1. Single units and evoked potentials were recorded in the dorsal cochlear nucleus (DCN) of paralyzed decerebrate cats in response to electrical stimulation at two sites: 1) in the somatosensory dorsal column nuclei (together called MSN below for medullary somatosensory nuclei), which activates mossy-fiber inputs to granule cells in superficial DCN, and 2) on the free surface of the DCN, which activates granule cell axons (parallel fibers) directly. The goal was to evaluate hypotheses about synaptic interactions in the cerebellum-like circuitry of the superficial DCN. A four-pulse facilitation paradigm was used (50-ms interpulse interval); this allows identification of three components of the responses of DCN principal cells (type IV units) to these stimuli. The latencies of the response components were compared with the latency of the evoked potential in DCN, which signals the arrival of the parallel fiber volley at the recording site. 2. The first component is a short-latency inhibitory response; this component is seen only with MSN stimulation and is seen almost exclusively in units also showing the second component, the transient excitatory response. The short-latency inhibitory component precedes the evoked potential. No satisfactory explanation for the short-latency component can be given at present; it most likely reflects a fast-conducting inhibitory input that arrives at the type IV unit before the slowly conducting parallel fibers. 3. The second component is a transient excitatory response; this component is seen with both MSN and parallel fiber stimulation; it is weak and appears to be masked easily by the inhibitory response components. The excitatory component occurs at the same latency as the evoked potential and probably reflects direct excitation of principal cells by granule cell axons. The excitatory component is seen in about half the type IV units for both stimulating sites. With MSN stimulation, the lack of excitation in some units suggests a heterogeneity of cochlear granule cells, with some carrying somatosensory information and some not carrying this information; with parallel fiber stimulation, excitation probably requires the stimulating and recording electrodes to be lined up on the same "beam" of parallel fibers. 4. The third component is a long-lasting inhibitory response that is observed in virtually all type IV units with both MSN and parallel-fiber stimulation; its latency is longer than the evoked potential. Evidence suggests that it is produced by inhibitory input from cartwheel cells. The appearance of this inhibitory component in almost all type IV units can be accounted for by the considerable spread of cartwheel-cell axons in the direction perpendicular to the parallel fibers. 5. The evoked potential and all three components of the unit response vary systematically in size over the four pulses of the electrical stimulus. These results can be accounted for by two phenomena: 1) a facilitation of the granule cell synapses on all cell types that produces a steadily growing response through the four pulses, resembles presynaptic facilitation, and is seen with both MSN and parallel-fiber stimulation; and 2) a strong reduction in the granule cell response between the first and second pulse for MSN stimulation only. This reduction probably occurs presynaptically in the glomerulus or in the granule cell itself and could reflect inhibitory inputs. 6. The response components described above are seen in type IV units recorded in both the fusiform-cell and deep layers of the DCN; this suggests that both pyramidal and giant cells are activated similarly. The simplest interpretation is that both principal cell types are activated by the cerebellum-like circuitry in superficial DCN. Alternatively, because giant cells appear to make limited contact with the granule-cell circuits of superficial DCN, this finding may suggest the existence of currently undescribed granule cell circuits in deep DCN that are si