Anticipatory Saccades in Smooth Pursuit: Task Effects and ...a concurrent listening condition instead, to see if this mild distraction would degrade performance. Performance improved

Pergamon 0042-6989(94)00161-8

Vision Res. Vol. 35, No. 5, pp. 667--678, 1995 Copyright ~) 1995 Elsevier Science Ltd

Printed in Great Britain. All rights reserved 0042-6989/95 $9.50 + 0.00

Anticipatory Saccades in Smooth Pursuit: Task Effects and Pursuit Vector After Saccades PETER VAN GELDER,*t+ + SERGEY LEBEDEV,*t PETER M. LIU,I" WAI HON TSUI*t

Received 23 September 1993: in revised form 4 July 1994

The dramatic improvement in smooth pursuit performance seen while analyzing the pursuit target has been ascribed to attention enhancement. With a periodic constant velocity target trajectory we ran a concurrent listening condition instead, to see if this mild distraction would degrade performance. Performance improved somewhat with the listening task, suggesting that displacing attentionai effort from pursuit accuracy, rather than increasing it, brings better pursuit performance. Catch-up saccades were evenly distributed across tracking, listening, and target analysis conditions, but anticipatory and overshooting saccades were almost eliminated with target analysis. Thus the poor pursuit seems to have been caused by anticipatory and overshooting saccades, produced erroneously in the attempt to perform purposive smooth pursuit. Pursuit velocity immediately following anticipatory saccades was reduced such that the target would catch up with the point of gaze when it reached the endpoint of its trajectory, indicating a predictive goal other than instantaneous target foveation and velocity match.

Pursuit Saccades Prediction

To investigate smooth pursuit performance, we note the phenomenon well known in the psychiatric smooth pursuit literature whereby dramatic improvement is seen in pursuit performance when subjects (normals or patients) are asked to analyze some changing characteristic of the target, such as reading silently a changing letter or number which is the pursuit target (Holzman, Levy, & Proctor, 1976; Shagass, Roemer, & Amadeo, 1976; Spohn, Coyne, & Spray, 1988). In other tasks the subject presses a button (Iacono & Lykken, 1979; Iacono, Peloquin, Lumry, Valentine, & Tuason, 1982; Iacono, Tuason, & Johnson, 1981), counts (Van Gelder, Anderson, Herman, Lebedev, & Tsui, 1990b), notices (Clementz, Sweeney, Hirt, & Haas, 1990; Levin, Lipton, & Holzman, 1981; Lipton, Levin, & Holzman, 1980b), or receives no instructions whatever (Cegalis, Hafez, & Wong, 1983; Cegalis & Sweeney, 1981), regarding the changing pursuit target.

This phenomenon tends not to be mentioned lately (e.g. Abel, Levin, & Holzman, 1992) because target analysis is thought to be simply a novel stimulus or involving task that induces better performance in other- wise inattentive subjects. In this light, target analysis

*Nathan S. Kline Institute for Psychiatric Research, Orangeburg, NY 10962, U.S.A.

tDepar tment of Psychiatry, New York University Medical Center, 550 First Avenue, New York, NY 10016, U.S.A. [Email pvg (aipl .med.nyu.edu].

STo whom all correspondence should be addressed at New York University Medical Center.

tasks are often called attention enhancement tasks. This explanation is supported by a consistent body of evidence that the reverse situation holds: a concurrent demanding cognitive or attentional task, unrelated to the pursuit task, distracts subjects and degrades smooth pursuit performance by introducing large saccades and fixations (Acker & Toone, 1978; Brezinova & Kendell, 1977; Lipton, Frost, & Holzman, 1980a; Pass, Salzman, Klorman, Kaskey, & Klein, 1978).

It seems equally likely, however, that smooth pursuit is best performed automatically, in the service of target analysis, so that diverting attention away from the oculomotor act itself by redirecting it toward target analysis could enhance performance. To test this notion, the present study employs a concurrent listening task as a mild distractor. The attention enhancement explanation of the target analysis facilitation would predict a tendency toward worse performance with the listening task than with simple pursuit, while the attention diversion explanation would predict a tendency toward better performance. We also employ a target analysis condition, and a more difficult distraction condition to replicate the worse performance obtained by more severe distraction from the tracking task as a whole.

Poor pursuit performance tends to be reported by such global measures as smooth pursuit gain (pursuit velocity/target velocity) and root-mean-square (r.m.s.) error. Also typically seen are large saccades, unrelated to the small corrective catch-up saccades made by most subjects. Task differences in performance within subjects here allow us to differentially characterize the saccades

667

668 PETER VAN GELDER et a/.

made in good and poor pursuit, and to look at the interaction between saccades and smooth pursuit gain.

Poor performance with our trajectories was associated with anticipatory and, less often, overshooting saccades. In moving the point of gaze ahead of the target, subsequent pursuit showed a strong tendency to be reduced in an orderly way, providing an opportunity to observe predictive characteristics of pursuit apart from its function of instantaneous target foveation. We show this pursuit-after-saccade function for anticipatory saccades.

Although the incidence of anticipatory saccades in pursuit can be vastly reduced by more favorable target trajectories and by suitable instructions to subjects, their presence in our data allows us to suggest an alternative way to characterize the purposive smooth pursuit that occurs in the laboratory.

METHOD

Subje.cts

Twelve normal subjects were run in Expt 1, and two groups of 10 normal subjects each were run in Expt 2. The ages in Expt 1 ranged from 23 to 43 yr, with a mean of 29.8 yr. The ages in Expt 2 ranged from 23 to 36 yr, with a mean of 30.0 yr, in the first group and from 21 to 32 yr, with a mean of 25.4 yr, in the second group. None reported oculomotor, neurological, or psychiatric conditions likely to influence eye movement. All reported normal or corrected visual acuity at the normal reading distances required of their occupations (e.g. medical students, secretaries).

Apparatus

EOG recording was employed: a horizontal channel obtained with electrodes at the outer canthus of each eye, vertical channel obtained by electrodes above and below the dominant eye, and a mid-forehead ground. The differentially amplified signals were digitized in real time (250Hz) by the same computer presenting the stimuli on the subject's monitor (Van Gelder, Todd, & Tsui, 1979), so that subsequent analysis could match experimental events with response waveforms (Tsui & Van Gelder, 1979). The EOG waveforms of horizontal and vertical channels were displayed on an experimenter's monitor. Head restraint was achieved with a chin and forehead rest with additional head stops.

Conditions

Four pursuit conditions were run, in 40 sec trials that used either a constant or (where indicated below) sinusoidal velocity trajectory, and in either the horizontal or (where indicated below) vertical plane. Of primary interest are tracking, target analysis and listening conditions only. The conditions were as follows.

Standard tracking, to follow an upper-case "X", 0.34 deg wide × 0.6 deg high. Instructions were to "follow the target as closely as you can". Target analysis, where the target was a lower-case letter in the same font, changing to a new letter

every 0.5 sec. Separate randomizations were used for each trial, constrained to remove words formed by successive letters. Instructions were to "read the letters to yourself while following the target". Listening, with the same pursuit target as for the tracking condition, but the same letter series as for target analysis in auditory presentation. Subjects were asked to "pay close attention to the letters on the tape while tracking". Distraction, similar to listening but with a different high-frequency three-letter word embedded in the first, middle and last thirds of the trial. An example was given of a random letter sequence with a three-letter word embedded. Subjects were asked to find all occurrences of such words and repeat them to the experimenter at the end of the trial.

Experiment 1 running order

Two trials were run per condition, in the order: tracking, sinusoidal tracking, listening, target analysis, distraction, vertical tracking, vertical listening, vertical target analysis, sinusoidal tracking repeat, tracking repeat. The tracking condition was run at the beginning and end of the series as a baseline, paired with sinusoidal tracking (at the beginning and in reverse order at the end) to compare performance on the two trajectories. The order of tracking and sinusoidal tracking was counterbalanced across subjects; independently the order of listening and target analysis (along with vertical listening and vertical target analysis) was counterbalanced, for a total of four separate running orders with three subjects in each.

Experiment 2 running order

To obtain a more direct comparison of the three conditions of primary interest, Expt 2 was run, with constant and sinusoidal trajectories run in separate groups. Only one trial was run per condition, in the order: tracking, listening, target analysis, tracking, listening, target analysis, tracking in the first group and in the same order with sinusoidal trajectories in the second group. Again, the order of listening and target analysis was counterbalanced across subjects in each group.

Procedure

Each trial began and ended with a brief calibration trial. The 0.4 Hz tracking stimulus moved 30 deg horizontally or 22 deg vertically. Constant velocity trajectories moved 30deg/sec horizontally or vertically. A pause of 0.22 sec was included at each endpoint of the trajectory in the horizontal constant velocity conditions (illustrated in Fig. !) and, to preserve frequency and velocity, 0.50 sec in the vertical conditions.

For the auditory conditions, listening and distraction, three recordings of the two lists were made, with a new letter every 0.625 sec. These three recordings were varied across trials and subjects. All auditory stimuli were delivered over two speakers located behind the subject screen.

ANTICIPATORY SACCADES IN PURSUIT 669

Data analysis'

Invalid portions of the response waveforms were removed first. These were quite rare in the horizontal conditions, comprising mostly some occasional crosstalk from blinks. Blink artifact was frequent in vertical waveforms. Hardcopy plots of stimulus trajectories and response waveforms were then produced with saccades indicated, as a validity check on our calibrations and saccade detection algorithm.

Overall smooth pursuit gain (pursuit velocity/target velocity) was computed separately for each smooth pursuit segment longer than 100msec, excluding saccades, invalid data excluded interactively, and times when the target was stationary. A time-weighted average of these quotients was determined for each half cycle, and these were averaged over the trial. A r.m.s, error measure was computed over all data except invalid data excluded interactively, expressed in degrees. The r.m.s, error indicates angular distance between point of gaze and target, so reflects both pursuit and saccades.

Saccades were classified by computer with respect to three mutually independent relationships to the target trajectory: saccade direction (toward or away from the target), starting position (on, behind or ahead of the target) and ending position (also on, behind or ahead of the target). Only 10 out of the 18 saccade types generated by this partition are physically realizable; the result has intuitive face-validity and lends itself to an objective, automated classification procedure. We will give primary consideration to the three most commonly observed saccade types: catch-up (toward target from behind, landing on target), anticipatory (away from target from on target to ahead of target) and overshoot (toward target from behind, landing ahead of target).

Parameters of each saccade were saved automatically for subsequent analysis. For task effects analysis, saccades and a measure of the average velocity of the pursuit segments 60msec immediately preceding and following each saccade were determined automatically for all subjects.

For pursuit-after-saccade projection analysis, saccades and constant velocity pursuit immediately preceding and following each identifiable saccade were marked interactively, beginning at the edge of the saccade and extending as far as the pursuit velocity held constant. This manual waveform selection procedure was employed to acquire longer and more valid constant velocity pursuit segments surrounding the saccades than afforded by our automatic procedure.

These interactively determined saccades and pursuit segments were marked on horizontal constant velocity tracking and listening conditions on a subset of 12 subjects. Of the 12 subjects of Expt 1, four were selected for the present analysis as having the highest total saccade amplitude. Of the 10 subjects in Expt 2 who were run with a constant velocity trajectory, two were dropped from the interactive analysis for having noisy EOG waveforms.

RESULTS

Waveform observations

Figure I presents representative waveforms for a subject in Expt 2, for each condition in the order of running. Listening improves performance slightly, an effect that was enhanced in other subjects where listening followed target analysis rather than tracking (Van Gelder, Lebedev, & Tsui, 1990a). Anticipatory saccades are not seen at all with target analysis, but appear to characterize the relative performance deficit in the tracking and listening conditions. Some other subjects mixed these with overshooting saccades: after an initial delay, a large saccade would extend well ahead of the target. For either anticipatory or overshooting saccades, the subsequent pursuit gain tends to be reduced to produce a trajectory where the direction of gaze will meet the target at approximately the end point of its excursion. Catch-up saccades are seen at the beginning of most half-cycles of pursuit, independent of condition.

Subject distributions and group effects

Figure 2 shows distributions of performance on the reference tracking condition separately for the 12 subjects in Expt 1 (horizontal and vertical) and the 10 subjects in each of the two groups in Expt 2. Wide differences in performance are seen across subjects; some made almost no saccades except for small catch-up saccades. The subject in Fig. 1 had median performance in Expt 2, constant velocity.

FIGURE 1. Constant velocity stimulus trajectories and response waveforms for two cycles of each tracking condition for a subject in Expt 2, in running order. Saccades detected by computer algorithm are enclosed in rectangles. T, tracking condition; L, listening condition:

A, target analysis condition.

670 PETER VAN GELDER et al.

8

7

6

ILl

3

2

1

0

i i

I : i i

I I I

1 H l V 2 0

Data Group

8

! i

I

2 8

FIGURE 2. Distributions of r.m.s, error performance in the reference tracking condition for each data group. IH, Expt 1, horizontal constant velocity; IV, Expt 1, vertical; 2C, Expt 2, constant velocity;

2S, Expt 2, sinusoidal velocity.

The two groups of Expt 2 differ only in target trajectory, so they demonstrate directly an advantage for sinusoidal over constant velocity trajectory on this r.m.s. error measure [F(1,18)=7.8 , P =0.01]. This group difference is seen also in the other measures of Fig. 3: saccades/trial ( F = 5.2, P = 0.03), saccade amplitude (F = 6.4, P = 0.02) and pursuit gain (F = 4.8, P = 0.04).

Task effects

Figure 3 shows performance in tracking, listening and target analysis conditions separately for horizontal and vertical target directions in Expt 1, and for constant and sinusoidal trajectories in Expt 2. The saccade measures of count/trial and mean amplitude are shown, as well as smooth pursuit gain and r.m.s, error.

In general, each measure appears to show improvement in performance with listening, and further improvement with target analysis. Over all three horizontal pursuit data groups, the effect of condition was highly significant, with P < 0.001 for each measure separately. There were no group by condition interactions, but there nearly was for r.m.s. [F(4,58)= 2.5, P = 0.053]. Differ- ence contrasts showed target analysis better than listening on each measure: saccades/trial [F(3,29)= 13.4, P < 0.001], saccade amplitude (F = 9.4 P < 0.001) pursuit gain ( F = 5.3, P =0.005) and r.m.s. ( F = 6 . 9 , P = 0.001). Listening was better than tracking on saccade amplitude [F(3,29)=3.7, P =0.02] and r.m.s. (F = 3.8, P = 0.02), but not for saccades/trial or pursuit gain.

Vertical pursuit was analyzed separately, and showed an overall condition difference only for the saccade amplitude measure [F(2,22) = 11.7, P < .001]. Differ-

ence contrasts showed listening better than tracking [F(1,11) = 6.6, P = 0.03] and listening worse than target analysis (F = 6.9, P = 0.02) for this measure. Difference contrasts for other measures showed only listening worse than target analysis for r.m.s, error (F = 6.7, P = 0.03). Although the rather frequent blink artifact was removed from consideration in analysis, EOG at best provides a poor measure of vertical eye movement. Thus the vertical results should be considered only generally, as not inconsistent with the results of the horizontal conditions.

While statistical evidence of improvement in performance with a concurrent listening task was not seen for all measures, the trend is clearly toward improvement rather than degradation. The only reversals are in the vertical r.m.s, and in saccades/trial, which can actually increase with better performance (Puckett & Steinman, 1969), particularly as saccade amplitude decreases (Van Gelder et al., 1990), making saccade counts an unstable measure.

For sinusoidal velocity tracking in Expt I, difference contrasts (among tracking, sinusoidal tracking, listening and distraction conditions), showed sinusoidal tracking improved over tracking in saccades/trial [F(I, l l) = 21.8, P = 0.001], smooth pursuit gain (F = 5.5, P = 0.04) and r.m.s, error (F = 4.9, P = 0.049), but not for saccade amplitude (F = 0.24). Experiment 1 thus shows within subjects an advantage of the sinusoidal over the constant velocity trajectory similar to that seen across subjects in Expt 2.

The more demanding distraction condition of Expt 1 did not show the predicted worsening of performance. Performance in the distraction condition was better than in tracking for pursuit gain (F = 5.4, P = 0.04), and the same as in tracking for the other measures. Also, the same difference contrasts show listening better than distraction only in r.m.s, error (F = 7.4, P = 0.02), and the same as distraction for the other measures.

Task effects on types of saccades

Table 1 shows catch-up, anticipatory, overshooting and all other saccades separately for tracking, listening and target analysis, using all horizontal constant velocity data from both experiments for these conditions (data groups 1H and 2C of Fig. 3). Of the 10 saccade types of our saccade classification scheme, 71.1% of the 4580 saccades are in these three categories. There are fairly equal numbers of catch-up saccades in the three conditions, while anticipatory and overshooting saccades almost disappear with target analysis. Saccades in the "Other" category, which includes some variations on anticipatory and overshooting saccades, are also reduced by half with target analysis.

The improvement in performance with the concurrent listening task was seen in Fig. 3 to come from a reduction in saccade amplitude and, in horizontal trajectories, reduced r.m.s, error, which may have derived from the saccade amplitudes. Table 1 shows this reduction in saccade amplitudes to come primarily from anticipatory and overshooting saccades, in that saccade amplitudes of


anticipatory and overshooting saccades in the listening condition are less than in the tracking condition.

Classification results of hand-marked data Of the 1817 saccades marked for the pursuit-after-

saccade analyses given below, 71.5% are in our three categories of primary consideration. As shown in Table 2, pursuit gain before and after catch-up saccades is as we would expect: low gain before and high gain after the saccade, reflecting the prevalence of these saccades at the abrupt start of each half-cycle of pursuit• Pursuit gain surrounding anticipatory saccades shows the reverse, reflecting the tendency for lowered pursuit velocity after anticipatory saccades. Overshooting saccades tend to have low gain both before and after the saccade, reflecting their tendency to occur near the start

of the pursuit half-cycle: a delayed start of pursuit will combine functions of a catch-up and anticipatory saccade in a large saccade extending well ahead of the target instead of landing on the target.

Post-saccade pursuit: projection to trajectory endpoint

Figure 1 provided illustrations of anticipatory saccades followed by pursuit whose velocity is reduced such that the point of gaze reaches the endpoint of the trajectory when the target does. In some instances the pursuit continued to that endpoint; in others the pursuit was interrupted by a subsequent saccade, often corrective. In both types of instances the initial pursuit after the anticipatory saccades seemed to have the same velocity characteristic, so both are treated the same here. Note that the pursuit velocity itself varies, being

(a) 60

5 0

_ 4O ¢1 L.

-8 3o

II1 m 2o

1 0

0

1,0

0.g

,- 0.8

0.7

~ 0 .6

m 0 ,5

0,4

0,3

(c)

, ,

\

\ \

%% O .. . . . . . . ~

' - . .

I I I

T L A

O

I

I o .,.~ t

t

,,...@ t

I I I

T L A

Condition

7

'10

<

'10

86 O

(b) g

3

(d) 5

I I I

=,.

\ %%

"" ""'°° " '" '"" Q ' " " ' " " ° I X

"°"O

I I I

T L A

I I I

','7

rr

4

3

..*.-El

\ \

A . \

~ X

"'-, . " '-- , . . .

1 i i T L A

Condition

.~ ~. 1 H & . - - - - - - - . - & 2 0

B-- - - - B 1 V 13- . . . . . . . . . G 2 8

F I G U R E 3. Performance in tracking, listening, and target analysis conditions for each of the four data groups of Fig. 2. 1H, Expt 1, horizontal trajectory; IV, Expt 1, vertical trajectory; 2C, Expt 2, constant velocity trajectory; 2S, Expt 2, sinusoidal trajectory. Measures are count of saccades per trial (a), mean saccade amplitude in degrees (b), smooth pursuit gain (c), and

r.m.s, error (d).


TABLE 1. Types of saccades by condition

Catch-up

Saccade type

Anticipatory Overshooting Other

Measure T L A T L A T L A T L A

Saccades /trial Mean 9.01 9.91 10 .36 10 .69 10.66 0.57 2.78 2.34 0.23 9.14 9.32 4.64

Saccade amplitude Mean 4.62 4.72 3.97 9.12 7.70 6.53 16 .23 14 .12 13.82 5.23 5.01 3.40 SD 1.37 2.55 1.22 6.82 5.45 4.08 7.89 7.37 5.87 4.05 3.58 1.60

T, tracking condition; L, listening condition; A, target analysis condition.

determined by the endpoint of the saccade. On these waveform plots ideal examples would show the pursuit after anticipatory saccade in each case as the initial segment of a straight line between the saccade endpoint and the trajectory endpoint.

The trajectory endpoint is 30 deg from its starting point, so at the time the trajectory reaches its endpoint the initial pursuit-after-saccade vector of our ideal example would project to 30 deg. Distributions of actual pursuit vector projections are shown for each of the primary saccade types in Fig. 4. Anticipatory saccades come closest to the mark, with a mean (and SD) of 28.9 (2.4) deg. Catch-up and overshooting saccades fall shorter of the mark, and with more variability, at 26.3 (3.9) and 27.5 (3.8) deg, respectively.

A properly behaved catch-up saccade should acquire the target and follow it, so that the initial pursuit vector would project trivially to 30 deg. The counterintuitive lower mean and greater variability of the catch-up saccade pursuit vector distribution shown in Fig. 4 remained even when we removed the cascades of catch- up saccades, each followed by low pursuit gain, that some subjects produced in the first cycle or two. All such data are included in Fig. 4.

TABLE 2. Types of saccades, with surrounding smooth pursuit gain

Saccade type

Measure CUS AS OSS Other

Saceades/trial 8.20 9.80 3.29 Saceade ampftude

Mean 4.99 9.01 16.2 SD 1.87 5.71 7.25

Gain before saccade Mean 0.42 0.80 0.38 SD 0.24 0.25 0.23

Gain after saccade Mean 0.88 0.41 0.44 SD 0.20 0.29 0.36

8.49

5.70 3.46

For all hand-marked saccades. CUS, catch-up saccades; AS, anticipatory saccades; OSS, overshooting saccades. Mean and SD pursuit gain for the constant velocity pursuit segment immediately before and after the saccade are shown for the three principal saccade types.

Post-saccade pursuit: pursuit gain One would imagine alternative courses for pursuit

after an anticipatory saccade would be (a) to continue pursuit in parallel with the target, with unity gain; (b) to stop and wait for the target to catch up, with zero gain; or (c) some indeterminate slowing. Figure 4 suggests none of the above, but that pursuit gain would be a function of the time and position of the endpoint of the saccade. To specify the function, let ~ = distance between endpoints of the trajectory (in deg); T = half-cycle travel time (in sec); t~ = time of saccade end (in sec); E(t) = eye position at time t (in deg); G ( t ) = pursuit gain at time t; Ve(t ) = eye velocity at time t (in deg/sec); V~(t ) = target (stimulus) velocity at time t (in deg/sec). We use a common definition of pursuit gain, as

v~(t) G ( t ) -

vs(t)" A constant velocity trajectory is employed for the present data, so that

I / s ( t ) - Vs.

8 0 ' ' ' ' i , ' ' ' i ' , , , i , , , , i , , , ,

2O I I A 8 I

"" I

:/,,.;I ?\ / ½\,,

1 {5 2 0 2 5 3 0 3{5 4 0

Pursuit Vec to r Pro jec t ion (dog)

FIGURE 4. Distributions of pursuit vector projections to the time target reaches the end of its trajectory (30 deg), for catch-up saccades (CUS, solid line), anticipatory saccades (AS) and overshooting saccades (OSS). For example, a value of 29 means that if the initial pursuit after a saccade were to continue with the same velocity, the subject's point of gaze would be at 29deg, or 1 deg short, when the target

reached the end of its trajectory at 30 deg.


FIGURE 5. Hypothetical gain of initial pursuit after anticipatory saccades, as a function of the time and position of the saccade endpoint, assuming that pursuit is slowed to reach the endpoint of the trajectory concident with the target. The indicated function slices are

shown individually in Fig. 6.

I f the vector o f eye velocity points to the endpoint o f the trajectory, then this velocity is

f~ - E (t~) _ remaining distance V ~ t t , ) -

T - t~ remaining time

Gain o f this pursuit vector then becomes

f ~ - E ( t l ) G (tl) -

V, x ( T - tl)"

This hypothet ical function o f two variables is shown in Fig. 5, in the form of two-dimensional slices o f each variable at several levels o f the other. The function corresponds to the distributions in Fig. 4 in that if all data points were tightly aligned with this gain-after-

*We have also characterized these pursuit vectors from the time and angular position of the start of the saccade, adding the saccade's duration and amplitude to obtain its endpoint. We let t o = time of saccade start (in sec); ,4 = saccade amplitude (in deg); D = saccade duration (in sec) and note that

E(f i ) = E(to) + A t~ = to + D.

Then f ~ - ( E ( t o ) + A )

G(t l ) V, x (7" - (to+ 0))"

The formula is equivalent to the end-of-saccade function given above, but describes gain as a stochastic function of different parameters. To visualize this function we reduce its four variables (E(t0), A, D, to) to two by expressing duration as a function of amplitude and by assuming that the saccade begins on target, thereby determining E(to). This has the appeal of including main sequence parameters, producing a function of start time and amplitude, We obtain clear fits of pursuit vectors after anticipatory saccades to this function (Van Gelder, Lebedev & Tsui, 1992), as well as to the function in Figs 5 and 6.

saccade function, the corresponding distribution o f pursuit vector projections would center narrowly on 30 deg.

Rather than show the data in this space, we obtain a clearer view of the fit o f da ta to the function by breaking out each slice individually, in Fig. 6. All ant icipatory saccades o f Table 2 and Fig. 4 within the range o f these slices are plotted both ways in Fig. 6.

We see that in general gain after ant icipatory saccades is neither unity, with pursuit parallel to the target, nor zero, indicating a fixation. Rather, gain is a lmost always between these values. Further, there is a s trong tendency for the data points to lie close to the gain-after-saccade function. In particular, data points near the bo t tom of the plots, where gain after saccade is near zero, tend to occur with end position near 30deg. Fixations after ant icipatory saccades are thus a special case o f the general function, where the saccade had reached the endpoint already. Similarly, data points near the top o f the plots, where gain after saccade is near unity, tend to occur early in end time and end position, so that very little slowing is required for the target to catch up by the endpoint o f the trajectory. No t shown are 20 o f the 599 anticipatory saccades whose gain after saccade was negative.*

Logical constraints on the data should be noted. Figure 6(a) shows gain after ant icipatory saccades as a function o f end time of the saccade, for various values of saccade end position. The vertical dashed lines indicate for each plot a time later than which a properly defined anticipatory saccade endpoint would not occur. This is the time the target passes that position: for an anticipatory saccade to land ahead of the target, and to land at that position, it must do so before the target gets there. Similarly, Fig. 6(b) shows gain after ant icipatory saccades as a function o f end position o f the saccade, for various values o f saccade end time. Here the dashed lines indicate the position o f the target at the time of the plot, so at that time the saccade must have ended at a position beyond the target.

A looser constraint is that ant icipatory saccades tend not to begin until pursuit has begun, perhaps with a catch-up saccade, and some refractory period has elapsed. They end after another 50 msec or so o f saccade duration. The earliest saccade endpoint was at 367 msec.

DISCUSSION

Antic ipatory saccades were produced erroneously by some subjects with our target trajectories, degrading pursuit performance in the apparent effort to make it better. They can be differentiated from catch-up saccades in several converging respects, affording a sugges- tion as to their cause. Because they serve to move the eyes ahead of the target, pursuit following anticipatory saccades demonstra tes predictive effects apar t from the preservation o f an instantaneous velocity match.

T a s k e f f ec t s

Catch-up saccades were distributed evenly across task conditions, but ant icipatory and overshoot ing saccades

674 P E T E R V A N G E L D E R et al.

were virtually eliminated with target analysis. Our listening condition also produced some improvement, primarily by a reduction in amplitude of anticipatory and overshooting saccades. The improvement was not nearly as great as in the target analysis condition, but it was consistently in the direction of improvement over three subject groups, four response measures, sinusoidal

as well as constant velocity target trajectory, and for vertical as well as horizontal pursuit.

Our distraction condition did not in general lead to improved smooth pursuit, but it did not cause the predicted degradation either. Our distraction condition was not as severe as those distraction conditions pre- viously shown to degrade performance in normal

0.9 < 0 0 < CO cr 0,6 I11 I-

Z o 0.3

0.0 0,26

I I I

0.50 0.76 1.00

END TIME (sec) AT 13 DEG

1.2

I I

1.26 0.25 0,50 0,75 1,00 1.25

END TIME AT 15 DEG

I I I

0.25 0.50 0,75 1,00

END TIME AT 17 DEG

125

i i 1,2

0,9 < O O < 09 cc 0.6 W I- LL

Z

0.3

0,0 025

i

o o

I I i

0.50 0.75 1 D0

END TIME AT 19 DEG

I I 1

1.25 0.25 0.50 0.75 1.00 125

END TIME AT 21 DEG

/

o° : °

I I I; 0.25 0.50 0.75 1.00

END TIME AT 23 DEG

1 . 2 5

1.2

LU 0.9 £3 < O O ,< 03 ,-r 0.6

(D 0.3

0.0

, i i

o

;-o °.o. °~,

o o o I o

* to , . , !

0.25 0.50 0.75 1.00

END TIME AT 2/5 DEG

1.25 125 0.25

, o o

o o l d o * *°°:

o o t ° ° !

0.25 0.50 0.75 1,00

END TIME AT 27 DEG

° °i* ° * °o° o *o~ ..~d'~.

0.50 0,75 1.00 1.25

END TIME AT 29 DEG

FIGURE 6(a). Pursuit gain after anticipatory saccades as a function of end time of the saccade, for successive 2-deg ranges of end position of the saccade. At the midpoint of each range, as indicated for each scatterplot, the function is plotted indicating the gain after saccade required for the point of gaze to reach the endpoint of the trajectory coincident with the target. Dashed limit lines indicate the time the target passes that position, so that no anticipatory saccade should end at that position later

than that time.

p u e l !nsand q l o o t u s aAtsodand ~u.tanpoad j o s~Ismqns

OA~I aql sluosoad uo!l l .puoo ~u!ums!l a q I "l!nsand q l o o t u s

aMsodand se u o u a t u o u o q d , Q o l ~ a o q e I le.t3~j!lae s!ql

o l aajaa abA "l!nsand q l o o m s alean33e gu !onpoad j o ~lsel

aJ~U!S lz ql.lA~ laa.fqns e slttasaad uo.lll.puoa ~u.D[ol3al aq, L • otuos u! p~aaoju! luotuoAoadtu!

i en l3e j o aauap!Ao otuos ql!A~ ' u o ! l e p e a ~ o p qans ou

gU!A~OqS SUOp,!puo3 gu!pumuop SSO l U~a OSI~ uo!lepe2~op gU!A~OqS so!pnls oql jo lsolAI "(~;86I 'olpu~I ~g 'P~O~I 'tuo~lsmI~IAI) )Ism uo!leu!tugas!p qlguol-doo q Xaol!pn~ llna~!p e ao (8L6I '7 ° la SSed) uo!led!a!laed ;)A.ll::)t~ gu.u.mboa ~IS~l gu.lualSt.l o.tloqo.[p ~ '(e0861 "l n la uold.tq 'LL6I 'IPpuo)I 7~ eAOU!ZOa~ '8L61 'ouoo£ ~ a~IaV) ~UplOeal Ol!qA~ spaeA~lagq $ut.lunoa se qans 'sloo.fqns

g8

'uo[l!sod leql puoKoq popu~ l aA~q plnoqs op~3aes Xaol~d!ague aql os 'atu!l leql 1~ lag~l aql jo uo!l!sod oql alg, a[pu[ sau[l lltU[l paqsecI "lafl~l aql ql.t~ luap!au.to3 ,Qoloo.f~.q oql jo lu!odpuo aql qoea~ Ol az~ jo lu!od aql soj poa.mboa ap~oa~s JaOe u!~ aql ~u!l~o.tpu! POllold s! uo.tlaunj oql 'lold~ml~OS qo~a aoj poleo!pu! se 'o~ue~ qa~a jo lu.lodp!tu oql 1V "~pea3~s oql jo otu!l pua jo so~tte,.l aas- l '0 aA!SSaaans .ioj 'ap~oo~s Oql jo uotlt.sod puo jo UO!lOUnj 1~ s~ sap~oo'es ~.lol~d!a.lltI1~ .iOljl~ ull~ l!ns.m d (q)9 El~lfl£)Ict

03S' ~L lV NOILIS'Od (]NB

OC g~ O~ gL OL g

I ' 1 ~ol ~ll'l I I I I

go g go

OB£ L'L L¥ NOLLISOd ON3

O0 g~ O~ gL OL

'.k o 1 , ,

m o l

I I t

oq8 0 7. J_V NOIJJSC)d ON9

OC g~ O~ gL OL i i i i

i

g 00

O'0 0 _> z

g'o :~ £0

o o c~

6'0 m

gc

0L g I

098 6 £ V NOIZI$OdGN3

OC g~ O~ gt e : i I I I

i

g~ OL @ gC I

038 8" I V NOlil$Od ONB

Oe g~ OE gt - = I I I

03£ Z' IV NOIZISOd aN3

oc g~ o~ gt o~ • I I I I

I

g 0'0

_> z

90 :~ > © © >

6o ~

go

09£ 9' IV NOl±lSOcl ON]

OC g~ O~ gL Ot - : : i " o r I I

° °

. . . . L \

g£

O38 g IV NOtJ_IS'Od C]N3

oo g~ o~ ~L OL I I I I

. ! I I I I I ,

g£

038 17' ZV (0ep) NOIll£Od 0N3

OC g;3 0;~ gL Ot

o o

. o

°\ .o

O0

I I I I e iO

g 0'0

12'0 © >.

9'0 :a (.0 >, O O > O

60 m

~'L

~L9 ZlflS~Ifld NI S~]OVDDVS A~IOIVdlDIINV


listening to the random letter series. This divided attention condition entails a reduction of attentional effort in the smooth pursuit subtask, moving the pursuit process closer to its skilled automatic mode. Smooth pursuit outside the laboratory is similar to a target analysis task, where attentional effort is devoted to target analysis exclusively. In this case of real-world pursuit, investi- gation of the moving target is an integrated perceptual and visuomotor act, with the pursuit proceeding automatically and optimally. Consistent with this view, subjects report total lack of awareness of a performance difference between tracking and analysis conditions, in our laboratory and others (Holzman et al., 1976; Iacono & Lykken, 1979; Shagass et al., 1976). Further, the improvement with target analysis occurs for all subjects (except for ceiling effects in the case of some excellent subjects) and persists over time, here and in earlier work (Van Gelder et al., 1990).

Any form of mild distraction ought to provide the same implicit requirement to remove attention from the act of tracking. Each of three types of alternative listening materials produced similar improvement in pilot work. Of course, when providing instructions to subjects, explicit is generally better than implicit. A colleague who wishes to study smooth pursuit when it is optimal obtains dramatic improvement in performance when necessary by asking subjects to "relax, don't try so hard" (R. Fendrich, personal communication, May 1988). We may in fact bias subjects toward this attention shift by our instructions, which emphasize listening to the tape rather than pursuit accuracy. Whether by instruction or by task or both, our intent was to produce the attention shift, rather than a pure case of auditory/ visual interaction.

Effect of target trajectory

Pursuit of a constant target in a sinusoidal velocity trajectory was better than in a constant velocity trajectory, clearly across subjects in Expt 2 and on three of our four measures within subjects in Expt 1. Other studies have run both sinusoidal and constant velocity target trajectories, but with little heed to this comparison (Bahill & McDonald, 1983; Buizza & Schmid, 1986; Collewijn & Tamminga, 1984; Rashbass, 1961). The comparison has been secondary to other concerns in those studies, but, as it presumably varies the effort to achieve accurate pursuit, it is of central concern here.

Pursuit of a constant velocity target, with its abrupt stops and starts at the extremes of its trajectory, objec- tively is more difficult to follow accurately and subjec- tively seems more difficult. These are two separable considerations: a small catch-up saccade is made by most subjects at the start of each constant velocity target excursion, independent of task condition, reflecting the intrinsic difficulty of this trajectory. The task-dependent anticipatory and overshooting saccades, more prevalent in constant than in sinusoidal velocity trajectories, seem to be the subjective difficulty component, coming from a compensatory increase in effort to track accurately.

In the constant velocity trajectory, some subjects have

a characteristic strategy of avoiding the initial catch-up saccades by rounding the corners of the response waveform with gradual decelerations and accelerations. This may be more common in trajectories that have no pause between sweeps across the screen (Boman & Hotson, 1992).

Dual-mode pursuit and pursuit gain

Catch-up saccades are corrective, while anticipatory saccades are not. This, and seeing task differences with anticipatory but not with catch-up saccades, suggests that the corrective catch-up saccades are the saccadic component of dual-mode control in pursuit (Stark, 1971, 1983), while the anticipatory saccades are a product of some task-specific attentional component. This was the obvious intent of the dual-mode notion, that saccades are produced as necessary to correct for less than perfect pursuit gain, so it is nice to see catch-up and anticipatory saccades differentiated by these particular task effects. The present data suggest, then, that the dual-mode notion is still viable, but restricted to data containing only catch-up (and undershooting) saccades and smooth pursuit.

These considerations highlight a reciprocal relationship between catch-up and anticipatory saccades, that catch-up saccades are caused by low pursuit gain, while anticipatory saccades cause low pursuit gain. The improvement in smooth pursuit gain seen with target analysis thus suggests a cautionary note in the interpret- ation of pursuit gain results: slowed pursuit after anticipatory saccades will artificially reduce it.

Predictive pursuit

In dual-mode pursuit as classically conceived with catch-up but no anticipatory or overshooting saccades, the corrective catch-up saccades would land on target, followed by pursuit that remains on target and matches instantaneous target velocity. Thus the enhanced gain seen in pursuit of a predictable trajectory has been thought to come from some acquired match of this velocity (Becker & Fuchs, 1985; Mitrani & Dimitrov, 1978). This acquired velocity is seen by interrupting the target and noting the gradual decay of the response velocity. In the extreme, brief pulses of moving target are shown to illustrate both acquisition and decay of this velocity (Barnes & Asselman, 1992). When the pulses are placed at trajectory midpoints, the required trajectory reversals without an explicit target have been taken to indicate an estimate of periodicity of the trajectory (Barnes & Asselman, 1991; Barnes, Donnelly, & Eason, 1987). When a continuous trajectory is interrupted, the response continues through what would have been the reversal of its trajectory, tracing the learned model of the trajectory (Eckmiller & Mackeben, 1978; Hughes & Fendrich, 1992; Whittaker & Eaholtz, 1982).

So the predictive acquisition of both the target velocity and its trajectory are seen in the subject's attempt to continue the trajectory of an interrupted target. In the present study anticipatory saccades provide a naturalis- tic experiment to show the subject continuing pursuit but


after moving his eyes off target. With a choice between parallel pursuit to continue the velocity match or fixation to allow the target to catch up to the point of gaze, the mechanism of velocity computat ion often does neither. It determines a velocity consistent with overall periodicity and amplitude of the trajectory, but irrele- vant to instantaneous target velocity or position. Fur- ther, the saccade-to-pursuit transition seems as abrupt as for catch-up saccades, suggesting that this new pursuit velocity is precomputed in the same manner as for catch-up saccades that acquire the target. Off-target pursuit after anticipatory saccades thus provides evidence that instantaneous velocity match of the target and overall match of trajectory parameters are separable components of predictive pursuit.

Generali ty ~ f results

Clearly, the pervasiveness of anticipatory and overshooting saccades in our data should not be expected in all pursuit data. First, the performance improvement in sinusoidal over constant velocity trajectories suggests to us that perceived difficulty in a pursuit task may increase the incidence of anticipatory and overshooting saccades. Very few anticipatory saccades were reported among normal controls in one study using the same sinusoidal trajectory as ours (Clementz et al., 1990). Second, our 30deg/sec target trajectory has a higher velocity than most. It is presumably more difficult to follow, which may produce anticipatory saccades. We find that as target velocity is increased, good subjects tend to have a velocity threshold where anticipatory saccades are first seen, increasing thereafter, while poor subjects tend to show anticipatory saccades independent of target velocity (Van Gelder, Lebedev, Liu, & Tsui, 1993). A recent report dealing explicitly with saccade types used a 5 deg/sec target velocity and did not report anticipatory saccades (Friedman, Jesberger, & Meltzer, 1991), although another with the same velocity did (Kaufman & Abel, 1986). There also seemed to be a greater tendency for fixations than slowed pursuit after anticipatory saccades in that study, presumably owing to the 6 sec duration of each target sweep. Third, a periodic trajectory, which differentiated those two studies, may encourage anticipatory saccades. Thus anticipatory saccades themselves, and not just the smooth pursuit segment immediately following them, may be associated with predictivity in pursuit. Finally, anticipatory saccades diminish with practice, consistent with impli- cations of effortful pursuit, so experienced subjects should not be a good source of anticipatory saccades.

CONCLUSIONS

Our finding that pursuit improves with the mild distraction imposed by a concurrent listening task suggests that the further improvement in pursuit during target analysis comes from further diversion of explicit task focus from the oculomotor act, diverting attentionai effort from the literally misguided at tempt to make accurate smooth pursuit. In this new light, target analy-

sis is not an attention enhancement condition but an attention diversion condition.

This notion implies that the anticipatory and overshooting saccades seen with our trajectories in the original tracking condition are produced in the effortful at tempt to produce smooth pursuit, since target analysis removes anticipatory and overshooting saccades from the response waveform. They are produced erroneously and generally without awareness in the artificial effort to produce purposive smooth pursuit. Lacking brain mechanisms for producing purposive smooth pursuit, some subjects involuntarily recruit mechanisms more typically employed to produce purposive saccades, to produce here intrusive anticipatory and overshooting saccades.

With respect to pursuit velocity after anticipatory saccades, it seems from the present data that some internally maintained model of amplitude and periodicity of the overall target trajectory provides a predictive component of pursuit velocity, separable from the acquired velocity match. Real-world trajectories are in general different from this and from each other, so the set of trajectory attributes must surely be larger than this. In pursuit outside of the laboratory, without the anticipatory and overshooting saccades generated in the at tempt to produce purposive pursuit, accurate target foveation and velocity matching could follow from this internal model.

REFERENCES

Abel, L. A., Levin, S. & Holzman, P S. (1992). Abnormalities of smooth pursuit and saccadic control in schizophrenia and affective disorders. Vision Research, 32, 1009- 1014.

Acker, W. Q. & Toone, B. (1978). Attention, eye tracking and schizophrenia. British Journal of Social Psychology, 17, 173- 181.

Bahill, A. T. & McDonald, J. D. (1983). Smooth pursuit eye movements in response to predictable target motions. Vision Research, 23, 1573-1583.

Barnes, G. R. & Asselman, P. T. (1991). The mechanism of prediction in human smooth pursuit eye movements. Journal ~[" Physiology, 439, 439-461.

Barnes, G. R. & Asselman, P. T. (1992). Pursuit of intermittently illuminated moving targets in the human. Journal of Physiology, 445, 617 637.

Barnes, G. R., Donnelly, S. F, & Eason, R. D. (1987). Predictive velocity estimation in the pursuit reflex response to pseudo-random and step displacement stimuli in man. Journal of Physiology, 389, 111 136.

Becker, W. & Fuchs, A. F. (1985). Prediction in the oculomotor system: Smooth pursuit during transient disappearance of a visual target. Experimental Brain Research, 57, 562 575.

Boman, D. K. & Hotson, J. R. (1992). Predictive smooth pursuit eye movements near abrupt changes in motion direction. Vision Research, 32, 675-689.

Brezinova, V. & Kendell, R. E. (1977). Smooth pursuit eye movements of schizophrenics and normal people under stress. British Journal of Psychiatry, 130, 59 63.

Buizza, A. & Schmid, R. (1986). Velocity characteristics of smooth pursuit eye movements to different patterns of target motion. Experimental Brain Research, 63, 395 401.

Cegalis, J. A. & Sweeney, J. A. (1981). The effect of attention on smooth pursuit eye movements of schizophrenics. Journal ~[" Psychiatric Research, 16, 145-161.

Cegalis, J. A., Hafez, H. & Wong, P. S. (1983). What is deviant about deviant smooth pursuit eye movements in schizophrenia. Psychiato' Research, lO, 47 58.


Clementz, B. A., Sweeney, J. A., Hirt, M, & Haas, G. (1990). Pursuit gain and saccadic intrusions in first-degree relatives of probands with schizophrenia. Journal of Abnormal Psychology, 99, 327-335.

Collewijn, H. & Tamminga, E. P. (1984). Human smooth and saccadic eye movements during voluntary pursuit of different target motions on different backgrounds. Journal of Physiology, 351, 217 250.

Eckmiller, R. & Mackeben, M. (1978). Pursuit eye movements and their neural control in the monkey. Pflugers Archiv, 377, 15 23.

Friedman, L., Jesberger, J. A. & Meltzer, H. Y. (1991). A model of smooth pursuit performance illustrates the relationship between gain, catch-up saccade rate, and catch-up saccade amplitude in normal controls and patients with schizophrenia. Biological Psychiatry, 30, 537-556.

Holzman, P. S., Levy, D. L. & Proctor, L. R. (1976). Smooth pursuit eye movements, attention and schizophrenia. Archives of General Psychiatry, 33, 1415-1420.

Hughes, H. C. & Fendrich, R. (1992). Reversals of smooth pursuit direction in the absence of a visible stimulus. Investigative Ophthal- mology & Visual Science, 33, 1359.

Iacono, W. G. & Lykken, D. T. (1979). Electro-oculographic recording and scoring of smooth pursuit and saccadic eye tracking: A paramet- ric study using monozygotic twins. Psychophysiology, 16, 94-107.

Iacono, W. G., Tuason, V. B. & Johnson, R. A. (1981). Dissociation of smooth-pursuit and saccadic eye tracking in remitted schizophrenics: An ocular reaction time task that schizophrenics perform well. Archives of General Psychiatry, 38, 991-996.

lacono, W. G., Peloquin, L. J., Lumry, A. E., Valentine, R. H. & Tuason, V. B. (1982). Eye tracking in patients with unipolar and bipolar affective disorders in remission. Journal of Abnormal Psychology, 91, 35-44.

Kaufman, S. R. & Abel, L. A. (1986). The effects of distraction on smooth pursuit in normal subjects. Acta Otolaryngologica (Stock- holm), 102, 57 64.

Levin, S., Lipton, R. B. & Holzman, P. S. (1981). Pursuit eye movements in psychopathology: Effects of target characteristics. Biological Psychiatry, 16, 255-267.

Lipton, R. B., Frost, L. A. & Holzman P. S (1980a). Smooth pursuit eye movements, schizophrenia, and distraction. Perceptual Motor Skills, 50, 159-167.

Lipton, R. B., Levin, S. & Holzrnan, P. S. (1980b). Horizontal and vertical pursuit eye movements, the oculocephalic reflex, and the functional psychoses. Psychiatry Research, 3, 193 203.

Malmstrom, F. V., Reed, L. E. & Randle, R. J. (1983). Restriction of pursuit eye movement range during a concurrent auditory task. Journal of Applied Psychology, 68, 565-571.

Mitrani, L. & Dimitrov, G. (1978). Pursuit eye movements of a disappearing moving target. Vision Research, 18, 537-539.

Pass, H. L., Salzman, L. F., Klorman, R., Kaskey, G. B. & Klein, R. H. (1978). The effect of distraction on acute schizophrenics' visual tracking. Biological Psychiatry, 13, 587-593.

Puckett, J. W. & Steinman, R. M. (1969). Tracking eye movements with and without saccadic correction. Vision Research, 9, 695 703.

Rashbass, C. (1961). The relationship between saccadic and smooth tracking eye movements. Journal of Physiology, 159, 326 338.

Shagass, C., Roemer, R. & Amadeo, M. (1976). Eye tracking performance and engagement of attention. Archh~es q/'General Psychiatry, 33, 121 125.

Spohn, H. E., Coyne, L. & Spray, J. (1988). The effect of neuroleptics and tardive dyskinesia on smooth-pursuit eye movement in chronic schizophrenics. Archives of General Psychiatry, 45, 833 -840.

Stark, L. (1971). The control system for versional eye movements. In Bach-y-Rita, P. & Collins, C. C. (Eds), The control qfeye movements (pp. 363 428). New York: Academic Press.

Stark, L. (1983). Abnormal patterns of eye movements in schizophrenia. Schizophrenia Bulletin, 9, 55 72.

Tsui, W. H. & Van Gelder, P. (1979). A general-purpose data extraction language. Behavior Research Methods & Instrumentation, 11, 199 204.

Van Gelder, P., Lebedev, S. & Tsui, W. H. (1990a). The role of large saccades in smooth pursuit. Investigative Ophthalmology & Visual Science, 31, 602.

Van Gelder, P., Lebedev, S. & Tsui, W. H. (1992). Saccades determine smooth pursuit gain. Investigative Ophthalmology & Visual Science, 33, 1153.

Van Gelder, P., Todd, J. & Tsui, W. H. (1979). Data acquisition software for high data rate experiments. Behavior Research Method~ & Instrumentation, 11, 192-198.

Van Gelder, P., Lebedev, S., Liu, P. M. & Tsui, W. H. (1993). Effects of pursuit velocity, frequency and pause time on anticipatory and catch-up saccades. Investigative Ophthalmology & Visual Science, 34, 1500.

Van Gelder, P., Anderson, S., Herman E., Lebedev, S. & Tsui, W. H. (1990b). Saccades in pursuit eye tracking reflect motor attention processes. Comprehensive Psychiatry, 31, 253-260.

Whittaker, S. G. & Eaholtz, G. (1982). Learning patterns of eye motion for foveal pursuit. Investigative Ophthalmology & I/isual Science, 23, 393 397.

Acknowledgement Peter M. Liu was supported by MH35976, A. J. Friedhofl, Principal Investigator.

Anticipatory Saccades in Smooth Pursuit: Task Effects and ...a concurrent listening condition instead, to see if this mild distraction would degrade performance. Performance improved

Documents