1. The purpose of this investigation was to determine the axes of eye rotation generated by oculomotor burst neuron populations and the coordinate system that they collectively define. In particular, we asked if such coordinates might be related to constraints in the emergent behavior, i.e., Listing's law for saccades. 2. The mesencephalic rostral interstitial nucleus of the medial longitudinal fasciculus (riMLF) was identified in four monkeys with the use of single-unit recording, and then explored with the use of electrical microstimulation and pharmacological inactivation with the inhibitory gamma-aminobutyric acid (GABA) agonist muscimol. Three-dimensional (3-D) eye positions and velocities were recorded in one or both eyes while alert animals made eye movements in response to visual stimuli and head rotation. 3. Unilateral stimulation of the riMLF (20 microA, 200 Hz, 300-600 ms) produced conjugate, constant velocity eye rotations, which then stopped abruptly and held their final positions. This is expected if the riMLF produces phasic signals upstream from the oculomotor integrator. 4. Units that burst before upward or downward saccades were recorded intermingled in each side of the riMLF. Unilateral stimulation of the same riMLF sites produced eye rotations about primarily torsional axes, clockwise (CW) during right riMLF stimulation and counterclockwise (CCW) during left stimulation. Only small and inconsistent vertical components were observed, supporting the view that the riMLF carries intermingled up and down signals. 5. The torsional axes of eye rotation produced by riMLF stimulation did not correlate to external anatomic landmarks. Instead, stimulation axes from both riMLF sides aligned with the primary gaze direction orthogonal to Listing's plane of eye positions recorded during saccades. 6. Injection of muscimol into one side of the riMLF produced a conjugate deficit in saccades and quick phases, including a 50% reduction in all vertical velocities and complete loss of one torsional direction. CW was lost after right riMLF inactivation, and CCW was lost after left inactivation. 7. The plane that separated the intact torsional axes from the missing axes correlated with the orientation of Listing's plane. Thus, during left or right riMLF inactivation, the vertical axes of intact horizontal saccades were abnormally aligned with Listing's plane. The orientation of these axes was not correlated with external anatomic landmarks. 8. As suggested by their alignment with Listing's plane, the intact vertical axes of horizontal saccades following riMLF inactivation were orthogonal to torsional riMLF stimulation axes.(ABSTRACT TRUNCATED AT 400 WORDS)