1. In order adequately to control eye movements, oculomotoneurones have to be supplied with both an eye-velocity signal and an eye-position signal. However, all the command signals of the oculomotor system are velocity signals. Nowadays, there is general agreement about the existence of a brainstem network that would convert velocity command-signals into an eye-position signal. This circuit, because of its function, is called the oculomotor neural integrator. The most obvious symptom of its eventual failure is a gaze-holding deficit: in this case, saccades are followed by a centripetal post-saccadic drift. Although the oculomotor neural integrator is central in oculomotor theory, its precise location is still a matter for debate. 2. Previously, microinjections of kainic acid (KA) into the region of the nucleus prepositus hypoglossi (NPH) and of the medial vestibular nucleus (MVN) were found to induce a horizontal gaze-holding failure both in the cat and in the monkey. However, the relatively large volumes (1-3 microliters) and concentrations (2-4 micrograms microliters-1) used in these injections made it difficult to know if the observed deficit was due to a disturbance of the NPH or of the nearby MVN. These considerations led us to inject very small amounts of kainic acid (50 nl, 0.1 microgram microliter-1) either into the rostral part of the MVN or into different sites along the NPH of the cat. 3. The search coil technique was used to record (1) spontaneous eye movements (2) the vestibulo-ocular reflex (VOR) induced by a constant-velocity rotation (50 deg s-1 for 40 s) and the optokinetic nystagmus (OKN) elicited by rotating an optokinetic drum at 30 deg s-1 for 40 s. 4. In each injection experiment, the location of the abducens nucleus of the alert cat was mapped out by recording the antidromic field potentials evoked by the stimulation of the abducens nerve. Two micropipettes were then glued together in such a way that when the tip of the recording micropipette was in the centre of the abducens nucleus the tip of the injection micropipette was in a target area. The twin pipettes were then lowered in the brainstem until the recording micropipette reached the centre of the abducens nucleus. Kainic acid was then injected into the brainstem of the alert cat through the injection micropipette by an air pressure system. 5. Carried out according to such a protocol, KA injections into the NPH or the rostral part of the MVN consistently led to specific eye-movement changes.(ABSTRACT TRUNCATED AT 400 WORDS)