From this sketch it follows that the arrangement seen in the spinal cord, would result from the cranial arrangement if this third system of lateral roots were left out. Further, since the cranial system is the oldest, we must search in the invertebrate ancestor for a tripartite rather than a dual system of nerve-roots for each segment; a system composed of a dorsal root supplying only the sensory nerves of the skin-surfaces, a lateral mixed root supplying the system connected with respiration with both sensory and motor fibres, and a ventral root supplying the motor nerves to the body-musculature. COMPARISON OF THE APPENDAGE NERVES OF LIMULUS AND BRANCHIPUS TO THE LATERAL ROOT SYSTEM OF THE VERTEBRATE. If the argument used so far is correct, and this tripartite system of nerve-roots, as seen in the cranial nerves of the vertebrate, really represents the original scheme of innervation in the palæostracan ancestor, then it follows that each segment of Limulus ought to be supplied by three nerves--(1), a sensory nerve supplying its own portion of the skin-surface of the prosomatic and mesosomatic carapaces; (2), a lateral mixed nerve supplying exclusively the appendage of the segment, for the appendages carry the respiratory organs; and (3), a motor nerve supplying the body-muscles of the segment. It is a striking fact that Milne-Edwards describes the nerve-roots in exactly this manner. The great characteristic of the nerve-roots {158}in Limulus as in other arthropods is the large appendage-nerve, which is always a mixed nerve; in addition, there is a system of sensory nerves to the prosomatic and mesosomatic carapaces, called by him the epimeral nerves, which are purely sensory, and a third set of roots which are motor to the body-muscles, and possibly also sensory to the ventral surface between the appendages. Moreover, just as in the vertebrate central nervous system the centres of origin of the motor nerves of the branchial segmentation are distinct from those of the somatic segmentation, so we find, from the researches of Hardy, that a similar well-marked separation exists between the centres of origin of the motor nerves of the appendages and those of the somatic muscles in the central nervous system of Branchipus and Astacus. In the first place, he points out that the nervous system of Branchipus is of a very primitive arthropod type; that it is, in fact, as good an example of an ancient type as we are likely to find in the present day; a matter of some importance in connection with my argument, since the arthropod ancestor of the vertebrate, such as I am deducing from the study of Ammocoetes, must undoubtedly have been of an ancient type, more nearly connected with the strange forms of the trilobite era than with the crabs and spiders of the present day. His conclusions with respect to Branchipus may be tabulated as follows:-- 1. Each ganglion of the ventral chain is formed mainly for the innervation of the appendages.