analogon’ (1. c., p. 383).

With the essential nature of the Appendicularian house thus defined as a single, but differentiated, cuticular envelope, the next problem is clearly to determine, if possible, the original form of that envelope by a comparison of the various types. Here we are at once confronted with the fact that the structure, development and mode of functioning of the house is adequately known only in the case of Oikopleura. In his original account (1899) Lohmann cautiously limited himself to the statement that the vesicle of Fritillaria ‘stands lowest’ in the series, that the ‘simple’ houses of Ko walevskia and Appendicularia come next, and that the elaborate houses of Oikopleura furnish the climax; but until the first three of these types are known as accurately as the last, it must remain quite problematic whether their simplicity is real or superficial, primary or secondary. The matter can only be provisionally settled by appeal to isolated features and indirect evidence.

In a later account (1909) based on essentially the same facts of structure and development as his earlier account, Lohmann’s standpoint shows an appreciable change. The emphasis shifts from the cuticular nature of a ‘Haus’ subservient to protection and locomotion as well as nutrition, and is concentrated on its significance as a ‘Fangapparat’: ‘So tritt vor allem hervor, wie die Bildung eines Fangapparates den Ausgangspunkt und sogleich den Kernpunkt derselben bildet’ (1909, p. 217). The Stufenfolge of his previous account is now definitely put forward as a phyletic succession based on this idea. The gelatinous vesicle suspended in front of the mouth of Fritillaria is regarded as equivalent not to the whole ‘Haus’of Oikopleura, but to its ‘Fangapparat’ alone, the ‘Haus’ having arisen later ‘durch weitere Komplikationen’ of its enveloping jelly. The houses of Kowalevskia and Appendicularia are interpreted in the same way, but remain at a lower stage of evolution than the house of Oikopleura. In them the ‘Fangapparat’ is regarded as having undergone a ‘glockenförmig’ expansion, in the centre of which the Appendicularian is suspended like the clapper of a bell, not behind it as in Friti 11 aria and Oikopleura.

Finally, in their recent report (1926), Lohmann and Bückmann have completed this change of front by attaching a surprising significance to certain peculiarities in Tectillaria fertilis (the Fritillaria fertilis of Lohmann’s earlier description, 1896). This is a form in which the dorsal hood is produced in front of the mouth by a forward shift of its base, so that the oikoplastic epithelium, which in other forms is restricted to the floor of the hood-cavity, is carried upwards on to its roof (1. c., fig. 46). They suggest, with a somewhat formal reservation, that this may be the primitive condition of the oikoplastic epithelium, i.e. the lining of a more or less tubular multicellular gland, originally limited to the head region. On such a theory the evolution of the oikoplast area has consisted in the gradual opening out of the gland, and the extension of its cells over an increasing area of the dorsal and lateral surface of the body, culminating in Oikopleura in the maximum extension of the secretory epithelium and the disappearance of the hood altogether. It gives circumstantial detail to Lohmann’s later view, -which has been endorsed by Ihle (1. c., p. 472), that the ‘Haus’ began as a mere food-trap over the mouth, its enlargement into a complete envelope being a secondary and later development.

Quite apart from the difficulty of comprehending how the filtering capsule could begin in the first instance as the secretion of a gland, or be transformed subsequently into an enveloping house, this theory seems to me to invert the natural order of events altogether. It marks a complete withdrawal from the point of view hitherto consistently advocated by Lohmann himself, according to which the formation of the house was essentially a ‘Häutungsprozess’ or ecdysis—a view which I venture to submit is still the correct one.

On the latter point, in view of my preliminary remarks, I need only refer again to Lohmann’s own arguments and discoveries (1889, p. 207; 1899, p. 382; 1905, p. 357; 1926, p. 119), which seem to me unanswerable; but against the case presented for the primitive nature of the vesicle of Fritillaria additional objections can be urged on many points.

(1) In histological characters Fritillaria is even more profoundly reduced and specialized than Oikopleura.

(2) In general structure Fritillaria has admittedly been modified secondarily (‘in der höchsten Ausbildung’) for a floating life by gelatinization of the blastocele, &c. (Lohmann and Bückmann, 1926, pp. 101, 105), and the characters of the jelly-vesicle have been claimed by Lohmann as ‘eine überraschende Anpassung’ to the same end (1902, p. 27).

(3) As a cuticular deposit is primarily a clothing of the body (cf. Lohmann, 1909, p. 219: ‘Die typische Bedeutung der Cuticula ist ja zweifellos die einer Schutzhülle’), and as the vesicle of Fritillaria is formed from an oikoplastic area homologous with that of other Appendicularians, and in a similar way, its projection in front of the body is clearly a secondary adaptation and not a primitive disposition.

(4) The same conclusion obviously applies to the peculiar elastic mechanism which causes the vesicle to collapse and spring back under its protective hood when not in use.

(5) The general architecture of the vesicle of Fritillaria is more, not less, elaborate than that of the houses of Kowa1 e v s k i a and Appendicularia. The latter (Appendicularia?) are provided with a single large median aperture which serves both for the entrance and exit of the water circulated; but the vesicle of Fritillaria is admitted to contain two opposite apertures for entrance and exit respectively, as in Oikopleura longicauda, as well as an oral tube (Lohmann, 1909, p. 217).

(6) If the ‘Fangapparat’ of Fritillaria arose first, and the ‘Haus’ of Oikopleura is a secondary extension of it, the relations of the mouth to the ‘Fangapparat’ should be primary in all cases. Lohmann has shown, however, that the primary oral tube in Oikopleura gives rise to the circular valve round the exhalant (oral) aperture of the ‘Haus’, and that the connexion of the mouth with the ‘Fangapparat’ is secondary, although he was unable to determine how it was brought about (1899, pp. 384, 391, 392; Taf. Ill, fig. 10, a, b, c).

On all these grounds I find it impossible to accept Lohmann’s theory of the evolution of the Appendicularian house. It leads only to a tangle of inconsistencies, and is based on the same confusion between simplicity and primitiveness which has characterized the prevalent interpretation of Appendicularians themselves. Lohmann’s achievement is to have shown that every kind of Appendicularian house, whether it encloses the body or not, contains an apparatus which subserves the filtration and collection of extremely fine food particles. He has also shown that the mode of formation of this apparatus in Oikopleura ‘ist ein uber Erwarten komplizierter, und lasst auf eine sehr lange und wechselvolle phylogenetische Geschichte schliessen’ (1899, p. 377). To trace all the steps of that ‘long and changeful history’ needs more knowledge than at present we possess, but Lohmann’s own contributions have pointed the way.

Assuming that the ‘Haus’ with its ‘Fangapparat’ has been derived by modifications of a pre-existent cuticle, a ‘Haus’ which encloses the body is more likely to retain primitive arrangements than one which does not, and the vesicle of Fritillaria must be regarded as the end-result of a process of reduction and specialization. The primary functions of protection and locomotion subserved by the house have been gradually discarded, and the secondary feeding functions retained and specialized, in obvious correlation with the adaptation of Fritillaria itself to a passively drifting life.

Strictly speaking, of course, the evolutional process has not been one of cuticular modification, but of the differentiation of the ectodermal matrix. Starting from a homogeneous ectoderm, like that of Ascidians, the primitive condition must have been that of a homogeneous semi-gelatinous investment. The differentiation of a food-trap in the walls of this investment after partial exuviation is the expression of a modification in the structure and activity of certain groups of cells in the anterodorsal region. But there are details in the complicated development of this apparatus as described by Lohmann which imply that a gelatinous tunic was not only a primitive feature in Appendicularian evolution, but was indispensable as a predecessor of the membranous parts of the food-trap. The parts of this apparatus which ultimately became hollow are at first filled with jelly, so that a gelatinous cuticle is really the basis on which the membranous parts are moulded. This applies to both parts of the ‘Fangapparat’, viz. the lateral entrance channels,¹ and the paired series of filter-tubes. The former are laid down as a successive pair of flat striated membranes over the anterior crescent of Fol’s group of oikoplasts: ‘Die Häute sind durch einen Zwischenraum von einander getrennt, der anfangs von gallertiger Masse erfüllt ist, aber späterleer erscheint’ (1899, p. 374; Taf. II, figs. 6, 7; Taf. Ill, fig. 3). The latter are formed as a connected series of fibrils, parallel to the surface, by exudation from the three rows of small prismatic cells behind Fol’s giant-cells (1. c., Taf. II, figs. 1, 5, Rsb); but the exudations continue, so that new series of connected fibrils are pushed up from the matrix beneath the first series and its successors, until they take the form of a tall row of filter-tubes rising vertically or obliquely above the matrix, and corresponding in number with the transverse rows of formative cells (ca. 27) on each side. Lohmann does not say that these tubes individually contain at first a gelatinous core, but the whole series on each side is supported, as it exudes, by a gelatinous mass on its anterior face, secreted by the well-known transverse row of giant-cells: ‘An seiner Oberfläche wird eine gänzlich structur-und formlose Masse ausgeschieden, die das Lumen der Flügel [i. e. the ‘wings’ of filter-tubes] erfüllt und durch ihre allmählige Zunahme die einzelnen Teile des Fang apparates in ihre definitive Lage zu einander bringt. Sie verschwindet spater vollkommen’(l. c., pp. 874, 375, Taf. II, figs. 6, 7, 9). Thus the gelatinous exudations of cells in both parts of Fol’s area serve as a scaffolding for the deployment of the membranous parts of the ‘Reusenapparat exactly as the notochord serves as a preparatory axis on which the cartilage of the vertebral column is moulded in the development of a Vertebrate. These relations seem to furnish conclusive proof that in the evolution of the Appendicularian house a gelatinous tunic came first and the membranes and fibres of the filtering apparatus are the product of a later differentiation of the epithelium.

We come back, then, to the problem of the original form of the Appendicularian cuticle, when it was still homogeneous, but with the problem simplified by the elimination of Fritillaria or its ‘Fangapparat’ from any claim to represent a primitive condition. The radial symmetry of the large house of Kowalevskis is equally anomalous, and is associated with a unique restriction of the oikoplastic epithelium (Lohmann, 1896, Taf. I, figs. 1-3), probably due to the approximation and fusion mid-dorsally of lateral groups of cells corresponding to Fol’s oikoplasts. It should probably be regarded, with Fritillaria, as a secondarily simplified form—a conclusion in harmony with the minute size of the animal and with its well-known pharyngeal modifications (loss of endostyle, &c.).

The most primitive Oikopleurid house is admittedly that of Oikopleura longicauda (=spissa of Fol). This house has no ‘filtering windows’, and retains two opposite apertures in front and behind for exit and entrance of water respectively. The latter corresponds with the posterior margin of the oikoplastic area; the former, as already remarked, is connected at first with the mouth of the Appendicularian by the primary oral tube. There can be no doubt that Lohmann is right in regarding the house of other species of Oikopleura as derivable from this type (see Lohmann and Buckmann, 1926, p. 125, fig. 36), since in them the primary inhalant aperture behind is present in a modified condition, closed by membrane, and serving the contained Appendicularian simply as ‘emergency exit’ (Text-fig. 8, EE).

The house of Appendicularia is the only remaining type, but is so imperfectly known that its exact position in the scale can hardly be determined. Lohmann has shown that it contains a paired food-collecting apparatus, remarkably similar at the outset to that of Oikopleura in general appearance (1896, Taf. I, fig. 6; 1899, Taf. Ill, fig. 7), and the single observed aperture of the house almost certainly corresponds with the posterior (inhalant) aperture of Oikopleura longicauda. The apparent absence of a primary oral aperture (exhalant) seems to associate it with Kowalevskia, but it is by no means so certain as in that case that the aperture is really absent. The circumoral tract of the oikoplastic area in Oikopleura, from which the primary oral tube is formed, is greatly reduced in Kowalevskia (Fol, 1. c., Pl. x, fig. 5), and Lohmann figures the house arising as a purely dorsal structure (1896, Taf. I, figs. 1-3), which is presumably slipped over the mouth subsequently to its formation. The absence of an oral (anterior) aperture to the house can thus be accounted for, while the large size of the structure when expanded (35 × 20 mm.), and Fol’s experimental study of it, leaves no room for doubt as to the facts. But the house of Appendicularia is the smallest known (2·6 mm.) and direct observation on this point must be wellnigh impossible. The corresponding aperture in Fritillaria would probably still be unknown if Fol had not had the large capsules of living Fritillaria megachile (10×8 mm.) for observation. Appendicularia is in many respects a connecting link between Fritillaria and Oikopleura, and both these genera possess the anterior aperture; the circumoral zone is well developed; and there can be no doubt that the house-rudiment, which Lohmann figured in 1899, approximates much more nearly to the Oikopleurid than to the Kowalevskia type. Thus, if the aperture exists, the architecture of the house must be essentially similar to that of Oikopleura longicauda, even if the filtering mechanism is more simple; and if it does not exist, its absence is almost certainly secondary, and marks a beginning of the series of retrograde modifications which culminate in Kowalevskia.

From this examination it results that in all probability the primitive type of Appendicularian ‘Haus’, from which all others have been derived either by elaboration or reduction, consists of a hollow, oval or spherical capsule with two apertures at opposite extremities, one oral, and the other formed round the boundary zone of the oikoplastic epithelium, which in primitive cases is limited by the base of a dorsal hood or veil. Assuming this to mark the limits of a former atrio-cloacal cavity, the primitive house must have been a cast-off test, with opposite oral and cloacal apertures. Such a test could only be formed by a pelagic Tunicate of the Doliolid type; and since Doliolum, at least Doliolum mülleri, is known to possess the power of total ecdysis (Uljanin, 1. c., p. 14), there is no need to say more than that the case is thus complete. There can be no question of the fact or of the correspondence in essential details. Uljanin observed the process under the microscope under conditions which enabled him to describe it as ‘eine vollkommen normale Erscheinung and to assert ‘dass unter der abgeworfenen beschmutzten Cuticula eine sehr dünne neue immer schon ausgeschieden wird’. He points out, moreover, that in Doliolum the test throughout life remains hyaline and structureless, like that of the youngest larval stages of ordinary Ascidians, and contains no cellulose; and that, as in Appendicularians alone among Tunicata, there is no involution of test-substance within the oral aperture (or within the cloacal aperture). These are precisely the conditions required for the beginnings of the evolution of the Appendicularian house.

The modifications introduced into the house for purposes of food-collection form a subsequent story which does not now concern us. I will therefore merely point out that if my criticism of previous views concerning the relations of ¹ Haus’ and ‘Fangapparat’ is valid, it is in the development of the houses of Oikopleura longicauda and Appendicular i a that we may best expect some primitive features by which to interpret the complexities of formation revealed by Lohmann in Oikopleura albicans.

Thus, after a fairly exhaustive survey of Appendicularian peculiarities, there can be no doubt that their simplicity is more apparent than real. When the structure and working of their pharynx and endostyle are examined closely and comparatively, these organs alone yield conclusive proof of a derivation from the more normal Tunicate type. Moreover, in every respect—position and number of gill-slits, structure and functioning of endostyle, form of the intestinal loop, position of the anus, and torsion of nerve-cord—Doliolum has been shown to provide the key to their phylogeny. Finally, the architecture of the house, and its relations to the oikoplastic epithelium, have confirmed our view that the absence of a cloaca and of peribranchial cavities is a secondary phenomenon, and that the ancestor of the Appendicularians was essentially a Doliolum with a barrel-shaped test subject to periodic exuviation. In the dorsal veil or hood of Oikopleura longicauda, Appendicularia, Fritillaria, and Kowalevskia we even seem to have a hypertrophied remnant of the ancestral cloacal aperture with the dorsal outgrowth (=atrial languet ?) that in Doliolids carries the creeping stolon (A n c h i n i a) or its equivalent the succession of migrant buds (Doliolum itself).

The remarkable thing about these features of resemblance to Doliolum is that they are all essentially adult, as distinct from larval, characteristics, and may be thought to give a final refutation of the neotenic theory which I began by defending, since only the tail is left as a relic of larval conditions. As was remarked, however, at an earlier stage (p. 95) Ascidian larvae do not feed, and have no visceral organization of their own distinct from that of the adult phase which, in varying degrees, they seek precociously to develop. The same conditions must have prevailed among primitive Doliolids, before the larva was finally boxed up in an inflated egg-capsule serving as a float. Even if, as I believe, the Doliolids are derivable from some more central stock than the Clavelinid Synascidians, the abandonment of fixation, by obliterating the original boundary between larva and adult, must have facilitated just such a combination of larval and adult characters as we find permanent in Appendicularians. The young nurse-form (oozooid) of Doliolum (Korschelt and Heider, fig. 67; Sedgwick, iii, fig. 48, both after Uljanin) offers a remarkable blend of these characters, and, with certain qualifications, can well be regarded as an Appendicularian in the making. Whether it be called adult precocity, or larval persistence, is immaterial. It is essentially the breakdown of a former separation between the two stages of a metamorphic life-history, and only requires one step further in the process of simplification, viz. the abandonment of budding and consequent reacquisition of sexual powers by the oozooid, to produce a primitive Appendicularian, with larval tail and simplified adult body.

The dorsal hood, in association with the opening of an atriocloacal cavity, points probably to the existence in the ancestors of Appendicularians of a dorsal outgrowth, similar to that which in Doliolids is associated with the development of the buds, but it is not in itself conclusive of a former reproduction by budding. In the ‘pharyngeal packets’ of Fritillaria, however, and the ‘oral glands’of Oiko pleura we have peculiar organs which seem to be the modified vestiges of the Ascidian epicardium, and thus to complete the story.

Before submitting the evidence under that head, however, which will carry us away from Appendicularians altogether, I must briefly refer to two points in Appendicularian structure which prima facie may seem to conflict with the conclusions already arrived at. These are the differences between Appendicularians and Ascidian tadpoles in regard to the structure of the tail and the arrangement of its muscle-cells.

The tail of Appendicularians is devoid of test-substance, its fin is produced by keel-like extensions of the true cellular wall, and its lateral muscles are provided by a single row of cells on each side of the notochord which, from their large surface and transverse lines of suture, have given rise to the much-discussed appearance of segmentation.

The tail of Ascidian larvae, though similar superficially, is composed quite differently, and consists of two elements, (1) a central axis, circular in section, continuous with the body of the larva, and (2) a tail-fin composed entirely of test-material and continuous with the test covering the body (see my figures 4, T, and 5, CF; cf. Seeliger, 1900, and in Bronn, figs. 173-4, Taf. IV, V; and Martini, 1909, p. 300, figs. 1-7). Although the tail of Doliolum is not functional, and appears to be destitute of test-material, it corresponds with the tail of Ascidian larvae in other respects, being circular in section and provided with several longitudinal rows of muscle-cells on each side, which have the same alternating arrangement as in Ascidian larvae (cf. Uljanin, 1884, Taf. VII, fig. 3; and Neumann, 1913, Taf. XI, XVII; with Seeliger in Bronn, fig. 174).

These differences may again raise the idea of the retention by Appendicularians of a more primitive condition than Doliolum itself affords. But when the special points of resemblance between the Appendicularian tail and that of larval Ascidians are borne in mind, these two points of difference are seen to be of minor significance, and are readily explained as part of the many adaptations which have fitted the Appendicularian tail—whatever its origin—to the specialized functions it discharges in relation to the ‘Haus’. As Lohmann has well said (1896, p. 6), ‘Beide Merkmale [i.e. tail and house] hängen auf das Engste zusammen und haben sich offenbar gleichzeitig ausgebildet’. These acquired functions consist not only of the creation of currents through the ‘Haus’ (or into the vesicle, as the case may be) by a mode of undulation recognizably different from that used during free locomotion, but also of manipulative powers, by which the tail assists in extending the house-rudiment over the body and in enlarging and shaping its internal spaces. Clearly a tail which has acquired a mobility and versatility of this order cannot be expected to have retained the primitive features of its origin. Its free articulation with the body, its width, length, and flexibility, its smooth surface, sensitive tip, and specialized neuromuscular apparatus, must all be the result, more or less, of special modifications, whether the ancestor was Amphioxine or Ascidian.

Assuming, as we must, that the Doliolid ancestor possessed a locomotive larva with a tail-fin of test-material like that of Ascidians, it is obvious that a fin of this structure, sufficient for the brief career of a tadpole, would be quite inadequate for a creature that depends lifelong on the efficiency of its tail, especially as the cuticular fin would presumably be shed at every ecdysis and leave its possessor helpless. To substitute for this a dorsal and ventral keel-like extension of the tail itself would be no difficult operation, especially as the opening up of the atrio-cloacal cavity would bring the tail into immediate contact with a tunic-free area, so that a slight change in the cell-mosaic of the gastrula would enable the tail to be clothed by the same tunic-free epithelium as the cloacal area itself.

As regards the arrangement of muscle-cells I can readily agree with Martini (1909, p. 306) that the Appendicularian type is theoretically capable of derivation either from the fully segmented myotomes of Amphioxus or from the unsegmented arrangement of the Tunicate larva, but the question is, which is the simplest hypothesis consistent with the facts ? In effect Martini rejects the second alternative by his opposition to the theory of neoteny, but he seems not to have realized that in expounding the first he was merely substituting one larva for another, and that neoteny (in its wide sense) was involved in either case. His proposition is that by ‘Eutelie’ the myotomes of Amphioxus may have been reduced in Appendicularians to the form of single cells, by successive stages of reduction. He begins with an Amphioxus embryo, in which the somites are still undivided, and in which the muscle-plate consists only of six cells (fig. 5). By assuming that the loss of a coelom would leave these little groups of cells in situ, and their metameric arrangement intact, it is of course ‘sehr leicht’ for ‘Eutelie’ to convert ‘den Querschnitt der AmphioxusLarve in den der Appendicularie’ (fig. 6). But what of the metamerism ? It is impossible to admit that the metamerism of the muscle-plates of Amphioxus would survive the loss of somite-formation. As soon as coelom-formation ceased to be a feature of the ancestral development, the Tunicate condition with wandering mesenchyme would be at once established, and an unsegmental arrangement of muscle-cells, as in Ascidian larvae, would follow. On his own theory, therefore, a condition essentially similar to that of Ascidian larvae would necessarily intervene between the ancestral Amphioxus larva and the Appendicularian. Moreover, so far as ‘Eutelie’ is concerned, it is surely a simpler hypothesis to take the arrangement of muscle-cells actually found in Tunicate larvae as the startingpoint, and reduce the number of parallel longitudinal rows from three to one on each side, in order to produce the Appendicularian arrangement.

The histological difference between the muscle-cells of Appendicularians and those of Ascidian larvae is relied upon by Martini in support of his Amphioxus theory, but a last stand can hardly be made upon this. If Appendicularians have been derived from Amphioxus, then Ascidians have been derived from Appendicularians, and the muscle-cells which in the latter have a one-sided arrangement of their fibres must have been converted into those of Ascidian larvae with fibres on all sides. This again is surely the wrong way round. The arrangement of fibres in the Ascidian muscle-cells is simply the generalized, primitive, arrangement characteristic of mesenchymatous muscles everywhere; while that in Appendicularians and Amphioxus is the specialized arrangement associated with expansion over flat surfaces, epithelia, &c. The former is primitive, the latter secondary. In fact the muscles of an Appendicularian, so far from pointing to an origin from fully segmented ancestors, are just the primitive muscles of an Ascidian larva reduced in number and specialized for their peculiar and more extended tasks.

Under the term ‘Eutelie’, which has been used above, Martini (1909, 1910) attempted to distinguish a special type of phylogenetic simplification, of which the Appendicularians, in his opinion, furnish one of the chief examples. The other principal illustration is supplied by the Nematoda. Both these groups display phenomena of ‘cell-constancy i. e. the presence in definite number and position of individual histological units, which may correspond to entire organs in other groups, and are equally characteristic of all the individuals of a species. Associating this phenomenon with the direct or determinate mode of development followed by the same groups, and assuming that it is the end-result of a process of phylogenetic simplification, Martini defines ‘Eutelie’ as the accelerated attainment of a biological goal by the utmost simplification of the organization and the most precise developmental method. I use the latter phrase to express the author’s ‘praeziseste Arbeit’, ‘Praezisionsarbeit’, and ‘präzisen Entwicklungsmechanismus’, but exactly what Martini means by these peculiar expressions it is not easy to understand. As an equivalent of the first he gives in brackets ‘determinierte Entwicklung’ (p. 298), and qualifies this term as being used ‘im deskriptiven, nicht experimentellen Sinne’ (p. 292)—by which one is led to understand him to mean simply the elimination of ‘palingenetischen Umwegen’, i. e. direct non-recapitulative development (cf. p. 295). As direct development in this sense is equally characteristic of cases hitherto classed under ‘Neoteny whether total (e.g. Axolotl) or partial! (e. g. Perennibranchiate Amphibia), clearly ‘Eutelie’ would seem to be distinguishable from the latter merely by the greater extent of the retrogressive process (viz. cellreduction).

To meet this difficulty Martini adopts two lines of argument: (1) Neoteny, being essentially an arrest of development, either as a whole, or with reference to particular organs, cannot be applied to cases of phylogenetic simplification which retain more primitive features than do the larval forms of their group. As an example of this he gives only the case of the Appendicularian muscles, the weakness of which I have already exposed. Moreover, if the example were valid, it would only tend to show that Appendicularians were neotenic Amphioxi instead of neotenic Ascidians. (2) For a similar reason, the term ‘Neoteny’ should be restricted to cases of ‘total Neoteny since in such cases there is no doubt that an actual larval form has been arrested in development by precocious maturity. There is something to be said for this proposal of Plate’s; but the absence of a name for the large class of ‘partial Neotenies’ would not remove their existence, or render ‘Eutelie’ in any sense a suitable substitute.

Martini remarks that ‘die partielle Neotenie nur ein Modus eines phylogenetischen Rückbildungsprocesses ist’ (p. 296). It must be equally apparent that ‘Eutelie’ is only a mode of partial Neoteny. Its distinctive features are neither phylogenetic retrogression, nor arrested metamorphosis, nor direct development, all of which it shares with partial Neoteny; but simply the extension of the process of simplification into the cellconstitution of the various organs. As the term ‘Eutelie’, with its emphasis on ‘direct development’, ‘going straight to an end’, is not distinctive, and is applicable to a far wider range of cases than its author intended, it should clearly disappear, and the phenomenon of cell-reduction, which is of great interest and importance, should be distinguished under some such descriptive title as ‘Katacytosis’, or ‘Oligocytosis’.

With Plate and Martini, on the other hand, one can appreciate the advantage of a distinctive term for cases of total Neoteny, which present a special problem, and seem in general to arise abruptly under particular conditions in the course of individual life-histories. Reserving ‘Neoteny’ for these cases, Giard’s ‘Progenesis’ would cover its extension into still earlier phases of the ontogeny.

When individual larval features alone are retained, however, in combination with a simplification of adult characters, and a réadaptation of the whole, we have a different problem, which, as Martini has well emphasized, demands a long phyletic process of gradual change. The process of simplification and adaptation may proceed along different lines in different cases, but all are intelligibly covered by the term ‘Paedomorphosis’ (Garstang, 1922), while a strict use of ‘partial Neoteny’ would leave many significant cases outside the pale (e. g. Holo pus).

Pharyngeal Packet of Eritillaria.—This structure is a solid group of protoplasmic cells wedged in the floor of the pharynx immediately behind the endostyle, and projecting backwards between the spiracles so as often to fill the space between the pharyngeal wall and the ectoderm in this region: ‘Unter Pharyngealpaketen sind eng aneinanderliegende, plasmareiche, grosse Zellen zu verstehen, die sich vom Hinterende des Endostyls nach hinten erstrecken und sich auch der Innenwand der Kiemengänge anlegen können. Die Verbindung mit dem Endostylist sehr bezeichnend. Die Bedeutung der Pakete ist nicht bekannt’ (Lohmann u. Bückmann, 1926, p. 164, figs. 48, 49. Also cf. Lohmann, 1896, Taf. II, fig. 8; IV, fig. 1; VII, fig. 2).

This mysterious ‘packet’ finds a ready explanation on the Appendicularian theory here maintained, since from this standpoint it can be nothing but the modified vestige of the Ascidian epicardium. It is clearly no functional organ necessary to the life of Fritillaria, for it varies from species to species in the most extraordinary and erratic fashion, and finally becomes resolved, especially in the haplostoma series, into a loose arrangement of cells varying in number from five or six in Fritillaria drygalski (Lohmann und Bückmann, fig. 53) to three in Fritillaria borealis (Lohmann, 1896, Taf. VIII, fig. 2), or divided into a pair of lateral groups adjoining the inner margin of the spiracles, and undergoing the same diminution in number from groups of two to three cells in Fritillaria helenae (Lohmann und Bückmann, fig. 52) down to a single cell on each side (ibid., fig. 47) or vanishing altogether (Fritillaria haplostoma). A striking feature of the packet, in addition to its relations to the endostyle, is its tendency to an asymmetrical position or development on the left side (Fritillaria venusta, pellucid a, helena e—reversed only in Fritillaria scillae).

A good idea of the size of the ‘packet’in Fritillaria pellucida is conveyed by a section given by Salensky (‘Études 1904, iii, Pl. xiii, fig. 22), and a full account of it is given by Martini in his cytological study of the same species (1909, pp. 126-9, Taf. II, fig. 35). Ihle and Martini, who confirm its glandular properties, follow Salensky and term it ‘branchial gland’ (‘Kiemendrüse’). In young individuals of this species it consists, according to Martini, of four cells, two large and two small, but the latter tend to atrophy and disappear in later life. Adjacent parts of each cell together constitute a patch which forms the floor of the pharynx between the endostyle and the left spiracle, but the peripheral regions of the gland bulge outwards beneath the basal membrane of the pharynx: ‘So resultiert in der Pharynxwand eine zusammenhängende Stelle, die wir als die weite Mündung der Drüse auffassen können’ (Martini, p. 129). The function of the gland is quite problematic, although the fact that it secretes fluid droplets into the pharynx is beyond question. Salensky suggested that the secretion might have a food-entangling function, to supplement deficiencies of the small endostyle, a theory to which there are fatal objections, based on the position of the gland, apart from the reason given by Martini for rejecting it.

In considering the homologies of this organ, the special peculiarities of Appendicularian cytology have to be borne in mind, for reduction in size has been accompanied in this group by an extraordinary modification of normal Metazoan structure. Neither cells nor nuclei seem to have the same elementary or unitary values as in ordinary animals: each cell is almost equivalent to a syncytium, but with a single branching and lobulate nucleus, instead of a whole series of normal nuclei inside it. The pharynx, which in a normal Ascidian or its larva is constituted by a regular multicellular epithelium, is here nothing but a basement membrane deposited by a thin sheet of cytoplasm under the control of seven nuclei located at strategic corners. Suppose the ancestor of Appendicularians to have had the constitution of a Synascidian larva, with a pair of epicardial tubes, separate or united distally, what form would these tubes take under the conditions of Appendicularian cytology, especially if we can assume that they were no longer wanted either for budding or for the septation of blood channels ? Sooner or later of course they would disappear, as in most Appendicularians apparently is the case, but in the few Fritillarias in which these ‘packets’ persist, their form, as suppressed diverticula from one or both sides of the pharynx immediately behind the endostyle, and their variability, as vestigial organs, fulfil all the conditions required. The glandular transformations which this modified epicardium has undergone (if the homology be accepted) is probably not a specific functional adaptation, but something more akin to ‘fatty degeneration’, by which the individual attempts, metaphorically speaking, to get rid of a useless heirloom. A closer parallel is probably to be seen in the conversion of tongue-bars, gillpouches, and other pharyngeal vestiges into endocrine glands in Vertebrata. In any event a glandular modification of the epicardium in Ascidians is not without precedent, since Ritter has described it (1903) in the remarkable Polyclinid Euherdmania, and under conditions which are at least consistent with a theory of ‘glandular degeneration’. Euherdmania is unique in possessing two long, completely separate epicardial tubes, of which the left is feebly developed, though thickwalled and glandular. As the pharyngeal apertures of both tubes are closed, the glandular secretion cannot escape. Nothing is yet known of the details of the budding process, but the glandular tube is almost certainly sterile. From such a modification it is but a further step to the condition realized in Molgulids according to Damas (1902). Here the epicardium has been saved from extinction by taking over a renal function, comparable to that of the renal vesicles in Ascidians. The peculiar relations of the heart to this epicardial kidney, together with special ontogenetic details, render Damas’s interpretation irresistible.

The oral glands of Oikopleura are a pair of conspicuous round bodies lying beneath the integument on either side of the endostyle in many species of this genus. They were shown by Fol (1. c., p. 25) to open to the exterior and to secrete a fluorescent (or phosphorescent) material which decorates the surface of the ‘Haus’ (cf. Lohmann, 1899, pp. 208, 213; Lohmann u. Biickmann, 1926, pp. Ill, 115). They seem to be unicellular, with a lobulate or branched nucleus (Lohmann, 1896, p. 18), and I can confirm this, so far as Oikopleura dioica is concerned. Salensky has claimed them to be multinucleate, though cell-outlines are not recognizable in them (1903, p. 6; 1904, p. 71). In any case the cell-reduction which Appendicularians have undergone renders the distinction between uni- and multi-cellular of little moment: the predecessors of these glands, if they existed in the ancestors of Appendicularians, were almost certainly multi-cellular organs (see account of endostyle, p. 106).

Salensky homologizes these glands with the adhesive suckers of larval Ascidians, which he believes may have been derived from them. Ihle objects that it is not easy to compare two oral glands with three oral suckers, and that they are absent in Oikopleura longicauda and other species believed to be primitive. Although, as we shall see in a moment, if there is any relationship between these structures, the descent must be traced the other way round, it is worth noting that Ihle’s first objection at any rate is not fatal to a possible homology, for two reasons. Firstly, in Kowalevskia Lohmann has shown the existence of three¹ grosse Hautdrüsen’ in the pharyngeal region, two lateral (?=the oral glands of Oikopleura) and one mid-ventral (1896, Pl. I; cf. my Text-fig. 4, a); and secondly, that in Botryllus at any rate the innervation of the larval suckers is by a pair of nerves, one to each lateral sucker, the median sucker merely receiving a twig from the right lateral nerve (Grave and Woodbridge, 1924). Thus the primitive larval condition may have been one with a single pair of suckers, the median one being a later addition; or if three are primitive, the median one may be represented in Oikopleura by the simpler ‘Kehldriise’ which Lohmann associates with the oral glands, significantly enough, as possibly serving ¹ zur Befestigung des Tieres’’ inside the ‘Haus’ (1899, p. 380, Taf. II, fig. 15).

But all this is overshadowed by Delsman’s remarkable observations on the development (1910 and 1912), a full account of which is given by Ihle (1. c., pp. 476-7, figs. 27, 28). The oral glands are of endodermal, not ectodermal origin. They are said to be derived from a Y-shaped mass of endoderm, which makes its appearance below the gut, immediately behind the endostyle, which is lodged in the fork of the Y.

Delsman gives his account with some reserve, but believes this mass to originate from remnants of the subchordal gut. The details he gives, however, are of such an extraordinary character that further evidence is necessary before they can be regarded as established. Leaving the question of origin sub judice, there appears to be no uncertainty about the later stages. The Y-shaped mass shrinks and becomes U-shaped (Ihle, fig. 28): this divides into right and left halves, and each half, moving forward, fuses with the ectoderm on each side of the endostyle, and acquires an opening to the exterior.

There is thus a possibility that the Y-shaped mass does not really migrate forwards out of the tail, but arises locally by differentiation from the floor of the pharynx. In that case it would be comparable with the pharyngeal packet of a Fritilia r i a, but more symmetrically constituted, each half, instead of remaining as an appendage of the pharynx, having severed its connexion with the latter and acquired an outlet to the exterior. The matter, however, must necessarily remain in abeyance until Delsman’s observations have been confirmed by sections or corrected, and the question of a, further relic of the Ascidian epicardium hangs on the issue.

But in one respect Delsman’s observations are sufficient even now to bring these oral glands into a highly interesting relation with another problematic group of organs, the cement-glands of larval Teleostomes (Graham Kerr, 1919, pp. 79, 178, &c.), which Graham Kerr has shown also to be of endodermal origin. Reserving, however, my discussion of this resemblance, as well as of the relations of both organs to Ascidian suckers and epicardia (see p. 144), let us turn to Amphioxus, for I must now claim the mysterious ¹ club-shaped gland’ as the modified vestige of a similar pair of structures.

The Club-shaped Gland of Amphioxus, once regarded’by Van Beneden and Julin as representing the intestine of Ascidians, and by Willey (after his discovery of its later internal aperture) as the first gill-slit of the right side, undergoes two phases of development. Arising early, just before the mouth and first primary gill-slit, as a shallow groove in the floor of the pharynx, immediately behind the transverse rudiment of the endostyle, it becomes constricted off as a closed vesicle and then opens to the exterior beneath the antero-ventral margin of the left-sided mouth. It retains this structure, from which its name is derived, through the long period with one gill-slit. Although it lies mainly outside the actual right wall of the pharynx, its morphological position of origin, like that of the long left limb of the endostyle in front of it, and of the primary gill-slit behind it, is essentially left-sided.

Then comes the second phase of its development, which provides its blind extremity with a new internal opening into the pharynx, immediately above (i. e. morphologically to the right of) the hinder part of the short right limb of the endostyle (Willey, 1894, figs. 74 and 75). The gland now ceases to be club-shaped and is converted into a transverse tube with an internal orifice on the right of the endostyle and an external orifice on the left of it (Goodrich, 1909, Pl. xv, figs. 32, 33). According to Van Wijhe, the internal orifice, though not formed simultaneously with the gland, may occur as early as the first larval stage with a single gill-slit (1914, p. 69, in confirmation of Legros, 1897).

Willey’s interpretation of this puzzling organ as the first gillslit of the morphological right series was based mainly on the grounds that, without it, the first primary slit (i. e. of the left series) had no antimere, and that the final internal orifice was definitely on the right side; but the objections to this view are surely insuperable. Regarded as a gill-pouch, the bulk of the gland is derived from the left, not the right, side, and the right internal orifice is obviously not the original orifice of évagination. The same objections apply to Van Wijhe’s theory of it.

It would be easier to regard the gland as a fused pair of gillpouches, of which the right member was retarded in development; but again neither the original nor the final internal orifice falls into sequence with the rudiments of the primary and secondary slits, their close proximity to the endostyle being quite anomalous and at variance with the antagonistic functions normally belonging to gill-slits and endostyle. The absence of an antimere for the first primary gill-slit cannot count for much in the development of Amphioxus, especially as this slit itself is merely transitory.

The simplest interpretation is plainly to. regard the gland as formed from a pair of pharyngeal diverticula, with separate internal orifices, but fused distally into a single cul-de-sac, which acquires an external opening by fusion with the ectoderm. This Y-shaped tube has then been rendered asymmetrical by a retardation in the formation of its right-sided component, and by early severance of the original left-sided connexion with the pharynx (Text-fig. 11, d, e, f). I need scarcely point out how directly this interpretation points to the origin of the gland as a vestigial epicardium which has undergone a glandular modification and secondarily acquired an external opening. The epicardium of most Synascidians is a Y-shaped tube with a terminal cul-de-sac (Text-fig. 11, b), and its original orifices of evagination become closed in later life (Seeliger in Bronn, p. 566). The conditions of asymmetry in Amphioxus account in principle not merely for a retardation in the appearance of the right internal orifice, but also quite naturally for a persistence of this second aperture after the first (left) one has closed. This necessary difference between right and left sides in the time at which the internal orifices open and close has however been secondarily exaggerated by the assumption of a definite glandular function by the organ. I have already cited the case of E u h e r d m a n i a as an unequivocal example of a glandular modification of the epicardium. The pharyngeal packet of Fritillaria affords another, if more debatable, illustration. In each case the glandular modification appears to have no functional importance but to be merely a step in the retrogressive metamorphosis of the organ. But in the larval Amphioxus the suggestion presents itself that the vestigial epicardium, while undergoing ‘glandular degeneration’, has been drawn into the service of a larva prematurely hatched and inadequately provided with a normal feeding mechanism. This aspect of the asymmetrical phase of Amphioxus is more fully discussed below (p. 151). In the meantime it can be pointed out that the club-shaped gland is a transitory organ, limited to the asymmetrical larval phase, and with no homologue at all in the larval or adult Balanoglossus, but naturally falling into line with the metendostylic glands in Appendicularians just discussed. The club-shaped gland first directs its products to the outside, like the oral glands of Oikopleura, but when its second internal orifice arises (by simple inheritance) this orifice is retained as a means of diverting the secretion directly into the pharynx. The external orifice has a very significant position in relation to the mouth, and the second internal orifice disappears with the gland itself as soon as the pharynx has become normally constituted and its endostyle fully developed. Thus both the position of the external orifice and the duration of the second internal orifice have probably each been adaptively modified to secure the passage of the secretion into the pharynx of the larva (cf. Van Wijhe, 1914, p. 70).

These simple explanations fit the facts so completely that, after a temporary amazement, I cannot doubt their validity, and must therefore take courage to submit the inevitable corollary. The endodermal ‘cement organs’ of larval Teleostomes are also relics of the Ascidian epicardia.

—These ‘very interesting and puzzling structures’ (Graham Kerr, 1. c., p. 182), which have been variously regarded as modified gillpouches (like the club-shaped gland), or coelomic cavities (like the Ascidian epicardia), are just paired outgrowths of the pharyngeal endoderm, which become constricted off, fuse with the ectoderm, and then open to the exterior, their glandular surfaces forming the adhesive tips of the suckers (Kerr, 1. c., fig.101). InPolypterus the suckers are situated at the corners of the mouth, but are subsequently carried forward into a preoral position (1. c., fig. 100). It is clearly possible that their original position may have been postoral, a change comparable with that which the premandibular cavities have admittedly undergone. Kerr, indeed, regards them as ‘really homologous’ with the postoral cement-organs of Dipnoi and Amphibia, in spite of the difference in their final positions and the apparent absence of endoderm in the organs of the latter groups (1. c., p. 181). Like the epicardial outgrowths of Ascidians, the endodermal cement-organs show a marked tendency to fuse into median structures. But for the uncertainty already mentioned as to the origin of the endodermal element in Oikopleura, the cement-glands of Teleostomes and the oral glands of Appendicularians are almost identical (cf. Text-fig. 12).

It may be urged, on the other hand, that if Salensky is right in homologizing the oral glands with the suckers of larval Ascidians, the ectodermal suckers of Ascidian larvae, as well as the oral glands of Oikopleura, may have been derived from the endodermal suckers of Teleostomi, the former by loss of the separate endodermal element (cf. Amphibian suckers), the latter by loss of the suctorial function.

To this method of putting the facts my reply would be threefold: (1) that, in view of the peculiarities common to all the glands under discussion, it is not permissible to detach the clubshaped gland of Amphioxus and the pharyngeal packets of F r i t i 11 a r i a from simultaneous treatment with the oral glands of Oikopleura; (2) that (with reservations already mentioned in the last case) all these glands are readily derivable from epicardial structures, but those of Frit ilia ria and Amphioxus are not easily derivable from endodermal suckers; and (3) that, as the Vertebrate pharynx shows vestiges of a former Protochordate structure (endostyle, subnotochordal rod, tongue-bars, see p. 60), it is more consistent to consider the endodermal element in the suckers of Teleostomes as derived from vestigial epicardia.

It should be borne in mind that even if we could be satisfied with an interpretation of the Protochordate glands as modifications of Teleostome suckers, we should merely be substituting one mystery for another, since it is the uncommon complexity of the Teleostome organs which renders them, in Graham Kerr’s phrase, so ‘interesting and puzzling Thousands of animals have invented ectodermal cement-glands, but only Teleostomes have turned endodermal diverticula inside out for this simple purpose. I can well believe with Professor Kerr that there has been an unbroken series of Vertebrate larvae with ventral suckers from primitive Teleostomes to modern Amphibia, but I should express the evolutional change which has occurred along this series rather as an example of the universal tendency to simplification of development. The Teleostomes may have used endodermal glands at the outset, not because endoderm was necessary for suckers, but because glands of that character were already in existence and could be utilized as such from the beginning. Their subsequent simplification affords the best possible evidence that their original complexity was related to some antecedent function which had been superseded.

Although the budding of Ascidians is commonly ignored in discussions of Vertebrate origins, or regarded as a secondary incident of their general degradation, it is difficult to believe, as already remarked, that budding can have arisen de novo at the Chordate level of organization. Assuming then that in this respect Ascidians have retained, and Amphioxus has lost, a primitive Protochordate feature, it is not surprising to find traces of the characteristic Ascidian organ of budding in Amphioxus and in the lower Vertebrates, which must be admitted to have had a Protochordate origin (cf. Van Wijhe, 1914, p. 74).

While the theory of a vestigial epicardium seems to afford a perfectly satisfactory explanation of all these pharyngeal glands, as it does of the reno-pericardial glands of Molgulids (Damas, 1902), it is not irrelevant to add that, so far as I am aware, there is no other structure that can be brought into any conceivable relation with them, unless it be the pair of gastric diverticula in Actinotrocha, which Masterman (1897) interpreted as a pair of lateral notochords. The peculiar position of the droplets secreted by these glands, at the very tips of the cells, and outside the nuclei, makes it highly probable, however, that their secretion is discharged into the blastocele. These glands stand consequently in a class apart, and plainly discharge some important larval function, which it would be interesting to determine more precisely. In view of the observations of Ritter and Davis (1904) on the part played by the gastric epithelium in discharging material into the blastocele of Tornaria, it is possible that the glands of Actinotrocha are also indirectly glaeogenous, both larvae being highly gelatinous, and undergoing the same marked diminution in size at the metamorphosis, when this particular glandular activity ceases. In any case, except for a certain similarity in position, the glands of Actinotrocha seem to possess no features which throw any special light on the origin of the glands we have been discussing.

It should be noted that the reaction of Masterman’s ‘vacuoles’ to osmic acid, in consequence of which De Selys Longchamps interpreted them as fat droplets (1902, p. 581), has been denied in the case of the American species by Brooks and Cowles (1905, p. 89).

I need scarcely add that the presence of epicardial vestiges alone is quite consistent with the previous evidence which points to a Doliolid ancestry of Appendicularians, for although the stolon of D o 1 i o 1 u m contains cloacal as well as pharyngeal diverticula, the loss of the cloacal cavity in Appendicularians necessarily carries with it the loss of any trace of cloacal outgrowths.

The case of Amphioxus is more intricate since in the formation of its endostyle and gill-slits it retains features which are more primitive than those in any known Tunicate. The identification of its club-shaped gland as a vestigial epicardium is even embarrassing, since it points to this organ as already an essential feature of the earliest Protochordates, in which case its absence in certain sections of Tunicata must be secondary. This question will therefore require examination, but before proceeding with it, the larval asymmetry of Amphioxus needs somewhat closer study, for not only does it affect the interpretation of the club-shaped gland, but has itself been given a phylogenetic significance which is inconsistent with the views here maintained.

When Amphioxus hatches it does so at an exceptionally early stage, the neural plate being still largely exposed, and only two or three pairs of coelomic segments established. This stage may be regarded as that of a Dipleurula larva, but it has none of the characteristic features of such a larva, being uniformly coated with long flagella and mouthless. It is in fact a mere embryo, prematurely hatched. When the mouth appears, it already exhibits the characteristic asymmetry, as also do the other pharyngeal organs, the right gill-slit of the first pair being absent, the right limb of the endostyle much shorter than the left (or altogether absent ?), and the ‘epicardium’ correspondingly limited almost entirely to its left portion, the ‘clubshaped gland’. Apart from this asymmetry, however, its grade of organization (except for the absence of atria) now corresponds with that which we might expect in a primitive Ascidian tadpole, viz. mouth, single pair of gill-slits, peripharyngeal band with ventral loop; but, instead of actively swimming, like a tadpole, its physiological condition is as abnormal as its morphological. It possesses numerous coelomic segments but no functional muscles; still drifts passively, suspended vertically by its general coating of long flagella, and, according to all accounts, continues to do so during the whole of this phase of its larval career (14 days). The significance of these peculiarities seems hitherto to have escaped notice, but their abnormality is quite as great as that of the structural asymmetry, and an explanation which accounts for all of them together is more likely to be correct than any ad hoc speculation which is applicable only to one or two special points.

Dealing with the asymmetry alone, Korschelt and Heider (1893, p. 1461), followed by MacBride (1909), have regarded it as a relic of a hypothetical ‘pleuronectid’ ancestry, a view which, apart from other objections, is at variance with the mass of facts, already discussed, which demand a common ancestry for Tunicates and Amphioxus up to a stage of organization well beyond that when the pharyngeal asymmetry is first manifested.

Willey’s suggestion (1891, p. 214; 1893, iii; 1894, p. 159) that it is due to the dislocation of an originally dorsal mouth by the anterior extension of the notochord can hardly be taken seriously, since it rests on the impossible assumption that the dorsal position of the mouth in Ascidian larvae is primitive. This feature, if not a mere anticipation of adult character, admits of a simple explanation in the great development and forward direction of the ventral lobe of fixation, which is obviously a functional adaptation.

The ‘Tremostome’ theory of Van Wijhe solves the problem of asymmetry in accordance with some remarkable facts, but leaves a more difficult problem in its place, viz. why the original mouth should have closed. The only suggestion offered by Van Wijhe is that, in connexion with the rotation of the flagellated embryo observed by Hatschek, ‘the first gillslit of the left side occupied the most favourable position for an ingestive aperture’ (1907, p. 70). Apart from the fact that Hatschek’s description of the rotation ¹ von rechts nach links’ is ambiguous, and that both adult Amphioxus (Franz, 1924) and larval Ascidians (Grave, 1920) rotate in a direction opposite to that which Van Wijhe appears to have assumed (viz. in these forms, clockwise as viewed from behind, or from left to right as seen from above), the ‘advantage’ of a left-sided mouth is surely dubious. The larva is advancing all the time that it is rotating, so that a mouth in front would get as much, if not a greater, advantage from the spiral rotation as would an opening on the side. As a matter of fact a ciliary feeding current is capable of maintenance just as well, and usually much better, in the absence of any locomotive movements at all. A rotating scoop, if efficient, would flood the pharynx with more water than it could filter; but there is no evidence that the larval mouth is fitted to act in this way.

The explanation now offered is based on the relative yolklessness of the Amphioxus egg. In general it is not impossible that a yolkless egg may be primitive; but in the groups under immediate consideration it is a complete anomaly, and falls into the same category as the small pelagic eggs of Teleosts, the little-yolked eggs of Passerine Birds, and those of Mammals. If, therefore, we assume that the yolklessness of Amphioxus eggs is a secondary feature—a change from a more highly yolked condition in the Protochordate ancestor—a chain of consequences is set up which accounts both for the muscular feebleness of the larva and its structural asymmetry. For though the thin and diffuse distribution of yolk throughout the egg, and then throughout the germ-layers, ensures a rapid and even differentiation of the latter up to a certain point, the slightness in its amount limits the process, and the embryo hatches long before it is equipped with the normal functional mechanism for feeding and swimming. Like a Passerine chick, or a Marsupial babe, it is precocious in form, but retarded in function, and, as with them, its defective condition sets up an immediate demand, at the expense of everything else, for a provisional feeding mechanism. In the parallel cases mentioned, the demand is met by parental attentions and maternal modification; but here the larva is dependent on itself, and, metaphorically speaking, must improvise something out of the rudiments of an apparatus functionally adapted to a different set of conditions. At any rate a study of the asymmetrical pharynx reveals the possibility of an explanation of it on these lines.

Exactly how this abnormal pharynx works we do not know from observation and experiment, but certain peculiarities of structure indicate its main principles. The mouth, unlike the mouth of the adult, or of Tunicates at any stage, is powerfully and peculiarly ciliated (Willey, 1891, p. 211), the gill-slits apparently less so, at least in the earlier stages (cf. Goodrich, 1909, Pl. xv, figs. 32, 36). This means that the current is produced by the mouth, and not, as is usually the case, by the cilia of the gill-slits. If a strong stream were to be drawn through the first and each successive gill-slit, food particles and water together would sweep almost uselessly through the pharynx, since the compression of the body is so extreme that there is no room for a filter of mucilage between mouth and gillslits, as in Appendicularians, and there is no long endostyle, as in the adult, to coat the walls themselves with slime. From the arrangement of parts it is probable that the long, deepseated, oral cilia (Willey, 1. c.) beat the water in a thin diffuse stream against the inner left wall of the pharynx upwards, downwards, and sideways, but especially downwards (where food particles are specially noticeable in preserved specimens), so that food particles can be collected by the peripharyngeal bands which encircle the mouth at a slight distance within the pharyngeal cavity.

If the current were to set strongly in against the right wall of the pharynx, it is difficult to understand how any food particles could be kept from dropping at once into the exhalant streams converging towards the gill-slits. But if the mechanism works in the way suggested, and indeed in whatever way it works, so long as the oral cilia are the chief current-producers, the larger and more numerous the outlets, the gentler and more diffuse will be the streams; so that the dilatation and weak ciliation of the gill-slits and their multiplication posteriorly are exactly the adaptations required for a successful expedient in the absence of a filter across the intake. Presumably the endostyle at this stage works as in Appendicularians by directing a flow of mucilage along the peripharyngeal bands, especially the lower (left) one, which, like the limb of the endostyle continuous with it, is better developed than the upper (right) one, and lies like a trap along the narrow floor of the pharynx between mouth and gill-slits in these early stages (Goodrich, 1. c., fig. 33). Later on, when the endostyle has grown backwards from the apex of its convergent limbs, and the median flagella have been developed (see p. 90), the dominant direction of its ciliary activity presumably changes from longitudinal to lateral, as in the adult.

More than this we can scarcely infer, except, as already suggested, that the club-shaped gland almost certainly plays an important part in supplementing the function of the imperfectly developed endostyle, as Goldschmidt (1905) and Van Wijhe (1914) have previously maintained. We can point out that the gland produces and contains material which is probably utilizable, and that, when the larva is vertically suspended and drifting slowly with the surrounding water, the position of its external orifice near the anterior end of the lower lip is ideally situated for supplying the edge of the mouth with mucilage, and indirectly—through the action of the oral cilia—the important prebranchial zone between mouth and peripharyngeal bands which cannot be served by mucus from the endostyle. The special anterior tufts of cilia described by Willey (1. c., pp. 211-12, figs. 20, 21, 23), together with the sub-oral tract of small cilia, may play an important part in sweeping the secretion into the mouth. The subsequent internal aperture, which is peculiarly long and slit-like, must first supplement, then replace, the external pore in function, until the asymmetry is in course of rectification, but by that time the structural conditions are too complex for any safe opinion on the actual direction of the currents or the part played by secretion from the gland.

If this is a correct exposition of the larval feeding mechanism, it is no more necessary to force a phylogenetic meaning into it than in the corresponding cases of the chicks of Sparrows and the embryonic babes of Marsupials, and even the most ardent Haeckelian has not yet ‘explained’ their abnormalities in terms of naked, blind, and suctorial ancestors.

A certain asymmetry, or left-sidedness, is noticeable both in Rhabdopleura (Schepotieff, 1904) and the Tunicata, so that a slight tendency in this direction may be an inheritance from prechordate ancestry; but its extreme development in Amphioxus larvae I regard as a larval adaptation, consequent on the secondary yolklessness of the egg. The mouth is enlarged to colossal proportions and specially ciliated to collect food. It cannot be confined to its normal place, so it invades the side to which there is perhaps an inherited tendency, and displaces to the right side the gill-slit that belongs to the left, all the normal right-sided organs being temporarily suppressed in development. As the mouth enlarges the gill-slits, weakly ciliated (see Goodrich, 1909, Pl. xv), multiply and dilate, thus avoiding any intensification of the oral stream by the formation of ‘flues Without these compensatory adjustments in the early stages, feeding would be impossible, or at least involve great waste of food; but, with these arrangements, and by postponement of all muscular activity and unnecessary organformation, such as atrial folds, the larva grows in bulk, the mouth turns round into its proper place, the gill-slits follow suit, the two limbs of the endostyle unite and grow backwards, supernumerary gill-slits of the precociously developed left series are absorbed, and epibranchial flaps eventually grow downward over the gill-slits of both sides to form the belated atrium. Adaptation and self-regulation could hardly go farther, but there is nothing in the whole chain of events which is not sequential on the original disturbing factor, the secondary reduction of yolk in the egg.

Nevertheless, although the ultimate product of this extraordinary development is a bilaterally symmetrical creature which is universally accepted as the simplest type of Vertebrate normality, the morphologist, who believes in the possibility of reconstructing phylogeny, may well ask the question: Has all this chain of larval modifications left no effect on the adult form and structure ? Is the final product identical in essential features with the adult form of that earlier life-history which began with a more normal provision of yolk in the egg ?

The analogies which I have given of yolk-reduction do not help us here, for the Sparrow and the Marsupial are carried over the difficult post-embryonic period not by conspicuous modifications of their own structure, but by the attentions and modification of their parents. But in Amphioxus the structure of organs, and their relations to one another, are definitely changed in the larva, and these changes are inherited as truly as any adult changes may be inherited. The mouth develops from the outset on the left side, and the left gill-slit on the right side, though the larva is freely and vertically suspended. The subsequent rectification of these relations masks, but does not explain, the morphogenetic problem involved. If Amphioxus could reproduce before the rectification took place, a new type of organism would have come into existence, the distinctive features of which were consequential on the primary mutation of yolk-reduction. Indeed, if Goldschmidt (1905) is right, that Amphioxides is capable of reproduction, this has already happened (see, however, Gibson, 1910, p. 239).

It is therefore possible that certain distinctive features of the adult Amphioxus may be primarily due to the provisional larval adaptations we have described, and that they have been retained, either because rectification was physically impossible, or because, in association with changed habits of life, they yielded a new viable combination which needed no rectification. One such feature, indeed, has long been recognized as a result of Van Wijhe’s observations (1889), viz. the innervation of mouth and buccal (pre-oral) cavity exclusively by nerves of the left side. To this two points in particular may be added: the form of the atrium, and the multiplication of gill-slits. We have already given reasons for the opinion that the formation of the atrium is secondarily retarded, and that the multiplication of gill-slits forms a vital part of the provisional larval feeding mechanism, so essential that a number of the firstformed gill-slits are subsequently absorbed, when the right series begins to appear. A bionomical condition of both these features is undoubtedly the peculiarly passive drifting life of the larva, for no larva could safely combine such extensive fenestration of its body-wall with active habits of muscular locomotion.

It follows that the unusual multiplication of gill-slits in A m phioxus, i.e. its elongated pharynx, and the single ventral atrium, may be secondary consequences of the larval modifications induced by the diminution of yolk. And when to this possibility we add the evidence derivable from the history of the club-shaped gland, we submit that in all probability the ancestors of Amphioxus were essentially primitive sessile Ascidians, with an active muscular larva of a generalized Ascidian type; and that Amphioxus owes much of its actual organization to secondary modifications consequent on the elimination of fixation, metamorphosis, and budding from its life-history, and on the reduction of yolk in its egg.

In its possession of protonephridia, however, Amphioxus displays a primitive feature unique among Chordata, but the analogy of Actinotrocha renders it probable that we should regard this as the retention of a larval character, not without possible relation to the absence of genital ducts. We know that in Teleostei the secondary reduction of yolk and the introduction of pelagic spawning have been followed by diminution and eventual atrophy of the oviducts, and the reduction of genital ducts in Appendicularians is another comparable illustration. Although Amphioxus is not exactly a pelagic animal, its organization for burrowing is limited to the adaptation or extension of certain features previously in existence (notochord, atrium, incipient metamerism, general shape), and some of those pre-existent features (neuro-muscular metamerism, notochord) point emphatically to an active larval life in the past before the period when yolk-reduction had set in. Another larval character of the adult Amphioxus is the absence of a pericardium, which is a constant feature in adult Tunicata, even in the degenerate Appendicularians.

On all points we reach the conclusion that the organization of A m p h i o x u s is essentially a paedomorphic transformation of a sessile Protascidian type, due primarily to a prolongation of the originally active larval life, with the elimination of fixation and metamorphosis, and modified by direct and indirect consequences of a reduction of yolk in the egg, and by minor adaptations to a sand-burrowing existence.

This view of the relationship between Amphioxus and the Tunicata finds a somewhat striking confirmation in a comparison of the caudal regions. Hitherto the Ascidian tail has been generally regarded as equivalent to the post-anal region of Amphioxus, specialized for larval life by degeneration of the myotomes. As a matter of fact the young Amphioxus larva possesses a highly characteristic caudal fin long before the segmented ‘tail’ of the adult begins to develop (Korschelt and Heider, figs. 875-6); and its peculiar striations, which simulate fin-rays, are strikingly similar to those in the cuticular tail-fin of larval Ascidians, and to those of the youngest larval Teleosts, as already remarked upon by Lankester and Willey (1890, p. 457).. On the histological nature of these resemblances, which may well be clinching, I hope to report later. In the meantime it should be noted that at the metamorphosis, when the formation of myotomes and tail is completed, this larval fin is pushed off and replaced by the adult fin, which possesses a radically different structure (Lankester and Willey, l.c., figs. 1, 4; Korschelt and Heider, figs. 877—8). Now, although histological details are lacking, it is remarkable that in Asymmetron the larval fin is borne almost entirely on an unsegmented caudal process which contains extensions of notochord and nervecord, but no myotomes. At the metamorphosis this fin appears to migrate forwards (Andrews, figs. 5, 3, 4), but is more probably replaced by a new one in the segmented region, while the appendage persists in the adult as a finless ¹ urostyle’. This appendage, and the larval fin of Amphioxus, must be homologous with the Ascidian larval tail.

Thus two phyletic stages are represented in the ontogeny of the Cephalochordate tail—a primary stage, comparable with that of Ascidian larvae, and a subsequent definitive stage associated with the secondary metamerism of the body. In Asymmetron these two stages are distinct and successive; in Amphioxus they are blurred by suppression of the primary caudal appendage altogether. In other words the Ascidian tadpole is not a reduced Amphioxus, but Amphioxus (with certain obvious reservations as regards the coelom) is an Ascidian tadpole which has lost its tail and undergone a secondary elongation and segmentation of its body. The postanal region of Amphioxus, with its myotomes and caudal fin, is a new and later differentiation, unrepresented in Ascidians—an adaptation of its body to the retention of a free life beyond the originally restricted larval stage.

The original nature of the pharyngeal diverticula which fuse to produce the stolonial septum of the Clavelinid Synascidians is still unsettled, and there are legitimate differences of opinion whether these structures arose as such in connexion with the budding process, or became secondarily involved in the.process after previously discharging a purely somatic function (cf. Seeliger in Bronn, pp. 564, 921). The situation at the present time may be defined, I think, as follows.

It is fairly clear that the occasional relations of the epicardia to the heart (e. g. in Clavelina), to which the pharyngeal diverticula owe their name (Van Beneden and Julin, 1887), are secondary. The recurrence of these relations in the renopericardial organ of Molgulids is all the more noteworthy (Damas, 1902). In Ciona the pericardial rudiments have long been known to arise independently (Willey, 1893, i; Damas, 1899; De Selys Longchamps, 1901); and both in Clavelina and Dis tap lia Julin himself finally asserted the complete developmental independence of the two organs (1904, p. 545). The epicardial diverticula, when present, always arise later than the pericardial. In other words Van Beneden and Julin’s conception of ‘procardial’ outgrowths, which subsequently become separated into epicardial and pericardial elements, has gone, together with the theoretical superstructure based upon it. The interpretation of the epicardia as modified coelomic diverticula, possibly ‘collar cavities’ (Seeliger in Bronn, p. 925; MacBride, 1914), loses an important argument in its favour by the withdrawal of all evidence of a developmental connexion with the pericardium (cf. Sedgwick, p. 16).

Similarly the expansion of the epicardia in Ciona to form a pair of perivisceral cavities seems to have no greater significance than the similar expansion of peribranchial diverticula in Botryllus and the Polystyelidae or Amphioxus itself. The fact that ectodermal diverticula can discharge a pseudocoelomic function very similar to that of the epicardial sacs in C i o n a seems to show that in both cases the ¹ perivisceral’ function is a secondary one, and lends no support to a theory of their coelomic origin.

Finally, the fact that the epicardial diverticula by distal fusion (Clavelina, Polyclinids, &c.) or by prolongation of the left one alone (Ciona) undoubtedly function as a septum between afferent and efferent blood-channels, either in the abdomen alone or in stolonial or pseudo-stolonial extensions, is paralleled by a similar modification of peribranchial extensions in Archiascidia according to Julin (1904), in Polystyelids (De Selys Longchamps, 1917; Oka, 1926), and possibly in Perophora (Lefevre, 1898), though the last case rests on indirect evidence and is at variance with observations by Kowalevsky (see De Selys, 1901, p. 514) which have not yet been disproved.

This allelomorphism of endodermal diverticula of the pharynx and ectodermal diverticula of the atrium subservient to the same physiological functions is also recognizable in the types of budding among Ascidians, which with a few insignificant modifications fall into one or other of two classes according to the origin of the inner layer of the triploblastic bud rudiment. These types can be distinguished as pharyngeal and peribranchial respectively and are characteristic of different natural groups of Ascidians. In the Clavelina-Distomid-Polyclinid series the inner vesicle of the bud-rudiment is derived from diverticula of the pharynx (i. e. endoderm), but in the BotryllidPolystyelid series from diverticula of the peribranchial epithelium (i. e. ectoderm). It is not known whether Julin’s Archiascidia reproduces by some form of budding or not, but if it does—and this is highly probable—the inner vesicle is almost certainly derived from the abdominal septum, as in Polyclinids; but this septum, as already remarked, is formed in Archiascidia from peribranchial, not pharyngeal, diverticula if Julin’s remarkable account is to be accepted. It is greatly to be regretted that Julin gave no illustrations to support his account of the development of the septum in this rare and remarkable animal, especially as the sections he figures of the adult structure point on the face of things to a simple origin of the septum from posterior diverticula of the pharynx as in Clavelina (cf. his figs. 8 and 9), and the peculiar postero-dorsal series of stigmata (figs. 1, 2) require an extension of the peribranchial cavity on each side exactly corresponding with the extensions which, according to his account, atrophy and disappear after giving rise to the abdominal septum.

However, apart from the doubtful cases of Archiascidia and Perophora, the contrast between the two types of budding is clear and distinct, and corresponds with wellestablished systematic differences between the two series of Ascidians based on other characters. The cleavage between the two series on grounds both of adult structure, larval characters, and mode of budding is indeed so sharp and clear that the group of Ascidiacea (i. e. the fixed Ascidians) must be regarded as consisting of two perfectly distinct stocks, related to one another only in the sense that both are derivable from some third stock which is still unknown, and possibly quite extinct.

These two stocks, which have not hitherto been formally recognized, I propose to term Endoblastica and Periblastica respectively. The recognized groups fall into them as follows:

—Pharynx produced behind the endostyle into a pair of diverticula (‘epicardia’) which separately or by distal fusion form a septum (which may be distinguished as the ‘endophragm’) between afferent and efferent bloodchannels in the abdominal region or in the branches of a single ectodermal outgrowth of mid-ventral origin (stolon and testvessels); endophragm absent in solitary forms only and then replaced by folds of the walls. Larva with cerebral eye and otolith, and a ventral adhesive organ with three suckers, but no peripheral processes (except Didemnidae). Peribranchial cavities developed before the cloaca. Brain with subneural gland. Gonads unpaired. Buds, if produced, deriving their inner vesicle by outgrowth from, or by fragmentation of the endophragm or epicardia (with rare modifications).

1. Aplousobranchia,¹Lahille, 1890 (but more correctly Haplobranchia), amended by Garstang (1895) by elimination of Doliolidae and Pyrosomidae = K r i k o branchia, Seeliger, 1907.

2. Phlebobranchia, Lahille, 1890 (and Garstang, 1895) =Dictyobranchia, Seeliger, 1907.

—Pharynx without epicardia or endophragm, or at most with modified vestiges (Molgulids). Larva with one cerebral sense-organ only, with or without ventral suckers, but with a ring of ectodermal papillae around the adhesive area. Peribranchial cavities formed as outgrowths from cloaca. Brain with supraneural gland (rarely lateral or subneural). Gonads usually paired. Buds, if produced, deriving their inner vesicle by outgrowth from the outer peribranchial epithelium or from a pair of posterior peribranchial diverticula. Centrifugal test-vessels supplemented or replaced by peripheral, all without a septum (constricted or ‘doubled’ in Molgulids).

1. Stolidobranchia, Lahille, 1890 (and Garstang, 1895) =Ptychobranchia, Seeliger, 1907.

(?) 2. H e mi bran chi a, nom. nov. for Julin’s Archiascidia if its abdominal diaphragm should prove to be of peribranchial origin, as alleged (i. e. an ‘ectophragm’).

The inclusion of Archiascidia, however, would involve omission of the special larval peculiarities of the Periblastica, since its own larva in those respects is typically Clavelinid.

[3. Aspiraculata, Seeliger, 1907. This suborder, which was proposed for the single species Hexacrobylus psammatodes of Sluiter, falls in here; but as this type is admittedly of Molgulid extraction, and closely related to Bourne’s Oligotrema, the group should properly be suppressed and replaced by a subfamily of the Molgulidae, e. g. Oligotrematinae, to include Oligotrema and Hexacrobylus.]

It will be noticed that the first group (A) includes both longbodied and short-bodied forms, and that the second group (B), again with the uncertain exception of Archiascidia, includes only short-bodied forms. Sluiter’s attempt to found a primary subdivision on this basis (as Merosomata and Holosomata) broke down owing to the cross-relations of such forms as Ciona, Rhopalaea, Ecteinascidia, and D i a z o n a. As now arranged the cleavage between the two primary groups is sharp and absolute, with regard both to somatic, larval, and blastogenetic characters, although, as in other phyletic classifications, it is impossible to draw a hard and fast line between these two series on the basis of any one selected character. In group A the test-vessels are septate and radiate from a central point (i. e. centrifugal), which coincides with the base of the stolon when this is present; in group B the centrifugal system is supplemented or replaced by test-vessels having a diffuse or peripheral origin, septa are absent, and budding, when it occurs, is parietal instead of stolonial, although without any direct relations to the peripheral ring of test-vessels. The gonads are unpaired in (A), usually paired and parietal in (B).¹

Clearly in this contrast we have come upon a crucial point in Ascidian evolution, and it is important to find an explanation. The clue once more is provided by Doliolum, or rather by the details of the budding process in the Thaliacea generally, for in these forms (with the possible exception of Anchinia) there is neither epicardial budding as in the Endoblastica, nor peribranchial budding as in the Periblastica, but transverse division of a median stolon which contains both pharyngeal and cloacal diverticula. To Julin (1904, p. 547) belongs the credit of first pointing out the significance of these relations: ‘les deux types connus du bourgeonnement des Ascidies doivent dériver d’une forme primitive, réalisée chez Doliolum’; but the paper in which he promised to develop his ideas seems never to have been published.

In the meantime De Selys Longchamps (1916 and 1917), by his work on the Polystyelids, has not only widened the basis of comparison, but has reached the same conclusion as Julin, with a slight, but important, qualification on the phylogenetic question: ‘Il faut bien admettre que les Thaliacés, encore que secondaires au point de vue de la structure des adultes et de leur embryologie, sont primitifs en ce qui concerne la blasto-génèse (formation du stolon) … et … que la forme ancestrale commune aux Thaliacés et Ascidiacés était douée d’un mode blastogénique pharyngopéribranchial, différant en réalité fort peu d’une simple division transversale, avec laquelle la blasto-génèse des Thaliacés a conservé tant d’analogie’ (1917, p. 271). De Selys, however, refrains from any development of these phylogenetic considerations, and raises them merely to emphasize his view of the secondary nature of the ‘stolonial’ (= epicardial) mode of budding in Tunicata, as against the primitive role which Korschelt and Heider were inclined to ascribe to it (1. c., pp. 269, 272).

There the matter rests; and the question is now whether this new orientation towards the facts helps us to comprehend the evolution of budding in Tunicata and thereby, in conjunction with other considerations, form some idea of the form and habits of the primitive Tunicate from which both Ascidiacea and Thaliacea have been derived. The evidence derived from the modes of budding, which seems to make the Thaliacea more primitive than the Ascidiacea, has to be reconciled with the unequivocal evidence already given that the Thaliacea have been derived from fixed ancestors (position of cloaca, structure and functioning of gill-slits and endostyle, degeneration of oozooid).

When the modes of budding in Tunicata are traced back to a process which ‘differs in reality very little from one of simple transverse division’, it is nevertheless to be noted that the process in question differs materially from true fission as exemplified, for example, in Phoronis (Harmer, 1917), in the fact that the functional somatic organs are not themselves divided. In Phoronis both limbs of the intestinal loop are cut through in the process, and there is regeneration on each side of the dividing plane. In Thaliacea the body is not divided, but only a process of the body which contains special prolongations of the organs immediately adjacent to it. It is as if we had before us, not an adult Phoronis, but an Actinotrocha in which the ventral invagination had been everted, and then invaded by special prolongations of the descending and ascending portions of the gut instead of by the gut itself. This would constitute the equivalent of the Thaliacean stolon, and in the transverse division of such a stolon we should have a real parallel to the mode of budding now regarded as primitive in Tunicata.

This comparison will serve, I think, to show that the interpretation of Tunicate budding as in any sense related to transverse fission is superficial and incorrect, the segmentation of the stolon being a secondary feature having no connexion with such a process. Korschelt and Heider (1893, p. 1365) long ago pointed out that the axial relations of the first four ascidiozooids to the stolon in Pyrosoma are completely at variance with any theory of transverse (somatic) fission, but easily interpreted in terms of stolonial budding.

The difference between the typical Thaliacean stolon and that of Ascidians in the order of development of the buds is also readily correlated with a change from sedentary to pelagic conditions. The stolon of a Perophora or a Clavelina provides anchorage for the colony as well as nutrition for the buds, and distal growth is a necessity of the situation. But the stolon of a pelagic Tunicate is free from this restriction, and the establishment of a proximal zone of growth follows inevitably from the conversion of an undivided, leisurely branching, stolon into an unbranched organ devoted to the rapid production of free buds.

Thus, apart from differences in its internal constitution, the monaxial or trailing stolon of pelagic Tunicates can be consistently regarded as an adaptive modification of the creeping stolon, and some confirmation of this view is provided by Anchinia. Although data are still lacking for a final conclusion on certain points, we know that in Anchinia a slender stolon-like thread is present along the dorsal side of the colonial stalk to which the free buds are found to be attached (Barrois, 1885, p. 203, Pl. viii, figs. 1, 3). The stalk itself almost certainly corresponds to the dorsal stalk of Doliolum, which is represented in the gonozooids of Anchinia by a long tentacle-like process from the dorsal edge of the cloacal aperture. In our ignorance we can only suppose that oozooid and gonozooid are more or less alike, and that the ventral stolon in the oozooid bends up one side of the body and grows dorsally and backwards along the cloacal tentacle, which becomes the colonial stalk. Something similar to this creeping growth of the stolon over the parental body takes place, as is well known, in Salp a (Korschelt and Heider, 1893, p. 1399, fig. 849), but there the stolon glides along a track carved out for it in the test substance, and then trails freely behind, and the process is easily reconciled with proliferation from the base. It is probable, therefore, that the stolon of Anchinia, by creeping over the parental body, is repeating, in a very novel way, the original M 2 mode of growth which characterized its fixed ancestors, for presumably the whole of the stolon rests on a substratum provided by the parental body. In this mode of growth we have the probable explanation, as Barrois suggests, of the remarkable migration of the free pro-buds of Doliolum, which repeat the process; but, although these pro-buds are produced by fragmentation from the free-end of the stolon, which must therefore grow from its base, it by no means follows that this rule also obtains in A n c h i n i a, in which the slenderness of the stolon, in relation to the length of its journey, seems to require retention of a terminal growing-point as in Clavelina—unless the whole stolon is literally dragged along by amoeboid phorocytes, as the individual buds are transported in Doliolum (cf. Korotneff, 1884; Korschelt and Heider, 1910, fig. 610).

These various considerations all point to one and the same conclusion, that the ancestors of Thaliacea were sedentary forms which reproduced asexually by budding from a ventral stolon. The differences between Pyrosomata and Myosomata (Doliolidae and Salpidae) in regard to the form of the pelagic colony are considerable at first sight, and might seem to justify a case for the independent origin of these two tribes from the ancestral sedentary stock, but the cross-relationships are too intimate to render such a view tenable. If the ancestor of Pyrosoma was a colonial Tunicate having the form of a Cyathocormus (Oka, 1913), the primary stolon must have had the form of a wreath of buds around the larval body (cf. Dip loso ma, Lahille, 1890, fig. 64), and such a wreath, by further elongation and more extensive proliferation, may well have given rise to the winding stolon of an Anchinia or a Salp. Similarly if the stolon of Doliolids was directly preceded by a creeping stolon in an ancestor of the Social Ascidian type, this ancestor was nevertheless as closely related to the other as a Clavelina is to a Distaplia. An independent origin of the two sections from genera more remote from one another than those is precluded by their possession of too many peculiar points of resemblance.

But apart from the constitution of the stolon, there are several obstacles to the view that the Thaliacea are derivable from the Clavelinid group of Ascidians (Haplobranchia). As pointed out by me in 1895, the structure of the internal armature of the Pyrosoma pharynx is quite definitely Botryllid, not Clavelinid, in character; and in the present paper I have drawn attention to the peculiar way in which the gill-slits of Doliolum apparently develop as pairs of protostigmata (p. 113), a method which is as remote as possible from the Haplobranchiate type, but closely approaches that of the Stolidobranchia, though still more primitive. Equally significant are the special resemblances between Salpa, Doliolum, and the Stolidobranchia as regards the mode of development of the cloaca, to which my daughter and I have drawn attention in our account of the development of Botrylloides.

In these facts, then, we have good grounds for the belief that the Thaliacea are the pelagic representatives of an early stock of sessile Tunicata, antecedent to all known types of Ascidians, and in which the ventral stolon of the Haplobranchs was combined with the pharynx of a primitive Botryllid. To these features we must add that in all probability the constitution of the stolon was already Thaliacean in the sense that it contained both pharyngeal and peribranchial diverticula. The origin of this latter complexity remains for consideration.

As budding from a ventral stolon characterizes both the Thaliacea and the Endoblastic Ascidians, it is reasonable to assume that it characterized the earliest Tunicates, provided the absence of such a stolon in the Periblastica can be interpreted as an example of secondary loss (see p. 173). If this be admitted, a connexion with Pterobranchia is at once afforded by the correspondence in position of the ventral stolo prolifer in both groups. Except for the presence or absence of an endodermal septum, there is complete agreement between Rhabdopleura and Clavelina in their methods of asexual reproduction: both have ventral stolons from which the summer colonies arise, and both form resting (‘winter’) buds by a distinct process of stolonial fragmentation (Schepotieff, 1904; Kerb, 1908). Moreover, in view of the relations now established between the coenoecia of modern Pterobranchia and those of the ancient Graptolites (Schepotieff, 1909), the wide variation of the latter, both in structure and in conditions of fixation and flotation, suggests many possibilities affecting the phylogenetic outlook. A drifting sicula with its Monograptid or Digraptid chain of buds must have been something remarkably similar to a Salpa-nurse with its trailing stolon. The shape of the sicula is often different from that of the rhabdothecae. ‘The sicula itself ceases to grow, as a rule, after the first thecae are budded, and sometimes it becomes obsolete or absorbed’ (Zittel in Eastman, 1913). Clearly Graptolites showed all the phenomena we discuss in pelagic Tunicates—differentiation between oozooid and blastozooid, arrested development of oozooid, precocious budding; and it should be remembered that neither in Cephalodiscus nor in Rhabdopleura can we yet assert that the larva completes its development after fixation into a * normal’ adult individual resembling the blastozooids.

In view of these many resemblances, there is clearly a possibility that budding from a ventral stalk of fixation is a feature which Tunicata have inherited from Pterobranch ancestors, and we can test this hypothesis in the usual way, by assuming it and seeing whether it works. The first result of assuming it is that we obtain a series of stages in the varying constitution of the stolon as follows:

1. The stolon encloses a pair of coelomic diverticula alone (Pterobranchs).

2. The stolon encloses mesoderm together with paired pharyngeal and cloacal diverticula (Thaliacea).

3. The epithelial diverticula are reduced either to pharyngeal outgrowths alone (Endoblastica) or to peribranchial outgrowths alone (Periblastica), but in the latter case the stolon as a single median outgrowth has disappeared, or undergone disintegration, a phenomenon not without parallel in the Endoblastica (e.g. D i s t a p 1 i a, Didemnidae).

We have to bear in mind the uncertain nature of the stolon of Anchinia (Thaliacea), which according to Barrois (1. c.) contains only an axial core of loose cells which he regards as endodermal. These may possibly represent an epicardium, but more probably result from the breakdown and fusion of both pharyngeal and peribranchial elements. If the latter view should be confirmed, and for Doi ch ini a also (Korotneff, 1891), the Thaliacean stolon, in contrast to that of Ascidiacea, would be consistently ‘amphiblastic’.

In the stolon of Pterobranchs there appears to be no special endodermal diverticulum from the parental body. The existence in Rhabdopleura of a central endodermal core between the coelomic diverticula was maintained by Fowler (1905) but denied by Schepotieff (1904, 1907), who derives the endoderm of the bud from wandering cells which collect together on either side of the median septum (i.e. mesoderm). In Cep halo discus the endoderm of the bud is definitely produced by a special invagination of the ectoderm as in larval development (Masterman, 1898). This layer in Pterobranchs, behind the proboscis region at any rate, is relatively unspecialized.

In Tunicates, on the other hand, the ectoderm of the bodywall is everywhere specialized for the production of test material, and is thus rendered incapable of playing a formative part (Hjort, 1896). We have thus to account for the entrance into the stolon of special pharyngeal and cloacal derivatives which could take over the development of the internal organs. It seems natural to connect this problem with the phenomena of periodic degeneration and recrescence which are widespread among small plankton-feeding organisms, and well known in Ascidians themselves (see especially Caullery, 1895). The whole body may degenerate and leave only a dormant residue of undifferentiated material; or, as in D i a z o n a (Della Valle, 1884; Korschelt and Heider, 1893, p. 1360), the thoracic region only may atrophy, leaving the intestinal loop and heart intact (cf. Aphanibranchion, Oka, 1906, and the remarks of Seeliger and Hartmeyer on it in Bronn, pp. 1229, fig. 229). Caullery has described very similar phenomena in Didemnidae, degradation of the thorax, followed by regeneration, being a constant feature of Diplosoma gelatinosum, while the abdomen remains usually unaffected (1. c., pp. 18, 59, 60). It is in this latter type of thoracic atrophy and regeneration that I think we find the key to our problem. Caullery has shown the intimate relationships between budding and regeneration in Ascidians, and that the epicardium is as much an organ of partial regeneration as it is of complete budding both in Polyclinidae, by experiment (1. c., p. 118), and in Didemnidae naturally (p. 133). In these cases, however, the epicardium is already established as ‘le tissu régénérateur’, and I wish to consider the facts in relation to an antecedent condition, before the epicardium had become an organ capable of regenerating everything. It is likely that its powers of regeneration were more restricted at the outset.

These remarks with regard to the epicardium are equally applicable to the peribranchial epithelium in the contrasted group (Periblastica). In P o 1 y c a r p a, at any rate, it is known that after spontaneous evisceration, which occurs in various genera, the whole alimentary canal is regenerated by folds of the peribranchial wall (De Selys Longchamps, 1915). It is obvious that the history of regeneration in Tunicates has run a course parallel to that taken by budding, so that in the end we are confronted with two alternative modes of wellnigh total regeneration, viz. in Endoblastica from the epicardium, in Periblastica from the atrium. Clearly the common ancestor of these two groups could not have regenerated in either of these exclusive ways: it must have combined both processes, each with a more restricted scope. The primitive mode of regeneration can therefore have been little more than one of specific tissue-repair, gut repairing gut, and atrium repairing atrium, exactly as in Asymmetron according to Andrews: ‘Here, as in the tadpole or salamander, there seems to be a regeneration of each tissue to form new tissue of its own kind’ (1893, p. 27. Cf. Franz on Amphioxus, 1925, p. 438).

Let us imagine the primitive Tunicate as a little sac-shaped organism fixed by a short ventral stalk to a creeping stolon, like a creature intermediate between Rhabdopleura and Cep halodiscus, but with a complete gelatinous cuticle and the general organization of a young C i o n a. After a period of feeding activity, during which the stalk became filled with reserve food-stuffs, the pharyngeal region died down, leaving only a few patches of undifferentiated material at the base of the former pharynx and atrial cavities. It is well known that these are the sites of active growth in an Ascidian, so that it is precisely here that undifferentiated material is normally concentrated—on either side of the retro-pharyngeal tract in the pharynx, and at the base of each atrium. To meet this periodic phase of atrophy and renewal, short diverticula or pockets from these regions would be a natural provision as in other cases, and would enable the process of regeneration to begin without delay. Judging from Oka’s figs. 4 and 6 (1906), it seems probable that in Aphanibranchion (=Diazona?) atrium and pharynx are still regenerated from independent rudiments, although explicit statements on this point are lacking.

In Clavelina, before the process of degeneration has extended into the abdomen, there is a well-marked stage when the atrium is reduced to a pair of small vesicles alongside a pair of ‘epicardial horns’ on either side of the degenerating pharynx. If we imagine the stolonial extension of the two epicardia (i. e. the endophragm) to be absent, this condition is strikingly suggestive of the two pairs of regenerative pockets which I have postulated as an early provision for purely thoracic regeneration (see J. S. Huxley, 1926, p. 12, Pl. Ill, fig. 5 b).

From such a stock of fixed, budding, and regenerating Prototunicates the Thaliacea would be derivable, possibly through ancestors that, like many Graptolites, were fixed to floating objects, thus facilitating the transition to a free pelagic life. The normal atrophy of the pharynx in old Doliolid nurses and Appendicularians may well be a relic of the periodic degeneration characteristic of their sessile predecessors, the original sequel of rejuvenation having been biologically replaced in Doliolum by an intensification of the budding process, and in Appendicularians by precocity of sexual reproduction.

A provision of regenerative pockets at the junction of thorax and abdomen would yield all the elements required for the complex Thaliacean stolon. The pairs of pharyngeal and peribranchial diverticula, originally developed for use in simple regeneration, could readily extend downwards into the stalk, and thus meet the needs of the stolon, as the ectoderm became too specialized to play its original major part in the budding process.

In support of this suggestion, I need only point to the salient facts in the history of the Thaliacean stolon. In Pyrosoma, which admittedly retains the most primitive condition, the budding process is entirely based on the co-operation of specific organ-regenerators: * Alle Stolostrange … sind abgeschniirte Fortsatze der entsprechenden Organe des Muttertieres … um als “Strange”in den Stolo einzutreten und wieder die entsprechenden Organe der neuen Knospe zu bilden’ (Neumann in Bronn, 1913, p. 150). In Salpa, although certainty is out of the question, it is probable that the specificity of some of the stolonial strands has broken down: the endodermal tube is derived from the pharynx and reproduces a pharynx; but, if Korotneff is right, it also proliferates the peribranchial tubes, and has thus enlarged its primitive scope, like the epicardium of Endoblastica (Korschelt and Heider, 1910, p. 780). Finally in Doliolum, if one dare refer to so controversial a type, it would seem that all remnants of original specificity have disappeared: the epicardial tubes, no longer united, give rise apparently to the reproductive organs, while the cloacal tubes, on the other hand, double inwards on themselves to form secondary strands which unite to constitute the pharynx (Neumann; ci. Korschelt and Heider, 1. c., p. 809). Thus in Thaliacea the constituents of the stolon may be regarded at the outset as special regenerators of the particular organs from which they arose, but in the course of evolution they have acquired formative powers having little or no relation to their morphological origin. This means that they have become indifferent, or generalized, budding organs, controlled, we know not how, by epigenetic factors. The ultimate result of such a change should be a simplification of machinery by a reduction of elements, but Salpa and Doliolum show no numerical reduction. Nevertheless a qualitative, if not a quantitative, change in this direction is recognizable even in them, since certain of the constituent strands appear to have taken over a preponderating role in the budding process. It can hardly be without significance (if the facts have been correctly determined) that in one of these cases the Endoblastic element, and in the other the Periblastic, shows signs of predominance. Apparently pelagic life has not favoured any great development of these tendencies to concentration, for the full expression of which certain conditions of fixation seem to have been required, as will appear later.

From the same Prototunicate stock the Endoblastic Ascidians would be derivable by retention of the median stalk of fixation as the apex of the stolon, and of the pharyngeal diverticula (epicardia) which ran into it, but the lateral or peribranchial diverticula were dispensed with. Originally, as in the larva, the intestinal loop would lie behind the stalk of fixation, and somewhat dorsally, as in Pterobranchia, so that the stalk itself and the pharyngeal diverticula into it may have been quite short. A remnant of this condition persists in such short-bodied forms as Perophora; but in general those Phlebobranchiate Ascidians which have dispensed with budding tend to lie over on their right (Cor el la) or left side (Ascidia), which gives them a broader area of attachment, and facilitates an extension of the pharynx behind the intestinal loop altogether. On the other hand, the Haplobranchiate Ascidians, which are plainly derivable from a primitive Cionid stock, but have developed, rather than reduced, their early powers of budding, are all vertically elongated, and the intestinal loop has descended into the stalk of fixation in a manner very similar to that exemplified in Phoronis, though less abruptly. This descent of the gut has converted the proximal part of the stolon into a secondary abdominal cavity, which has to be traversed by the epicardia—or the endophragm resulting from their fusion—before they can reach the base of attachment where budding actually proceeds; and the extension of the epicardia has naturally been utilized for the more efficient separation of afferent and efferent bloodchannels both in the stolon itself and in the abdominal cavity formed from its proximal portion. Such a use involves no specialization of its tissue.

Thus on the assumption of a primitive relation of stolon to body similar to that in Pterobranchia, every stage in the further evolution of parts characteristic of Endoblastic Tunicates can be satisfactorily accounted for, both morphologically and physiologically, whether budding has been retained as in the Haplobranchs or discarded in favour of greater individual development as in the Phlebobranchs. In the latter section the characteristic ‘test-vessels’ of C i o n a with their dividing septum represent simply an adaptation to somatic ends of the former stolon with its epicardial axis, while the ‘double-vessels’ of Ph alius i a and other simple Ascidians (Seeliger in Bronn, pp. 246, 552, 575), which have a similar origin and distribution, can only consistently be regarded as the same vessels simplified by failure of the epicardium to develop, and readapted for the double circulation by longitudinal constriction of their walls. That the epicardial septum should have replaced such an arrangement is unthinkable.

In contrast to the Endoblastica, the Periblastica, though derivable from the same original stock, have undergone a secondary change in the conditions of fixation which has profoundly affected their processes of growth, budding, and general symmetry of organization. It may be roughly described as a change from the centripetal concentration of organs to their centrifugal dispersal, and is typically associated with a flattening of the body instead of its vertical elongation. The most primitive stock on the whole consists of the Botryllids which still retain the original median group of larval suckers, but these are combined with a peripheral ring of ectodermal papillae, which in the remaining groups tend to replace the original suckers altogether. The result of this arrangement is that on fixation the original median area of attachment, which forms the extremity of the stolon in Endoblastic Ascidians, is enclosed and completely buried within the larger disc of adhesion outlined by the peripheral ring of papillae, so that budding from a median stolon is effectively precluded. (Excellent figures of this condition for Botryllus and Styelopsis in Damas, 1904, Pl. xxi.) It is natural, therefore, that with the loss of the median stolon should go also the loss of the inner or pharyngeal pair of bud-diverticula, while the outer or peribranchial pair are retained in connexion with the lateral or parietal budding, that has been substituted for median budding.

It is remarkable that a ring of ectodermal papillae surrounding the primitive stalk of fixation is also found in the larvae of many Didemnids (Lahille, 1. c., fig. 64; Oka and Willey,’ Quart. Journ. Mier. Sci.’, xxxiii, fig. 7), and doubtless serves the same function in widening the base of fixation. They are not necessarily homologous in the strict sense, because in each case they represent simply a precocious development of ectodermal prolongations characteristic of the adult phase, and mainly subservient to test-production. But it is interesting to note that the absence of a median stolon in Didemnids is explicable as a consequence of the same larval phenomenon as in Periblastica, and may well have been the chief factor in bringing about the remarkable modification of epicardial budding which is a distinguishing feature of this family. If there are any to whom this mixture of budding and regeneration seems to present an incipient stage in the development of budding from regeneration, they should refer to Brien’s illuminating study of the budding (abdominal strobilization) of Aplidium (1925), which is shown to form ‘le pont réunissant la blastogénése des Aplidiens á celle des Didemnidae’, and carries the story back to the post-abdominal (stolonial) fission of Polyclinidae, in which, as in the stolonial budding of Clavelina, the epicardium bears the entire burden of organ-formation.

The differences between endoblastic and periblastic budding are undoubtedly substantial, and may seem at first sight to imply complete independence of origin, or at least to involve a sharp mutation. In the preceding sketch, though necessarily condensed, I have tried to show that a complete continuity of budding may have been preserved through all the main lines of Tunicate evolution, even from pre-Tunicate ancestors, in spite of great changes in the actual mode of budding.. In this connexion we have to remember the extraordinary adaptability of the budding process, as of regeneration itself. It is difficult enough to trace the evolution of a particular organ, and impossible fully to understand it, without considering it in relation with the evolution of the body generally. This difficulty is increased tenfold in the case of the budding process, and renders it essential that it should be considered in connexion with the general phyletic relations of the groups involved. As Caullery has put it: ‘Le bourgeon porte au plus haut point l’empreinte d’une épigénèse, c’est-à-dire d’une évolution dépendant avant tout des conditions environnantes’ (1. c., p. 137). Hence we find great differences in the mode of budding even in genera so closely related as Clavelina and Distaplia, although in this case there can be no question as to the continuity of budding itself. In Clavelina there is a median stolon traversed by a single epicardial septum, and the buds arise leisurely after fixation as outgrowths from it. In Distaplia there is neither stolon nor septum, but a pair of minute epicardial diverticula, from one of which alone a vesicle is constricted, which clothes itself with ectoderm as it migrates through the parental skin, and divides into fragments, each of which develops into a complete zooid. This is not parietal budding, but it is a step in that direction, for the median stolon has gone, and the source of the bud is distinctly lateral. What are the ‘environing conditions’ that have provoked this change ? Simply the formation of a solid colony, instead of a diffuse one, and the provision of yolky eggs. So we get precocity of bud formation in the larva instead of in the adult; early detachment under cover of the parental test; and finally fragmentation from a single epicardium before the latter has fused with its fellow.

Again, none can question the continuity of budding in the Thaliacea, although there are considerable differences between Pyrosoma, Anchinia, Doliolum, and Salp a in the constitution of their stolon and the developmental processes in the buds.

With these examples before us I submit that the peculiarities of budding in the Periblastica are perfectly consistent with the descent of this group from early Tunicates still provided with a ventral median stolon, the change in their mode of budding being adaptive to recognizable changes in their mode of fixation and orientation. That Molgulids retain modified vestiges of the original paired epicardia has already been mentioned (Damas, 1902).

It results from this study that epicardial and peribranchial outgrowths subservient to budding have probably arisen within the Tunicate group itself and are derivable from corresponding pockets of undifferentiated material which originally served as regenerative centres for the renewal of pharynx and atrium after degeneration. Assuming that the ventral stalk of fixation was already in existence in the ancestors of Tunicates, and was the site of budding processes after fixation, it became transformed into the stolo prolifer of Tunicata by the extension into it of prolongations from the regenerative pockets just mentioned. The earliest type was probably ‘amphiblastic’ with paired diverticula both from pharynx and atrium, a type which has survived only in the pelagic Thaliacea, but with loss of the terminal suckers of fixation. In the Ascidiacea, however, the diverticula were soon reduced to one pair or the other, the Endoblastica retaining the pharyngeal outgrowths (as ‘epicardia’) in conjunction with terminal suckers and the adoption of an upright attitude with extended intestinal loop (secondarily lost in Phlebobranchia which discarded budding altogether), while the Periblastica retained only the peribranchial psckets for lateral budding in adaptation to the lateral extension of a compact body, with widened base of fixation, and more or less horizontal attitude. Budding once established from the posterior peribranchial diverticula, there was of course no obstacle to its extension over the whole parietal wall of the peribranchial cavity, which undergoes no special differentiation.

In many types, both of Endoblastica and Periblastica, budding has disappeared altogether, the loss of budding bringing in its train increase of size, longer duration of life, and greater complexity of pharyngeal structure (Ascidiidae, Molgulidae, Cynthiidae); but the fact that the families of large solitary Ascidians are each related intimately to small, budding forms of more primitive structure, points strongly to the absence of budding in those cases as secondary, e. g. Rhopalaea leading to Ciona, Perophora to Ascidia, Botryllus to the Molgulids, and Polystyelids to Cynthiids. On the current view, according to which the solitary Ascidians are primitive, and budding is regarded as a secondary acquisition, the existence of such structures as the endodermal septum of the testvessels in Ciona remains completely unintelligible.

In conclusion, it is desirable to see how far this independent study of budding in Tunicates accords with the conclusions already reached with regard to Amphioxus. We have seen that in the character of its U-shaped gill-slits and endostyle, as well as in its protonephridia, it retains more primitive features than any Tunicate, but that in its elongated form, multiplication of gill-slits, peculiar atrium, and lack of fixation and budding, it has deviated from the primitive Protochordate stock, which in these respects resembled the Tunicata. The common ancestor was essentially a small stalked Ascidian with the pharynx of a young, but symmetrical Amphioxus having only two to three pairs of U-shaped gill-slits. To these characters we now add that the stalk of fixation arose ventrally behind the endostyle, and at first contained mesoderm only, as in Pterobranchia. Subsequently, but before the origin of any existing type of Tunicate, there ran into it also two pairs of regenerative diverticula from the pharynx and lateral atria respectively. It is to this stage in the phylogeny that we refer the origin of Amphioxus, of which it may be regarded as the paedomorphic representative. It is therefore quite consistent to find that Amphioxus possesses a vestige of the epicardial diverticula as previously suggested, in its club-shaped gland. What is more remarkable, perhaps, and certainly unexpected, is to find that Lankester’s pair of ‘atrio-coelomic funnels’ arise from the atrium in exactly the spot where the pair of peribranchial diverticula into the stolon ought to arise if they had persisted, viz. alongside the hinder extremity of the pharynx where the ‘base’ of the lateral atria would be in a primitive sessile Tunicate ! This coincidence is all the more striking because in Amphioxus the atrium extends far behind this primitive limit, like the perivisceral extensions in a Botryllus, so that one can hardly avoid the conclusion that the coincidence is significant. Amphioxus in that case possesses homologues of the four principal constituents of the Thaliacean stolon, and incidentally corroborates the theory of the primitive nature of that type of stolonial constitution. The extreme dislocation between the pharyngeal and the peribranchial pair, with the whole length of the pharynx between them, is readily explained on the one hand by the precocious specialization of the pharyngeal pair and its fixation anteriorly in the larva as the clubshaped gland, and on the other by the retarded development of the atrium. There is, of course, nothing but circumstantial evidence to support the homology suggested for the atrial funnels, but, in the absence of any evidence of a function for these structures, it is permissible to regard them provisionally as mere vestiges. It should be added that Lankester’s expectation of the existence of a coelomic pore at the apex of the funnels has been definitely negatived by Franz (1925, p. 449) after special investigation.

Since this essay was sent to press two recent papers by Van Wijhe (1926 and 1928) have come to my notice which bear closely on some of the problems I have been discussing. Van Wijhe has observed that the larva of Amphioxus, during its asymmetrical growth-period, is capable of temporary fixation, as previously noted by Orton (1914, V); but, whereas Orton considered the means of attachment to be the secretion of the club-shaped gland, Van Wijhe traces it to a pair of glandular patches near the anterior ends of the future ‘atrial folds’, and identifies them, together with the unpaired ‘mandibular papilla’ in front of the mouth, as homologues of the three adhesive papillae of Ascidian tadpoles. If this comparison can be sustained, it will obviously add support to the views here maintained, since the papillae in any particular Ascidian are all similarly constituted, while in Amphioxus their supposed homologues are distinctly heteromorphic, and the posterior pair become drawn out into the long ‘tape-like.’ stripes of ‘glassy’ cells which have been so often figured (and variously interpreted) along the edges of the metapleural folds (Lankester and Willey, 1890, p. 466, figs. 1, 2, 7, 13, 14; MacBride, 1900, p. 356, figs. 9, 10; Gibson, 1910, p. 228, figs. 8-12).

Nevertheless, it cannot be assumed without further investigation that the arrangement of these papillae in a triangle, with one in front and two behind, is either typical or primitive in Ascidians. In Botryllus and Cor ell a, apparently also in Distaplia and Clavelina, the condition is just the reverse (viz. a pair of papillae in front and a median papilla behind), and in Amaraecium all three papillae lie consecutively in a median row. The innervation of the papillae in Botryllus rather indicates that, morphologically, one left-sided papilla is balanced by two on the right side—a condition also suggested by Kupffer’s figure of the larva of C i o n a. As this condition is completely reversed in Amphioxus, it will plainly be no easy matter to establish individual homologies between them.

The relation of the posterior ‘glandular’ patches to the ‘pterygial folds’, however, has already led Van Wijhe to assert the complete independence of the atrial cavities of Amphioxus and the Tunicata. A great deal depends therefore on the cogency of Van Wijhe’s evidence, of which a fully illustrated account is promised.

The further discovery is announced by Van Wijhe (1926) of a transitory funnel-like communication between the pre-oral (buccal) cavity and the anterior extremity of the pharynx, and this duct is now substituted for Hatschek’s pit as the ‘autostome’ or primitive mouth. From the brief details given in the preliminary note, it is impossible to assess the significance of this new aperture. From the fact that it arises late and is said to be ‘filled with a slimy mass … most likely derived from the pre-oral organ’, the duct may be merely the result of a closure of the ‘mandibular groove’ between Hatschek’s pit and the mouth—in which case it would be a secondary passage, comparable to a nasal duct, rather than the relic of a ‘primitive mouth’. But it may be recalled that Goodrich (1909, p. 197) has already reported the existence of a communication in one of his larvae between the buccal cavity and the duct of Hatschek’s nephridium (indirectly therefore with the pharynx), and it will be of great interest to learn whether this new duct has any relation to that isolated organ.

In his latest paper (1928, p. 997) Van Wijhe has anticipated me in drawing attention to the contrast between Appendicularians and other Tunicates in regard to the twist of the intestinal loop, but by overlooking the exceptional cases of Doliolum and Anchinia he has been unfortunately led into an untenable hypothesis. Regarding the mouth and club-shaped gland as the first pair of gill-slits in Amphioxus, and homologous with the pair of spiracles in Appendicularians, he now suggests that the right intestine of Appendicularians and the left intestine of other Tunicates together represent the second pair of gill-slits in Amphioxus, of which the left can be regarded as more potent than the right! The existence of D o 1 i o 1 u m, however, with its median gut and anus, destroys the basis of this curious proposition, since it prohibits the ‘sharp division’ of Tunicates into ‘Dextricolica’ and ‘Laevicolica’ on which Van Wijhe was relying. I may add that, whatever may be thought of the eccentric mouth of the larval Amphioxus, there is really no adequate reason for disputing the homology of the Tunicate intestine with that of Amphioxus. The blastopore is closed over alike in the embryos of both types, and there is a secondary anal outgrowth in Amphioxus as in Ascidians, so that the difference between them in this respect is only one of degree. If a gill-slit were needed to make an outlet for one, it would be just as necessary for the other. The presence of ‘subchordal’ cells in the Ascidian tail does not affect this question if I am right in my conclusion that this region is represented in Amphioxus only by the greatly reduced larval fin (p. 155).