198 DU. W. N. F. WOODLAND ON THE 



intrn.eelliilai' cliannels liave been so tliorouglily studied hj the 

 anthers just named, I shall not refer to them in any detail. I 

 will merely say that since, accoi-ding to Jaeger (47), gas glands 

 are able, when active, to swell to three times their normal size, 

 and that the finer ducts only become visible when filled with 

 fiuid, it is not surprising that they are only visible in very few 

 preparations. 



One or two other histological features to be observed in serial 

 sections of the gas gland and associated structures of Gobius niger 

 and other types remain to be mentioned. In transvei'se sections 

 across the anterior end of the rete mirabile, and only at its 

 anterior end, where the arteries are easily distinguishable from 

 the veins, all the ai'teries possess a peculiar endothelium. This 

 endothelium (PI. TV. fig. 32) contains many more nuclei than usual, 

 and these, instead of being flattened in the usual way, are more or 

 less spherical, and, each being contained in a small mass of 

 cytoplasm, project into the vascular lumen in the manner shown 

 in the figure. This peculiar endothelium, in the case of the 

 smaller arterial capillaries, results in such a thickening of the 

 wall (and incidentally in some of the smallest capillaries a blocking- 

 up of the lumen) as to cause the capillary to bear a strong 

 resemblance to a bile- or pancreatic duct ; indeed, were it not for 

 the presence of blood corpuscles these small capillaries would be 

 almost unrecognisable as such. This type of vascular epithelium, 

 which must be well-known to histologists, in all probability results 

 from the contraction of the pulsatile arteries in forcing the blood 

 through the rete, the narrowed circumference of the endothelium 

 causing the cells both to assume a globular form and to protrude 

 into the lumen of the vessel. 



Concerning the posterior extension of the difi"use pancreatic 

 acini among the arteries and veins anterior to the formation of 

 the rete mirabile (fig. 25, e. g.) there is little to note other tha,n 

 the fact itself, which I have not seen recorded by writers on the 

 teleost pancreas (see list of references to j^ancreas literature 

 below). When first studying teleost "red bodies" I was misled 

 into supposing that this extraordinary extension of the pancreatic 

 acini through and right into the bladder wall in many teleost 

 genera rejDresented a new gland specially developed in connection 

 with the rete mirabile. I suggested (78) that the purpose of this 

 supposed new gland was the abstraction from the venous blood 

 stream of the globules and granules resulting from the breaking- 

 up of the red blood corpuscles referred to above, and in my sections 

 (stained by the picro-indigo-carmine method described in Ap- 

 pendix A) the zymogen granules present in the pancreas cells 

 often strongly resemble erythrocyte globules which have been ab- 

 stracted from the blood stream [cf. text-fig. 56 and PI. III. fig. 21). 

 Adopting the view of Jaeger (fully discussed in Part II.) that 

 the cells of the gas gland produce a toxin for the purpose of 

 breaking up the erythi'ocytes and so enable themselves to absorb 

 more easily the oxygen in the blood stream which they sub- 

 sequently pump into the gas bladder, I suggested that the purpose 



