Three dimensional kinematics of rapid compensatory eye movements in humans with unilateral vestibular deafferentation

Abstract

Saccades executed with the head stationary have kinematics conforming to Listing's law (LL), confining the ocular rotational axis to Listing's plane (LP). In unilateral vestibular deafferentation (UVD), the vestibulo-ocular reflex (VOR), which does not obey LL, has at high head acceleration a slow phase that has severely reduced velocity during ipsilesional rotation, and mildly reduced velocity during contralesional rotation. Studying four subjects with chronic UVD using 3D magnetic search coils, we investigated kinematics of stereotypic rapid eye movements that supplement the impaired VOR. We defined LP with the head immobile, and expressed eye and head movements as quaternions in LP coordinates. Subjects underwent transient whole body yaw at peak acceleration 2,800 degrees /s(2) while fixating targets centered, or 20 degrees up or down prior to rotation. The VOR shifted ocular torsion out of LP. Vestibular catch-up saccades (VCUS) occurred with mean latency 90 +/- 44 ms (SD) from ipsilesional rotation onset, maintained initial non-LL torsion so that their quaternion trajectories paralleled LP, and had velocity axes changing by half of eye position. During contralesional rotation, rapid eye movements occurred at mean latency 135 +/- 36 ms that were associated with abrupt decelerations (ADs) of the horizontal slow phase correcting 3D deviations in its velocity axis, with quaternion trajectories not paralleling LP. Rapid eye movements compensating for UVD have two distinct kinematics. VCUS have velocity axis dependence on eye position consistent with LL, so are probably programmed in 2D by neural circuits subserving visual saccades. ADs have kinematics that neither conform to LL nor match the VOR axis, but appear instead programmed in 3D to correct VOR axis errors.

Fig. 1

Typical VCUS initiated with a centered target in subject with right UVD who was rotated ipsilesionally. a VCUS was identified by marked change in horizontal eye velocity from the VOR slow phase. b VCUS was identified by discontinuity in horizontal eye position from the VOR slow phase. c VOR slow phase torsion matched neither the LL value, nor head torsion. However, during the VCUS eye torsion paralleled calculated ideal Listing's torsion. d Data in horizontal-torsional quaternion plane shows that the VOR slow phase (black solid dots) deviated from LP, but that the VCUS (grey) roughly paralleled LP, maintaining the non-Listing torsion generated by the VOR slow phase

Fig. 2

Typical VCUS initiated with a target 20° up in subject with right UVD who was rotated ipsilesionally. a VCUS was identified by marked change in horizontal eye velocity from the VOR slow phase. b VCUS was identified by discontinuity in horizontal eye position from the VOR slow phase. c Slow phase VOR torsion matched neither the ideal Listing's torsion, nor head torsion. However, during the VCUS, eye torsion maintained a constant difference from, and thus paralleled, Listing's torsion. d Data in horizontal-torsional quaternion plane shows that the VOR slow phase deviated from LP, but that the VCUS roughly paralleled LP

Fig 3

Effect of starting vertical eye position on velocity axis tilt of horizontal VCUS in the subject demonstrated in Figs. 1 and 2. The tilt angle ratio measured during the middle of the VCUS is plotted against its corresponding vertical eye position (relative to Listing's primary position, which differs from straight ahead because Listing's plane is tilted vertically). The slope of 0.51 is consistent with the ideal Listing's law value of 0.5

Fig. 4

A typical secondary VCUS in the same subject of Fig. 1, who had right UVD. The subject was rotated ipsilesionally while fixing a central target. a Primary and secondary VCUSs were identified by corresponding peaks in horizontal eye velocity. b In the horizontal-torsional quaternion plane, trajectories of both primary and secondary VCUS paralleled LP, while the VOR slow phase added non-LL torsion

Fig. 5

Abrupt deceleration (AD) of the VOR slow phase during viewing of central target by representative subject with left chronic UVD rotated contralesionally. a AD was identified by marked change in horizontal eye velocity from the VOR slow phase. b AD was identified by discontinuity in horizontal eye position from the VOR slow phase. c Slow phase VOR torsion did not match head torsion. Torsion during the AD deviated from ideal Listing's torsion. d Failing to match the head, slow phase eye torsion remained in Listing's plane (LP). During the AD, eye torsion trajectory was perpendicular to LP, and brought the eye close to the inverse of head torsion indicated by the grey arrow

Fig. 6

Abrupt deceleration (AD) of the VOR slow phase during initial fixation of a target 20° upward by representative subject with left chronic UVD who was rotated contralesionally. Positive values indicate upward, rightward, and clockwise rotations. a The AD rapidly reversed horizontal velocity. b AD corresponded to a discontinuity in horizontal eye position from the VOR slow phase. c VOR slow phase eye torsion did not match head torsion. Torsion during the AD deviated from ideal Listing's torsion. d Failing to match the head, slow phase eye torsion remained close to LP. During the AD, eye torsion trajectory was approximately perpendicular to LP, and caused eye torsion to exceed the negative of head torsion (gray dots)

Fig. 7

Mean trajectory angles relative to Listing's plane (LP, 0°) in horizontal-torsional quaternion plane for central (open bar) and vertically eccentric (striped bar) initial eye positions. Data for 20° up and down initial eye positions were pooled as they did not differ statistically. * P < 0.005 for comparison of initial and final eye angle to either vestibular catch-up saccade (VCUS) or abrupt deceleration (AD) trajectory angles. a During ipsilesional rotation, where VCUSs frequently occurred. b During contralesional rotation, where ADs frequently occurred

Fig. 8

Mean latency and duration of vestibular catch-up saccades (VCUSs) and abrupt decelerations (ADs). a Latency distribution of all primary VCUSs and ADs. The mode was 40–60 ms for VCUSs, but 120–160 ms for ADs. b Mean latency of VCUSs and ADs for central and vertical eccentric starting positions. Mean VCUS latency was significantly shorter than AD latency (P < 0.005), but neither latency was significantly influenced by vertically eccentric starting position. c Mean duration of VCUSs and ADs in central and vertically eccentric starting positions. Mean VCUS duration was significantly longer than AD duration for vertically eccentric initial eye positions, but there was no significant difference for the central initial eye position. Mean duration was significantly influenced by vertically eccentric starting position for ADs but not VCUSs