The participation of the dentate nucleus (DN) in the initiation of a voluntary movement was investigated in five baboons (Papio papio). In these experiments, we have analyzed the effects of excluding the DN on the latency (reaction time, RT) of a learned goal-directed movement. Two techniques were for excluding the DN. In three animals, the structure was cooled with a probe, stereotaxically implanted on the side ipsilateral to the active hand. In two others, a partial electrolytic destruction of the DN ipsilateral to the operant hand was undertaken. In one further animal, both DNs were destroyed electrolytically. A comparison was made of the effect of DN inactivation on the latency of stereotyped goal-directed movements of constant amplitude and direction, and of goal-directed movements whose amplitude and/or direction were varied in random fashion. The exclusion of DN brought about a prolongation of RTs in all animals. This prolongation was not accentuated by variation of different characteristics (amplitude and/or direction) of the impending goal-directed movement. A recovery of the RTs to their prelesion values was observed after irreversible unilateral DN lesion, but no so easily after bilateral destruction. These results show that in the monkey DN is concerned with the initiation of a goal-directed movement, but is not critically implicated in the encoding of direction and amplitude parameters. These findings are discussed in view of the role that is usually attributed to the neocerebellum in programming voluntary movements.