Although slow melting favors the generation of basaltic melt from a mantle matrix with large radioactive disequilibrium between two actinide nuclides (MCKENZIE, 1985a), it results in long residence time in a magma chamber, during which the disequilibrium may be removed. An equilibrium melting model modified after MCKENZIE (1985a) is presented here which suggests that, for a given actinide parent-daughter pair, there exists a specific melting rate at which disequilibrium between these two nuclides reaches its maximum. This melting rate depends on the decay constant of the daughter nuclide concerned and the magma chamber volume scaled to that of its source. For a given scaled chamber size, large radioactive disequilibrium between two actinide nuclides in basalts will be observed if the melting rate is such that the residence time of the magma in the chamber is comparable to the mean life of the daughter nuclide. With a chamber size 1% in volume of the melting source, the melting rates at which maximum disequilibrium in basalts is obtained are 10(-7), 2 x 10(-7), and 3 x 10(-6) y-1, respectively, for U-238-Th-230, U-235-Pa-231, and Th-230-Ra-226. This implies that, while large disequilibrium between U-238-Th-230 and between U-235-Pa-231 may occur together, large Th-230-Ra-226 disequilibrium will not coexist with large U-238-Th-230 disequilibrium, consistent with some observations. The active mantle melting zone which supplies melt to a ridge axis is inferred to be only about 10 km thick and 50 km wide. The fraction of melt present in such a mantle source at any time is about 0.01 and 0.04, respectively, if melting rate is 10(-7) and 10(-6) y-1. The corresponding residence time of the residual melt in the matrix is 10(5) and 4 x 10(4) y.