The effect of the volume of the liquid sample, the degree of turbulence in the liquid, and the rate of cooling upon the probability of nucleation has been studied for water and aqueous solutions. Nucleation rates were measured for droplets nearly instantaneously cooled to a predetermined temperature. Also threshold temperatures for nucleation were determined in stirred externally cooled vessels. The liquids studied (tap water, distilled water, and orange juice) did not contain particles causing nucleation at supercoolings less than 6°C. Nucleation at smaller supercoolings is attributed to the nucleative action of surface irregularities of the vessel walls. Provided no air bubbles are dispersed in the liquid, the rate of stirring (degree of turbulence) appeared to have no influence upon the nucleation. The experimental results are inconsistent with the stochastic hypothesis of Bigg7), the singular hypothesis of Langham and Mason8) and the pseudo-singular hypothesis of Vali and Stansbury10). They could, including those reported in literature7, 8, 10) be explained by the assumption of three distinct probabilities of nucleation: namely, the probability per unit time that a particular particle can cause nucleation at equilibrium conditions; the probability per unit time that equilibrium is established; and the probability that a particle with characteristic properties is present in the sample.