INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Visual stabilization mechanisms and labyrinthine reflexes have evolved to complement each other in maintaining visual acuity. The primary function of the vestibuloocular reflexes is thus to provide short-latency compensatory eye movements early into the motion, before visual tracking mechanisms come into play. Accordingly, rotational vestibuloocular reflexes (VOR) originating from the semicircular canals have been shown to operate at very short latencies (~7 ms) (Maas et al. 1989; Minor et al. 1999; Tabak and Collewijn 1994). In primates and humans, in particular, preattentive visual mechanisms that sense the pattern of optic flow during translational motion have been shown to elicit compensatory eye movements at latencies of ~60 ms in monkeys and ~85 ms in humans (see Miles 1998 for a review). If visual acuity is to be maintained during translational as it is during rotational stimuli, translational VORs should also operate at very short latencies. Other than a recent study during free-fall (Bush and Miles 1996), all previous reports have suggested that translational VORs are characterized by long latencies (>30 ms) (Bronstein and Gresty 1988; Bronstein et al. 1991; Gianna et al. 1997; Snyder and King 1992). If true, this fact would question the utility of these labyrinthine reflexes in maintaining visual acuity during fast perturbations. In the present study, we have used abrupt, transient translational stimuli and fixation at near targets in an attempt to estimate the response latency of the VORs during both lateral and fore-aft motions.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Four juvenile rhesus monkeys were chronically implanted with a head restraint platform and dual coils on each eye. Binocular three-dimensional (3-D) eye movements were recorded inside a magnetic field (CNC Engineering), then calibrated and expressed as rotation vectors (relative to straight-ahead; leftward was positive) (for details see Angelaki 1998). The motion was delivered by a whole-body displacement on a sled (Acutronics) either along the lateral or fore-aft direction. Translational stimuli consisted of a steplike linear acceleration profile, followed by a short period of constant velocity (peak linear acceleration: 0.5 G; steady-state velocity: ±22 cm/s). The stimulus waveform had a frequency content of <50 Hz. The present analyses are concentrated on horizontal eye movements elicited during the first 50 ms (~0.25 cm). An identical, precalibrated search coil that was secured on the head implant measured negligible head rotation (<0.5°/s) during motion. For lateral movements, the animals (n = 4) fixated a centered target (at distances of 40, 30, 20, 15, and 10 cm from the eyes). For fore-aft motion, the animals (n = 3) fixated one of two targets at horizontal eccentricities of approximately ±6 cm from the right eye and a distance of 20 cm (~17°).¹ Each trial was initiated under computer control only when the animal had satisfactorily fixated (binocularly) the target. Within a random time of 2-20 ms after the target light was turned off, positive, negative, or no motion stimuli were presented in a pseudorandom fashion (similar results were also obtained when the target remained on and space-fixed during motion). Responses and the output of a linear accelerometer mounted on the head were low-pass filtered (200 Hz, 6-pole Bessel) and digitized (833.33 Hz, 12- or 16-bit resolution). Eye position was differentiated with a Savitzky-Golay quadratic polynomial with a 1-point forward and backward window (<5% attenuation for frequencies <50 Hz). Both actual and ideal responses (see next paragraph) were further processed through digital notch filters (60 and 120 Hz).

For the rotational VOR, latency is usually computed as the difference between the onsets of eye velocity and head velocity. For the translational VOR, however, the choice might not seem as straightforward. A comparison of the onsets of eye position/velocity and linear acceleration is not considered appropriate because of different waveform and scaling of the two signals (Figs. 1-3). Thus it is necessary to estimate translational VOR latency by comparing the actual eye movement with an "ideal" (i.e., simulated) response (computed from the head displacement waveform; see APPENDIX). The onset of the eye movement was then calculated for each experimental run in each of the animals, as follows. First, the preresponse baseline was determined by computing the mean (±SD) during the 50-ms period immediately preceding sled motion. Response onset was then computed either by fitting a line to the data (starting at the point at which the profile first exceeded the mean baseline by 3 SDs and ending 17 ms later) (Bush and Miles 1996) or as the time after which the next 5-15 consecutive data points exceeded the mean baseline by 3 SDs. Because both methods gave similar results, numbers reported here are based on the latter method.

View larger version (44K): [in this window] [in a new window]   Fig. 1. Lateral motion: mean (±SD) right eye position (A) and velocity (B) from at least 40 experimental runs are compared with ideal responses (dark vs. gray thick lines) during rightward and leftward motion stimuli (top and bottom traces, respectively). Positive eye movement is to the left. Thin vertical dotted lines mark 10-ms intervals. The single thick vertical dotted line in each plot marks the start of the stimulus (linear acceleration has been also superimposed in A). Histograms of latency computed from each individual run illustrate the mean and median values (thick solid and dashed lines, respectively). Data from animal 2 at a viewing distance of 10 cm.

View larger version (26K): [in this window] [in a new window]   Fig. 2. Mean horizontal right and left eye position during rightward and leftward motion (top and bottom traces, respectively) at different viewing distances illustrates that viewing distance dependence was seen with the shortest latency. Dotted lines: stimulus linear acceleration (peak of 0.5 G). Data from animal 2 (middle target). Notice the conjugate eye movements that are elicited for viewing distances of 20-40 cm, as compared with rather disjunctive eye movements for viewing distances of 10 and 15 cm. The abducting eye (e.g., left eye during rightward motion and right eye during leftward motion) was always characterized by larger responses compared with the adducting eye.

View larger version (28K): [in this window] [in a new window]   Fig. 3. Fore-aft motion: mean horizontal eye position (A) and ideal position (B) during forward (top part of each plot) and backward (bottom part of each plot) motion for fixation at ~17° to the left (left plots) or to the right (right plots; animal 2). Thin vertical dotted lines mark 10-ms intervals. The single thick vertical dotted line in each plot marks the start of the stimulus. The initial portion of stimulus acceleration is shown as a reference.

As a measure of early reflex gain during lateral motion, we computed initial acceleration for each run by fitting a line to the first 17 ms of both actual and ideal eye velocity (shifted by the corresponding latency). Initial acceleration gain was computed as the ratio of the slopes of the two lines. Similar values were also obtained by fitting average response profiles. Because of smaller response amplitudes, only average response profiles were fitted for fore-aft responses.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Translational VOR latency

Lateral motion response latencies were estimated by comparing the onset of eye position and velocity to that of the respective ideal responses. Computed latencies for each run varied between 8 and 22 ms with a mean of 11.9 ± 5.4 ms (mean ± SD; Table 1). Similar latency values were also obtained by examining average response profiles (Fig. 1, A and B). Even the earliest components of the response scaled with target distance (Fig. 2). Horizontal eye velocity for viewing distances of 40 and 10 cm differed by 1 SD as early as 12.8 ± 3.4 ms (right eye) and 12.6 ± 3.8 ms (left eye).

Forward and backward responses were asymmetric, with latencies being significantly smaller during forward compared with backward motion (Table 2; F(1,766) = 48.2, P 0.01). During forward motion, latency averaged 7.1 ± 9.3 ms (same for both eyes). During backward motion, latencies were longer for the abducting eye (i.e., left eye during left fixation and right eye during right fixation) compared with the adducting eye (18.9 ± 9.8 ms and 12.5 ± 6.3 ms, respectively; F(1,766) = 27.1, P 0.01). As seen from Fig. 3, the abducting eye often lagged behind during backward motion and did not start moving until several milliseconds after the adducting eye.

Initial acceleration gain

Initial eye acceleration gains were generally higher than unity and consistently larger during forward compared with backward or lateral motion stimuli (Table 3). The large initial acceleration gains tended to compensate for the response latencies such that the early eye movement response approached, albeit consistently incompletely, that required for maintaining visual acuity during the movement. In addition, initial acceleration gains were systematically and consistently higher for abducting than adducting eye movements. During lateral motion, for example, gains were significantly higher for the left eye during rightward motion (eliciting a leftward eye movement) and the right eye during leftward motion [rightward eye movement; F(1,398) = 34.7; P 0.01]. During forward/backward motion where the direction of the elicited eye movement depends on gaze direction, gains were also higher when the left eye rotated to the left and the right eye to the right (Table 3).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

These results suggest that translational VOR latency might be as short as that of the rotational VOR. Because of the inherent difficulty associated with converting a head translation to an eye rotation, we have considered it more appropriate to reference latency estimates to an "ideal" eye rotation rather than to head acceleration. For lateral motion and fixation at the nearest target, latencies averaged 11.9 ± 5.4 ms. The longer latencies previously reported during lateral motion in humans (~34-60 ms) could be at least partly due to shallower stimulus profiles and fixation of far targets (Bronstein and Gresty 1988; Bronstein et al. 1991; Gianna et al. 1997). The present values for horizontal eye movement latencies during lateral motion could be considered similar to those recently reported for vertical eye movements during free-fall (Bush and Miles 1996). Long latencies for the translational VOR were also implied from experiments using eccentric rotations (Crane et al. 1997; Snyder and King 1992). Because of small linear acceleration stimuli in these eccentric rotation studies, early changes in eye velocity could have been difficult to resolve. Moreover, even though it is likely that these rotation axis-dependent components were at least partly controlled by otolith signals, they might not actually represent a linear superposition of the translational VORs.

Surprisingly short latencies were also observed during forward movements (7.1 ± 9.3 ms). Response latencies were usually longer for backward motion. This could reflect a broader adaptation toward forward-directed eye-head coordination tasks that make up the bulk of visually guided behavior. An unexpected asymmetry was also found in the movement of the two eyes. Initial acceleration gains were consistently larger for abducting than adducting eye movements. In contrast, latencies were longer for the eye in abduction (albeit this difference in latency was only significant for backward head movements). Although a functional significance for these differences is unclear, they might reflect neural constraints in the yet unknown neural elements generating these responses. The higher initial acceleration gains for abducting eye movements, for example, could reflect higher gains in otolith-ocular pathways to the abducens compared with the medial rectus motoneurons.

Despite these asymmetries, the present data suggest that the short-latency labyrinthine signals could provide for a fast compensation and improved visual acuity during the early stages of translatory movements when optic flow visual mechanisms are inoperative.