Gap-overlap effects on latencies of saccades, vergence and combined vergence-saccades in humans

~

Pergamon

0042-6989(95)00073-9

Vision Res. Vol. 35, No. 23/24, pp. 3373-3388, 1995 Copyright © 1995 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0042-6989/95 $9.50 + 0.00

Gap-overlap Effects on Latencies of Saccades, Vergence and Combined Vergence-Saccades in Humans MINEO TAKAGI,*t ELLIOT M. FROHMAN,* DAVID S. ZEE*~: Received 2 December 1994; in revised form 8 March 1995

We examined the effect of gap-overlap stimuli on the distribution of latencies for pure saccades, pure vergence and combined saccades and vergence in three normal subjects. With the gap stimulus, a distinct peak of"express saccades" occurred, both with and without associated vergence, but a distinct "express vergence" response was not identified. Nevertheless, with the gap stimulus there was a decrease in vergence latencies (17 msec), but less so than for saccades (41 msec). In the combined paradigm the gap effects on saccades and vergence resembled those for each component made alone. In addition, the latencies of the saccade and vergence components were linearly correlated with an average slope of 0.5. To explain these results we suggest that there is common signal processing at an early stage of saccade and vergence initiation, which is followed by activity that builds in separate trigger mechanisms that can be influenced by the conditions of fixation. Gap-overlap effect Saccade Vergence Fixation Human

INTRODUCTION The latency of a saccade to a target appearing in the visual periphery at an unexpected time and in an unexpected location is affected by the timing between the extinction of the fixation target and the appearance of the new peripheral target (Saslow, 1967). If the fixation target is extinguished before the new target appears (the gap condition), the time to initiate the saccade to the new target is less than when the fixation target disappears and the peripheral target appears at the same time (simultaneous condition). If the fixation target is not extinguished until after the new target appears (the overlap condition), the time to initiate the saccade is greater than in the simultaneous condition. Another feature of the gap condition is the appearance of "express saccades". Fischer and Ramsperger (1984) reported that many human subjects show a bimodal distribution of saccade latencies in the gap condition, with the earlier peak being at about 110 msec. These saccades were called "express saccades". They have been the subject of considerable study in both human beings and monkeys (Fischer & Boch, 1983), and models have

been proposed to explain the role of various brain structures in their generation (reviewed by Fischer & Weber, 1993). The emphasis in prior studies of express saccades and of the effects of gap-overlap stimuli on saccade latencies has been on horizontal saccades made to new targets located across the visual field, but without a change in depth. In natural circumstances, however, many saccades are combined with a change in vergence. Because the behavioral characteristics and the neural circuitry underlying saccades and vergence are quite different, we investigated gap-overlap effects on gaze shifts in three-dimensional space. We measured the distribution of latencies for pure saccades, pure vergence and for both the saccade and vergence components when they were combined. METHODS

Subjects Three normal men, MT (33 yr old), DZ (49 yr old), and EF (35 yr old) were subjects in this experiment. Each had normal corrected vision, normal eye motility, and normal ocular alignment. Viewing a small target at 48.4 cm, MT and DZ had an exophoria (5 and 4 deg respectively) and EF had an esophoria (7 deg). The interpupillary distance was 64 mm for MT and 59 mm for DZ and EF. The head was immobilized with a dental bite bar. Each subject gave informed consent before participating in these experiments.

*Department of Neurology, The Johns Hopkins Hospital, 600 North Wolfe Street, Baltimore, MD 21287, U.S.A. tDepartment of Ophthalmology, Niigata University School of Medicine, Niigata 951, Japan. :~To whom all correspondence should be addressed. 3373

MINEO TAKAGI et al.

3374

Target arrangement Visual targets were small light-emitting diodes (LEDs) with a diameter of 2 ram. There were two stimulus configurations as shown in Fig. 1: one a Plus paradigm to elicit pure saccades and pure vergence (Expt 1), and the other an X paradigm to elicit saccades combined with vergence (Expt 2). For Expt 1, targets were arranged in a Plus shape at eye level so that the target could j u m p from the center to any of four target locations (right, left, far, near) to elicit either pure vergence or a pure saccade. The center target was place at 48.4 cm distance from the subject's eye and called for 7.6 deg of convergence for subject M T and 7.0 deg for subjects D Z and EF. The far target was at 124 cm and the near target was at 33.5 cm. The far target called for 4.6 deg divergence (relative to the vergence angle for the center target) for subject M T and 4.3 deg for subjects D Z and EF. The near target called for 3.3 deg convergence (relative to the vergence angle for the center target) for M T and 3.1 deg for D Z and EF. The right and left targets were 5 deg to the side of the center target. For Expt 2, targets were arranged in an X shape at eye level so that the target could j u m p from the center to any of four target locations, each of which called for a combination of saccades and vergence. The center target was at the same location as the center target in the Plus paradigm. The four eccentric targets were either right 5 deg or left 5 deg from the straight ahead direction. The two far targets were placed at 124 cm distance and the two near targets were placed at 33.5 cm. Thus, the distance in depth and the lateral eccentricity of the targets were comparable for the Plus and X paradigms. In the X paradigm far targets called for a combination of saccades

of 5 deg and divergence of 4.6 (MT) or 4.3 deg (DZ and EF). The near targets called for a combination ofsaccades of 5 deg and convergence of 3.3 (MT) or 3.1 deg (DZ and EF).

Stimulus conditions In each experiment the timing between the offset of fixation target and the onset of new target was varied to produce a gap, simultaneous or overlap stimulus. We used two gap periods of 200 and 75 msec, and a single overlap period of 200 msec. All the experiments were performed in a completely dark room. Long gap stimulus. The fixation target went off 200 msec before the onset of the stimulus target. During the gap period the r o o m was completely dark. Short gap stimulus. The fixation target went off 75 msec before the onset of the stimulus target. During the gap period the r o o m was completely dark. Simultaneous stimulus. The fixation target went off at the same time as the new target was illuminated. Overlap stimulus. The fixation target went off 200 msec after the onset of the stimulus target. During the delay period both targets were visible. The target jumps away from the center of the Plus or X were nonpredictable in location and timing (1-2 sec), while the return j u m p was always predictable, toward the center. We only analyzed the latencies of eye movements for jumps away from the center target.

Measurement of eye movements Eye movements were recorded using the magnetic field search coil method (Robinson, 1963; Collewijn, Van der M a r k & Jansen, 1975). A scleral annulus was mounted on

EXPERIMENT 1 (Plus paradgim) Pure Saccades 0

EXPERIMENT 2 (x paradigm)

Pure Vergence

Combination

•

5 deg 0"4 P O

FIGURE 1. Target configurationsfor Expts 1and 2. The circlesrepresent the positions of the LEDs across the horizontal meridian and in depth, as seen from above, relative to the head of the subject, which is depicted at the bottom of the figure. In Expt 1 five targets were arranged in a Plus shape. In Expt 2 five targets were arranged in an X shape. The center target of the X is the same as that in Expt 1. The exact placement of the target stimuli are detailed in the Methods.

GAP EFFECT ON SACCADES AND VERGENCE

3375

SACCADES WITH CONVERGENCE 6

Left Eye

OI 4

Right Eye

ms

lOO

2001

300

400

Saccade Component

Saccade Velocity Vergence Velocity

Vergence Component

Convergence Onset

Saccade Onset

( > 3 deg/s)

( > 15 deQ/s )

FIGURE 2. Sample trace from Expt 2 (X paradigm). The eye movement response to a target jump from the center to the near right target is shown, calling for a combined 5 deg rightward saccade and 3.3 deg of convergence. The time scale is on the abscissa, eye position is on the ordinate. At 0 msec (beginning of the trace) the stimulus target went on and the fixation target went off. From the top, the traces are left eye position, right eye position, saccade (conjugate) component, saccade (conjugate) velocity (thin trace), vergence position component and vergence velocity (thin trace). Upward deflections indicate rightward for conjugate traces and divergence for vergence traces. On the ordinate, one division is 2 deg for eye position, 200 deg/sec for saccade velocity and 40 deg/sec for vergence velocity. The vertical, dotted lines indicate the onset ofvergence (velocity > 3 deg/sec) and the saccade (velocity > 15 deg/sec).

each eye of the subject. The output signals from the phase detectors were filtered with a bandwidth of 0-90 Hz, sampled by a digital computer at 500 Hz with 12-bit resolution, and then stored to disk for later off-line analysis. System noise limited resolution to about 0.05 deg. The coil signal was calibrated by requiring the subject to fix upon five horizontal targets, separated by 5 deg, centered around the straight-ahead position at 124 cm from the subject.

Procedure We collected data using the Plus paradigm and the X paradigm on different days. Viewing was always binocular and subjects wore their corrective glasses. Experimental sessions lasted about 40 min. There was a brief pause every 10 min. About 40 trials o f each type were obtained. Subjects were instructed to move their eyes to

the LED as soon as it became visible but not before. During the experiments, subjects were encouraged to make accurate saccades and to refrain from making premature responses.

Data analysis The latencies of saccades and vergence were measured using a computer-assisted procedure, in which individual trials were displayed on a video monitor. A computer algorithm detected the onset of the saccade and vergence by velocity criteria using a digital filter with 3 dB point at 112 Hz (for saccades) or at 45 Hz (for vergence). A conjugate component was calculated using the average of the calibrated left eye and right eye horizontal signals, and a vergence component using the difference between them. The onset of saccades was determined at the point when the eye velocity of the conjugate component exceeded

3376

MINEO TAKAGI et al.

15 deg/sec (Fig. 2). The onset ofvergence was determined at the point when eye velocity exceeded 3 deg/sec (Fig. 2), which corresponds to 2.5 times the value of the SD of the noise for vergence velocity. The correctness of these marks was verified by the experimenter. Trials were rejected (a) if the amplitude of the saccade or the vergence response was less than half of the expected value, (b) if the saccade or vergence was in the wrong direction, or (c) if the record could not be marked because of blinks or other artifacts. Seven percent of trials had to be rejected using these criteria. The values for the latencies of saccades and of vergence were displayed in histograms using 10msec bin widths. Saccades with latencies < 80 msec were excluded from further analysis, which is the limit for visually-guided saccades determined by Wenban-Smith and Findlay (1991). We also used 80 msec as the criterion for excluding premature vergence responses, based upon the appearance of the pure vergence histograms and the decoupling of vergence and saccades (vergence occurring much earlier) in the combined paradigms. For statistical analysis, the distributions with each stimulus were compared using the Kruskal-Wallis one-way analysis of variance on ranks. When there was a significant difference each distribution was compared to that with the simultaneous stimulus using Dunn's method.

small, uncalled-for changes in eye alignment during the gap period, which were divergent in most cases for subjects MT and DZ and convergent for subject EF. These were less frequent with the short gap stimulus and

L A T E N C I E S OF P U R E S A C C A D E S 30

Gap(200ms)

20

10

'

o

11oo

200

30

'

300

m s

400

Gap(75ms)

20

shorter shorter

shorter

10

0

RESULTS

!1O0

200

300

400

b

Experiment 1 (Plus paradigm) Saccades. Figure 3 shows the distribution of latencies for the pure saccades in the Plus paradigm. Latencies for rightward and leftward saccades from the center, which were both nonpredictable, were combined. For comparing results among subjects the individual values for each subject at each time bin are connected by a solid line. With the long gap (200 msec) stimulus subject MT showed a bimodal distribution of saccade latencies; the earlier peak was at 110 msec. Subject DZ showed a sharp single peak, also at 110 msec. These early peaks with the long gap stimulus correspond to the latency of so-called express saccades. Subject EF showed a later peak, at about 140 msec. With the short gap (75 msec) stimulus the earlier peak in subject MT became less prominent, while the later peak became more prominent. The single peak in subject DZ became less sharp, Thus, in these two subjects, with the short gap stimulus the proportion of express saccades decreased. With subject EF, however, an early peak appeared at about 120 msec. For all subjects, with the simultaneous and overlap stimuli the latencies became longer and their distributions broader. For all subjects there were significant differences in the median values for the different stimulus conditions (KruskalWallis one-way analysis of variance on ranks, in all cases P < 0.01). Compared to the simultaneous stimulus, saccade latencies were significantly shorter (using Dunn's method) with the long gap and the short gap stimuli and significantly longer with the overlap stimulus. With the long gap stimulus there were often (44-67%)

30

Simultaneous

20

i

MT

i

" DZ

10

o 30

~8

{~

~

20

~

8 ms

Overlap longer longer

'

10

0

~1O0

200

300

400

FIGURE 3. The distribution of the latencies for pure saccades in Expt l (Plus paradigm). The panels are long gap (200 msec) short gap (75 msec) simultaneous and overlap stimuli. In each panel the abscissa is saccade latency and the ordinate is the frequency (absolute number of trials). The histograms for all three subjects are superimposed using different line types as shown in the inset on the right side of the "'simultaneous" panel. The statistical significance of the differences between the distributions of latencies with the gap and the overlap stimuli and the distributions with the simultaneous stimulus are given in each panel in the order MT, DZ and EF from the top. "Shorter", the distribution of latencies was significantly less than that of the simultaneous stimulus (P < 0.05); "longer", significantly greater (P < 0.05). The vertical dashed line demarcates a latency of 80 msec.

GAP EFFECT ON SACCADES AND VERGENCE

3377

L A T E N C I E S OF PURE V E R G E N C E A

convergence

divergence

B

Gap(200ms)

Gap(200ms) lO

10

i

horter hotter horter

0

;100

200

300

shorter shorter shorter

~L

0 ,

......: m s 400

, 0

; i

;100

; ...... A ; m s 300 400

200

Gap(75ms)

Gap(75ms) lO

10

:

shorter

shorter shorter n.s.

shorter

0

' ms 100

0

200

300

. . . . . . . .

I' .....

o

400

i

. . . . . . . .

1100

200

Simultaneous

:

ms

400

10

DZ EF

1

i

DZ

5

. . . . . . . . . ~ . . . ^ . . ~ . '.,.- '~~ , ~ , ,~~ -

o

. . . . . . . .

Simultaneous

? i!

10

:

30o

o

8

,¢(:3

C',I

:

8 ms

~

o

8

~

.............. ~ ................................ .~..........

8

8

200

300

8 ms

'~"

Overlap 10

10

i

.................................

i! ~

,

i+LJi:,

0

i100

...............................

..s.

n.s.

..................................

.....................................................

...........................

!

2OO

300

4O0

0

1100

III~

400

FIGURE 4. The distribution of the latencies of pure vergence in Expt 1. The scheme for depiction of distributions and statistical comparisons between stimuli is identical to that for Fig. 3. (A) Data for pure convergence (when the target jumped from center to near), (B) data for pure divergence (when the target jumped from center to far). n.s., not significant.

only rarely seen with the simultaneous or overlap stimuli. There were, however, no significant differences in saccade latencies between those trials with and without a preceding vergence response ( M a n n - W h i t n e y rank sum test). Vergence. Figure 4 shows the distribution o f latencies for the pure vergence trials in the Plus paradigm. The results o f the convergence trials are shown in Fig. 4(A).

The distributions o f the latencies were n a r r o w e r than those for saccades, forming sharper peaks. With respect to the width o f the distributions, for all subjects and all stimuli, the m e a n differences (+_ 1 SD) between the 25% and 75 0Vovalues were 45 -t- 21 msec for pure saccades and 23 _ 6 msec for pure convergence, which were significantly different using the M a n n - W h i t n e y rank sum test (P < 0.01).

3378

MINEO TAKAGI et al.

As was the case for pure saccades, for all subjects there were significant differences in the median values for the different stimulus conditions (Kruskal-Wallis one-way analysis of variance on ranks, in all cases P < 0.01). With the long and short gap stimuli convergence latencies were significantly less than with the simultaneous stimulus except in one case. In no subject, however, was there a significant difference in the latency of convergence between the overlap and the simultaneous stimulus. Finally, we did not find a distinct peak of "express convergence". Figure 4(B) shows the distributions oflatencies for pure divergence. Again, the peaks in the histograms were sharper than those for saccades. With respect to the width of the distributions of the latencies the mean differences between the 25% and 75% values were 45 +_ 21 msec for saccades vs 30 ± 17 msec for pure divergence, which were significantly different using the Mann-Whitney rank sum test (0.01 < P < 0 . 0 5 ) . With the gap stimuli, and especially the long gap stimulus, subjects MT and DZ showed premature vergence responses which were usually divergent. As was the case for convergence, for all subjects there were significant differences in the median values for the different stimulus conditions (Kruskal-Wallis oneway analysis of variance on ranks, in all cases P < 0.01). With the long and short gap stimuli the divergence latencies were significantly less than with the simultaneous stimulus except for one case. In no subject, however, was there a significant difference in divergence latency between the overlap and the simultaneous stimuli. For subject EF, however, the distribution of divergence latencies with the overlap stimulus showed a long tail extending to nearly 300 msec. Finally, as was the case for convergence, we did not find a distinct peak of "express divergence". Compared with the other two subjects, subject EF, who had an esophoria, showed shorter latencies in the convergence trials and longer latencies in the divergence trials (Fig. 4). For subjects MT and DZ, divergence latencies were significantly shorter than convergence latencies for all stimuli; for subject EF convergence latencies were significantly shorter than divergence latencies for all stimuli (Mann-Whitney rank sum test, P < 0.01). The comparison of the gap-overlap effect on pure saccades and on pure vergence is also shown in Fig. 6 and Fig. 8 (panels labeled "pure saccades" or "pure vergence"). The gap-overlap effect was stronger for saccades. In all three subjects the difference of the median values for saccade latencies with the short gap and the overlap stimulus was around 80 msec. The average difference between the median values with the short gap and the simultaneous stimuli was 41 msec for saccades and 17 msec for vergence. There was, however, no significant difference between the vergence latencies with the overlap and the simultaneous stimuli. To summarize the main results from Expt 1: for pure saccades, each subject developed express saccades with the gap stimulus. With the overlap stimulus there was an increase in saccade latency compared to the simultaneous

stimulus. For pure vergence movements there was a gap effect, but it was smaller than that for saccades, and there was no clear "express vergence". There was no difference between vergence latencies with the simultaneous and the overlap stimuli.

Experiment 2 (X paradigm) A typical response in the X paradigm, eliciting saccades combined with vergence, is shown in Fig. 2. The onset of the saccade component was easily distinguished from any ongoing vergence by using the conjugate trace (Fig. 2, right-hand, vertical, dotted line). In most trials the vergence component began before the saccade. In some cases, however, the vergence component appeared to start at the onset or during the saccade. Only rarely did the vergence response begin after the saccade was completed (0.8% of trials). Because of the inherent changes in eye alignment that occur during pure horizontal saccades it is difficult to discern exactly when a superimposed vergence begins if its onset is at the beginning or during the saccade. Accordingly, these trials were eliminated for the analysis of vergence latencies. For saccades combined with convergence, the percentage of trials eliminated varied from 11% to 21% (MT); 3% to 36% (DZ) and 0% to 6% (EF), depending upon the stimulus. For saccades combined with divergence, the percentage of trials eliminated varied from 0% to 4% (MT), 6% to 8% (DZ) and 41% to 60% (EF), depending upon the stimulus. In other words, for the subjects with exophoria (MT and DZ), more of the divergence responses began before the saccade, and for the subject with an esophoria (EF), more of the convergence responses began before the saccade. With the long gap stimulus the initial vergence movement could be in the wrong direction, similar to the premature vergence in the Plus paradigm. These misdirected responses were seen primarily on the convergence trials for subjects MT (50%) and DZ (88%), and on the divergence trials for subject EF (30%). These trials were also eliminated in the analysis of vergence latencies in the combined paradigm.

Saccade component in combined responses. The distributions oflatencies of the saccade component for the combined movements are shown in Fig. 5. Figure 5(A) show data for saccades combined with convergence and Fig. 5(B) for saccades combined with divergence. As for pure saccades, for all subjects there were significant differences in the median values for the different stimulus conditions (Kruskal-Wallis one-way analysis of variance on ranks, in all cases P < 0.01). Compared to responses to the simultaneous stimulus, saccade latencies were significantly shorter with the long gap (200 msec) and the short gap (75 msec) stimuli and significantly longer with the overlap stimulus. Subject MT still showed express saccades with the long gap stimulus for trials associated with either convergence or divergence. For subject DZ the express saccade peak became less prominent in the combined paradigm (cf. Figs 3 and 5). For subject EF, in both the pure and the combined paradigms, the

GAP EFFECT ON SACCADES AND VERGENCE

early saccade peak was around 130-140 msec. Finally, with the long gap stimulus there were occasional premature saccades, which were not observed in the paradigm. The distribution of latencies for pure saccades and for saccades in the combined paradigm are compared in Fig. 6. For all three subjects the magnitude of the

Plus

LATENCIES A

gap-overlap effect was roughly the same for both paradigms. The gap-overlap difference was a little tess in the combined paradigm, largely because the differences in the distributions of latencies between the simultaneous and the overlap stimuli were always smaller in the combined paradigm. The mean difference between the median values with the short gap and the simultaneous

OFSACCADECOMPONENT g

C O M B I N E D WITH C O N V E R G E N C E i

20

C O M B I N E D WITH D I V E R G E N C E

20

Gap(200ms) 15

3379

~ ~

,~

10

:

Gap(200ms)

i

shorter shorter shorter

15

shorter shorter shorter

1o .

5

A

IA.

.

II

5

0

'

',

0

1100

'

........ " ....

200

:ms

300

o

400

0

20

100

200

300

20

P-=n(75ms)

Gap(75ms)

15

~~~A ,A

15

~rter ~rter ~rter

10

10

5

?~ ~

5

0

:. . . . . . . . .

0

1100

200

"ms

300

0

400

-

:

0

i

.

.

1100

.

.

shorter shorter

shorter 300

400

20

i

~

Simultaneous

15

Simultaneous 15

!

~

MT

MT 10

10

--"-----" DZ L " .......................EF]

~

5

5

0

;

N~ =

0

o

o

8

8

o ms

8

o

8

8

8 ms

O3

"q"

20

!~

Overlap

Overlap

15

15

i !

10

A / ~,

!

longer

longer ,~q~,\longer

i

~

10

5 0

~ms

~' ..............

200

I

i

20

20

400

~

~I1~,~

longer

longer

5

........

,............. 0

i :100

...........

.......

200

300

, 400

........

0

0

;100

200

300

400

FIGURE 5. The distribution of the latencies of the saccade component when combined with vergence in Expt 2. The scheme for depiction of the distributions and statistical comparisons between stimuli is identical to that for Fig. 3. (A) The data for the saccade component combined with convergence (when the target jumped from the center to near right or near left), (B) data for the saccade component combined with divergence (when the target jumped from the center to far right or far left).

MINEO TAKAGI et al.

3380

pure saccades

saccades with convergence

300

sim

250

MT

*

over

E

saccades with divergence

200

150

sgap

lOO

i!i

Igap 78 ms

78 ms

60 ms

50 300 t~

E 250

DZ

200

J

150

lOO

79 ms

69 ms

74 ms

50 30O

25O

EF

..................... / ...... ...............

2OO

i;

150

IOO

50

88 ms

58 ms

67 ms

FIGURE 6. Comparison of the distribution of latencies of saccades between the pure condition and the combined condition. Median (O), 75% (top end of vertical bar) and 25% (bottom end of vertical bar) are shown. Bars are arranged in the order of long gap of 200 msec (lgap), short gap of 75 msec (sgap), simultaneous (sim) and overlap (over) from left to right. The horizontal dotted lines show the median value for the distribution of latencies in response to the simultaneous stimulus. The gap-overlap difference [differences of median values between short gap (75 msec) and overlap stimuli] are shown at the bottom of each panel. The asterisks denote that the distribution is significantly longer in the combined condition than in the pure condition: *0.01 < P < 0.05; **P < 0.01.

stimuli was 41 msec in b o t h the p u r e a n d the c o m b i n e d p a r a d i g m . T h e m e a n difference b e t w e e n the m e d i a n values with s i m u l t a n e o u s a n d the o v e r l a p stimuli was 41 msec in the p u r e p a r a d i g m a n d 27 msec in the c o m b i n e d p a r a d i g m s . T h e d i s t r i b u t i o n s o f latencies in the c o m b i n e d p a r a d i g m were c o m p a r e d to those with the c o r r e s p o n d i n g stimuli in the p u r e p a r a d i g m . I n 16 o f 24 cases latencies were significantly l o n g e r in the c o m b i n e d p a r a d i g m ( M a n n - W h i t n e y r a n k s u m test). In eight cases there was n o significant difference.

In no case was the latency s h o r t e r in the c o m b i n e d paradigm.

Vergence component in combined responses. T h e d i s t r i b u t i o n o f latencies o f the convergence c o m p o n e n t in E x p t 2 are s h o w n in Fig. 7(A). A s was the case for p u r e convergence, the g a p - o v e r l a p effect was n o t as p r o m i n e n t as it was for saccades. A distinct p e a k o f express convergence was n o t observed. The results in subject E F were difficult to i n t e r p r e t because his eyes

G A P E F F E C T ON SACCADES A N D V E R G E N C E

tended to converge spontaneously during the gap period. For all subjects there were significant differences in the median values for the different stimulus conditions (Kruskal-Wallis one-way analysis of variance on ranks, in all cases P < 0.01). With the long and short gap stimulus the latencies were significantly less than with the simultaneous stimulus except in one case. In all three

3381

subjects the overlap stimulus had no significant effect on convergence latencies. The distribution of latencies of the divergence component are shown in Fig. 7(B). With the long gap stimulus there was more premature divergence (73%) than occurred in the Plus paradigm with pure vergence stimuli (29%). No distinct peak of express vergence was

LATENCIES OF VERGENCE COMPONENT A

B

CONVERGENCE COMPONENT

DIVERGENCE COMPONENT =

i

;

20

Gap(200ms)

i

20

I

i

shorter .......

10

n,s.

10

Gap(20Orns)

i !

shorter shorter shorter

_ ~

................................

0

!100

200

300

400

0

;100

200

.......................

300

400

i i

i

20

Gap(75ms)

i

' i !

15 10

shorter n.s. shorter

20

°.°,"°s)

i

lO

shorter

5 0

......................

:................. o

!1oo

.........................

200

.............................................

300

o 0

400

i

1100

200

300

400

i J

20

20

Simultaneous

1

i r~ 10

.T

10

0

0

0

1100

200

300

400

0

i100 =

2~

3~

4~

t

i

20

1

-~

i i

Overlap

2o

Overlap

i i

/

i i

| 10 ~-

i

//

i! 0

1100

~'~ !"P:t,~

200

n.&

n.s.

n.s.

300

lO

400

o

J

11oo

200

300

400

F I G U R E 7. The distribution of the latencies of the vergence component combined with saccades in Expt 2. The scheme for the depiction of the distributions and statistical comparisons between stimuli is identical to that for Fig. 3. (A) Data for the convergence component (when the target jumped from the center to near right or near left), (B) Data for the divergence component (when the target jumped from the center to far right or far left), n.s., not significant.

3382

MINEO TAKAGI et al.

pure convergence

pure divergence

convergence with s a c c a d e s

300

30O

250

250 sim

divergence with saccades

over ............. ~ . , _ . . . . ~ . .

200

200

..........s.agp . ~ . . .

MT

/

150

150

100

100

19 ms

27

ms

50

5O

3OO

300

250

250 ;

200

14 ms

34 ms

==

200

DZ 150

150

100

100

20 ms

0 ms

EF

50

50

300

300

250

250

200

2OO

150

........~ _

~

....

~

100 50

"

16 ms

32 ms

**

57 ms

:¢

150 100

30 ms

10 ms 50

38 ms

FIGURE 8. Comparison of the distribution of latencies of vergence between the pure condition and the combined condition. The scheme for the depiction of the distributions and the statistical comparisons between stimuli is identical to that for Fig. 6. Also, the gap-overlap difference (differences of the median values between the short gap and the overlap stimuli) are shown at the bottom of each panel. The asterisks denote a significant difference from the result in the pure condition: *0.01 < P < 0.05; **P < 0.01. When the asterisk is place above the bars it means values are significantly longer in the combined condition than in pure condition, and vice versa.

apparent. The data from the long gap stimulus were difficult to interpret because subjects MT and DZ tended to diverge during the gap period and many of the responses of subject EF began at the same time as the saccades. As was the case for the convergence component, however, for all subjects there were significant differences in the median values for the different stimulus conditions (Kruskal-Wallis one-way analysis of variance on ranks, in all cases P < 0.01). With the long and the short gap stimulus the latencies were significantly less than with the simultaneous stimulus except in one case. In all three subjects the overlap stimulus had no significant effect on divergence latencies. As was the case for the Plus paradigm with pure vergence stimuli, subject EF showed shorter latencies for the convergence component and

longer latencies for the divergence component when compared to the other two subjects. Likewise, for all stimuli, divergence latencies were significantly shorter than convergence latencies for subjects MT and DZ and significantly longer for subject EF. A comparison of the distribution of the latencies for pure vergence and for the vergence component in the combined paradigm is shown in Fig. 8. The short gap-overlap difference was larger in the combined paradigm in all eases. With the simultaneous and the overlap stimuli, the latencies of the vergence component in the combined paradigm were significantly longer in 7 of 12 cases, while with the short and long gap stimuli they were sometimes longer (2 out of 12 cases) and sometime shorter (5 out of 12 cases).

GAP EFFECT ON SACCADESAND VERGENCE

Latencies and the width of their distribution.

Fig. 9(B). This includes all the data shown in Fig. 8. In this case all the data were analyzed together regardless of the stimulus. Again, there was a significant positive correlation between them. Just as was the case for saccades, as vergence latencies became longer their distributions became broader.

The relationship between the median value of saccade latencies and the width of the distribution (estimated as the range from 25% to 75%) is plotted in Fig. 9(A). This includes all the data shown in Fig. 6, i.e. results from both the pure and the combined conditions. Because with the long (200 msec) gap stimulus the distributions could show more than one peak, we treated them as a separate group (shown by + in the figure) and excluded them from the analysis of correlation. While the distribution also might have been bimodal with the short (75 msec) gap stimulus, there was still a significant positive correlation when the data from the short gap, simultaneous and overlap stimuli were lumped together. This means that as the saccade latencies became longer their distributions became broader. The relationship between the median values ofvergence latencies and the widths of the distributions is plotted in

A

100

LATENCIES

Relationship between the initiation of saccades and of vergence in the combined paradigm. The relationships between saccade latencies and vergence latencies in the combined paradigms are shown in Fig. 10. For this figure trials in which vergence appeared to start during the saccade were eliminated. The regression lines were calculated with saccade latencies arbitrary considered as the independent and vergence latencies as the dependent variable. Note that in each panel the regression lines for the responses to the different stimuli lie close to each other. The correlation coefficients

v s W I D T H OF D I S T R I B U T I O N

saccades

+

O

0

80

O

sgap,sim,over

+

Igap regression

60

3383

0

0

4 +-I-

0

+

00/t o

40

R=0.39" (n:27)

0

o

20

0

#

+ median value I

i

I

100

2OO

3OO

100

vergence all stimuli

80

O

regression ~3

=.

60

qD

R=0.40**(n:48)

0

@

40

20 O

O m e d i a n value

0

i 100

i 200

I 300

FIGURE 9. The relationship between the median value of latencies (abscissa) and the width of the distributions estimated as the range from 25% to 75% for each stimulus for each subject (ordinate). (A) Saccadiclatencies (correspondsto Fig. 6). O Data for the short gap, simultaneous, and overlap stimuli. The data for the long gap stimulus (+) were excluded from this analysis (see text). (B) Vergencelatencies (corresponds to Fig. 8). Data for all stimuli, including the long gap, were plotted using O. The correlation coefficient,number of data points (in parentheses) and the regression line are shown in each panel. *0.01 < P < 0.05; **P < 0.01.

0

100

20O

0

100

300,

0

B

o

A

e-

vE

8

E~

8

E

200

100

•

••

300

I

saccade (ms)

O

0

10o

o

8

E

I

20O

lOO

+

I

+

MT

• •

_

0

300

I

I

•

400

saccade (ms)

•

lo o

2O0

3OO

E

"~

I

100

FIGURE l~caption

200

100

4-

I

+

DZ

DZ

I

SACCADE LATENCIES vs DIVERGENCE LATENCIES

I

I

O

MT

SACCADE LATENCIES vs C O N V E R G E N C E L A T E N C I E S

300

I

300

I

r

400

saccade ( m s )

on faeh~g page.

200

I

saccade (ms)

I00

200

300

IO0

200

300

8

E

8

e-

>

E~

8t-(D

E

v

100

100

e+

"

200

~

EF

200

•

- o

~P

•

•

300

I

I

400

saccade (ms)

300

saccade (ms)

0

• overlap

o simultaneous

EF

+ gap(75ms)

• gap(20Oms)

>

Z ©

GAP EFFECT ON SACCADESAND VERGENCE

3385

TABLE 1. Correlation coefficients, slope and y-intercept of the regression line between the latencies of the saccade component and those of the vergence component in Expt 2

Saccades-convergence MT DZ EF

Saccades-divergence

Long gap

Short gap

Simultaneous

Overlap

Long gap

Short gap

Simultaneous

Overlap

0.55** (36) 0.50x + 53 0.81"* (14) 0.70x + 34 0.21 (46) 0.19x- 23

0.64**(44) 0.48x + 68 0.81"*(47) 1.0x- 40 0.36** (71) 0.24x + 91

0.54**(68) 0.50x + 71 0.58**(57) 0.5Ix + 70 0.40**(76) 0.28x + 126

0.43**(63) 0.45x + 77 0.61"*(75) 0.45x+ 79 0.55**(74) 0.28x+ 102

0.52* (15) 0.37x+ 51 0.40 (6) 0.11x + 79 0.79**(15) 0.82x+ 7.9

0.42** (41) 0.30x + 56 0.70** (61) 0.61x+ 17 0.64**(25) 0.77x+ 14

0.41"*(78) 0.33x+ 64 0.67**(66) 0.63x + 29 0.75**(34) 0.80x+ 10

0.48**(78) 0.30x+ 75 0.83**(70) 0.78x - 6.2 0.58**(40) 0.51x + 69

Saccade-convergence, between saccadecomponentand convergencecomponent;saccade divergence,between saccadecomponentand divergence component. The number of data points are given in parentheses. The degree of significance of the correlation is indicated by asterisks: *0.01 < P < 0.05; **P < 0.01. are shown in Table 1, and indicate the tight relationship between the onset of saccades and vergence. In only two cases were the correlations not significant, and in one of those there were only six data points (long gap for subject DZ), in one case the correlation was weakly significant (0.01 < P < 0.05) and in the other 21 cases they were highly significant ( P < 0 . 0 1 ) . The slope and the y-intercept of the regression lines are also shown in Table 1. For all trials and all subjects the mean value of the correlation coefficients was 0.57. The mean value of the slopes of the regression lines was 0.50. To summarize the main results from Expt 2 in which saccades and vergence were combined: the effect of the gap stimulus on saccade and vergence latencies was similar to the effect on latencies for saccade and vergence when made alone. Saccade and vergence latencies were highly correlated with the slopes of the regression lines being, on average, 0.5. DISCUSSION Two main findings emerged from this study. First, vergence eye movements, either alone or when combined with saccades, were less influenced by the timing relationships between the extinction of the fixation target and the onset of the peripheral target (gap-overlap effects) than were saccades. We also could not identify express vergence. Secondly, when vergence and saccades were combined, gap-overlap effects were similar to those in the pure condition. In addition, the onsets of saccades and vergence were highly correlated in the short-gap, simultaneous, and overlap stimulus conditions, with the slope of the relationship being about 0.5. We will first discuss the gap-overlap effects on vergence and saccades. Thereafter, we will present a hypothesis to explain our main findings.

Different effect of gap-overlap stimulus on pure saccades and on pure vergence With the long gap stimulus, all three subjects developed express saccades. These findings demonstrate that the

conditions we used were appropriate for eliciting express saccades. There was not, however, a bimodal distribution of vergence latencies with the same stimulus. Thus, we have no convincing evidence for the existence of express vergence in response to the same stimuli that elicit express saccades. There was, however, a small but significant effect of target timing on vergence latencies in the short gap vs the simultaneous stimulus (on average 16 msec), albeit much smaller than the effect on saccades (on average 41 msec). A similar, relatively small, gap effect has also been found for smooth pursuit (Krauzlis & Miles, 1995), and like vergence, no distinct population of short-latency pursuit responses was found. Therefore, it appears that the gap stimulus can influence the initiation--albeit to different degrees--of each of the visually-driven subclasses of eye movements: saccade, vergence, and pursuit. The smaller effect of the gap stimulus on pure vergence, and the lack of a significant difference in vergence latency between the overlap and the simultaneous conditions suggest that the same hierarchical system that acts to modify saccade latency does not influence vergence. The influence of the conditions of visual fixation, presumably mediated by higher cortical processes, appears to be much weaker on vergence than on saccades (see below).

Factors that influenced vergence latencies The effects of phoria. For the long gap (200 msec) stimulus, vergence often appeared to begin prematurely. These responses were usually divergent for subjects MT and DZ (who had exophorias) and convergent for subject EF (who had an esophoria). They were elicited more often in the X (combined) than in the Plus (pure) paradigm, which may be related to the 100% chance for having a vergence movement in the X paradigm but only 50% in the Plus paradigm. We also found that in both the Plus and the X paradigms the latencies for convergence were greater than for divergence for subjects MT and DZ, and less than divergence for subject EF. This was true not only for the gap stimulus but also the simultaneous and

FIGURE 10 (opposite).Scatter plots oflatencies of the saccadecomponentand vergencecomponentsin individual trials for Expt 2. (A) Relationship between latencies of saccadecomponent(abscissa) and of convergencecomponent(ordinate); (B) Relationship between latencies of saccade component (abscissa) and of divergence component(ordinates). Stimulus conditions are indicated using different symbols as shown in the top right. The regression lines for each stimulus are also displayed. VR 35/23-24~H

3386

MINEO TAKAGIet al.

overlap stimuli, in which cases there was rarely any drift towards the phoria position prior to the onset of the vergence response. Therefore, the difference in latencies between convergence and divergence in our subjects is not an artifact of the gap stimulus, in which the dark period predisposes toward a drift towards the inherent phoria. Rather, the inherent phoria itself appears to affect the initiation of vergence responses. The simultaneous onset of vergence and saccades in the Xparadigm. The vergence component in the X paradigm sometimes appeared to start at the same time as the saccade. Since it was difficult to separate the onset of vergence from the inherent changes in horizontal alignment that occur naturally during horizontal saccades (Enright, 1984; Collewijn, Erkelens & Steinman, 1988), we had to eliminate these trials from the analysis of vergence latencies. This reduction of the data could have caused a bias in the analysis, skewing the distributions of latencies toward lower values. The elimination of some trials could also have affected the correlation between the saccade component and the vergence components in the X paradigm. The correlation, however, was very tight, even when the proportion of trials that was excluded was small, and would almost certainly have been even tighter if these excluded vergence movements could have been included in the analysis. Thus, the elimination of any trials from the analysis did not obscure the main finding that the onsets of vergence and saccades were tightly correlated in the X paradigm.

Comparison between pure and combined paradigm Gap-overlap effect on the pure and combined paradigms. In each subject the magnitude of the gap-overlap effect on saccades was almost the same for responses in the pure and combined paradigms. The gap-overlap effect was also much less on vergence than on saccades in the combined paradigm. There was no overlap effect on vergence in either the pure or combined paradigm. Overall, these results suggest that the mechanisms that generate the vergence and the saccade commands can be influenced quite independently even when the two movements occur together. Longer latencies of saccade andvergence in the combined paradigm. In 16 of 24 cases, saccade latencies were greater in the combined paradigm than in the pure paradigm as seen in Fig. 6. In 9 of 24 cases, vergence latencies were also larger in the combined paradigm as seen in Fig. 8. One possible explanation for the increase in latency in the combined paradigm is that it takes additional time to calculate the average (conjugate) position when the images of the new target fall on disparate positions on the retinas. Findlay and Harris (1993) showed that when the images of two targets were presented separately to each eye, used tachistoscopic viewing, the resulting saccade demonstrated not only spatial averaging, bringing the eyes to a position between the two targets, but also an increase in latency. The latency was least (215 msec) when the left eye and right eye targets fell on corresponding points of the retinas in the two eyes. The latency increased,

however, with disparity at a rate of 2.5 msec for each degree. This phenomenon could explain some increase in saccade latencies in the combined condition in our experiments. For our subjects, the disparities in the X paradigm were 3.1-4.6 deg, which would be expected to produce an increase in saccade latency of 8-11 msec. In our experiments, though, the average increase in the median value of the distribution of latencies between the Plus and the X paradigms was almost two-fold larger, 17.7 msec. Another important difference between our experimental conditions and those of Findlay and Harris (1993) is that we used natural targets with real differences in depth so that the subjects knew that they had to make a vergence movement. In the experiments of Findlay et al., a vergence response was not explicitly requested nor was one often made. Therefore, another factor for the increase in latency in the combined paradigm may be that it simply takes more time to program both saccades and vergence together.

Relationship between saccade and vergenee latencies in the combined paradigm Another clear result in our study is that the latencies of the saccade and the vergence components in the X paradigm were tightly correlated. Moreover, in most instances, the slopes of the lines relating saccade and vergence latencies were similar in all stimulus conditions. Saccades and vergence are normally considered as distinct subclasses of eye movements, with considerable differences in their control systems characteristics. There may be, however, an interaction at the brain stem level which accounts for the acceleration of the rate of change in eye alignment that occurs when saccades and vergence occur at the same time (Zee, Fitzgibbon & Optican, 1992). The relationship between the initiation of vergence and saccades suggested in our study is different from this lower level interaction in the dynamics of vergence. Because vergence and saccades still begin at different times during the X paradigm, the interaction for initiation is assumed to be at a higher, probably cerebral level. A model for gap-overlap effects on saceades and vergence. Any hypothesis to explain the gap-overlap effects on saccades and vergence must account for at least six findings. (1) The distribution of latencies for responses to a given stimulus was broader for saccades than for vergence. Hence the peaks of the vergence latency distributions were sharper than those for saccades. (2) The distribution of latencies to a given stimulus was broader when the median value of latencies was longer. (3) Gap-overlap effects were similar in the pure and the combined paradigms. (4) Saccade and vergence latencies were highly correlated in the combined paradigms. (5) The slope of the relationship between saccade and vergence latency was only about 0.5. (6) In most cases, for a given subject, the slope remained about the same for the short gap, simultaneous, and overlap stimuli. One possible scheme to explain these results is shown in Fig. 11. Visual information from the retina of each eye

GAP EFFECT ON SACCADES AND VERGENCE

3387

expresspathway

~

SAC.~7__trigger Saccades ......

Target Jump - - "

I visual analysis

signal

." fixation "'; ',, condition ,,, i,"........ °"

27-

p~

Vergence

VERG trigger FIGURE 11. Block diagram for initiation of saccades and vergence. Information from the retina is analyzed to determine the direction and depth of the target of interest. Once a decision to generate an eye movement toward the visual target is made activity is initiated in two independent trigger circuits, one for the saccade system (SAC trigger) and the other for the vergence system (VERG trigger). Activity builds at a specified rate (reflected in the slope of the line) until a threshold (dashed line) is reached when vergence and saccades are triggered. Latencies can be altered, depending upon the stimulus conditions, level of attention, etc., by changing the slope and/or the thresholds within the individual trigger circuits. A potentially separate express pathway is also shown for saccades.

is processed centrally to create a precept of the location of the target in three-dimensional space. A decision to generate an eye movement to a new target is made for saccades and vergence at the same time. This would account for the high correlation between the latencies of saccade and vergence in the combined paradigms. Thereafter, the activities are passed to parallel vergence and saccade trigger mechanism. Within the two trigger pathways it takes some time for activity to build to the necessary threshold to trigger the premotor circuits to generate the actual saccade and vergence premotor commands (Carpenter, 1988). We assume that the rates at which activity rises in the vergence and saccade trigger pathways are different or that their thresholds are different. Assuming some biological noise in the system common to both pathways, the slope of the relationship between saccade and vergence latencies in the combined paradigm would be different from 1.0. As an example, if the rates of rise to threshold were in the ratio of 1:2 (corresponding to a lower rate of rise for saccades), any noise that was common to the threshold levels of the two pathways would lead to a distribution of latencies in which the relationship between saccade and vergence latencies would have a slope of 0.5. The distributions would be wider for saccades than vergence, as well as when the overall value of latencies to a given stimulus was greater (Carpenter, 1988). The fact that changing the stimulus conditions (e.g. from gap to overlap) affected the median values and the distribution of latencies in a specific way, could be explained by assuming a change in the threshold levels, or a change in the rates of rise in activity in the triggers. This would be a way that the particular circumstances of fixation could influence the triggering of saccades and vergence, and lead to similar

slopes of the regression lines for the gap, simultaneous, and overlap stimuli. The absence of express vergence and the presence of express saccades suggests a separate pathway for the express saccade (see also below). Anatomical substrate for gap-overlap effects. Fischer and colleague (Fischer, 1987; Fischer & Weber, 1993) have proposed a three-loop model to describe the process of saccade initiation: disengagement of attention, decision making, and computation of the saccade metrics. Each of these processes has been attributed to a different neural pathway; it has been hypothesized that the component related to a shift of attention is mediated in the parietal cortex, and the decision-making component by the frontal eye fields (Fischer & Boch, 1990; Fischer & Weber, 1993). The fixation neurons in the superior colliculus (SC) could be a site where saccade initiation is influenced by the conditions of fixation and perhaps also by attention (Munoz & Wurtz, 1993). The express saccades themselves are presumably mediated through an oligosynaptic pathway from the primary visual cortex to the SC. To prevent this reflex-like saccade from occurring in most natural circumstances, when cognitive processes must have time to determine the salience of a target, cerebral structures probably have a suppressive effect on the SC. When the loops of attention and decision are bypassed and these inhibitory actions are removed, express saccades could be triggered by the SC via a direct pathway that bypasses higher-level, cerebral, decisionmaking circuitry (Fig. 11). Whether or not there is a comparable pathway for express vergence is not yet settled. Busettini, Miles and Krauzlis (1994) could elicit vergence responses in humans at extremely small latencies, of around 80 msec, when using disparity steps of large textured scenes imposed immediately after a

MINEO TAKAGI et al.

3388

saccade. They assumed that this response was mediated by a rapid pathway from visual cortex to the vergence premotor centers in the brain stem. Our experimental conditions apparently did not allow access to such a short-latency pathway, perhaps because the target stimuli were smaller, and the responses required an analysis of the position of the stimulus, both in depth and across the visual panorama. The successful completion of such a task may be incompatible with an extremely short-latency response. Gap-overlap effects on saccades combined with other types of movements. Finally, it will be useful to consider interactions between saccades and other types of motor movements. Zangemeister and Stark (1982) also reported that the relationship between the latencies of saccade and head movements during combined rapid gaze shifts was very tight, but the slope of the regression line relating them was almost 1.0. These results suggest that the processes that trigger saccades and head movements are common or more closely related than are those for saccades and for vergence. Fischer and Rogal (1986) reported that the correlation between latencies of saccades and those of hand-reach movements is high with the overlap stimulus and almost absent with the gap stimulus. They proposed that the execution of the reach movement depends on the completion of the preparation of the saccade but not vice versa. In our data the correlation between saccades and vergence was always tight regardless of the stimulus condition, which supports a common command hypothesis for saccades and vergence. REFERENCES Busettini, C., Miles, F. A. & Krauzlis, R. J. (1994). Short-latency disparity vergence responses in humans. Neuroscience Abstracts, 20, 1403. Carpenter, R. H. S. (1988). In Movements of the eyes (pp. 78-79). London: Pion. Collewijn, H., Erkelens, C. J. & Steinman, R. M. (1988). Binocular coordination of human horizontal saccadic eye movements. Journal of Physiology, London, 404, 157-182. Collewijn, H., Van der Mark, F. & Jansen, T. C. (1975). Precise recording of human eye movements. Vision Research, 15,447-450.

Enright, J. I. (1984). Changes in vergence mediated by saccades. Journal of Physiology, London, 350, 9-31. Findlay, J. M. & Harris, L. R. (1993). Horizontal saccades to dichoptically presented targets of differing disparities. Vision Research, 33, 1001--1010. Fischer, B. (1987). The preparation of visually guided saccades. Reviews in Physiology, Biochemistry and Pharmacology, 106, 1-35. Fischer, B. & Boch, R. (1983). Saccadic eye movements after extremely short reaction times in the monkey. Brain Research, 260, 21-26. Fischer, B. & Boch, R. (1990). Cerebral cortex. In Carpenter, R. H. S. (Eds), Vision and visual dysfunction, Vol. 9. Eye movements (pp. 277-296). London: Macmillan. Fischer, B. & Ramsperger, E. (1984). Human express saccades: Extremely short reaction times of goal directed eye movements. Experimental Brain Research, 57, 191-195. Fischer, B. & Rogal, L. (1986). Eye-hand-coordination in man: A reaction time study. Biological Cybernetics, 55,253-261. Fischer, B. & Weber, H. (1993). Express saccades and visual attention. Behavioral and Brain Sciences, 16, 553-610. Krauzlis, R. J. & Miles, F. A. (1995). Similar changes in the latency of pursuit and saccadic eye movements observed with the "gap paradigm". In Delgardo-Garcia, J. M., Godaux, E. & Vidal, P. P. (Eds), Information processing underlying gaze control (pp. 269 277). Oxford: Pergamon Press. Munoz, D. P. & Wurtz, R. H. (1993). Fixation cells in monkey superior colliculus. 2. Reversible activation and deactivation. Journal of Neurophysiology, 70, 576-589. Robinson, D. A. (1963). A method of measuring eye movements using a scleral search coil in a magnetic field. IEEE Transactions on Biomedical Engineering, BME-IO, 137-145. Saslow, M. G. (1967). Effects of components of displacement-step stimuli upon latency for saccadic eye movement. Journalof the Optical Society of America, 57, 1024-1029. Wenban-Smith, M. G. & Findlay, J. M. (1991). Express saccades: Is there a separate population in humans? ExperimentalBrain Research, 87, 218-222. Zangemeister, W. H. & Stark, L. (1982). Gaze latency: Variable interactions of head and eye latency. Experimental Neurology, 75, 389-406. Zee, D. S., Fitzgibbon, E. J. & Optican, L. M. (1992). Saccade-vergence interactions in humans. Journal of Neurophysiology, 68, 1624-1641.

Acknowledgements--This work was supported by a grant from the National Institute of Health EY-01849 (DSZ), a Manpower Award from Research to Prevent Blindness (DSZ) and a postdoctoral research fellowship from Uehara Memorial Foundation (MT). The authors thank Mr A. G. Lasker and Mr D. Roberts for their technical assistance, and Dr W. Wolf for helpful comments.