1. Intracellular recordings were obtained from neurosecretory terminals (endings of Herring bodies), axons and pars intermedia cells in the isolated neuro-intermediate lobe of the rat. Responses to current injection and stimulation of the neural stalk (NS) were examined at 34 degrees C. Cellular identity was verified following injection of Lucifer Yellow. 2. Neurosecretory terminals were identified by a constant-latency action potential response to NS stimulation and an appropriate collision test. This response was not blocked by Ca2(+)-free solutions. Hyperpolarization of the terminals could block the generation of a local spike. Injection of Lucifer Yellow into eleven units confirmed that such responses were recorded from neurosecretory terminals. 3. Neurohypophysial nerve terminals had a resting potential of -60.4 +/- 1.1 mV and displayed a spike amplitude of 72.4 +/- 1.9 mV. The local spike threshold was -41.6 +/- 1.9 mV. Voltage-current relations were linear near resting potential, but displayed strong outward rectification positive to -55 mV. While terminals could fire at high frequencies during NS stimulation, repetitive activity could not be evoked by prolonged depolarizing current pulses. 4. During the initial 1-3 s of a train of brief depolarizing pulses or NS stimuli, nerve terminals showed a progressive broadening of their action potentials. At steady state, the duration of these impulses increased logarithmically with firing rate, showing a maximum near 25 Hz. Spike broadening in nerve terminals was reversibly abolished by superfusion of the neural lobe with Ca2(+)-free, Mn2(+)-containing solutions. In the absence of external Ca2+, action potentials were smaller, and lacked a prominent shoulder on their repolarizing phase. 5. Sustained (greater than 5 s) repetitive stimulation at 10-20 Hz led to a gradual increase in the latency for invasion and eventual failure of the spike-generating mechanism within the terminal. This effect required several seconds to recover. In contrast, action potentials recorded within the axon followed continuous repetitive stimulation and did not show any frequency-dependent changes in duration (0.7 +/- 0.1 ms). 6. Cells of the pars intermedia (PI) displayed an input resistance of 215.7 +/- 47.4 M omega and fired a single action potential in response to current injection. The amplitude of this current-evoked spike decreased during repetitive stimulation, but its duration was not affected. In 87% of the PI cells tested, stimulation of the NS evoked a Ca2(+)-sensitive synaptic response which reversed near -40 mV, but no cell was directly activated by the stimulus.