Journal of Vestibular Research, Vol. 3, pp. 123-139, 1993 Printed in the USA. All rights reserved.

0957-4271/93 $6.00 + .00 Copyright © 1993 Pergamon Press Ltd.

THREE DIMENSIONAL EYE MOVEMENTS OF SQUIRREL MONKEYS FOLLOWING POSTROTATORY TILT

Daniel M. Merfeld, * Laurence R. Young, * Gary D. Paige,t and David L. Tomko:\:

*Man-Vehicle Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts; tDepartment of Neurology, University of Rochester, Rochester, New York;

tVestibular Research Facility, NASA Ames Research Center, Moffett Field, California Reprint address: Dr. D.M. Merfeld, Man-Vehicle Laboratory,

Massachusetts Institute of Technology, Cambridge, MA 02139

o Abstract - Three-dimensional squirrel monkey eye movements were recorded during and imme­diately following rotation around an earth-vertical yaw axis (1600/s steady state, l000/s2 acceleration and deceleration). To study interactions between the horizontal angular vestibulo-ocular reflex (VOR) and head orientation, postrotatory VOR alignment was changed relative to gravity by tilt­ing the head out of the horizontal plane (pitch or roll tilt between 15° and 90°) immediately after ces­sation of motion. Results showed that in addition to post rotatory horizontal nystagmus, vertical nys­tagmus followed tilts to the left or right (roll), and torsional nystagmus followed forward or backward (pitch) tilts. When the time course and spatial ori­entation of eye velocity were considered in three di­mensions, the axis of eye rotation always shifted toward alignment with gravity, and the postrota­tory horizontal VOR decay was accelerated by the tilts. These phenomena may reflect a neural pro­cess that resolves the sensory conflict induced by this postrotatory tilt paradigm.

::J Ke~'worrls - vestibulo-oeula" reflex; eye movements; otolith organs: semicircular canals: monkey.

Introduction

When an upright subject is rotated at a con­stant velocity about an earth vertical axis and then rapidly decelerated, a postrotatory hor­izontal nystagmus occurs which then gradu-

ally decays. Studies have shown that the time course of the horizontal postrotatory response is dramatically affected by the orientation of the subject relative to gravity (1-3).

Recent studies in monkeys (4), humans (5), and cats (6) have also demonstrated the pres­ence of a decaying vertical nystagmus as part of the postrotatory response following off ver­tical axis rotation (OVAR). In each of these studies, the vertical component of the eye movement shifted the plane of the response toward the earth-horizontal. Earlier studies (7,8), using low-velocity centrifuges, also re­ported the presence of a vertical nystagmus when the upright humans were facing the mo­tion or had their backs to the motion. When these paradigms were repeated with monkeys, the axis of eye rotation clearly shifted toward alignment with gravitoinertial force (9,10).

Since a common neural element (11) has been hypothesized to explain the time course of optokinetic afternystagmus (OKAN) and :c yield horizontai pcstrot2,LOry nystagmu~ whi:!: outlas:, th~ enc o~g:ar. r~sp:mse: c: s:m­ilar influence of gravity might be expected foi­lowing optokinetic stimulation. Research has confirmed this prediction in monkeys by showing that the axis of eye rotation during OKAN shifted toward alignment with gravity when the subject and optokinetic stimulus were tilted with respect to gravity (12), while human experiments have not confirmed this large axis shift (13).

RECEIVED 7 May 1992; REVISED MANUSCRIPT RECEIVED 16 September 1992; ACCEPTED 23 September 1992.

123

124

While many of these studies have indicated the presence of a shift in the axis of eye rota­tion toward alignment with gravity, none have measured three dimensional (3-D) eye orien­tation, corrected for cross-talk between the eye position signals, quantified the extent of the axis shift, and quantified the time course of the angular VOR decay following constant velocity rotation. In this study, we quantified ~he effects of tilt on the angular 'lOR in an :u:empt :c :n'lestiga~e :he interac:ior.. 1)1' se!1-sory information from ~he semicircular canals and the otoliths. 3-D eye movements were re­corded during and after reorientation of the monkey immediately following rapid deceler­ation from constant angular velocity around an earth-vertical axis. Postrotatory tilts were performed in both roll (left ear down/right ear down) and pitch (nose up/nose down), and the shift in the axis of eye rotation and the de­cay time constants were determined.

Methods

Experiment Design

Monkey preparation. Five adult male squir­rel monkeys (Saimiri sciureus) weighing be­tween 0.9 and 1.0 kg were the subjects. [These monkeys were also used in other studies (9,10,14).] A stainless steel bolt was used to stabilize the head of the monkey while allow­ing repeatable head positioning. The bolt was anchored to the skull using dental acrylic and miniature stainless steel screws which were in­verted and secured in keyhole slots drilled in the skull. Right eye position was measured using a search coil technique (15). Prefabri­cated coils made of insulated stainless steel wire were implanted on the right eye. An 11-mm diameter 3-turn coil was sutured to the sclera concentric to the limbus in the eye's frontal plane, and an 8-mm diameter 4-turn coil was sutured to the globe laterally and pos­teriorly to the frontal coil in the eye's sagittal plane. The leads left the orbit such that they did not interfere with normal movements of the eye. See Paige and Tomko (14) for a com­plete description of all surgical procedures.

D. M. Merfeld et al

Experimental paradigm. The upright monkeys were accelerated (1000/S2) about an axis aligned with gravity and the monkey's spinal axis to a constant velocity (160 0/s) using a servo-controlled rotatory device at the NASA Vestibular Research Facility. The monkey's head was positioned such that the plane of the horizontal canals was perpendicular to the axis of rotation and such that the axis of ro­tation was directly between the ler't and iight -·/estibulcS. When [he :lorizomai :mgular 'lOR was nearly extinguished (>60 seconds), the monkey was quicklY decelerated (100 0/5 2) ro a stop. While the postrotatory ocular response was still strong (1 to 3 seconds after stopping), the monkey was manually tilted (1 to 2 sec­onds) 15, 32, 45 or 90 degrees in one of the four primary directions: nose up, nose down, left down, and right ear down. The lights were turned on between trials and extinguished just prior to the start of a trial. Three of the mon­keys were tested on more than one occasion, with one monkey (B) tested five times. The or­der of the trials was randomized. To maintain monkey alertness, d-amphetamine (0.3 mg/ kg) was administered intramuscularly 15 min­utes before the beginning of each test session.

Eye movement recording. The eye measure­ment system was calibrated using two orthog­onal search coils similar to those implanted in the monkeys. The reference coils were mounted on a model eye secured to a non­metallic three-axis calibration jig, which was used to set the measurement sensitivities. The accuracy of this calibration procedure has pre­viously been verified (14,16).

Data Analysis

Eye kinematics. Using two search coils we can accurately estimate the 3-D orientation of the eye. For eye movements that do not deviate far from the primary eye position, the detec­tor circuits accurately indicate yaw, pitch, and roll of the eye. However, as the eye position deviates from the primary gaze position a number of nonlinear errors become important (15). These errors are enhanced by any mis-

3-D Eye Movements Following Postrotatory Tilt

alignment of the search coils on the globe (17). An electronic correction scheme had been de­veloped (14,16) to minimize these errors. As long as coil placement errors and eye displace­ments were limited to less than 10 degrees for each axis of eye rotation, the worst-case eye position measurement error was less than 100/0, and the mean error was less than 3%.

The electronic correction method proved inadequate for this study since eye displace­ments sometimes exceeded 20 degrees, lead­ing to worst case errors of greater than 20%. To minimize these errors, we developed and applied here an exact solution algorithm that corrected for coil placement errors (9,18). With eye displacements of up to 20 degrees, this algorithm yields worst case measurement errors of less than 10% (18).

Coordinate system. Dynamic knowledge of the three-dimensional eye orientation allowed us to calculate a three dimensional represen­tation of eye velocity (18). Eye velocity is a vector quantity and may be represented in a number of different coordinate frames. We chose to represent eye velocity in a head-fIxed Cartesian coordinate system to allow us to compare the eye responses directly to the motion stimuli. We defined the head-fixed, right-handed coordinate system such that the x-axis aligns with the naso-occipital axis, the y-axis aligns with the interaural axis, and the z-axis is perpendicular to the x- and y-axes. The positive direction is forward for the x-axis, leftward for the y-axis, and upward for the z-axis. In this Cartesian representation. eye veiocity 1:- repre<.::'mec by a ve·:tor wiIi'; components . I:. eacr: ~{' :~e ~hree· ·::i:-e:tian~ (x, y. and z). With the eye in the primary po­sition, horizontal eye velocity is represented by a vector aligned with the head fixed z-axis (w z ). Similarly, vertical (Wy) and torsional (wx) eye velocities are presented by vectors aligned with the monkey's y-axis and x-axis, respectively (Figure 1). This representation is different from the commonly used Fick rep­resentation, where eye velocity is represented by three quantities, called Euler rates, which represent rates of ocular yaw, pitch, and roll. Euler rates are a reasonable approximation to

125

eye velocity when eye movements are small «5 degrees) or when responses are limited to a single plane (yaw or pitch or roll), but are a poor choice during 3-D eye rotation because the Fick coordinate system moves with the eye, because the Euler rates are nonorthog­onal (except in the primary position), and because the representation of eye velocity de­pends upon eye orientation.

Using the Cartesian representation, the pri­mary post rotatory response following yaw ro­tation (around the z-axis) is a horizontal eye movement (wz ; also around the z-axis). To evaluate the extent to which postrotatory tilts caused the axis of eye rotation to shift away from the z-axis, we calculated the shift in the axis of eye rotation in both the roll (coronal) and pitch (sagittal) planes (Figure 1). The roll shift in the axis of eye rotation (Or) represents the angle by which the response shifts from the z-axis (8r = 0) toward the left or right ear (y-axis). This occurs when the response in­cludes a head vertical component a well as the predominant head horizontal component. The pitch shift in the axis of eye rotation (Op) rep­resents the angle by which the response shifts from the z-axis (Op = 0) toward or away from the nose (x-axis). This occurs when the re­sponse includes a torsional component as well as the predominant head horizontal compo­nent. To keep the axis of eye rotation aligned with gravity following postrotatory roll tilts of the head, the axis of eye rotation, measured relative to the head, must shift in roll (Or), with no change in pitch (Op)' Similarly, to keer the axis of eye rotation aligned with grav -

i:y foliowing DOS~iO:a:O:-y pitch tiltf of the je2.: th: c:.).:i~ .): e:': :-ota:io:: ::J.:..!S! shi:: ir:. pitch (Op), with no change in roll (Or)'

Fast phase removal. Since we were interested in the slow phase eye velocity as representa­tive of the VOR, we used a computer algo­rithm to identify and remove the fast phases of the nystagmus. The algorithm used low­order digital filters to calculate three-dimen­sional eye velocity and eye acceleration. The movement was marked as a fast phase and re­moved if the acceleration of the eye exceeded a threshold set by the user. All edited data

.~l

I.

~

.... I\) m

Right Ear Down Postrotational Tilt

i ~~~ ~;.=== ...... \ I ~ j - 100 \ i a: Ii o 50 li > " ii 0 !i

~ -50 1-100

-1 50 11--__ --L_.:...-_--1-. __ ---l

20

i 10 0

:!!. -10 a: -20 0

> -30 B -40 i! CD -50 >

-60

40 ~ 30 ! 20 a: 10 0 > 0 ii -10 c: a -20 l -30

-40

.......... ---~

...... VIew

~!n~ II'IHlVIrw

~ l;;' - ~ .. Eft VELOCITY

s ... v ....

Nose UI) I'osi. utational Tilt

200 ~I ---,'

~ 150 GI

:!?. 100 50

a: g ii 0 ~ -50 N

'~-100 z -150 ,-I ____ •

50

i 40 30

:!?. 20 a: 0 10 >

0 ii .2 -10 1:

:!: -20 -30 I ..1 .. _

40, 1--~ 30 CJI GI 20 :2. a: 10 g 0 ~ -10 .~ -20 {!. -30

-40

f\) -.J

'a 40 ~ GI 70 :Eo

30 60 .. .!! Rau. OF BYB .OTAllON AXJB ~ 20 ~ 50 IIOIIIIDI1aL

i 1 g rr ------c 40 r 32 tiagroo!lj

-~-t 30 - -.---- _ . ! .... WLCICIIY 0 , a: -10 20 '-a:

.,. mva.ocm f- -20 . GI 10 w : ,..

'0 -30 i w

I'rHlVllw '0 0 ................ i··--········· .. ·· .... · .............. · .... ·· .... ··· Or =tan-1(ro,/IDz) 1) -40 : '0 -10 --- a: a:

Ci 70 ! GI

:Eo 40 .. 60 . !l 30

I PITCH OF BYB ROTAllON AXIS ~ :~ hC~~2d~' •• '

~ 20 c 0 c 10 '; .2

t s 'Ii 0 0 0 , a: 20 a: -10 , '- GI f- IVE VlLocrn Jj' 10 w -20 '0 0 '0 -30 BIdcVIew

!1 -10 .s= -40 Op=tan-1( ro.,/roz) a: 0 50 100 150 J:! 150 a: 0 5(l . 100 Time [5) limq [!l)

Figure 1. Post-rotatory tilt trillls for '311"jPct R. Left column shows data obtained for a 32-degree postrotatory tilt to the right. Right column shows data obtained for a 32-degree postrotatory tilt backwrord l)illgrams In the center column define the quantities plotted. The angular yaw stimulus as well as the horizontal ocular response are shown In the first row . 'hp vertical and torsional ocular responses are shown In the second and third rows, respectively. The roll shift (II.) and the pitch shift (lip) in the axis of ey .. rotlltion are shown In the fourth and fifth rows, respectively. Dashed vertical lines mark the postrotatory tilt. Discontinuities In the fourth and fifth rows result from l1llcp.rtainty In the calculation of the axis of eye rotation when the horizontal response nears zero.

128

were manually reviewed and compared to the original data to verify the integrity of the al­gorithm and to remove any fast phases which were missed. This simple algorithm removed more than 95"70 of the fast phases; a limited number of fast phases were manually removed. All gaps were filled with a zero-order hold of the eye velocity preceding the saccade. [Data processing details including eye kinematics, coordinate systems, and fast phase removai are ;:>resemed ~lsew!1ere (9;~ 3) . :

Curve fits. Least squares regression l i 9) ':Vas used to perform linear fits to the data. To de­termine the dominant decay characteristics of the VOR responses, we estimated the decay time constant (1') of the responses by fitting the curve,

y = K + Ae-tlr, [1]

to the data (K, A, and l' are the parameters determined by the curve fit). This fit was ob­tained by minimizing the mean square error between the data and the curve [Levenberg­Marquardt algorithm (20)]. K and A were nec­essary to obtain an acceptable curve fit, but were not analyzed further.

To minimize the effect of the axis shift on the estimated time constant we began the curve fit after the primary off-axis response (vertical for roll tilts; torsional for pitch tilts) had reached its maximum magnitude. This curve fit ignores the complex dynamics that occur during the build-up of the off-axis re­sponse and concentrates upon the decay of the response.

Large secondary responses (for example, horizontal and vertical VOR in the left column of Figure 1) were relatively rare, but to be sure that these responses were not systematically affecting our determination of the dominant time constant we also fit the curve,

to some of the data showing significant sec­ondary responses. No large or systematic dif­ferences between the dominant time constant determined by the single exponential curve fit

D. M. Merfeld et al

(equation 1) and the double exponential curve fit (equation 2) were observed, so we chose to fit all the data with the single exponential.

Results

VOR Response

ReI! :,~;ds. A ::epresentative trial showing a 32-deg!"':!:: ;Jostfotatory roll Lilt co the right is shown in the plots in :he left I.:olumn of Fig­ure 1. The -15°/s spike in the torsional VOR at approximately 70 seconds is a compensa­tory response to the roll rotation. The hori­zontal response decays immediately following the tilt, but then recovers slightly before de­caying more gradually with a time constant of 7.3 seconds. Approximately 2.2 seconds after the tilt, a strong vertical ocular response ap­pears and decays with a time constant of 6.4 seconds. In this trial, the vertical slow phase velocity reached a peak magnitude of approx­imately 500/s. Note that the horizontal and vertical responses reverse at nearly the same time (approximately 90 seconds), indicating that the responses may be loosely coupled.

The last two rows in Figure 1 show the time course of the shift in the axis of eye rotation. Since the primary horizontal and vertical re­sponses are both negative, the roll shift of the axis of eye rotation is positive with a peak magnitude of approximately 45 degrees while the peak magnitude of the pitch shift is less than 10 degrees. The secondary horizontal and vertical responses following reversal are both positive, which again yield a positive shift in the axis of eye rotation. In this case the roll shift during the secondary response almost ex­actly equals the roll tilt of 32 degrees, while the secondary pitch shift is near zero. This re­sponse is characteristic of the roll tilt trials with the roll shift of the axis of eye rotation tending to align with gravity following the postrotatory tilt.

Pitch trials. A representative trial showing 32-degree nose up tilt following rotation is shown in the plots in the right column of Figure 1. The 400/s spike in the vertical response at ap-

3-D Eye Movements Following Post rotatory Tilt

proximately 70 seconds occurs as a compen­satory response to the pitch rotation. Shortly after the tilt, a strong torsional ocular re­sponse appears and decays with a time con­stant of 2.4 seconds, while the horizontal ocular response decays with a time constant of 4.4 seconds. In this trial, the torsional slow phase velocity reached a peak of approxi­mately 15°/s.

The last two graphs in the column show the shift in the axis of eye rotation. Since both the horizontal and torsional responses are nega­tive, the pitch shift of the axis of eye rotation is positive with a magnitude of approximately 20 degrees, while the roll shift is near zero. This pitch shift tends to align the axis of eye rotation with gravity following the nose up postrotatory pitch tilt. Comparable results were obtained during nose down tilts, with the sign of the torsional response always opposite that obtained during nose up trials. The re­versed sign of the torsional response led to a reversal in the calculated pitch of the axis of eye rotation, once again aligning the axis of eye rotation with gravity.

VOR Spatial Characteristics

The shift in the axis of eye rotation was de­termined for all trials for each monkey by cal­culating the average shift in the axis of eye rotation for one second after the peak verti­calor torsional response was obtained.

We chose to average over a full second to minimize the affect 0; noise on the estimated angle. We chose the first second after the off­axis peak because the measurement becomes very sensitive to noise as the responses, par­ticularly the horizontal, decay toward zero. The tilt angle measurement was relatively in­sensitive to alternate parameterizations, since both the horizontal and off-axis responses were simultaneously decaying. For example, we compared the chosen measure with one­second averages, starting one second and two seconds after the peak. No consistent or sig­nificant differences (always < 10"70) existed in these three measures. As an example, once the torsional velocity in the right column of

129

Figure 1 reaches a peak value and begins to decay, the pitch of the eye rotation axis re­mains relatively constant at approximately 20 degrees until the horizontal velocity decays to near zero. At this point, significant uncer­tainty in the calculated angle is evident in the increasing oscillations in the pitch of the eye rotation axis.

Roll trials. Figure 2 Panel A shows a plot of the roll shift (Or) of the axis of eye rotation for a single monkey (B). The slope of the least squares fit, shown as the thin line, is 0.96. The pitch shift (Op) of the axis of eye rotation is also plotted for these roll tilt trials. The slope of the least squares fit, shown as the thick line, is -0.03. Prior to head tilt, gravity and the axis of eye rotation are aligned. Following head roll tilt, the axis of eye rotation, relative to the head, shifts in a direction opposite the head roll. This axis shift with respect to the tilted head tends to maintain the orienta­tion of the angular VOR response relative to gravity.

Perfect alignment of the axis of eye rota­tion with gravity would require that the re­sponse roll shift equal the head roll tilt (slope of plus one), while the pitch shift remains zero (slope of zero). More generally, a shift toward alignment with gravity requires that the roll shift (Or) demonstrate a positive correlation with roll tilt. These data show that the axis of eye rotation shifts toward alignment with gravity following roll tilts. The roll shift in the response shows a significant (P < 0.001) pos­itive correiation (r = 0.984) with the head ro ll ~ilt, while ~he pi,.::h da~c. are no: significantl~

more, since the slope 0; th, iinear fit to the roll shift (0.96) is not significantly different from one, we can state that the axis of eye ro­tation closely aligned with gravity.

Pitch trials. Figure 2 Panel B shows a plot of the pitch shift (Op) of the axis of eye rotation as a function of head pitch for a single mon­key (B). The slope of the least squares fit, shown as the thick line, is 0.46. The response roll shift (Or) is also plotted, and the slope of the least squares fit, shown as the thin line, is -0.030.

130

i ~ ell

~ a ~ c ~

I.l ::-, ~ ... 0

~

40

30

20

10

D. M. Merfeld et al

Shift for Roll Tilts

y = 1.7+ 0.96· x ~, / ,/

+,/

+ A

or: :-; ," x x

i . -·~'--·~-1lI'-··'--'~--:---'~---"'''"·''''''-~·~'·~'-1 do. '"'.'N~ ' •. , ... ' ' w.<,«uu_<., ....

X

-10

-20

-30 /'

40 -40 -30

(Left Ear Down)

-20 -10

y :0.3 - 0.03 * .,J '

o Tilt [deg]

Shift for Pitch Tilts

10 20 30 40 (RighI Ear Down)

25r-----~------,_------r_----~------~----~~----~----__.

20

15

10

o

-5

-10

-15

-20

+

+ +

+

B

y = 3.0 + 0.46 • x

+

.-.................. .... ...... J........ . ....... ! ..... --......... . + +

y = 2.8 - 0.03 • x +

-25~----~------~------~----~------~------~----~----~ -40 -30 -20 -10 o 10 20 30 40

(Nose Down) (Nose Up) Tilt [deg]

Figure 2. Shift in the axis of eye rotation. Panel A shows the calculated shift in the axis of eye rotation for monkey B during postrotatory tilts to the left and right. The roll of the axis of eye rotation (8, from Figure 1) for each trial (+) Is plotted versus the magnitude of the tilt. Thin line represents the least squares fit. The pitch of the axis of eye rotation (8p from Figure 1) for each trial (x) Is also plotted. The thick line represents the least squares fit. Panel B shows the calculated shift in the axis of eye rotation for monkey B during tilts forward and backward. As before, the roll of the axis of eye rotation (8,) for each trial (+) Is plotted versus the magnitude of the tilt. The pitch of the axis of eye rotation (8p) for each trial (x) Is also plotted. [Panel A after Merfeld and Young (18).]

3-D Eye Movements Following Postrotatory Tilt

Analogous to the above head roll tilt exper­iments, alignment of the axis of eye rotation following head pitch requires that the re­sponse pitch shift (Op) equal head pitch (slope of one), while the response roll shift (Or) show no correlation (slope of zero). More gen­erally, a shift toward alignment with gravity requires that the pitch shift (Op) demonstrate a positive correlation with head pitch. Again, the data show that the axis of eye rotation shifted toward alignment with gravity. Statis­tically, the pitch response shows a significant (P < 0.001) positive correlation (r = 0.985), while the roll response is not significantly (P > 0.4) correlated (r = -0.124) with the pitch tilt. Unlike the roll experiment, the re­alignment of the ocular response axis with gravity following head pitch was incomplete and was roughly half (0.46) that required for alignment with gravity.

Comparison of pitch and roll responses. Ta­ble 1 shows a summary of the axis shift anal­ysis for each monkey during roll tilts and pitch tilts. During roll tilt, data for each monkey clearly indicate that the axis of eye rotation shifted toward alignment with gravity, since

131

the roll shift of the axis of eye rotation for each of the five monkeys has a high positive correlation (r> 0.96) with the roll tilt. Simi­larly, the pitch shift data from each monkey during pitch tilt show a high positive correla­tion (r > .98), again indicating that the axis of eye rotation shifted toward alignment with gravity.

Data from three of the five monkeys (B, E, F) tested with postrotatory roll tilts indicated that the axis of eye rotation closely aligned with gravity since the slopes of the linear fit to the roll shift were greater than 0.95. In con­trast, none of the four monkeys (A, B, E, F) tested with postrotatory head pitch displayed a shift in the axis of eye rotation that fully aligned with gravity. The slope of the linear fit never exceeded 0.9. In fact, data from three of the four monkeys (B, E, F) during pitch tilts showed that the slope of the linear fit was less than 0.5. On the whole, the pitch shifts of the axis of eye rotation following pitch tilts were less than the roll shifts of the axis of eye rotation following comparable roll tilts. Only one monkey (A) had comparable response axis shifts for roll and pitch tilts of the head.

Table 1. Axis Shift Regression Analysis

8p versus angle of tilt 8r versus angle of tilt

Monkey n s r p s r

Roll Tilts (as in Figure 2A)

A 9 -0.007 -0.059 0.16 P> 0.400 0.823 0 .989 B 15 -0 .026 -0 .202 0.74 P > 0.400 0 .955 0 .984 0 12 -0.025 -0.975 13.84 p> 0.001 0 .773 0 .962 E 3 -0.105 -(.985 5.73 P> 0 400 0 .998 0.998 F 3 0.02:' C.796 1.31 p> (, .400 O.96L 0 .996

Pitch Tilts (as in Figure 2B)'

A 1 • 0 .870 0 .997 37.46 P < 0 .001 0 .133 0 .605 B 15 0.462 0.985 20.78 P < 0.001 -0 .030 -0.124 E 3 0 .399 0.988 6 .31 P < 0 .050 0 .090 0 .720 F 3 0.422 0.991 7.46 P < 0 .050 0 .060 0 .490

• Monkey 0 not included because torsion coil failed before forward/backward tilt trials were completed. Notation:

n is the number of trials. s is the slope of the linear regression. r is the sample regression coeffiCient. t is the calculated t-value representing the significance of the correlation. P represents the statistical significance of the correlation. . Op and Or represent the axis shift of the eye as shown in Figure 1 .

17.54 20.01 11.28 17.84 i -; .07

2 .28 0.45 1 .04 0.56

p

P < 0.001 P < 0 .001 P < 0 .00 1 P < 0.050 P < C.10C

P> 0 .05 P> 0.40 P> 0 .40 P > 0.40

132

Time Course of the VOR Decay

Roll trials. The time course of the VOR de­cay was also affected by the postrotatory tilt. Figure 3 shows the effect that the postrotatory tilt has on the decay of the VOR. In Figure 3 Panel A, the horizontal time constants are plotted for varying head roll tilts. Despite a great deal of variability, the data indicate that the time constant decreases as the angle of head tilt increases. The dashed line shows the :east ,quares :inear :':r. Pus :iena 'NLlS ,llgruy

significant t? < 0.001). Since the mlsaiign­mem of gravity :md [he POSt rotatory head aus (z-a'{is) will vary with the sine of the magni­tude of the tilt angle (Ixl), we also used lin­ear regression to fit the equation:

y = yeO) + A sine Ixl). [3]

The solid curve shows this fit. This simple nonlinear curve provides a better fit to the data as measured by a decreased mean square error and an increased regression coefficient (Table 2).

Figure 3 Panel C shows the vertical decay time constants versus the horizontal decay time constants during roll tilt. The time con­stants are positively correlated (r = 0.82) with high significance (P < 0.(01). The slope of the regression (s = 0.77) is somewhat less than one, indicating that, in general, the vertical re­sponse decayed a little more rapidly than the horizontal response.

Pitch trials. Figures 3 Panel B shows the hor­izontal VOR time constants for pitch tilts. Once again the data indicated that the hori­zontal time constant decreased as the angle of tilt increased to 90 degrees. Similar to the data for roll tilts, these data indicated that the non­linear fit is better than the simple linear fit (Table 2) and that the downward trend was highly significant (P < 0.001).

Figure 3 Panel D shows the torsional de­cay time constants versus the horizontal de­cay time constants during pitch tilts. Again the time constants were positively correlated (r = 0.78) with high significance (P < 0.001), and the slope of the regression (s = 0.76) was,

D. M. Merfeld et al

again, somewhat less than one, indicating that, in general, the torsional response de­cayed a little more rapidly than the horizon­tal response.

Comparison oj pitch and roll responses. The similarity of the linear regression for roll tilts (Figure 3 Panel C) and pitch tilts (Figure 3 Panel D) and the similarity of the effects of tilt on the time course of [he horizontal an­gular VOR (Figure 3) suggest that the plane 0f r:he head tilt does not dramatically affect :he ~:me course of ~he response. This is m .:on­[rast cO the spatial charQcteristics, which show that the magnitude of the shift in the axis or' eye rotation is less for pitch than for roll tilts.

Summary

These experiments quantified eye move­ment responses by determining the spatial and temporal characteristics of 3-D ocular re­sponses following roll and pitch postrotatory tilt and showed that whole body tilts affect the postrotational angular VOR responses of monkeys. Most dramatically, the axis of eye rotation was shown to shift toward alignment with gravity following tilt. Quantitative anal­ysis indicated that the axis shift was greater for roll than it was for pitch tilts (Figure 2 and Table 1). These results are similar to the pre­vious quantitative findings showing that post­rotatory eye movements following OVAR shifted toward an earth-horizontal plane (4-6).

The time course of the decay of the hori­zontal angular VOR was also affected by the postrotational tilt in both pitch and roll (Fig­ure 3 and Table 2). Specifically, the decay time constant was shown to decrease as the angle of tilt increased, independent of the plane (pitch or roll) of the postrotatory tilt. This is consistent with previous research (2,3) that did not take the axis shift into account when de­termining the decay time constants. Further­more, the decay time constant of the off-axis VOR response (vertical for roll tilt; torsional for pitch tilt) was correlated with the decay time constant of the horizontal VOR response with the slope of the correlation less than one. This indicates that the decay of the off-axis

w w

30, Horizontal Time ConSIllnt for Roll Tilts 20

1

Time Constant Correlation for Roll Tilts , , , A C

25 x I I 15 " )(

20 ~

.-'-)( ...... )(

~ S XX .-' !O .-'-

:< g '--Xx u )( x x..lS-X-;

U 10 - _ _ _ _ _ 1I jY = 17 - 0.17· x .§ ~,l'

u x x -....._ - -x __ E-o -'C x ~ )C --..._ B 5 x l! ·G 'So' - "x 5 14 x - x

14 > ,1( -- .-,-' x y=0.4+0.77*x Y = 18 - 13.R ... ~in()() --

0' ~-~

00 0 10 20 30 ,10 50 60 70 80 90 5 10 15 20

Tilt rdr~rees] Horizontal Time Constant (seconds]

30 Horizontal Time Constant for Pitch Tilts

201

Time Constant Correlation for Pitch Tilts

B D

25 j

I ! 15 x

20 ~ .-' u x ~ g _.-j(

~ " 10 x '( U

g u x

_. __ !l<. ___ jY = 17 - 0.19 * x ~ )( )(,-' )(

U "'XV 1\.1

1 )( -~ .5 5 )( ,.-E-o II - ,- x

5 ,7( )( lE .-,-')()( y=0.5+0.76*x

'- E-o )(

0' r= 17 - ~tt.4'" ~i~('t) - -- 0

0 10 20 30 1n 50 60 70 80 90 0 5 10 15 20

Tilt fdf'grees] Horizontal Time Constant [seconds]

Figure 3. Effect of postrotatory tilt 011 von time constants. Panel A shows a plot of the decay time constant for the horizontal ocular response versus the post­rotatory tilt angie for roll tilts, while Pnnel B shows a piot of the decay time constant for the horizontal ocular response versus the postrotatory tilt angle for pitch tilts. Dashed lines show the le~!;t c:q1HHPo; linear fits. Solid curves show the least squares fit to the equation, y = y(O) + A sin(lxl), where Ixl is the magnitude of the tilt angle. Panel C shows the clpr.ny time constant for the vertical ocular response plotted against the decay time constant for the horizontal ocular response for postrotatory roll tilts. Panel 0 show!; the decay time constant for the torsional ocular response plolted against the decay time constant for the horizontal ocular response for postrotatory pitch tilt!' I inp.s represent the least square linear fits to the data •

.,..

134 D. M. Merfeld et al

Table 2. Correlation of Horizontal VOR Time Constant with Tilt

Line fit [y = y(O) + s·x) Sine fit [y = y(O) + A·sin(\xl))

s y(O) MSE r P A y(O) MSE r P

Roll tilts -0.17 174.48 -0.696.50 P<0.001 -13.8 184.23 -0.737.308 P<0.001 Pitchtilts -0.19 17 4.37 -0.70 6.13 P<0.001 -14.4 17 4.10 -0.74 6.923 P<0.001

Notation: s is the siope of the linear regression. y(O) is the y·intercept. A is the amplitude of the sine correlation. ,'viSE is ihe mean souare error. : is ihe samole regression .:oeriiclem. : is the calculatec :'lJalue ,epresenting the significance of 'he ;orreiation. P represents the slaiistical significance oi the correlation.

response may be somewhat faster than the de­cay of the horizontal VOR.

Discussion

VOR Spatial Characteristics

While the time course of the horizontal VOR was not strongly dependent on the plane (pitch or roll) of the postrotatory tilt, the plane of rotation did significantly affect the direction of the eye movement responses, since large vertical responses followed roll tilt and large torsional responses followed pitch tilt. In all trials the off-axis responses shifted the axis of eye rotation toward alignment with gravity. Previous studies are consistent with this finding. For example, vertical responses were observed during eccentric yaw rotations of humans about an earth-vertical axis (7,8), When this paradigm was repeated with mon­keys, the vertical component of the response was shown to always shift the axis of eye ro­tation toward alignment with gravitoinertial force (9,10). Studies have also shown that postrotatory eye movements shift toward an earth-horizontal plane immediately following OVAR stimulation (4-6). This is consistent with the 3-D measurements, showing that the axis of eye rotation aligns with gravity.

We propose that the shift in the axis of eye rotation indicates that the eNS resolves this sensory conflict by shifting a central estimate of rotation toward alignment with gravity

(Figure 4) (9,10,18). In order to discuss this hypothesis we must clearly and unambigu­ously define what we mean by "central esti­mate of rotation" and "sensory conflict".

We define the "central estimate of rotation" as a neural representation of angular velocity that is derived from the sensory afference. It has long been known that the time course of angular VOR responses differs significantly from the dynamics of the semicircular canals, and that visual stimuli can modify the VOR responses (11,23,24). For these reasons, Rob­inson explicitly hypothesized that the angular VOR reflects a central estimate of self-rota­tion (24), which is induced by a process of sen­sory integration that includes visual as well as semicircular canal signals. Since the neural processing required to transform retinal im­ages to an estimate of self-motion exceeds the neural processing required to transform oto­lith afference, it is not difficult to believe that otolith cues could playa similar role in de­termining a central estimate of self-rotation. OVAR studies, which show that the rotation of gravity can induce a bias response compo­nent that partially compensates for the actual rotation (4-6,25,26), and postrotatory tilt studies reveal that otolith signals may influ­ence the central estimate of rotation.

We, like Guedry (21,22), define "sensory conflict" to be an incongruity between at least two sensory systems. In this case, the sensory conflict occurs because the semi circular ca­nals and the otoliths provide conflicting infor­mation regarding self motion. Immediately

3-D Eye Movements Following Post rotatory Tilt

following deceleration the otolith signal indi­cating gravity is aligned with the canal signal indicating rotation, but during the postrota­tory tilt the semicircular canals tilt with the body and continue to indicate rotation about the body's z-axis, while the otoliths indicate the constant orientation of gravity which is now tilted with respect to the body. If, in fact, the tilted body were rotating about the body's z-axis (yaw) as indicated by the semicircular canals, then the otoliths would be expected to

135

measure the rotation of gravity about the yaw axis. Conversely, if gravity were truly fixed with respect to the body, then rotation could only occur about an axis aligned with gravity. Therefore, the sensory signals indicating the state of motion are incongruent and produce "sensory conflict."

We also define two ways to resolve the sen­sory conflict. Temporally, the conflict can be resolved by reducing the central estimate of rotation (Figure 4 Panel B). This resolution

Conflict Resolution

A

Central estimate of rotation (~) not aligned with gravity (g) , Sensory Conflict

B C

+~

~ " 19

No central estimate of rotation

~~ ~~} I ~' ... . ,,~ f "f

Central estimate of rotatior. (~) aligned with gravity (g)

t + Sensory Concurrence Sensory Concurrence

Figure 4. Sensory conflict resolution. Panel A demonstrates the sensory conflict created when the central estimate of rotation (~) doesn't align with the head-fixed orientation of gravity (g) following postrotatory tilt. Sensory conflict results since the central estimate of rotation should lead to a change In orientation of gravity with respect to the head, while the otoliths just as clearly indicate that the orientation of gravity with respect to the head is constant. One way to resolve the conflict is to eliminate the central estimate of rotation (Figure 4 panel B). Another way to resolve this conflict Is to align the central estimate of rotation with gravity (Figure 4 panel C), since the rotation no longer leads to a change in the orientation of gravity. The two resolu­tion processes are not mutually exclusive.

136

process may manifest itself as a weaker VOR response or as a more rapid decay of the re­sponse. Spatially, the conflict can be resolved by modifying the central estimate of rotation such that the axis of the central estimate aligns with gravity (Figure 4 Panel C). This resolu­tion process may manifest itself by realigning the axis of the VOR with gravity.

We might expect that Ihe size of the eye re­sponse axis shi ft would jepend upon 'N.hether :he ·Jff-axis response ~s verricai or WfSlOnal, since the gain of the vertical and ~orsionai VORs are known to differ (27). By quantita­tively analyzing the 3-D axis of eye rotation, we found that the shift in the axis of eye rotation was less for pitch than for roll tilts (Figure 2; Table 1). For roll tilts (left ear down/right ear down), where the off-axis re­sponse was primarily vertical, the axis of eye rotation aligned closely with gravity in the ma­jority of the monkeys. For pitch tilts (for­ward/backward), where the off-axis response was primarily torsional, the axis of eye rota­tion never aligned with gravity. We hypothe­size that this difference in the axis of eye rotation, with roll shift responses greater than pitch shift responses, might at least in part be due to the fact that the vertical angular VOR is 30070 to 100070 larger than the torsional an­gular VOR (27). Since the horizontal and ver­tical angular VOR gains are comparable (27), the measured axis of eye rotation might pro­vide an accurate estimate of the axis of self ro­tation for roll tilts. On the other hand, since the torsional VOR gain is less than the hori­zontal VOR gain, the measured axis shift for pitch tilts may not quantitatively represent the shift in the central estimate of rotation.

While the magnitude of the axis shift may not always represent the magnitude of the shift in the central estimate of rotation, the time course of the axis shift might still repre­sent the time course of the shift in the central estimate of rotation. In the absence of percep­tual or other physiological measures of the ap­parent veltical, the time course of this axis shift may provide the best estimate for the time course of the shift in the central estimate of rotation and, in turn, of the time course of the shift in the central estimate of gravity (18).

D. M. Merfeld et al

Axis shifts similar to those measured follow­ing postrotatory tilt have also been measured following OVAR (2,4-6). The qualitative sim­ilarity of the responses following off vertical axis rotation (OVAR) and following postro­tatory tilt (3,9,18) indicates that the exact par­adigm used to reach a tilted postrotatory orientation might be less important than that such an orientation be obtained. In other words the rapid decay and axis shift depend onlY JPon the presence and magnitude of the competing otolith signal.

Axis shifts have also been noted following optokinetic stimulation when monkeys and the optokinetic stimulus were tilted with re­spect to gravity (12). Raphan and colleagues (23) hypothesized that optokinetic stimulation "charges" a velocity storage mechanism which is responsible for optokinetic nystagmus, op­tokinetic afternystagmus, and vestibular nys­tagmus. It has also been suggested that canal and visual cues might be merged to yield a sin­gle central estimate of head angular velocity (24). If we assume that the velocity storage mechanism is charged by both canal and vi­sual cues and that a central estimate of head angular velocity results, then the response in­duced during tilted OKAN should be similar to that induced during postrotatory tilts and, therefore, should yield the observed shift in the axis of eye rotation.

Human studies have not confirmed the presence of a strong shift in the axis of eye ro­tation following optokinetic stimulation (13) or following postrotatory tilt (28). Other stud­ies, using eccentric rotation stimuli, have shown that the axis shift is more gradual and weaker for humans (7,8) than it is for mon­keys (9,10). Similarly, the axis shift following constant velocity OVAR stimulation has been shown to be weaker in humans (5) than in monkeys (4). This, along with that fact that OKAN is weaker in humans than in monkeys, may partially explain these species differences. Furthermore, some of the paradigms involve another difference. The monkey postrotatory tilts were passive, while the human postrota­tory tilts were active. Perhaps the active pro­cess of tilt interferes with the axis shift due to information from either the neck receptors in-

3-D Eye Movements Following Post rotatory Tilt

dicating tilt of the head with respect to the torso or a copy of the efferent signals sent to the neck muscles.

Time Course of the VOR Decay

Horizontal VOR. Earlier studies, while not noting the shift in the axis of eye rotation, documented the effect of gravity on the hori­zontal postrotatory response by showing that the postrotatory nystagmus elicited by earth­vertical and earth-horizontal rotation dif­fered. Specifically, horizontal postrotatory nystagmus following a velocity trapezoid was shown to be weaker and to decay more rap­idly when subjects were rotated around an earth-horizontal axis (1,2,25). Guedry (21,22) suggested that the difference was due to the sensory conflict induced following constant­velocity earth-horizontal rotation, since the otoliths correctly indicate the absence of ori­entation changes with respect to gravity while the postrotatory response of the semicircular canals indicates rotation after all motion has stopped (Figure 4 Panel A). Guedry proposed that the weaker response and more rapid de­cay were direct effects of the conflict, since the sensory conflict would be reduced or more rapidly eliminated by diminishing or eliminat­ing the central estimate of rotation.

Benson (3) provided evidence supporting Guedry's hypothesis when he tilted his sub­jects immediately following constant velocity rotation about a vertical axis. He found that the horizontal postrotatory response decayed more rapidl~ foli owing tiit. and that the rate of horizontal VOR deca~ W2." ciependen~ or. the amount of tilt. However, the shift in me axis of eye rotation was not yet documented, so Benson could not take this shift into ac­count while determining the decay time con­stant of the horizontal VOR. Our analysis (Figure 3) takes the axis shift into account by beginning the curve fit after the initial rapid decay (see Figure 1) and indicates that, in general, the time constant does decrease as tilt increases. It is important to note that the time constant did not always decrease with in­creased tilt, but the general trend is compel-

137

ling (Figure 3 Panels A and B). The data were similar for both roll tilts and pitch tilts, with the fastest decay occurring after tilts of 90°. These data are consistent with a process of sensory conflict resolution, since larger tilts, at least up to 90°, increase the sensory conflict induced by the graviceptor afference (29).

The data shown in Figure 1 provide exam­ples which qualitatively support this interpre­tation of the results. Note that immediately following tilt the horizontal response begins to rapidly decay. This is consistent with Gue­dry's hypothesized process for resolving the sensory conflict. However, the horizontal re­sponse recovers and decays much more grad­ually once the off-axis response (vertical-left column; torsional-right column) is initiated. This might indicate that the sensory conflict that initiated the rapid decay of the response was reduced once the axis shift was underway. If we assume that the aXis shift represents a manifestation of sensory conflict resolution and that the rapid decay is another manifes­tation of the resolution process, then the data shown in Figure 1 are consistent with both conflict resolution hypotheses (Figure 4 Panels B and C).

Guedry's hypothesis provides one explana­tion for the effect of gravity on the decay time constant of the VOR, but other mechanisms have been proposed to explain this effect. For example, the velocity storage models (30,31) provide good phenomenological explanations for the observed gravitational effects by as­suming that model parameters directly depend upon the orientation of gravity. In contrast to these models, another modeling approach (32) 2.ssume~ that :!1e signals frarr. tht- various sense organs are aynamicaiiy me;-geC: lC 17,; ;; ­

imize sensory conflict. This dynamic interac­tion of sensory signals appears more typical of our current knowledge of neural process­ing than is the idea of sensory signals dynam­ically and instantaneously changing the characteristics of the VOR neural network.

Off-axis VOR responses. The time constant of the off-axis response (vertical for roll tilts; torsional for pitch tilts) was highly correlated with the time constant of the horizontal re-

I i i

./

138

sponse (Figure 3), and the slope of the corre­lation (approximately 0.75) was similar for pitch and roll. A previous study, with tilted subjects rotated about an earth-vertical axis at constant velocity, showed a similar effect (33). Analysis showed that the vertical com­ponent of the response had a shorter time con­stant than the horizontal response. This was interpreted as indicating that the time constant ror vertical velocity storage is less than the :ime constanr :'or '1Corizo m a . c!CCW! 5IO r:J.ge even when both responses are m muimed simui­taneously. The same mechanism might explain the decay characteristics of the off-axis VOR responses following post rotatory roll tilt.

Summary

We measured three-dimensional eye move­ments and quantitatively showed that the axis of eye rotation always shifted toward align­ment with gravity following postrotatory tilt. We also showed that the time constant of the

D. M. Merfeld et al

horizontal postrotatory response was reduced depending on the magnitude of the postrota­tory tilt. These observations led us to hypoth­esize that the sensory conflict induced by postrotatory tilt might explain the VOR axis shift which has been observed in monkeys (4,9,10,12,18), humans (5), and cats (6). This process of conflict resolution may work in conjunction with the process described by Guedry (21 ,22) to reduce sensory conflict and :nay )r _ "we J. ,Jnyslc::-.i :xpianmion :'or :he ef­fect or" !5ravity on '/eloc:ry storage.

Acknowledgments-This research was supported by NASA contracts NASW-3651, NAG2-445, NAS9-16523, the NASA Graduate Student Researchers Program, the GE Forgivable Loan Fund, and by NASA Space Medicine Tasks 199-16-12-01 and 199-16-12-03. We also wish to thank the staff at the NASA Ames Vestibular Research Facility for their support, Sherry Modestino for her assistance through all phases of this project, and Dr. Mark Shelhamer for reviewing a preliminary draft of the manuscript.

REFERENCES

1. Benson AJ, Bodin MA. Comparison of the effect of the direction of the gravitational acceleration on post­rotational responses in yaw, pitch and roll. Aerospace Med. 1966;37:889-97.

2. Benson AJ, Bodin MA. Effect of orientation to the gravitational vertical on nystagmus following rota­tion about a horizontal axis. Acta Otol. 1966;61: 517-26.

3. Benson AJ. Modification of the per- and post-rota­tional responses by the concomitant linear accel­eration. In: 2nd Symposium on the Role of the Vestibular System in Space Exploration. Washington, DC: US Government Printing Office; NASA SP-U5. 1966:199-213.

-+. Raphan T, Cohen B, Henn V. Effects of gravity on rotatory nystagmus in monkeys. Ann N Y Acad Sci. 1981;374:44-55 .

5. Harris LR, Barnes GR. Orientation of vestibular nys­tagmus is modified by head tilt. In: The vestibular system: neurophysiologic and clinical research, Graham, Kemink, (ed) . New York: Raven Press; 1987:539-48.

6. Harris LR. Vestibular and optokinetic eye movements evoked in the cat by rotation about a tilted axis. Exp Brain Res. 1987;66:522-32.

7. Lansberg MP, Guedry FE, Graybiel A. Effect of changing resultant linear acceleration relative to the subject on nystagmus generated by angular acceler­ation. Aerospace Med. 1965;36:456-60.

8. Crampton GH. Does linear acceleration modify cu­pular deflection? In: 2nd Symposium on the Role of the Vestibular System in Space Exploration. Wash­ington, DC: US Government Printing Office; NASA SP-Il5. 1966: 169-84.

9. Merfeld OM. Spatial orientation in the squirrel mon­key: an experimental and theoretical investigation [ph.D. thesis]. Cambridge, Massachusetts: Massachu­setts Institute of Technology, 1990.

10. Merfeld OM, Young LR, Paige GO, Tomko DL. Spatial orientation of VOR to combined vestibular stimuli in squirrel monkeys. Acta Otol. 1991;481 (Suppl):287-92.

11. Cohen B, Henn V, Young LR. Visual-vestibular in­teraction in motion perception and the generation of nystagmus. 1980; Neurosciences Research Program Bulletin. Cambridge, Massachusetts: MIT Press; 18: 459-651.

12. Raphan T, Cohen B. Organizational principles of ve­locity storage in three dimensions; the effect of grav­ity on cross coupling of optokinetic after-nystagmus (OKAN) Ann N Y Acad Sci. 1988;545:74-92.

13. Lafortune SH, Ireland OJ, Jell RM. Suppression of optokinetic velocity storage in humans by static tilt in roll. J Vestibular Res. 1991;1:347-55.

14. Paige GO, Tomko DL. Eye movement responses to linear head motion in the squirrel monkey; 1: Basic characteristics. J Neurophys. 1991;65:1170-82.

15. Robinson DA. A method of measuring eye movement

3-D Eye Movements Following Post rotatory Tilt

using a scleral search coil in a magnetic field. IEEE Trans Biomed Electronics. 1963;BME-I0:137-45.

16. Haddad FB, Paige GO, Ooslak MJ, Tomko OL. Practical method, and errors, in 3-D coil eye move­ment measurements. Invest Ophthalmol Vis Sci. 1988;29{Suppl):342.

17. Ferman L, Collewijn H, Jansen TC, Van den Berg AV. Human gaze stability in the horizontal, vertical and torsional direction during voluntary head move­ments, evaluated with a three-dimensional scleral in­duction technique Vision Res. 1987;27:811-27.

18. Merfeld OM, Young LR. Three dimensional eye mea­surement following post-rotational tilts in monkeys. Ann N Y Acad Sci. 1992;656:783-94.

19. Larson RJ, and Marx ML. An introduction to math­ematical statistics and its applications. Englewood Cliffs, New Jersey: Prentice-Hall. 1986.

20. Press W, Flannery B, Teukolsky S, Vetterling W. Nu­merical recipes: the art of scientific computing. Cam­bridge: Cambridge University Press. 1986.

21. Guedry FE. Orientation of the rotation-axis relative to gravity; its influence on nystagmus and the sensa­tion of rotation. Acta Otolaryngol. 1965;60:30-49.

22. Guedry FE. Psychophysiological studies of vestibu­lar function. In: Contributions to sensory physiol­ogy. New York: Academic Press; 1965:63-135.

23. Raphan T, Matsuo V, Cohen B. A velocity storage mechanism responsible for optokinetic nystagmus (OKN), optokinetic after-nystagmus (OKAN) and vestibular nystagmus. In: Baker R, Berthoz A, eds. Control of gaze by brain stem neurons, developments in neuroscience, vol 1. Amsterdam: Elsevier/North Holland Biomedical Press; 1977:37-47.

24. Robinson OA. Vestibular and optokinetic symbiosis: an example of explaining by modelling. In: Baker R, Berthoz, A, eds. Control of gaze by brain stem neu-

139

rons, developments in neuroscience, vol 1. Amster­dam: Elsevier/North-Holland Biomedical Press; 1977:49-58.

25. Correia NJ, Guedry FE. Modification of vestibular responses as a function of rate of rotation about an earth-horizontal axis. Acta Otol. 1966;62:297-308.

26. Benson AJ, Bodin MA. Interaction of linear and an­gular accelerations on vestibular receptors in man. Aerospace Med. 1966;37:144-54.

27. Bello S, Paige GO, Highstein SM. The squirrel mon­key vestibulo-ocular reflex and adaptive plasticity in yaw, pitch, and roll. Exp Brain Res. 1991;87:57-66.

28. Fetter M, Tweed 0, Misslisch H, Hermann Vi, Koe­nig E. The influence of head reorientation on the axis of eye rotation and the vestibular time constant dur­ing postrotatory nystagmus. Ann NY Acad Sci. 1992; 656:838-40.

29. Udo de Haes HA, Schone H. Interaction between statolith organs and semi-circular canals on appar­ent vertical and nystagmus. Acta Otol. 1970;69: 25-31.

30. Hain TC. A model of the nystagmus induced by off vertical axis rotation. Bioi Cybern. 1986;54:337-50.

31. Raphan T, Cohen B. Multidimensional organization of the vestibulo-ocular reflex (VOR). In: Keller EL, Zee OS, eds. Adaptive processes in visual and ocu­lomotor systems. New York: Pergamon Press. 1986: 285-92.

32. Merfeld OM, Young LR, Oman CM, Shelhamer MJ. A multidimensional model of the effect of gravity on the spatial orientation of the monkey. J Vestibular Res. 1993;3: 141-161.

33. Hain, TC, Buettner UW. Static roll and the vestib­ulo-ocular reflex (VOR). Exp Brain Res. 1990;82: 463-71.

It

,I

I ,I .1

[, '.'