Motor cortex neurons were identified antidromically in anesthetized cats by their axonal projections to one of six targets: (1) somatosensory cortex, (2) opposite motor cortex, (3) red nucleus, (4) lateral reticular nucleus, (5) spinal cord, and (6) ventrolateral thalamus. Three inputs to motor cortex were tested for their influences on the identified cortical efferent neurons. The tested inputs originated from ipsilateral somatosensory cortex, opposite motor cortex, and ventral thalamus. Subthreshold effects of input pathways were detected by monitoring latency variations of antidromic responses. The three afferent sources, when activated by electrical stimulation, were not equally effective on motor cortex neurons. Ipsilateral corticocortical and thalamocortical excitation were found for the majority of neurons; the influenced proportions ranged from 55% to 100%, according to the target of the output neurons. Effects from the opposite hemisphere were found for only 5% to 30% of the neurons in the same projection classes. Many neurons (36 of 81, or 44%) were excited from more than one source, but few (5 of 37, or 14%) were influenced by all three possible sources of input, even in small regions of cortex innervated by all three of the inputs. Among 19 electrode tracks where all three inputs were present, there were only 2 tracks where all the neurons shared the same combination of inputs. Even for neurons in closest anatomical proximity ("clusters"), it was unusual (only 7 of 25 clusters) for all the neurons to have the same input pattern. Among the seven clusters where all the neurons shared the same input pattern, five of the clusters projected to the same target. These variable combinations of inputs to motor cortex neurons support the conclusion that efferent neurons could be recruited selectively from separate cortical layers or from within clusters of nearby neurons, according to the target of their axonal projection.