1. Isolated rat neural lobes were incubated in vitro and electrically stimulated to release vasopressin. The released vasopressin was assayed using a radioimmunoassay and there was a reasonably good correlation (r = 0.81) between results obtained with this assay and those obtained by bioassay with the rat blood pressure method.2. Regular stimulation at frequencies of 5, 10 and 20 Hz released progressively more vasopressin and the release could be blocked by addition of tetrodotoxin to the incubation medium.3. Stimulation with pulse patterns derived from tape recordings of phasically firing units in the supraoptic nucleus of dehydrated rats released more vasopressin than the same number of pulses regularly spaced in time. In the range 2-8 pulses/sec vasopressin release was related to the pulse frequency within the bursts (r = 0.90) and the number of short (< 100 msec) interpulse intervals (r = 0.92). Vasopressin released per pulse increased over the frequency range 3-6 pulses/sec, but above 6 pulses/sec vasopressin release per pulse tended to diminish.4. We conclude that phasic firing of vasopressin neurosecretory cells may enhance vasopressin release in vivo and that an important factor in determining release is the number of short interspike intervals.