1. Postsynaptic potentials were recorded in lumbar motoneurons of the frog in response to electrical activation of dorsal roots. After chemical synaptic transmission was blocked by replacing Ca2+ with Mg2+ in the superfusion medium, it was confirmed that the remaining electrical excitatory postsynaptic potentials (EEPSPs) recorded in motoneurons consisted of potential changes-produced by electrical coupling between the motoneurons and the stimulated axons. The EEPSPs could then be used as an assay to study the reliability of spike propagation into presynaptic terminals. 2. EEPSPs typically consisted of three components. The first was a small positive deflection (prespike or presynaptic volley) that could also be recorded extracellularly. The second component was a spikelike fast positive component and the third was a slow positive component that followed the second but had a distinct maximum and a slow decay. The amplitude of the fast component did not correlate with that of either the prespike or the slow component. 3. 4-Aminopyridine (0.1 mM), which widens action potentials by blocking K+ channels, increased the amplitude and width of EEPSPs. Heptanol (1-4 mM), which is known to be a blocker of electrical coupling, could block EEPSPs. 4. The amplitudes of EEPSPs evoked by dorsal root stimulation were compared at different temperatures (7.5-19.5 degrees C). A slight decrease of the amplitude of the fast component with increasing temperature (Q10 = 0.8) was within limits predicted by resistance-capacitance filtering of the presynaptic spike at the different temperatures, suggesting that the temperature does not affect propagation of the spike in this synapse. 5. The amplitude of the fast component of EEPSPs evoked by single-pulse and paired-pulse stimulation did not fluctuate more than the baseline noise in 37 experiments in which the SD of baseline noise was < 100 microV. We conclude that electrical synaptic transmission does not fluctuate intermittently in this system, and that branch points conduct or fail to conduct for periods of time longer than the longest period in the analyzed experiments.