The extraocular muscles, unlike the skeletal muscles, contain non-twitch muscle fibers. Recent experiments have located the non-twitch motoneurons. They lie around the periphery of the oculomotor, trochlear and abducens nuclei, separate from the more usual twitch motoneurons that cluster within the boundaries of the classical motor nuclei. The premotor inputs to non-twitch neurons were traced by the injection of rabies virus into the distal tip of the lateral rectus muscle. Retrogradely labeled cells were found in areas associated with the neural integrator, vergence and smooth pursuit premotor areas, but not the saccadic premotor burst neurons or the direct vestibulo-ocular pathways. The rabies tracing emphasizes for the first time that the central mesencephalic reticular formation (cMRF) and the supraoculomotor area exert direct premotor control over the non-twitch motoneurons. Because the two sets of motoneurons do not receive the same afferents, they must have different functions; these are not yet clarified. These results are not compatible with the concept of a single final common pathway from motoneurons to eye muscles. Putative sensory receptors, palisade endings, are located at the tips of non-twitch muscle fibers reminiscent of an inverted muscle spindle, which would make the non-twitch motoneurons, gamma-motoneurons. We propose that twitch motoneurons are the major source of tension used for eye movements, whereas non-twitch motoneurons are more important for fine alignment of the eyes. Furthermore, the non-twitch motoneurons could be controlled through sensory feedback networks (including perhaps proprioceptive signals from the palisade endings) that are relayed through the superior colliculus and via cMRF to the non-twitch motoneurons. The clinical repercussions of these hypotheses are discussed.