Gilman (1937), experimenting on normal dogs, observed that hypertonic solutions of sodium chloride, which cause cellular dehydration, gave rise to more drinking than hypertonic solutions of urea, which do not cause loss of cell water. He concluded that thirst is due to cellular dehydration.Experiments on dogs by Holmes & Gregersen (1950a, b) in which a variety of substances were given, confirmed that osmotically effective solutes cause most drinking and by implication most thirst, and Adolph, Barker & Hoy (1954) showed that sodium chloride was twice as powerful as urea in causing normal rats to drink. In both species, however, no simple quantitative relationship was found between the amount of substance injected and the amount of water drunk.In the present work it was thought that the analysis would be simplified if the kidneys were removed, because there could be no renal response to the substance, which could therefore act until neutralized by drinking, and the quantitative relation between the amount of substance injected and water drunk could thus be investigated. It was hoped that the cellular dehydration hypothesis of drinking would be tested in this way. METHODS Male albino rats, weighing between 150 and 400 g, were most commonly used in these experiments, but some female rats and a small number of hooded rats and ginger rats of both sexes were used also. Food (Diet 41, Bruce & Parkes, 1949) was available until the beginning of the experiment and water was available before and during the experiment from a drinking meter which enabled the time course of drinking to be recorded (Fitzsimons, 1958).Experimental procedure Each animal was anaesthetized with open ether and the external jugular vein catheterized by a soft polyvinyl chloride tube, outside diameter 1 mm. The tip of the catheter was advanced to the right atrium to miniimize the complication of dead space when injections were being made. The catheter was brought subcutaneously to the back of the neck and about 2 cm was left projecting through a stab wound in the skin. The catheter, which was filled with 0*9 % sodium chloride solution, was closed with a small plug.Bilateral nephrectomy was then carried out through a dorsal incision. A clean surgical technique was employed, but no aseptic precautions were taken because the experiments 36-2