123. If apolar cells are embryonic forms of cells which afterwards become multipolar, this interpretation will not suffice for the unipolar cells. They are not only abundant, but are mature forms in some organs, and in some animals; though in some organs they may truly be regarded as embryonic. Thus in the human embryo up to the fourth month all the cells of the spinal cord are said to be unipolar,[144] later on they become multipolar. But in birds, rabbits, dogs, and even man, the cells in the spinal ganglia are mainly (if not wholly) unipolar;[145] nor is there any difficulty in observing the same fact in the œsophageal ganglia of molluscs (see Fig. 22). Such are the observations. They have indeed been forced into agreement with the bipolar postulate, by the assumption that the single process branches into two, one afferent, the other efferent.[146] But before making observation thus pliant to suit hypothesis, it would be well to look more closely into the evidence for the hypothesis itself. For my own part, I fail to see the justification of the postulate; whereas the existence of unipolar cells is an observation which has been amply verified. [Illustration: Fig. 18.--_Supposed union of two nerve-cells and a fibre._ The processes subdivide into a minute network, in which the fibre also loses itself.] 124. Bipolar cells abound; multipolar cells are still more abundant; and these are the cells found in the gray substance of the neural axis. Deiters, in his epoch-making work,[147] propounded an hypothetic _schema_ which has been widely accepted. Finding that the large cells in the anterior horn of the spinal cord gave off processes of different kinds, one branched, the other unbranched, he held that the latter process was the origin of the axis cylinder of a nerve-fibre, whereas the branched process was protoplasm which divided and subdivided, and formed the connection between one cell and another. Gerlach has modified this by supposing that the minute fibrils of the branching process reunite and form an axis cylinder (Fig. 18). There is no doubt that some processes terminate in a fine network; and there is a probability (not more) that the unbranched process is always continuous with the axis cylinder of a motor nerve, as we know it sometimes is with that of a dark-bordered fibre in the white substances. This, though probable, is, however, very far from having been demonstrated. Once or twice Kölliker, Max Schultze, and Gerlach have followed this unbranched process as far as the _root_ of a motor nerve; and they infer that although it could not be traced further, yet it did really join an axis cylinder there. In support Of this inference came the observations of Koschennikoff,[148] that in the cerebrum and cerebellum, processes were twice seen continuous with dark-bordered nerve-fibres. But the extreme rarity of such observations amid thousands of cells is itself a ground for hesitation