—simplification or elaboration—development is taking place. For instance, it can hardly
be doubted that the branchial pocket of the Cryptobranchiatse is a specialization, and that
forms with the pocket are as a whole developed from those with none, though this need not
mean that existing forms with a pocket are derived from existing forms with none. Yet
change takes place in the inverse order. The Doridopsidas are cryptobranchiate as a
group, though the branchial pocket is often shallow, and Dondopsis rubra is norïnally
cryptobranchiate as a speciés, but in individuals the pocket atrophies and the branchiae
become completely exposed and non-retractile.1 The rhinophore pockets also atrophy in
this species and they disappear in a few Polyceridae and in all the Goniodorididas. This
last group seems to be in no sense primitive but rather an instance of extreme specialization.
Yet it rejects the specialized retractile rhinophores common in the Holohepatica
and converts them into simpler organs.
Again the structure of all other Gastropoda shows us that the ramified liver of the
Cladohepatica must be regarded as a specialization, but in Pseudovefmis paradoxus8 a
cladohepatic liver has become holohepatic, for in view of the animal’s general structure it
is hardly possible to regard it as an annectent type nearly related to the Holohepatica.
In the ascoglossan genera, Lobiancoia and Gyerce, there is no compact liver-mass but the
hepatic diverticula do not penetrate into the dorsal papillae, and, as the animals are in
most respects highly specialized, it seems probable that the diverticula have atrophied.
Bergh’s theory that the Holohepatica as a whole are derived from the Cladohepatica by a
gradual concentration of the liver is not likely to commend itself as a general explanation,
but such secondary concentration may have occasionally happened, as in the instances
mentioned. Although the highly specialized arrangement of the liver found, e. g., in
-ZEoli&s diverges widely from the normal gastropod type, yet the-same cannot be said of a
liver which lies within the body-cavity but consists of two or three portions (e. g. in
Dirona and such Dendronotoidea as have no hepatic diverticula in the cerata). For
the commonest form of liver in the Gastropoda and indeed in the Mollusca generally
is bilobed, and in the Fissurellidee, which are usually, considered an archaic family, there
are three hepatic ducts. In so far as the liver of the Holohepatica is an undivided mass
and opens into the stomach by a single duct it must be regarded as specialized. But the
various genera of this tribe show considerable variety both as to the traces of division
which the liver shows while remaining a solid mass and as to the number of ducts by
which it communicates with the stomach. Similarly the general structure of the
Gastropoda indicates that cerata are an instance of extreme specialization confined to a
particular class, but on the other hand some facts suggest that these papillae sometimes
tend to become effaced, so that forms without them may be derived from forms possessing
them. Thus the young Lomanotus eisigii has the appearance of an ^Eolid, but as it
becomes older a membrane grows up between the bases of the papillae and eventually they
1 The phenomenon has been noted in living animals as well as in preserved specimens. See
Eliot, in Linn. Soc. Journ. Zool., vol. xxxi, 1908, pp. 118—119. For a possibly similar instance in
Doris see Eliot in Proc. Malac. Soc., vol. vii, 1907, p. 357, on Btav/rodoris falklandica.
2 Kowalevsky, in Mém. Acad. Pétersbourg (8) Phys. Math., xii, 1901. I mast confess that I
cannot rid my mind of the suspicion that Ps. paradoxus is a larval form with abnormal sexual
activity. Its relationship to Ps. papillifera (sic) is very remarkable and just what might be expected
between two stages of a developing larva.
become little more than points on a lateral expansion of the back. So, too, in the young
Grosslandia1 the wings are bifid and subdivided but in the adult they become an undivided
and roughly triangular flap. Hence account must be taken of the possibility that Cladohepatic
forms without cerata have lost their cerata. All these instances (rhinophores,
branchige, hepatic diverticula, cerata) illustrate the same principle as is seen in the loss of
the shell. A special organ or form of an organ is developed; then variation begins in
another direction and the peculiarities acquired by a group as a whole are lost by its
more specialized members.
In considering the relationships of the Nudibranchiata, it must be remembered that we
have no fossil forms and that none are likely to be found. In the Cephalopods about 400
living and 8000 fossil species are known. If the Nudibranchs have become extinct at about
the same rate at least 20,000 species must have disappeared, and any attempt to relate
existing genera to one another must be made with the knowledge that they probably
represent a few specialized forms of the extinct host together with some chance survivals.
The principal recent authorities for the classification and phylogeny of the Nudibranchs
are Bergh and Pelseneer. The system of the latter, as already indicated, ignores the
difficulties presented by many* genera, and the phylogeny corresponding to it is open to
the same objection. It is not sufficient to represent Jarms and Pleurophyllidia as side
branches springing up together with the Solids from a parent stock represented by
Dendronotus. The perplexing thing is that these two genera share some of their characters
with the iEolids, and the rest with various members of the Holohepatica. But for all
that they cannot be regarded as unspecialized or annectent forms connecting the two
groups. Bergh, who had a longer and more varied acquaintance with the Opistho-
branchiata than any other zoologist, felt the difficulties of making any phylogeny, but the
suggestions which he offered are not very clear or convincing. He considered that the
Cladohepatica are connected with the Bullidge and the Aplysiidae through the Ascoglossa,
and that the Holohepatica are connected with the Bullidge through the Pleurobranchidse.2
This appears to be meant merely as a statement of resemblances, and not as an opinion that
the origin of the Nudibranchiata is polyphyletic, for he also considered that the JEolids
were derived from the Ascoglossa and passed into the TritoniidcB and Dorids by a gradual
reduction of their hepatic ramifications. But it is hard to believe that a highly specialized
group (the iEolids) can give rise to a comparatively generalized group (the Tritoniidaa),
and this again give rise to another and quite different highly specialized group. We have
not and are never likely to have sufficient materials for delineating the family tree of the
Nudibranchs, since we have no real knowledge of the trunk, still less of the root. The
families which we know represent (to continue the metaphor) little more than the topmost
boughs of a submerged forest which stand up above the water. Our data enable us to
discuss profitably whether a clump of boughs belongs to one submerged trunk or to more
than one, but only in the latest and highest branches can the points of origin and
division be seen.
• The line of division between the Holo- and Cladohepatica is undoubtedly to be found
near.the Tritoniidge. This family is clearly annectent and shows how the two main types
1 See Eliot in Liun. Soc.-’s Journ., vol. xxx, 1908.
2 Bergh 44, p. 996.