GAS GLANDS OF SOME TELEOSTEAN FISHES. 



199 



of the rete mirabile — a structure essentially consisting of the 

 intimate juxtaposition of the veins returning the blood from the 

 gas gland and carrying the hypothetical toxin, with the arteries 

 carrj'ing freshly-oxygenated blood to the gas gland — was to allow 

 the toxin in its venous capillaries to diffuse into its arterial 

 capillaries and so to disintegrate the oxygen-laden erythrocytes in 

 time for the oxygen to be available for abstraction by the gas 

 glaiid cells by the time the arterial blood reached the gas gland. 

 On this view the veins of the rete mirabile would be laden with 

 granular matter (as indeed they are — see fig. 32) which requires 

 to be eliminated, and I suggested that this pi'ocess of elimination 

 was the function of these pancreatic masses which are so closely 

 connected with the veins at the anterior end of the rete — the kind 



Text-fig. 56 ( X circ. 470). 



Vein suvroumled by modified acini (tlivee shown) of the pancreas 

 in Nerophis eequorius. 



of vessels and the identical position that the hypothesis would 

 suggest ; in other words, the hypothesis was supported by the 

 position of the gland, the modification of the pancreatic acini 

 surrounding the veins (described below), the great similarity 

 between the eryttirocyte globules and granules in the blood, the 

 zymogen granules in the panci'eatic cells and the granules in the 

 pancreatic ducts, and the a i^riori necessity for the elimination of 

 the erythrocyte granular matter. This hypothesis, however, con- 

 cerning the function of what I subsequently recognized as the 

 pancreas is obviously untenable, since the zymogen granules are 

 of course present in pancreatic cells not associated with the rete 

 mirabile veins and, as in higher Vertebrates, do not occur outside 

 the pancreas cells ; also the posterior extension of the teleost 



14* 



