Adaptation of the vestibulo-ocular reflex for forward-eyed foveate vision

Abstract

To maintain visual fixation on a distant target during head rotation, the angular vestibulo-ocular reflex (aVOR) should rotate the eyes at the same speed as the head and in exactly the opposite direction. However, in primates for which the 3-dimensional (3D) aVOR has been extensively characterised (humans and squirrel monkeys (Saimiri sciureus)), the aVOR response to roll head rotation about the naso-occipital axis is lower than that elicited by yaw and pitch, causing errors in aVOR magnitude and direction that vary with the axis of head rotation. In other words, primates keep the central part of the retinal image on the fovea (where photoreceptor density and visual acuity are greatest) but fail to keep that image from twisting about the eyes' resting optic axes. We tested the hypothesis that aVOR direction dependence is an adaptation related to primates' frontal-eyed, foveate status through comparison with the aVOR of a lateral-eyed, afoveate mammal (Chinchilla lanigera). As chinchillas' eyes are afoveate and never align with each other, we predicted that the chinchilla aVOR would be relatively low in gain and isotropic (equal in gain for every head rotation axis). In 11 normal chinchillas, we recorded binocular 3D eye movements in darkness during static tilts, 20-100 deg s(1) whole-body sinusoidal rotations (0.5-15 Hz), and 3000 deg s(2) acceleration steps. Although the chinchilla 3D aVOR gain changed with both frequency and peak velocity over the range we examined, we consistently found that it was more nearly isotropic than the primate aVOR. Our results suggest that primates' anisotropic aVOR represents an adaptation to their forward-eyed, foveate status. In primates, yaw and pitch aVOR must be compensatory to stabilise images on both foveae, whereas roll aVOR can be under-compensatory because the brain tolerates torsion of binocular images that remain on the foveae. In contrast, the lateral-eyed chinchilla faces different adaptive demands and thus enlists a different aVOR strategy.

Figure 1. Static eye position in response to whole-body tilt for 11 normal chinchillas

Each animal was rotated about the anterior–posterior axis (AP), inter-aural (IA) and superior–inferior (SI) axis. The zero position for AP and IA tilts was defined as the head orientation in which the horizontal semicircular canals were both Earth-horizontal; tilting 90 deg nose up from that position about the IA axis arrived at the starting zero position for tilts about the SI axis. The starting positions for SI, AP and IA were left-ear down (LED), right-ear down (RED) and nose down (ND), respectively. Only the eye position component that changes most (the head roll, pitch and yaw component for tilts about the AP, IA and SI axes, respectively) is shown. The three sinusoids represent the respective best fits to the eye position data. As the static tilt angle increases, static eye counter-rotation only partly compensates for the tilt. For example, when the animal rotates 90 deg about the AP axis counter-clockwise, the eyes rotate clockwise ∼10 deg.

Figure 2. Horizontal Vestibulo-Ocular Reflex During Transient High-Acceleration Whole Body Rotation

A, horizontal component of 3D eye and head rotational velocity during multiple leftward yaw head rotation impulses (accelerations at 3000 deg s⁻² reaching a peak velocity of 150 deg s⁻¹) in one chinchilla (CH1). Head trace inverted for comparison to eye. The slow phase eye velocity response is almost perfectly compensatory (opposite in direction to head velocity) during the first ∼60 ms after head rotation onset. Once the eye reaches the end of its oculomotor range, a quick phase, which is opposite in direction to the slow phase, brings the eye back to the centre of the oculomotor range. This cycle is repeated for the duration of the transient. B, quick phases have been removed from the data in A leaving only the slow phases for analysis. C, mean eye and head velocity traces (mean ± 1 s.e.m.) are calculated from the data in B. The velocity gain, G_V, is calculated by dividing the mean eye velocity by the mean head velocity during 200 ms of the velocity plateau shown in grey. D, the first 200 ms of the data in C are used to calculate the acceleration gain, G_A, which is the ratio of the eye to head velocity trace slopes during the linear part of the acceleration portion of the transient (grey region). Intersections of the line fits used to calculate G_A with the zero velocity axis are used to calculate latency.

Figure 3. 3D aVOR Misalignment: Chinchillas Outperform Humans and Squirrel Monkeys

A, the first panel shows pitch vs. roll eye velocity (with respect to head coordinates) during a RALP head rotation. Second panel shows pitch vs. roll eye velocity during a LARP head rotation, and final panel shows horizontal vs. roll eye velocity during a yaw head rotation. Two components (e.g. yaw vs. pitch) of instantaneous slow phase eye velocity prior to the first quick phase are plotted on axial view for one chinchilla (CH2). The 3D axis of eye rotation is the instantaneous eye velocity vector with the largest magnitude before the first quick phase. The maximum misalignment error, the angle between the eye and head rotation axes, for this animal was 5.3 ± 1.2 deg. B, mean axis of aVOR eye rotation responses for each of N = 11 chinchillas (black lines) during 3000 deg s⁻² transient head rotations about the yaw (+Z), pitch (+Y), roll (+X), LARP and RALP axes (⋄). Axes labelled LH, LA, LP, RH, RA and RP are the anatomic axes of the left (L, ○) and right (R, •) horizontal (H), anterior (A) and posterior (P) semicircular canals (from Hullar & Williams, 2006). The +/− sign indicates whether a right-hand rule head rotation about the vector denotes excitatory or inhibitory stimulation for that canal. The 3D axis of aVOR eye rotation aligns well with head rotation axis in all cases. C, mean ± 1 s.d. axis of eye rotation for the chinchilla (n = 11), squirrel monkey (n = 3, unpublished data from the Migliaccio et al. 2004 study) and human (n = 7, Cremer et al. 1998). The chinchilla rotational stimulus was the same as the squirrel monkey stimulus and similar to the human stimulus; however, the 3D axis of eye rotation deviates ∼50% less in the chinchilla compared to the squirrel monkey and human.

Figure 4. Responses to Sinusoidal Rotation Reveal Velocity-Dependent aVOR Gain

A, raw data from one chinchilla (CH1) during sinusoidal yaw rotations at 1 Hz with peak velocities of 20, 50 and 100 deg s⁻¹. The aVOR gain increases ∼20% when peak head velocity increases from 20 to 50 deg s⁻¹, and another ∼20% from 50 to 100 deg s⁻¹. B, each cycle of the stimulus and corresponding response are overlaid and then averaged to obtain the mean cycle response of the aVOR in the yaw, LARP and RALP planes. Hor, horizontal.

Figure 5. Average gain and phase plots of responses to 0.01–15 Hz rotations

Average gain and phase plots of responses to 0.01–15 Hz rotations given with vision occluded/in darkness, ± 20, 50 and 100 deg s⁻¹ in all animals (CH1–11). The rotations are in the yaw, pitch, roll, LARP and RALP planes. Error bars indicate 1 s.d.