We employed cholera toxin-B (CTB), an efficient retrograde tracer, to examine olivocochlear (OC) neurons in the cat. Our primary goals were (1). to determine whether the cat has two types of lateral OC (LOC) neurons as is found in certain rodents and (2). to document the morphology, number, and caudorostral distribution of OC neurons, bilaterally. Adult cats received injections of CTB through the round window of the left cochlea, and, after 3-6 days, the brains were sectioned transversely and CTB was revealed immunocytochemically in every section. In three cats, OC neurons were mapped, counted differentially according to cell group, and the numbers of each plotted bilaterally from caudal to rostral. In one cat, measurements were made on labeled LOC and medial OC (MOC) neurons. The results indicate that LOC neurons can be divided into two groups based on their proximity to the lateral superior olive (LSO): a more populous group of small neurons that have intimate contact with the LSO, designated marginal-LOC neurons, and a less populous, morphologically heterogeneous group, lying more distantly from the LSO, designated para-LOC neurons. Para-LOC neurons lying dorsal and rostral to LSO were significantly larger than marginal-LOC. We hypothesize that the cat marginal-LOC neurons and most probably the larger para-LOC neurons correspond to rodent intrinsic and shell LOC neurons, respectively, which have focal versus diffuse projections beneath the inner hair cells. Concerning MOC neurons, we confirm and extend previous observations on the clustering of these neurons near the rostral tip of the medial superior olivary nucleus and also show that MOC neurons differ in size according to cell group. Finally, we compare the present counts of OC neurons (mean total 1607, consisting of 1058 LOC neurons and 549 MOC neurons innervating one cochlea) and their proportional distribution ipsilaterally and contralaterally with those reported previously. Our estimate of the number of LOC neurons is somewhat higher than those previously obtained either by retrograde labeling with horseradish peroxidase or by counting unmyelinated axons in the olivocochlear bundle. In contrast, our estimate of the number of MOC neurons is very similar to those previously reported.