1. 1. Histological methods revealed that blood entering the gill filament may reach the efferent artery (1) through the lamellae, (2) through a central, lymphatic compartment and (3) directly from the afferent to efferent artery at the tip of the filament. 2. 2. In gill filaments immersed in saline, blood was shown to pass through the lamellae in the presence of adrenaline, and through the other vascular paths in the presence of acetylcholine. 3. 3. Inravascular injection of adrenaline caused a marked increase in the degree of blood oxygenation in the eel. 4. 4. The arterial oxygen tension was much higher in active than in resting eels, whereas the oxygen tension of efferent water was higher in resting than in active eels. The average difference in oxygen tension between blood and water was the same in resting and active eels. 5. 5. The expected oxygen uptake, as calculated from the anatomical dimensions of the gills and the measured blood-to-water oxygen tension difference, corresponded to the oxygen uptake which has been measured for active eels. 6. 6. It is concluded that the vascular arrangement of the gills of the eel permits variations in blood oxygenation, and thereby in oxygen uptake, by directing a varying proportion of the blood to circulate the respiratory lamellae while the rest is shunted through the alternate non-respiratory path of the central cavity. 7. 7. A double, intercommunicating vascular system, more or less similar to that of the eel, was demonstrated in all the other ten teleost species investigated, and we suggestthat it is common to all teleosts.