Slow eye movements to eccentric targets
Harry J. Wyatt and Jordan Pola
It is generally believed that a target offset from the direction of gaze can only be fixated with a
saccadic jump in eye position. By preventing saccadic eye movements from fixating a target, we
have observed slow eye movements to both stationary and moving eccentric targets. This
supports the view that target offset plays a role in guiding slow eye movements.
Key words: slow eye movements, pursuit eye movements, target offset,
retinal stabilization, open loop
o observe a smoothly moving object a person makes a smooth pursuit eye movement,
i.e., a slow eye movement with approximately the same angular velocity as the object. Motion of a target (real or perceived) is
usually thought to be the visual stimulus for
smooth pursuit and for slow eye movements
in general. In recent years, however, some
evidence has suggested that in addition to
target motion, target position (offset of the
target from the direction of gaze) can also
serve as a stimulus for slow eye movements.
In particular, if a subject attempts to look at a
target that is stabilized so it remains at a fixed
eccentricity from the fovea (independent of
the motion of the eye), vigorous slow eye
movements directed toward the target often
occur. 1 " 3 (The slow eye movements are interspersed with saccadic jumps in eye position, which apparently represent attempts by
the saccadic system, as well as the slow
movement system, to fixate the target.) There
From the State University of New York, State College of
Optometry, New York, N.Y.
Supported by National Eye Institute Grant EY02878,
the Research Foundation of State University of New
York, and the Optometric Center of New York
Foundation.
Submitted for publication Oct. 18, 1980.
Reprint requests: Harry J. Wyatt, Ph.D., State College
of Optometry, 100 E. 24th St., New York, N.Y. 10010.
are limitations on interpreting these findings,
however, since during slow eye movements
toward a target fixed at an eccentric retinal
position the target moves smoothly along with
the eye and thus moves (in head coordinates)
with respect to the subject. Once the slow eye
movement is underway it may be driven by
both target position and the subject's perception of target motion.4 Therefore it is difficult
to determine the extent to which target position is responsible for the movement.
To avoid this difficulty, perhaps the simplest experiment would be to ask a subject to
make a slow eye movement to a stationary
eccentric target (stationary in head coordinates). However, when most subjects elect to
look at a stationary target substantially offset
from the direction of gaze, they begin by
making a saccadic eye movement toward the
target. The saccade generally achieves fixation and the experiment is over. (If the target
is stabilized, the saccadic attempt fails and
the subjects subsequently make slow eye
movements as well as saccades.) These
findings suggest either that a substantially
offset target in the real world cannot elicit
slow eye movements, or that some built-in
system strategy mandates a saccade for the
first fixation attempt and slow eye movements can occur if the initial saccadic attempt
fails. To see if the latter could be true, we
have developed a target that is "open loop"
0146-0404/81/090477+07$00.70/0 © 1981 Assoc. for Res. in Vis. and Ophthal., Inc.
477
478 Wyatt and Tola
(retinally stabilized) for saccadic eye movements but that is "closed loop" and stationary
(fixed in head corrdinates) for slow eye movements. By denying target acquisition to the
saccadic system in this way, we have observed slow eye movements to stationary eccentric stimuli.
We have also asked subjects to visually
track a constant velocity ("ramp") target.
With this stimulus, Robinson5 found that subjects often make a saccade that falls short of
the target, followed by a "catch-up" smooth
pursuit eye movement faster than the target.
Since retinal target velocity is opposite to eye
velocity during catch-up movements, the occurrence of such movements suggests that
slow eye movements can occur in response to
target position under normal circumstances.
To explore this possibility we used ramp
target motion, but we stabilized the target
during saccades with the same technique
mentioned above for stationary targets. This
increased the amount by which saccades fell
short of the target and resulted in more vigorous and prolonged "catch-up" movements.
Methods and procedure
A 1.5 deg diameter round target was projected
on a rear-projection screen 114 cm from the subject. Horizontal target position was controlled by a
mirror on a penmotor galvanometer. Seated in a
dark room, subjects viewed the screen monocularly with the left eye. Head position was held
constant with the help of a dental impression
bite-bar. A signal of horizontal position of the left
eye was obtained with an infrared scleral reflection
device (Narco Bio Systems, Inc.) and was differentiated electronically to provide an eye velocity
signal. Target position, eye position, and eye velocity were recorded on a Grass polygraph. System bandwidth was about 0 to 75 Hz. Records
shown are tracings of polygraph records.
We used a computer (Digital Equipment Corp.)
to control the stimulus. The computer monitored
eye velocity to determine the presence or absence
of a saccade; for slow eye movements, eye velocity
is nearly always less than about 25 deg/sec,
whereas for saccades greater than 0.3 deg, eye
velocity is greater than 30 deg/sec.6" 7 During a
trial the stimulus remained stationary (or moved
with constant velocity for some trials) in head
coordinates as long as the subject used only slow
Invest. Ophthalmol. Vis. Sci.
September 1981
eye movements to achieve fixation. When a saccade started, the computer calculated the offset of
the target from the fovea immediately prior to the
saccade and jumped the target in the away-fromfovea direction by an amount equal to this offset.
At the end of the saccade the actual saccade length
was calculated and compared with the "assumed
length," and the target position was corrected if
necessary. This correction tended to be small, and
our results did not differ significantly if the correction was omitted. The delay from saccade onset
to target jump was less than 10 msec.
The eye position signal was calibrated before
and after each trial. The subject fixated five
equally spaced target positions along the horizontal meridian (straight ahead, ±7.5 deg, and ±15
deg) and signaled each fixation to the computer.
During the trial the eye position signal was converted to angle with a four-segment piecewise linear fit through the five data points taken during
the initial calibration. Trials were accepted if (1)
mean-square deviation from linearity was less than
2% for both initial and final calibrations, (2) the
slopes of the best linear fits to initial and final
calibration data were within 5% of each other, and
(3) the difference between the initial and final
calibration values for the straight-ahead fixation
point was equivalent to a difference of less than 0.3
deg. At least 20 trials were run in an experimental
session, and size and direction of target steps were
varied randomly. Subjects for these experiments
were men and women between 20 and 40 years
old.
Step stimuli. At the start of these trials the target
assumed its central position and the subject fixated
it, and after a short pause the target stepped horizontally away from the fovea. The subjects task
was to attempt to look at the eccentric target, but
whenever this attempt involved a saccade the
target jumped by an amount equal to the saccade.
The trial was initiated by the experimenter. Time
of initiation was random within a 2 sec window,
and many experiments were performed with direction of target motion randomized between left
and right. In some of the experiments we varied
the size of the target step (0.5, 1, 2, 4, or 6 deg).
Large steps (>4 deg) usually made it necessary to
use starting positions that were not at the center of
the screen, in which case direction of motion was
known before the start of the trial. This reduced
variability but did not alter the main characteristics of the eye movements. These experiments
were performed on 10 subjects, of whom seven
were naive.
Ramp stimuli. These trials were run in the same
Volume 21
Number 3
Slow eye movements to eccentric targets 479
10°/S
1 SEC
Fig. 1. A, Eye movement response to 3 deg target step off the subjects fovea, with target
position on retina stabilized against saccadic fixation attempts. T, Target position; E, eye
position; E, eye velocity. Calibration marks at right, 5 deg for T and E, 10 deg/sec for E.
Asterisk marks initial target step. Subject responded with saccade (S), which was compensated
for by stabilizing target jump (J). Smooth movement to target then commenced. Subject (J. P.)
was experienced. B, First-time response of a naive subject tested under the same conditions as
in A.
way as step stimulus trials, except that the stimulus was the sudden onset of constant velocity
target motion. On some trials the target jumped to
counteract saccadic attempts at fixation. Three
types of trials were used, randomly mixed. (1) The
stimulus was normal closed-loop ramp target motion and did not jump in response to saccadic eye
movements (jump-saccade ratio = 0). (2) The stimulus was ramp target motion but the target jumped
in response to a saccade by an amount equal to the
saccade (jump-saccade ratio = 1). (3) The stimulus
motion and jumps were as in (2), except that on
the first saccade the compensatory jump was twice
the size of the saccade (jump-saccade ratio = 2).
(For subsequent saccades the target jump equaled
saccade size.) The effect of the last condition was
to exaggerate the target offset from the fovea after
the first saccade (see Discussion). Trials were conducted with the subject initially fixating the target
at one or the other side of the screen. The target
started to move in ramp motion across the screen
and disappeared at the far end of the screen. Subjects were urged to fixate the target before it disappeared. Ramp speed was 10 deg/sec. Target
motion and eye movements were recorded on magnetic tape for subsequent computer analysis. (System bandwidth about 0 to 500 Hz.) This experiment was performed on four subjects.
Results
Responses to target steps. Fig. 1, A, shows
a typical result from a trial in which the target
stepped 3 deg horizontally away from the
fovea at the start of the trial and then remained stationary (except for a stabilizing
jump when a saccade occurred). The bottom
record shows stimulus position, the middle
record shows eye position, and the top record
shows eye velocity. The initial target step is
marked with an asterisk. The subject made a
saccade and this was compensated for by a
target jump. After the saccade the subject
achieved fixation in one continuous smooth
movement, during which the target remained stationary.
We observed considerable intersubject
variability in the responses to the "saccadestabilized" stimulus of Fig. 1, A, with some
subjects making more saccades, particularly
for larger values of target offset. The degree of
predictability of target behavior also affected
the responses. In general, greater predictability tended to produce responses with
shorter latency and higher smooth eye movement velocity. However, familiarity with target behavior was not a prerequisite for the occurrence of position-directed smooth movements. Fig. 1, B, shows a "first time ever"
response by a naive subject to a stimulus like
that of Fig. 1, A; smooth eye movements directed toward the stationary target are apparent. All 10 of our subjects showed some posi-
Invest. Ophthalmol. Vis. Sci.
September 1981
480 Wyatt and Tola
B
30 -
o
co
^,
<D
liors
o
o
20-
-J
LU
LU
LU
I
h-
1
SEC
o
o
10 -
STEP SIZE (deg)
Fig. 2. A, Response to a 6 deg target step. Compare with Fig. 1, A. B, Peak smooth eye
movement velocity as function of initial step size for subjects J. P. (•), H. W. (o), and A. R. (A).
Size and direction of steps were selected randomly. Error bars are standard deviations.
tion-directed slow eye movements; however,
in the responses of two of the naive subjects
these were often small or absent.
As we increased the size of the initial target
step we observed that faster slow eye movements occurred. Fig. 2, A, shows the response to a 6 deg step, and the response may
be compared with that of Fig. 1, A, where a
smaller step elicited a slower smooth movement. The velocity records of Figs. 1, A, and
2, A, show increasing smooth eye movement
velocity during the early part of the response
and decreasing velocity during the late part.
We have observed that this is a general feature of responses, whatever the initial step
size. For five subjects we examined the effect
of step size on the maximum smooth eye
movement velocity in the responses, we varied step size from 0.5 to 6 deg in both directions and measured maximum velocity. Results for the three experienced subjects are
shown in Fig. 2, B. The results lump together responses to leftward and rightward
steps of each size and are the average of at
least six trials for each step size for one subject (J. P.) and at least four trials for the other
subjects.
The responses of the five subjects tested
systematically with different target step sizes
all showed an increase in smooth eye movement velocity as step size increased. The
best-fit slopes for the subjects shown in Fig.
2, B, are 3.3 (J. P.), 1.2 (H. W.), and 1.0
(A. R.), where units of slope are (deg/sec of
maximum response velocity) per (deg of initial target offset). The two naive subjects had
slopes of 0.5 and 1.0. The positive correlation
between maximum velocity and step size was
significant for all five subjects at confidence
levels of greater than 95%.
Responses to target ramps. Responses to
constant velocity target motion under normal
closed-loop conditions and saccade-stabilized
conditions are shown in Fig. 3. For the normal closed-loop ramp (Fig. 3, A) the subject
usually made a pursuit movement followed
by a saccade and then a "catch-up" pursuit
(faster than the target) that achieved fixa-
Volume 21
Number 3
Slow eye movements to eccentric targets 481
1 SEC
10°/s
Fig. 3. Responses to ramp target motion. Target motion at 10 deg/sec began suddenly at the
start of the trial. When subject made an initial saccade the target was jumped by an amount
equal to the saccade length (jump-saccade ratio = 1) for saccade-stabilization as used with
target steps. On other trials the first target jump was twice the saccade length and subsequent
jumps were equal to saccade length (jump-saccade ratio = 2), or the target was not jumped at
all; that is, it was a closed-loop ramp (jump-saccade ratio = 0). A, Response to a stimulus with
ratio = 0. B, Response to a stimulus with ratio = 2. C, Average velocity records in responses
to ramp stimuli. Computer-processed smooth eye velocity is plotted as a function of time after
onset of ramp motion. Rightward velocities are shown in upper half of figure, leftward in lower
half. Dotted curves, Ratio = 0; dashed curves, ratio = 1; solid curves, ratio = 2. D, Same as
C, with responses from a second subject. Note that catch-up behavior is represented by
overshoot of velocity curves above final value.
tion. Experiments with saccade-stabilization
(jump-saccade ratio = 1 or 2, Fig. 3, B)
elicited similar responses, except that the
catch-up pursuits were faster and more prolonged. Oscillations in slow eye movement
velocity were often observed, as seen in Fig.
3, B. Although only one saccade occurred in
the response of Fig. 3, B, often two or more
saccades occurred for jump-saccade ratios of
1 or 2. These later saccades often occurred
near the positions of the maxima in the oscillatory velocity record of Fig. 3, B.
To see more clearly the differences between the slow eye movements to the closedloop ramps and to saccade-stabilized ramps,
we determined average slow eye movement
velocities for the different conditions (Fig. 3,
C and D). Each of the average slow eye movement velocities was obtained for three to four
responses recorded on magnetic tape. Saccadic spikes on the velocity records were removed and linear segments were inserted to
close the gaps. The average was smoothed
once (with a running 4-bin average, 5 msec
binwidth), and the result was run out on a
stripchart at the speed at which it was recorded, but with a high-frequency cutoff
added at 15 Hz to reduce the contribution
from 60 Hz noise.
It can be seen from Fig. 3, C and D, that as
the jump-saccade ratio was increased from 0
(normal ramp) to 1 to 2, the slow eye movement velocity increased. The difference was
only apparent after several hundred mil-
482 Wyatt and Tola
liseconds, corresponding to times after a first
saccade. (Prior to the first saccade the target
motion was a simple ramp regardless of the
value of the jump-saccade ratio.) In examining Fig. 3, C and D, it should be kept in mind
that the catch-up portion of the record is represented by the overshoot of the velocity
above its final steady value.
Discussion
Responses to target steps. It is apparent
from Figs. 1 and 2 that under the conditions
of our experiments subjects can make slow
eye movements that are directed toward a
stationary target. Furthermore, these movements become more rapid when the target is
farther from the fovea. These data support
the view that target position relative to the
fovea is a stimulus for slow eye movements.
Why does the simple experiment fail?
Normal subjects always make a saccade when
they choose to look at a stationary, substantially eccentric target. If target position can
guide slow eye movements, then, since it is
possible to choose to suppress saccades while
maintaining fixation of a stationary target8 or
following a ramp target,9 it might seem that
subjects could choose to suppress saccades
while making a slow movement to an eccentric target. We suggest that a built-in
strategy initiates a saccadic response when a
subject chooses to look at an eccentric target.
Because saccades are very fast and because
they do successfully fixate real-world targets,
such a strategy would be teleologically
sound.*
Responses to target ramps. We observed,
*It must be noted that one of us (J. P.) is capable of
executing voluntary slow eye movements in the absence
of a target. However, this requires a deliberate, conscious effort and was avoided for the purposes of this
paper. Additionally, this subject has not been successful
at using voluntary slow eye movements to achieve
smooth fixation of eccentric targets, since the voluntary
process is disrupted as the target is approached. The
steep slope shown by J. P. in Fig. 2, B, may be of relevance to his voluntary capabilities. The other author
(H. W.) cannot perform such eye movements voluntarily.
Invest. Ophthalmol. Vis. Sci.
September 1981
as did Robinson,5 catch-up pursuits in response to closed-loop ramp target motion.
With such targets, offset of the eye from the
target after a saccade was usually quite small—
about 1 deg or less. The effect of our "saccade-stabilization" technique (with jump-saccade ratios of 1 or 2) was to create larger
postsaccadic position errors. Like the smaller
errors, these larger errors were eliminated
by catch-up movements that were now more
vigorous and more prolonged (Fig. 3).
Given the similarity between the catchup pursuit movements under normal closedloop circumstances (ratio = 0) and the catchup slow eye movements in our ratio = 1 and 2
conditions, it seems reasonable to suggest that
the two movements are produced by a common mechanism. Thus the greater velocity in
the ratio = 1 and 2 conditions could result
from larger postsaccadic offsets. This suggests
that catch-up eye movements, including normal catch-up pursuits, could be accounted for
by the relationship between slow eye velocity
and target offset (Fig. 2).
Slow eye movements and saccades during
pursuit. An interesting aspect of our results is
that increases in slow eye movement velocity
often appear quite suddenly at the end of a
saccade (see Fig. 2, A). This might seem to
suggest that the slow eye movements we
have observed are "glissades," a type of eye
movement usually thought to result from
mismatches between the pulse and the step
of saccadic motor innervation.10"12 However,
the time course of a glissade is a decreasing
exponential as the globe shifts passively from
the position to which the pulse took it to the
position set by the step. In our experiments
we have often observed increases in slow velocity after saccades during the early part of
the response and decreases during the later
part, as noted above and as seen in Figs. 1, A,
and 2, A. Furthermore, oscillations in eye
velocity are common (Fig. 3, B). Unless the
"step" level of saccadic innervation is continuously adjustable in the absence of a
pulse, which contradicts most views of the
system, such responses cannot be explained
as pulse-step mismatches. (This is not to say
Volume 21
Number 3
that glissades are absent from our records;
however, they are usually quite distinct from
the slow eye movements under discussion.)*
Sudden increases in slow eye movement
velocity after saccades occur frequently during normal tracking and have been interpreted in various ways. In particular, Westheimer13 suggested that discrete changes in
the velocity of slow eye movements, often
accompanied by saccades, are adjustments to
tracking errors. Robinson's5 view was that
changes in slow eye movement velocity,
linked to saccades, represent a nonlinear interaction between the pursuit and saccadic
systems. We suggest another possibility,
perhaps related to these views, which is that
the decision to "look at" a target activates a
foveal position mechanism that employs both
slow eye movements and saccades to achieve
its goal. Support for this idea comes from experiments in which subjects tracked ramp
target motion while avoiding "looking at" the
target but attempting to match its velocity.9
These subjects did not make the usual saccades toward the target, but it was also the
case that their slow eye movements were less
vigorous than usual. (Their slow movements
fell behind the target rather than catching up
to it or even matching its velocity.) This si*We note that at least one other explanation besides the
one that we have proposed could account for some of the
slow eye movements during the "saccade-stabilized"
trials. Because the target is stabilized during saccades, it
jumps sequentially from one location in space to the
next, according to the number of saccades that occur.
This sequence of target jumps is perceived, to some extent, as target motion. It is possible that this perception
of target motion activates the pursuit system and thus
gives rise to some slow eye movements. However, a
substantial number of responses occur with only a single
saccade (as in Fig. 1, A), so that only one saccadestabilized target jump occurs. This is too brief a "sequence" to generate much perception of motion, yet if
anything the smooth movements on such trials are more
vigorous than on trials with the same stimulus, where
the responses show more saccades. A related point is
that catch-up pursuit eye movements are seen with ordinary closed-loop ramp target motion, in which case there
is no target jump and thus no jump-related percept of
motion. In fact, we have occasionally observed catch-up
pursuit responses with no saccades.
Slow eye movements to eccentric targets 483
multaneous change in both saccades and slow
movements could reflect the inactivation of a
shared foveal position mechanism. We have
proposed in previous reports that slow eye
movements are governed by a position
mechanism acting together with the traditionally recognized velocity mechanism.2' 3 It
seems possible, then, that when saccades are
"suppressed" the associated weakness of slow
eye movements could represent the performance of the velocity mechanism acting by
itself.
We thank Ann Romano for extensive assistance in
conducting these experiments, John Orzuchovvski for
help with and design of electronics, and Ralph Dippner
for help with the computer.
REFERENCES
1. Kommerell G and Taumer R: Investigations of the
eye tracking system through stabilized retinal images. In Cerebral Control of Eye Movements and
Motion Perception, Dichgans J and Bizzi E, editors.
Basel, 1972, S. Karger, p. 288.
2. Wyatt H and Pola J: Position and velocity: the stimuli for smooth pursuit eye movements. Invest
Ophthalmol Vis Sci 17(April Suppl.):139, 1978.
3. Pola J and Wyatt H: Target position and velocity: the
stimuli for smooth pursuit eye movements. Vision
Res 20:523, 1980.
4. Wyatt H and Pola J: The role of perceived motion in
smooth pursuit eye movements. Vision Res 19:613,
1979.
5. Robinson DA: The mechanics of human smooth pursuit eye movement. J Physiol 180:569, 1965.
6. Westheimer G: Mechanism of saccadic eye movements. Arch Ophthalmol 52:710, 1954.
7. Bahill AT, Clark MR, and Stark L: The main sequence, a tool for studying human eye movements.
Math Biosci 24:191, 1975.
8. Steinman RM, Haddad GM, Skavenski AA, and
Wyman D: Miniature eye movements. Science
181:810, 1973.
9. Puckett J de W and Steinman RM: Tracking eye
movements with and without saccadic correction.
Vision Res 9:695, 1969.
10. Weber RB and Daroff RB: Corrective movements
following refixation saccades: type and control system analysis. Vision Res 12:467, 1972.
11. Easter SS Jr: A comment on the "glissade". Vision
Res 13:881, 1973.
12. Dell'Osso LF, Robinson DA, and Daroff RB: Optokinetic asymmetry in internuclear ophthalmoplegia.
Arch Neurol 31:138, 1974.
13. Westheimer G: Eye movement responses to a horizontally moving visual stimulus. Arch Ophthalmol
52:932, 1954.