J.Phy8iol. (1967), 193, pp. 161-171 161 With 7 text-ftgure8 Printed in Great Britain PERIODIC EYE TRACKING IN THE MONKEY BY A. F. FUCHS From the Department of Medicine and Biomedical Engineering The Johns Hopkins University, Baltimore, Maryland, U.S.A. (Received 19 May 1967) SUUMMARY 1. Eye movements were measured in monkeys trained for visual tracking. 2. In response to periodic square wave target movements, monkeys do not show a significant reduction in the latency of saccadic movements. 3. Under similar conditions, human beings subconsciously reduce their latency and after several cycles are in step with the target. 4. In response to sinusoidal targets, monkeys show a latency or phase lag which increases monotonically with frequency starting at 0*3 c/s. Human beings can remain in phase with the target at frequencies up to 10 c/s. 5. Hence, monkeys do not exhibit the human predictive tracking response. INTRODUCTION In response to a target moving in a random or unpredictable manner, the human eye executes a combination of saccadic and smooth pursuit tracking movements. The delay before the saccadic response to a change in target position is about 250 msec; the delay before the smooth pursuit response to a change in target velocity is 130 msec. Because of these reaction times, the eye never succeeds in overtaking and accurately tracking a randomly moving target. If, however, the target motion is predictable, the amount of time by which the eye lags behind the target gradually decreases as the tracking proceeds until the eye actually overtakes the target. Thereafter, the eye continues to move accurately with the target with little or no time lag. This improvement in the eye movement response to periodic target trajectories is called predictive tracking. As early as 1930, Dodge, Travis & Fox observed the response of the eye to the periodic oscillations of a pendulum bob. At moderate frequencies (0.7 c/s) although the initial response was a rough, delayed inadequate pursuit of the bob, the eye gradually 'showed a much closer approximation of adequate pursuit with finer and relatively infrequent corrections'. II Physiol. I93
162 A. F. FUCHS More recent investigators have shown that the adaptation from random to predictive tracking occurs before a sinusoidal target trajectory has gone through one complete cycle (Stark, Vossius & Young, 1962; Dallos & Jones, 1963). After the adaptation is complete, Drischel (1958), Sunderhauf (1960) and Young (1962) demonstrated that for target frequencies between 0 3 and 1 c/s the eye may actually anticipate the target motion and lead it by a small amount (less than 1 of a cycle). The adaptation phase in response to a periodic square wave is considerably longer, requiring as much as 5 complete target cycles. As with the sine wave motion, the eye movement often precedes the change in target position and may lead it by as much as msec for target frequencies around 0 5 c/s. Monkeys execute both saccadic and smooth pursuit eye movements which are qualitatively similar to those of man (Fuchs, 1967). In response to random target trajectories, they also lag between 160 msec (the average smooth pursuit latency) and 240 msec (the saccadic reaction time) behind the target. When presented with a periodic target, however, as this study shows, monkeys do not exhibit the predictive tracking response of human subjects. When following a periodic square wave, the monkey never displays a gradual reduction in latency but continues to lag about one saccadic reaction time behind the target. A sinusoidal target trajectory elicits an eye movement which also lags behind the target by an amount which increases with the frequency of the stimulus. METHODS The animal experiments were performed on two monkeys (7-10 lb (3x18-4.54 kg) Macaca specio8a) trained for visual tracking. Eye movements were measured by the technique of a chronically implanted search coil in a magnetic field (Fuchs & Robinson, 1966). Three turns of fine wire are laid under the insertions of the four rectus muscles beneath the bulbar conjunctiva to form a coil which moves with the eye. When the monkey is exposed to two alternating magnetic fields in spatial and phase quadrature, a voltage is generated in the coil which is a function of the horizontal and vertical components of eye position. The eye movement transducer has a sensitivity of 1-5 min of arc and a linear range of + 200. Targets in the form of small circular spots subtending less than I' at the animal's fovea are projected on to a screen in front of the monkey. A mirror mounted vertically on a galvanometer movement is inserted into the optical path between the projector and the screen. A voltage signal to the galvanometer will then cause the light spot to move horizontally across the screen. Signals proportional to the horizontal and vertical components of eye position are compared to voltages proportional to the two components of target position. If the monkey follows within, say, + 1 of the target while at the same time pulling with his teeth on a bite bar to fi-x his head voluntarily, he is rewarded by a squirt of orange juice through the hollow centre of the bite bar. The training procedure to condition a monkey to visual tracking requires about 21 months and is described in detail elsewhere (Fuchs, 1966). Human eye movements were also measured in two naive human subjects by the electromagnetic technique described above. Neither subject was familar with the human phenomenon of predictive eye tracking, and each was given no other instructions but to follow the light spot as accurately as possible. The search coil which in monkeys can be permanently
PERIODIC EYE TRACKING 163 fastened to the eye is for human subjects embedded in a scleral suction contact lens which is held on the topically anaesthetized eye by 40 mm Hg negative pressure (Robinson, 1963). The amplitude of all the periodic stimuli presented to either man or monkey was ± 50 about the primary direction of gaze. RESULTS Human periodic tracking. To confirm the results of previous investigators, the human subjects were asked to track repetitive square waves at various frequencies between 03 and 1-5 c/s selected at random. It was Tara L T Eye ia 0 ~ 0.5C/S ~ ~1 sec 0~~~~~~~~~~~~~~~- C c/ bo 200_ B 300 400 1 0 1 2 3 4 5 6 7 8 9 Target cycle number Fig. 1. (A) A typical human eye response to a 0-8 c/s horizontal target. (B) The time course of adaptation of the saccadic latency to periodic square waves averaged over two naive subjects. Each point represents the mean of 7 runs; the brackets represent standard errors of the mean. The dashed curve is the average of 3 runs on one naive subject at 0 5 c/s under experimental conditions designed to simulate the monkey experiments. found that the latency to a change in target position decreases for all frequencies as the tracking run progresses. However, the rate of adaptation is greatest for frequencies between 05 and 0O8 c/s. Figure LA shows the ability of the eye to improve its lagging response when following a 0-8 c/s square wave. If at each target step the latencies to the resulting saccade are averaged over the two subjects, the time course of adaptation in Fig. 1B is obtained. After only 2-5 cycles of target motion corresponding to 5 I1-2
164 A. F. FUCHS saccadic responses, the humans have reduced a 300 msec latency by 60 % in response to 0 5 c/s square waves and by almost 90 % for a 0-8 c/s stimulus. At 0-5 c/s, the rate of adaptation decreases significantly, and the eye waits until the 6th target cycle before getting in step with the target movement. However, at 0-8 c/s the saccadic latency continues to decrease with each successive target jump so that the eye begins to lead or anticipate the target by the 7th saccade (or after 3*5 target cycles). At frequencies of 1.0 c/s, the adaptation rate resembles that at 0 5 c/s, indicating that, in man, the optimal frequency to elicit the most rapid adaptation is around 0-8 c/s. These results show good agreement with those of Stark et al. (1962) and Dallos & Jones (1963) on human subjects. The large standard errors of the mean attest to the lability of the adaptation process. A similar set of target stimuli were presented to the monkeys, but it soon became apparent that the animals were not exhibiting a predictive response. In order to give the monkeys the best possible opportunity to show adaptation, the same periodic stimulus was left on continuously for 10 min at a time, and the monkey initiated his own tracking run by pulling at the bite bar and following the target. To determine the human performance under a similar experimental situation, a naive subject was instructed to follow a continuously available target as accurately as possible. Between tracking runs the monkey often looked away from the oscillating spot to direct his gaze at other points in his visual field. To simulate this condition, the human subject was required to follow a flashlight beam around the edge of the screen between tracking runs. Under these circumstances the human rate of adaptation is significantly increased and is illustrated by the dashed curve of Fig. 1 B. In response to a train of 0 5 c/s square waves interposed in a random assortment of periodic targets human subjects do not lead the target until the 12th saccade (6th target cycle), whereas in response to a continuously present target, which he has previously treated, man can exhibit anticipation by the second saccade. Therefore, if monkeys do exhibit the phenomenon of predictive tracking in the manner of human beings then a continuously present target stimulus oscillating at 0-8 c/s should elicit it. Monkey periodic tracking. The curves of Fig. 2 are typical attempts by a monkey to track square waves at frequencies between 0 3 and 1-0 c/s. If a monkey is required to track longer than between 3 and 5 sec without a reward, he usually becomes impatient and gives up the task. Fortunately, one of the two monkeys (Adali), if rewarded every 3 sec, would continue to pull the bite bar and track the target while swallowing the orange juice, so that a number of repetitive rewards would induce him to track for over 12 sec. It is the longest of these responses which are plotted here.
L ~~~~~~~~~~~~~~~~~~Target PERIODIC EYE TRACKING 165 Even after as many as 11 cycles of the target, the monkey eye is still lagging a 0-8 c/s square wave by about one reaction time. The average latency for the 22 saccades is 258 msec, which compares to the typical saccadic reaction time of 241 msec to random targets (Fuchs, 1967). At T %~, 0-3 c/s 1 sec Eye T 0 5 c/s 1 sec Eye T - - _- _ '_ - -Target X ;;; ~~~~~~~~~~~~"_0Eye 0-8 c/sey 1 sec T.- - - ~ Target ~ ~~ ~ ~ ~ ~ Eye 1-0 c/s Fig. 2. Monkey periodic tracking of square waves at frequencies between 0 3 and 1-0 c/s. 10 c/s, the monkey is already beginning to experience difficulty in tracking and sometimes skips an entire cycle. At 0 5 c/s, the monkey shows occasional responses which coincide with or precede the target step (the 6th, 12th, 16th and 18th saccades in this record). However, such responses are isolated events which appear to be chance occurrences not associated with a gradual adaptation phase. At the lowest frequency studied (0.3 c/s), the average latency was 268 msec not including the apparent response after the 3rd saccade which was a startle reaction to a squirt of orange juice. Hence, adaptation also does not occur at frequencies either higher or lower than 0-8 c/s. Figures 3 and 4 show the saccadic latencies averaged individually over
166 A.F.FUCHS each monkey at frequencies of 0.5 and 0-8 c/s. At both frequencies, the latencies of the monkey Connie are relatively constant at values between 200 and 300 msec for at least 4 cycles of target movement. At 0-8 c/s, monkey Adali displayed an essentially constant latency for 8 target cycles. In response to a 0.5 c/s square wave, his saccadic latency remained between 170 and 250 msec for 6 target cycles. Thereafter, this plot is continued as the mean of three especially long tracking runs. Although the latency 10 cj 0 t~~~~~~~~~~~~. t c0cs 200 K0-5 c/s 0 / 300 400 I l l I l I 0 1 2 3 4 5 6 7 Target cycle number Fig. 3. The time course of adaptation of the saccadic latency to periodic square waves by the monkey Connie. Each point is the mean of at least ten values with the brackets representing standard errors of the mean. Points with no brackets are the last few latencies of one particularly long run. 0 S -I 0 ( 200 0-8 c/s 300 400 1 I 1 I I I I 0 1 2 3 4 5 6 7 8 9 10 Target cycle number Fig. 4. The time course of adaptation of the saccadic latency to periodic square waves by the monkey Adali. Each point is the mean of at least six and as many as nineteen samples. Brackets represent standard errors of the mean. Points with no brackets are the means of three especially long runs.
PERIODIC EYE TRACKING 167 is as small as 30 msec at 7-5 target cycles, the unplotted standard error of the mean is large (+ 230 msec). Furthermore, since at the 8th cycle the latency again returns to 190 msec it is impossible to conclude that any sustained adaptation has taken place. To further investigate the monkey's ability to track periodic targets, one of the animals was assigned the task of following a sinusoidal target stimulus. Most investigators of the human sinusoidal response agree that the duration of the adaptation phase is very short with the eye reaching its steady state within 1 cycle of the onset of target motion. In the steady state, human tracking is predominantely smooth with only occasional corrective saccades at frequencies up to about 0-8 c/s. Low frequency simian responses, however, are more liberally punctuated with saccades which occur in pursuit of the high velocity segments of the sine wave as may be seen in Fig. 5. Once again the target trajectory is constantly on display before the monkey for 10 min intervals during which time he can initiate his own tracking run. At the higher frequencies where human tracking becomes increasingly more saccadic in nature, the monkey exhibits smoother trajectories. Occasionally the monkey will demonstrate his ability to execute faster smooth movements than man (Fuchs, 1967) by responding exactly, except for a delay or phase shift, to i cycle of a high frequency trajectory. The earliest part of the response at 1-5 c/s shows the eye turning at a maximum velocity in excess of 45 deg/sec. Data from the sinusoidal target experiments are summarized in the gain and phase plots of Fig. 6. Since similar human plots are the result of a predominantly smooth response to sinusoidal targets, all the saccades were removed from the monkey's response and only the remaining smooth component was considered. The amplitude of the eye movement was estimated as the average of the magnitudes between the successive peaks of at least 11 'de-saccaded' responses at each frequency; the ratio of the eye movement amplitude to the target movement amplitude is the gain, G, which is expressed in decibels (db) as Gdb = 20 log G. The phase angle is the average of the differences in time, expressed in degrees, between the peaks of the eye and target trajectories. Had the records not been 'de-saccaded', the gain plot would be somewhat higher, but the phase plot would be unchanged. The simian gain characteristic falls monotonically from 0-3 c/s and is 3 db down at 1-25 c/s. Between 1 1 and 2-1 c/s, the gain decreases at an average rate of 4 db/octave so that at 2 1 c/s the gain is -6 db. Xt higher frequencies, the gain characteristic can be approximated by a linear asymptote with a slope of -24 db/octave indicating that the plot is characteristic of a fourth-order linear system. Except at two unrelated frequencies, the simian phase lag shows a monotonic increase with
168 A. F. FUCHS frequency. The monkey never gets in phase with the target even at low frequencies and still lags a 03 c/s sine wave by 10. Once again, the rather large standard errors of the mean calculated from a large number of samples demonstrate the variability of the monkey's periodic response. T I T I T I b-i sec-l T < Target NJc/ 1-1 sec-t Eye Target Fig. 5. 2-5 c/s Selected smoothest eye responses of the monkey Connie to sinusoidal periodic targets. DISCUSSION A comparison of Fig. 1 B with Figs. 3 and 4 shows that in response to the same periodic square wave monkeys do not exhibit the predictive tracking response of man. At both 0-5 and 0-8 c/s, human subjects show a gradual adaptation of their inherently lagging response and eventually overtake
PERIODIC EYE TRACKING 169 and accurately track the target, whereas both monkeys continue to follow approximately one saccadic reaction time behind the target. The chief objection to the conclusion that monkeys cannot predict is that perhaps the animals were not as motivated as their human counterparts to perform well. Both human subjects were naive as to the reason for Frequency (c/s) Frequency (c/s) 0-2 04 0-6 0-8 1-0 20 3 0 0-2 04 0-6 0-81-0 2-0 3*0 0 1 1 1111 Leadt I 1 11-4 -2 ~~~~~~~~~~0 Lag, 10 20 - ~~~~ -8~ ~ ~~~~4 - -10-12 P -14 III 30 50 60 70 1111 Fig. 6. Gain and phase characteristics of the monkey Connie. Each point is the average of at least forty values obtained as described in the text. Brackets represent standard errors of the mean. the experiment and were only instructed to follow the target as accurately as possible. At the conclusion of the experiment, neither subject was aware that he had adapted his eye tracking response. Hence, predictive tracking in naive humans is a subconscious phenomenon which should, therefore, also be elicitable in the thirsty monkey who is presumably tracking as well as possible to earn a liquid reward. Possibly the subconscious adaptation phenomenon is called forth in humans who are trying to give a good performance for a human experimenter, whereas a similar desire to perform well does not exist in the monkeys. At any rate, the question of unequal motivation of the two primates in the experimental situation described is unanswerable. No attempt was made to try to train the monkeys to predict by rewarding them for shorter latencies. The monkey was not penalized for his natural saccadic reaction time but was rewarded if he reached the correct position within 300 msec of the target change. The monkey's inability to predict can also be seen in his response to a sinusoidal target trajectory. Figure 7 shows a comparison of human and
170 A. F. FUCHS simian gain and phase characteristics for the same magnitude of target stimulus (viz., ± 50 peak-peak excursion). The human curve is a composite of the data of Young (1962) and Suinderhauf (1960). The bandwidth (-3 db) of the periodic tracking system is similar in man (140 c/s) and Frequency (c/s) Frequency (c/s) 0.1 0-2 0.5 1-0 2-0 50 0.1 02 05 1.0 2-0 50 6nLead 0an 0~ ~ ~ ~ + Lag 10-10 20-20 Fig. 7. Monkey _ 30 ~40 C" 50 Man ~~~60 Man -30 70 ~~~~~~~~~~Monkey Comparison of monkey and human gain and phase characteristics. monkey (1.25 c/s). Between 1-0 and 2-0 c/s, the human gain characteristic falls more rapidly than the monkey's owing to the fact that monkeys are able to turn their eyes smoothly at higher velocities than human beings. The result is that the monkey's response is not 6 db down until 2 1 c/s, whereas the human's is -6 db at 1-6 c/s. Although the monkey can follow the target up to 2 1 c/s with only a 2 to 1 diminution in amplitude, his eye always lags behind the target, as can be seen by the phase characteristic. Whereas the human phase characteristic exhibits a definite phase lead or anticipation of target motion between 0x1 and 1x2 c/s, the monkey lags more and more behind the target as the frequency increases, indicating his inability to predict yet another periodic stimulus. Finally, there is a piece of circumstantial evidence that monkeys are unable to predict. It is well know-n that the intrinsic visual feedback of the oculomotor control system can be altered by controlling the target position with a signal proportional to eye position. By doubling the amount of negative feedback, it should theoretically be possible to cause the system to oscillate. Robinson (1965), however, had to increase the amount of feedback by a factor of 5 to sustain oscillations, since after a few cycles of vain pursuit at less feedback his human subjects would simply stop tracking to wait for the apparently periodic target to return. He attributed this adaptation to the human ability to predict his own saccadic oscillations.
PERIODIC EYE TRACKING 171 Under similar conditions, simply doubling the amount of negative visual feedback to a monkey will cause him to break into machine-like square wave oscillations which will continue without pause until the animal tires (Fuchs, 1967). Hence, the animal shows no inclination to adapt in the manner of human subjects to the artificial conditions of increased negative feedback. This investigation was supported by a Special Fellowship of the National Institutes of Health (HSP-17,237). Most of the equipment and laboratory facilities were made available by Public Health Service Research Grant AM-05524 from the National Institute of Arthritis and Metabolic Diseases. The author is indebted to Dr David A. Robinson for his advice and encouragement and to Mr William J. Sullivan and Mr Samuel Ron for their participation in the tracking experiments. REFERENCES DAITOS, P. & JONES, R. (1963). Learning behaviour of the eye fixation control system. IEEE Tranm. autom. Control. AC-8, 218-227. DODGE, R., TRAvIs, R. & Fox, J. (1930). Optic nystagmus III. Characteristics of the slow phase. Arch8. Neurol. Psychiat., Chicago 24, 21-34. DRISCHEL, H. (1958). t-ber den Frequenzgang der horizontalen Folgebewegungen des menschlichen Auges. Pfluger8 Arch. ges. Physiol. 268, 34. FUcHS, A. (1966). A quantitative description of voluntary eye movements in the monkey. Ph.D. Thesis, Committee on Biomedical Engr. The Johns Hopkins University, Baltimore Md. FUCHS, A. F. (1967). Saccadic and smooth pursuit eye movements in the monkey. J. Phy8iol. 191, 609-631. FUCHS, A. & ROBINSON, D. (1966). A method for measuring horizontal and vertical eye movement chronically in the monkey. J. appl. Physiol. 21, 1068-1070. ROBINSON, D. (1963). A method of measuring eye movement using a scleral search coil in a magnetic field. IEEE Tran8. bio-med. Electron. BME-10, 137-145. ROBINSON, D. (1965). The mechanics of human smooth pursuit eye movement. J. Physiol. 180, 569-591. STARK, L., Vossius, G. & YOUNG, L. (1962). Predictive control of eye tracking movements. I.R.E. Tran8. hum. Factors Electron. HFE-3, 52-57. SPJNDERHAUF, A. (1960). Untersuchen uber die Regelung der Augenbewegungen. Klin. Mbl Augenheilk. 136, 837-852. YOU-NG, L. (1962). A sampled data model for eye tracking movements. Sc.D. Thesis. Dept. Aeronautics and Astronautics, Mass. Inst. Tech. Cambridge, Mass.