1. We cut serial sections through the medial part of the rat vestibular macula for transmission electron microscopic (TEM) examination, computer-assisted three-dimensional (3-D) reconstruction, and compartmental modeling. The ultrastructural research showed that many primary vestibular neurons have an unmyelinated segment, often branched, that extends between the heminode [putative site of the spike initiation zone (SIZ)] and the expanded terminal(s) (calyx, calyces). These segments, termed the neuron branches, and the calyces frequently have spinelike processes of various dimensions that morphologically are afferent, efferent, or reciprocal to other macular neural elements. The purpose of this research was to determine whether morphometric data obtained ultrastructurally were essential to compartmental models [i.e., they influenced action potential (AP) generation, latency, or amplitude] or whether afferent parts could be collapsed into more simple units without markedly affecting results. We used the compartmental modeling program NEURON for this research. 2. In the first set of simulations we studied the relative importance of small variations in process morphology on distant depolarization. A process was placed midway along an isolated piece of a passive neuron branch. The dimensions of the four processes corresponded to actual processes in the serial sections. A synapse, placed on the head of each process, was activated and depolarization was recorded at the end of the neuron branch. When we used 5 nS synaptic conductance, depolarization varied by 3 mV. In a systematic study over a representative range of stem dimensions, depolarization varied by 15.7 mV. Smaller conductances produced smaller effects. Increasing membrane resistivity from 5,000 to 50,000 omega cm2 had no significant effect. 3. In a second series of simulations, using whole primary afferents, we examined the combined effects of process location and afferent morphology on depolarization magnitude and latency, and the effect of activating synapses individually or simultaneously. Process location affects peak latency and voltage recorded at the heminode. A synapse on a calyceal process produced < or = 8% more depolarization and a 23% increase in peak latency compared with a synapse on a process of a neuron branch. For whole primary afferents, depolarization decreased 40% between simulations of the smallest and largest afferents. Simulations in which membrane resistivity and synaptic conductance were varied while afferent geometry was kept constant indicated that use of 5,000 omega cm2 and 1.0 nS produced results that best fit electrophysiological findings. Synaptic inputs activated simultaneously did not sum linearly at the heminode. Total depolarization was approximately 14% less than a simple summation of responses of synapses activated one at a time.(ABSTRACT TRUNCATED AT 400 WORDS)