link on the Annelids, as the most
primitive group of the three, to the unsegmented worms, and particularly
to the Turbellaria. His speculations on this matter may be summed up
somewhat as follows:--The common ancestor of all segmented animals is a
segmented worm-like form, not quite like any existing type, resembling
the Turbellaria in having two nerve strands on the dorsal side and no
oesophageal ring, potentially able to develop either the Vertebrate or
the Annelid mouth, and so to give origin both to the Articulate and to
the Vertebrate series. The common ancestor alike of unsegmented worms
and of all segmented types is probably the trochosphere larva, which in
the Vertebrates is represented by the simple _Keimblase_ or blastula.

The Annelid theory of Dohrn and Semper was perhaps not so widely
accepted as the rival Ascidian theory, but it counted not a few
adherents and gave a certain stimulus to comparative morphology. F. M.
Balfour, who pointed out about the same time as Semper the analogy
between the nephridia of Annelids and the mesonephric tubules of
Vertebrates,[405] while not accepting the actual theories of Dohrn and
Semper, took up a distinctly favourable attitude to the general idea
that Annelids and Vertebrates were descended from a common segmented
ancestor. Discussing this question in his classical work on the
development of Elasmobranch fishes,[406] Balfour came to the conclusion
"that we must look for the ancestors of the Chordata, not in allies of
the present Chaetopoda, but in a stock of segmented forms descended from
the same unsegmented types as the Chaetopoda, but in which two lateral
nerve-cords, like those of Nemertines, coalesced dorsally instead of
ventrally to form a median nervous cord. This group of forms, if my
suggestion as to their existence is well founded, appears now to have
perished."[407]

He held that while there was much to be said for the interchange of
dorsal and ventral surfaces postulated by Dohrn and Semper, the
difficulties involved in the supposition were too great; he preferred,
therefore, to assume that the present Vertebrate mouth was primitive,
and not a secondary formation.

His views as to the phylogeny of the Chordata and the genetic relation
of the various classes to one another are exhibited in the following
schema,[408] names of hypothetical groups being printed in capitals, names
of degenerate groups in italics:--

Mammalia. Sauropsida.