Objectives: To develop a hypothetical scheme to account for clinical disord
ers of vertical gaze based on recent insights gained from experimental stud
ies. Methods: The authors critically reviewed reports of anatomy, physiolog
y, and effects of pharmacologic inactivation of midbrain nuclei. Results: V
ertical saccades are generated by burst neurons lying in the rostral inters
titial nucleus of the medial longitudinal fasciculus (riMLF). Each burst ne
uron projects to motoneurons in a manner such that the eyes are tightly coo
rdinated (yoked) during vertical saccades. Saccadic innervation from riMLF
is unilateral to depressor muscles but bilateral to elevator muscles, with
axons crossing within the oculomotor nucleus. Thus, riMLF lesions cause con
jugate saccadic palsies that are usually either complete or selectively dow
nward. Each riMLF contains burst neurons for both up anti down saccades, bu
t only for ipsilateral torsional saccades. Therefore, unilateral riMLF lesi
ons can be detected at the bedside if torsional quick phases are absent dur
ing ipsidirectional head rotations in roll. The interstitial nucleus of Caj
al (INC) is important for holding the eye in eccentric gaze after a vertica
l saccade and coordinating eye-head movements in roll. Bilateral INC lesion
s limit the range of vertical gaze. The posterior commissure (PC) is the ro
ute by which INC projects to ocular motoneurons. Inactivation of PC causes
vertical gaze-evoked nystagmus, but destructive lesions cause a more profou
nd defect of vertical gaze, probably due to involvement of the nucleus of t
he PC. Vestibular signals originating from each of the vertical labyrinthin
e canals ascend to the midbrain through several distinct pathways; normal v
estibular function is best tested by rotating the patient's head in the pla
nes of these canals. Conclusions: Predictions of a current scheme to accoun
t for vertical gaze palsy can be tested at the bedside with systematic exam
ination of each functional class of eye movements.