By means of intraven. salt-water infusions, various degrees of hypertonicity of body fluids and of thirst, were produced in dog and man. A thirst threshold was measured under these conditions in terms of a calculated relative decrease in the cellular water content, this value being called [tau] . The value 100 [tau] is theoretically the percentual decrement in cellular water content at the thirst threshold. Calculations based on the osmometric equation are considered at present the most satisfactory means for investigating the cellular dehydration hypothesis of thirst. In the dog an avg. for 100[tau] of 2.15 [plus or minus] 0.64 is believed to be in excess of the "true" thirst threshold. In man 100[tau] averaged 1.23 [plus or minus] 0-48 which is thought to represent an upper limit of the "true" thirst threshold. With reference to an osmometric thirst diagram, graphic and algebraic interpretations of problems in thirst are presented. These make use of equations of thirst and of isodipsic parameters. Hyperdipsic, hypodipsic and adipsic states are distinguished. While no primacy can be given to increased effective osmotic pressure of body fluids or to decreased cellular water content in initiating thirst, it is believed that either or both, when present to.sufficient degree, can excite thirst. The cellular dehydration hypothesis is considered preferable to the "dry mouth" hypothesis and it is suggested that particular osmoreceptors, probably in the central nervous system, and similar to or the same as those postulated by Verney, lie on the afferent side of a thirst reflex.