Deubel (1995) Separate adaptive mechanisms for the ... - CiteSeerX

follow-up saccade the target was systematically displaced by 25% of the initial ... stepping target did not transfer to the saccades in the scanning situation, nor to ...
1MB taille 61 téléchargements 294 vues
~

Vision Res. Vol. 35, No. 23/24, pp. 3529-3540, 1995

Pergamon

0042-6989(95)00058-5

Copyright © 1995 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0042-6989/95 $9.50 + 0.00

Separate Adaptive Mechanisms for the Control of Reactive and Volitional Saccadic Eye Movements HEINER DEUBEL* Received 7 December 1994," in revised form 24 February 1995

Adaptive reduction of the gain of the saccadic system was induced by means of two basically different paradigms. In the first approach the subjects had to follow a step-wise moving target. During each follow-up saccade the target was systematically displaced by 25% of the initial step, into the opposite direction of the saccade. In the second approach the subjects scanned a display of six small items. During each scanning saccade the whole display was displaced by 25% into the opposite direction of the saccade. Both conditions lead to fast and consistent saccadic gain reductions. However, adaptation with the stepping target did not transfer to the saccades in the scanning situation, nor to delayed saccades in an overlap paradigm, nor to memory-guided saccades. Conversely, when saccades were adapted in the scanning situation, induced gain changes transferred to overlap and memory-guided saccades, but not to saccades following steps of a single target. The results suggest that two separate and largely independent mechanisms are involved in the generation of reactive, stimulus-triggered and volitional, internally generated saccades, respectively. Both types of responses can be selectively adapted. Eye movement Oculomotor

Saccade Selectiveadaptation

INTRODUCTION Except for eye movements of extremely large amplitudes ( > 50 deg), saccades can be safely described as being "ballistic" movements in the sense that they occur in a pre-programmed, open-loop manner. Such openloop systems are more sensitive to fluctuations of its parameters than a system controlled by a continuously operating feedback loop. Therefore, the maintenance of saccadic accuracy over lifetime presumes a continuous monitoring of saccadic performance and, when required, the ability to recalibrate the saccadic response. Evidence for the intriguing ability of the saccadic system to compensate, even for very profound dysmetrias, is provided by two basically different lines of research. In a first principal line of studies, long-term effects of altered properties of the peripheral oculomotor system have been investigated in patients and by means of animal experiments. It has been found, for instance, that patients suffering from partial eye muscle paralysis manage to adjust saccadic gain within a few days to account for the effect of the disease (Kommerell, Olivier & Theopold, 1976; Abel, Schmidt, DeU'Osso & Daroff, 1978). Optican and Robinson (1980) surgically weakened

Plasticity

the medial and lateral recti muscles of one eye in rhesus monkeys. Normal visual experience with the weakened eye then led to a gradual recovery of saccadic accuracy, provided the cerebellum was intact. Interestingly, this adaptive process not only included metrical adjustment of saccade size, but also the gradual compensation of postsaccadic drift eventually induced by muscle weakness. Thus, not only saccade metrics, but also the dynamical properties of the eye movement are under adaptive control. In a second line of investigations, saccadic dysmetria have been "simulated" psychophysically in the laboratory by means of consistent target shifts while the subject tried to acquire the target with a saccade (e.g. McLaughlin, 1967; Henson, 1978; Miller, Anstis & Templeton, 1981; Deubel, Wolf& Hauske, 1986; Deubel, 1987, 1991a,b). It is interesting to note that, provided the intrasaccadic shifts are not too large, the subjects completely fail to perceive them (Bridgeman, Hendry & Stark, 1975). In this "double-step" adaptation paradigm the subject has to track steps of a small target; while the eye follows with a saccade, the target is shifted systematically by a small amount, e.g. into the opposite direction of the saccade [see Fig. I(B)]. Typically, the saccadic system adapts to this situation quickly, by reducing saccadic magnitude to the required value. By means of this type of conditioning, the adaptive mechanism can largely compensate for the induced error

*Max-Planck-lnstitut fiJr Psychologische Forschung, Leopoldstrasse 24, D-80802 Miinchen, Germany [Email [email protected]. 3529

3530

HEINER DEUBEL

(A) 6

deg

................. -;o T ................. 3>

X

T .,

i

T .f- . . . . . . . . . . . . . . . . .

T

":-?. . . . . . . . . . . . . . . .

T

"Scanning"

(B)

i-°°-

.... i

e step adaptation

(c)

I I I ~ I D N I

6f

....

. . . . . . . ° . °

/

4

2

"Single step" test condition

o

(D)

f

6I 4

2

0 (E) 6

~---D !

~. .--

0~! ....

"Overlap" test condition

L

- ....

saccadic corrections should be expected after adaptation sessions when the precise foveation of small targets is required. This is not what we found when we occasionally made a subject scan objects in the normal environment outside the laboratory, measuring eye movements with EOG (Deubel, 1993b). Surprisingly, rough inspection of the eye movement traces showed that the saccadic responses had normal accuracy. This observation seemed to indicate that the induced adaptation effects are limited to the laboratory context in which the conditioning occurred, implying the existence of context-specific mechanisms and possibly the involvement of higher-level strategies. This intriguing finding was further elaborated in Deubel (1995). This study analysed the capability of saccadic gain control to adapt specifically to various context variables such as the orbital starting position of the saccade, the form and colour of the saccade target, the presence or absence of a background structure and the distinction between intentional and stimulus-elicited saccade generation. The results demonstrated that there is no fast position-specific learning, and that visual stimulus features do not trigger the use of specific parameter sets. However, evidence was provided for a selective modification of intentionally generated and stimulus-elicited saccades, suggesting the existence of individual gain adjustment mechanisms for stimulusguided and the internally triggered saccadic responses. The work presented here provides a more detailed experimental analysis of this kind of double dissociation between reactive and volitional saccade modes. Moreover, ! analysed the amount of transfer of selective adaptation to saccades in an overlap paradigm and to memory-guided saccades. Preliminary reports are presented elsewhere (Deubel, 1993b, 1994, 1995). GENERAL METHODS

Subjects 0

" - ' ' ' - "

FIGURE 1. Visual stimuli and experimental paradigms. (A) Visual stimulus in the Scanning condition. (B) Double step adaptation. (C) Single step test condition. (D) Overlap test condition. (E) Memorized target test condition.

within less than 200 training trials. Obviously, the retinal error signal forms the essential source of information for the adaptive process: post-saccadic errors are interpreted by the adaptive mechanism as reflecting internal miscalibrations. Interestingly, it has been observed that when, after adaptive gain reduction, single steps are provided, recalibration to the normal gain value may take longer than conditioning (Deubel et al., 1986; Deubel, 1987). The induced effects are so persistent that they sometimes even show up the day after the adaptation session (personal observations). Since the gain reductions that can be achieved are considerable, large and frequent

In total, eight subjects aged 21-32yr participated in this study, and between three and six subjects in the individual experiments. All had normal vision and were experienced in a variety of experiments related to oculomotor research. However, all subjects were naive with respect to the aim of the study.

Experimental set-up The subject was seated in a dimly illuminated room. The visual stimuli were presented on a fast 21 in. colour monitor (CONRAC 7550 C21) providing a frame frequency of 100 Hz at a spatial resolution of 1024 × 768 pixels. Active screen size was 40 x 30 cm; the viewing distance was 80 cm. The video signals were generated by a freely programmable graphics board (Kontron KONTRAST 8000), controlled by a PC via the TIGA (Texas Instruments Graphics Adapter) interface. The stimuli appeared on a grey background which was adjusted at a mean luminance of 2.2 cd/m 2. The luminance of the stimuli was 25 cd/m 2. The relatively high

SELECTIVE ADAPTATION OF REACTIVE AND VOLITIONAL SACCADES

background brightness is essential for avoiding the effects of phosphor persistence. In a physical measurement of luminance decay by means of a linear PIN diode we verified that the target luminance decayed to subthreshold values within less than 20 msec (Wolf & Deubel, 1993). Eye movements were recorded with a SRI Generation 5.5 dual-Purkinje-image eyetracker (Crane & Steele, 1985) and sampled at 400 Hz. The device projects a focused infrared light source into the eye, and tracks both the first Purkinje image (the reflection from the front surface of the cornea) and the fourth Purkinje image (the reflection from the back surface of the lens). As the eye rotates, the first Purkinje image moves into the same direction as the eye, while the fourth image, from the concave surface of the back of the lens, moves into the direction opposite the eye (relative to the optical axis). Thus, coincident movement of both images indicates head motion, while the difference between the two image motions indicates eye rotation. Special-purpose servomechanics allow a frequency response better than 250 Hz and a noise level equivalent to about 20 sec arc r.m.s. (Crane & Steele, 1985). Unlike earlier eye trackers, the fifth-generation device can follow saccadic movements of 15 deg or more without losing the eye. Head movements were restricted by a biteboard and a forehead rest. The experiment was completely controlled by a 486 PC. The PC also served for the automatic off-line analysis of the eye movement data in which saccadic latencies and saccade start and end positions were determined. By digital on-line differentiation of the sampled eye position signal, the computer derived a trigger signal indicating saccade onset. The saccade trigger was adjusted at high sensitivity: when instantaneous eye velocity exceeded 30 deg/sec, saccade-related sensory events such as target displacements were triggered. Early triggering is important because of a considerable delay in Purkinje-image eyetracker records due to slippage of the lens within the eye (Deubel & Bridgeman, 1995) and a display delay of up to 10 msec because of screen raster delays. The early triggering ensured that the presaccadic stimulus disappeared before the eye reached maximum velocity.

Calibration and data analysis Each session started with a calibration procedure in which the subject had to sequentially fixate 10 positions arranged on a circular array of 6 deg radius. The tracker behaved linearly within 8 deg around the central fixation. Overall accuracy of the eyetracker for static fixation positions was better than 0.1 deg. Dynamically, however, the eyetracker records considerable artifactual overshoots of the eye at the end of each saccade which can be ascribed to the movement of the eye lens relative to the optical axis of the eye (Deubel & Bridgeman, 1995). In order to determine veridical direction of gaze, an off-line program for the evaluation of saccade parameters searched the saccade record for the end of the overshoot and then calculated eye position as a mean over a 40 msec

3531

time window. For further data analysis, the evaluation program.calculated the starting and endpoints of each saccade in the record. In the experiments where the subject had to follow steps of a single target, saccadic gain was calculated as the ratio of saccadic amplitude and target step size. The time-course of gain adaptation was evaluated by fitting the experimental data with single exponential functions of the form: g(trial) = g o - g o x (1 - e x p ( - t r i a l / T ) ) . The variable g denotes the actual saccadic gain as a function of trial number, go is the gain of the pre-adaptive saccades and gc denotes the amount of gain change; the time constant T describes the speed of adaptation, given in trials, and corresponds to the time to asymptote.

Experimental paradigms The experiments were aimed at analysing the amount of transfer of adaptation induced in the traditional "double step" paradigm and in a scanning paradigm, respectively, to saccadic eye movements occurring to target steps, to targets in an overlap paradigm, to memorized targets and in a paradigm requiring the scanning of static items. Accordingly, different types of stimuli and stimulus sequences were used in various combinations. Figure 1 presents a summary of the applied experimental paradigms. Scanning. For the analysis of the metrical properties of saccades in a scanning situation, a display consisting of six items as shown in Fig. I(A) was used. In each trial the subject was asked to first fixate a small cross subtending 0.17 deg of visual angle that was presented on the screen. The cross appeared 3deg above and 3deg to the left, or in other trials, to the right of the screen centre. After a random delay of 800-1500msec, five more items appeared forming a rectangular configuration of 12 x 6 deg. These items had a horizontal separation of 6 deg and consisted either of the complete letter "T" or a version of the letter with 3 pixels missing. Upon appearance of the letters, the subject was given about 4 sec to completely scan the items in order to report, by means of a button-press, whether the number of complete "T"s in the display was 2 or 3, finally returning with the eye to the fixation cross. In cases where the cross appeared to the left of the centre, the subject was instructed to scan clockwise, otherwise anti-clockwise. Since the letters only subtended a visual angle of 0.14 deg, the task required the precise foveation of each of the items [the size of the items is considerably exaggerated in Fig. I(A)]. Only those trials were considered in the off-line data analysis in which the subject had produced the correct and complete sequence of saccades that was required. For these correct scanning trials the mean of the magnitudes of the four horizontal saccades was used for further data analysis. Double step adaptation. For the purpose of adaptation of stimulus-triggered saccades, a version of the "classical" double step paradigm was used. Here, the subjects had to follow a small cross (size 0.17 deg) on a video screen that performed a sequence of two steps, about every 3 sec

3532

HEINER DEUBEL

[Fig. I(B)]. In each such trial the subject initially fixated the target. After a random delay of 600-1200 msec, the target was displaced to the right or to the left, selected at random. The size of the initial displacement was 6 deg, in some of the experiments it was 6 or 8 deg. When the velocity of the primary saccade exceeded 30 deg/sec, the target was displaced by 25% of the initial step size (i.e. 1.5 deg), into the opposite direction of the first step. Due to the exogenous target displacement, the saccade normally overshoots the final target position, and a secondary, corrective saccade is typically elicited after about another 160 msec. The final target position after the double step sequence is then used as the starting position for the next trial. Due to the relatively small size of the intrasaccadic step, the subjects normally reported not to have perceived the second target displacement. Adaptation of scanning saccades. This type of paradigm aimed at the adaptation of saccades that occur during the scanning of the static items. For this purpose, the same stimulus material as in Fig. 1(A) was used. Again, upon appearance of the five items, the subject had to scan the items, clockwise or anti-clockwise, in an orderly manner. With the onset of each of the large horizontal saccades from one letter to the next, however, all six items on the screen were systematically displaced by 1.5 deg (i.e. 25% of the item separation), into the opposite direction of the saccade. As with the intrasaccadic displacements of the single targets, the subjects seldom perceived these stimulus shifts. Single step test condition. Before and after adaptation the gain of saccades elicited by steps of single targets was determined in a condition including a post-saccadic gap as shown in Fig. I(C). For this purpose, the saccade following the initial displacement of the target cross triggered a temporary blanking of the target. The target then reappeared after a gap duration of 400 msec. This type of paradigm was selected for testing actual saccade gain since a number of experiments not described here had demonstrated a considerable slowing of recalibration for conditions where no spatial target information was available immediately after the primary saccades (unpublished observations). Overlap test condition. The overlap test condition was of interest in order to investigate possible effects on saccadic gain of prolonged presence of the peripheral target before the saccade was actually elicited. In this condition, a saccade target appeared 6 deg to the right or the left of the actual fixation, while the central fixation target remained visible. The subject was instructed to wait with his/her saccade to the peripheral target until the offset of the actual fixation cross. As in the above single step test condition, the saccade target disappeared with the onset of the primary saccade and reappeared only 400 msec later. In order to obtain various amounts of overlap, the computer randomly varied the time O [see Fig. 1(D)] between peripheral target onset and central fixation offset in the range 0-400 msec. Memorized target test condition. This condition tested the amount of transfer of adaptation to saccades directed to memorized targets. For this purpose, the target was

presented for only 100 msec at the peripheral position [Fig. I(E)]. The subject was asked to saccade to this position only when the actual fixation cross disappeared. The time between peripheral target presentation and fixation cross offset was randomly varied between 100 and 1100 msec. To avoid rapid recalibration of saccadic gain during testing, the target then reappeared only 400 msec after the primary saccade. EXPERIMENTS

Experiment 1: double step adaptation does not transfer to saccadic scanning The first experiment investigated the amount of transfer of saccadic gain adaptation induced by means of double target steps on scanning saccades. Six subjects participated in this experiment. In an initial test phase, a block of 80 trials with the "Single step" test condition [Fig. I(C)] was flanked by two blocks containing 28 saccades with the "Scanning" paradigm [Fig. I(A)]. Subsequently, the subject experienced a total of 560 "Double step" adaptation trials [Fig. I(B)], interleaved with four blocks (10 trials each) of the Scanning test condition. Figure 2 presents a summary of the experimental results for each of the six subjects. The • in each graph display the amplitudes of the individual saccades directed to the stepping targets, given as a function of subsequent trial number. The O denote the mean magnitude of the four horizontal saccades elicited in the Scanning situation. Note that the conditioning double target steps started with trial No. 109. The solid and dashed curves represent the results of fitting the individual data with an exponential function. Initially, the subjects show a consistent undershooting behaviour of their saccades, with a tendency to produce larger saccades in the Scanning situation (mean amplitude and SEM: 5.65 __+0.08 deg) as compared to the stepping target situation (5.35 + 0.12 deg). With the introduction of the conditioning intrasaccadic target step, the magnitudes of the saccades to the stepping target decrease quickly and consistently in all subjects over time. The average final gain reduction was found to be 0.93 + 0.07, the mean of the time constants T of the adaptive process was 134 trials, its SE was 22.1 trials. Amazingly, however, the magnitudes of the interleaved scanning saccades are completely unaffected by the ongoing adaptation process, remaining approximately constant during the whole session. In consequence, the average amount of gain decrease of these saccades was only 0.10 ___0.04 deg, and final gain was not significantly different from the control condition. An analysis of variance with repeated measures (ANOVA) confirmed a highly significant interaction between the test conditions and the time of testing [F(1,5) = 81.7]. Post-hoc Newman-Keuls tests showed no significant gain change in the Scanning situation (P > 0.05), while the amount of gain decrease in the Double step condition was significant (P < 0.01). The

SELECTIVE A D A P T A T I O N O F REACTIVE A N D V O L I T I O N A L SACCADES 6.5

I

I

I

6.5

I

]

I

1

I

I

6.5

I

, ; ~ .,f * **** *

6.0

6.0

6.0 ~..* •

. . . . .

5.5 L.--a

. . . . . .

".-;~...'..7. " "" ~", ~~,.:' ~~; ~. ~, ' m. ' , . ; : :,-,,.

5.0

"

4.5

•-.." : . U , _ ~ . . , . ~

~d 4.0 o

4.5

"

~

~



3.0

3.0 I

I

I

300

450

600

*,

6.0

'-'~*

5.5

I

I

6.5

I

.

*

"

o

"

6.0

o

5.5

~ . - . " ~..':.~'- - -..~--." tl~:t ",~. t~K~6" * •..'...~,....I. h.;: :"*~. ' .""~-,--~,.-~.'%"

"o

5.0





: ..

:*.7 -'t.,-.

4.5

~'"

4.0





m.

"

~ • •

" *',,',i'4": ...),~

"

. . . .

".~"

,',~

• •

• :,.

3.5

03

3.0

.; . "." '~

0

I

"1

150

300

*

Trial -

I

I

450

600

#

~"



i

~

0-

4.5

., . . . . : .. : "%

... >. ~......:



.

o

=--." .,,..,-..,.....~ " . . . ..... ,_,,, .....%-',,~,Z. °, . . , - . , ~ . • :

o

4.0

.... *..•. •._.* *:.-..~ ~, . ' :

{3

.3.5

":.:.-- . : . . .

..,,. *..... ,',~.,~*~.o



.3.0 I

I

I

i

150

300

450

600

I

I

I

I

150

.300

450

600

2.5 750

t:,:

I

"

I

~

Trial -

.

I

6.5

o|

i





i

i

i

i

6,

',~•**"

55

";%~5.~Y.:.'z:" ~." ,, "~.~:'.:':-:, -1

5.0

4.5

45

4.0

4.0

.3.5

35

m

750

#

.,..% ." .. 6,0 1 _ ~ . . . .

5.0

2.5 750

i

*

--.£.-.~: .....



.•.

~



0

4

~'--~i~.." ."."-~"-Q--~.-.r.-.J~--! ,~,-_..-:X........."o,,. ... : ~ . ,

,.,.'~.~,~'.:~.L,,, . - ~

:. ~ "..-:,

:- T :¢..7t47.,-%',c, "~ :3..'. ".r..:,.,., -

%,*:T"$ . '.''::'~'..~ * . *,~ • " - " ~.." ~ ~

•.

.••.

,t, * "

1~'.

:4,L.

,,,."~t:,~'....&

30

3.0

2.5

I *

Trial - #

, .

5.0

~

.

#

f

• .

5.5

..

t

0

750

F!-

.

8

2.5

I

150

Trial -

• . ,.".,

t" ....

",.'0, .-*:~

~ •

N ~.5

6.5

": "

, . 5. < :2~ " ~. • :~.

3,5

0

o

,'.:" ::~:r " ".:.,',', •~

,.- : :...,

;-,~

.

O*

I

*

o

,

.=;.-- - " - ~ ' -

",, . ~ % : . . . :

5.0



o o

4-.0

"

2.5

8

5.5

,.~

3533

I

J

1

I

150

.300

450

600

Trial -

#

25 750

0

I

t

i

I

150

300

450

600

TNal-

750

#

F I G U R E 2. Experiment 1. Effect of double step adaptation on saccadic gain in a scanning paradigm. Data from the six subjects. The • in each graph display the amplitudes of the individual saccades directed to the stepping targets, given as a function of subsequent trial number. The O denote the mean magnitude of the four horizontal saccades elicited in the Scanning situation. The solid and the dashed curves represent the results of fitting the individual data with an exponential function.

mean amplitude of the initial saccade in the Scanning condition decreased slightly in the course of adaptation, but this decrease did also not reach statistical significance (t-test, P > 0.05). The mean and SD of the latencies of primary saccades in the Double step condition was 182 ___39 msec. It is interesting to compare this value with the latencies of the initial saccades that occurred after the onset of the scanning display; these were found to be considerably longer (310 + 141 msec).

Experiment 2: step adaptation affects fast but not delayed responses in an Overlap paradigm The first experiment indicates that adaptive gain reduction of reactive saccades, i.e. saccades that occur in immediate response to the onset of a peripheral target, does not transfer to the saccades with which stationary items are scanned. A major difference between these two conditions is that, for the scanning saccades, the oculomotor system has ample time to prepare and program the saccadic parameters on the basis of the complete visual information about all target positions. For the reactive saccades, however, processing time is limited to the short saccadic latency period. In order to investigate the effect of prolonged pre-saccadic target

availability on the saccadic accuracy, the effect of step adaptation on saccades in an "Overlap" paradigm was tested with the stimulus sequence shown in Fig. I(D). For this purpose, initially 180 Overlap test sequences [Fig. I(D)] with overlap durations between 0 and 400 msec were presented to the unadapted subjects who then underwent a long period of conditioning stimuli consisting of 480 Double step adaptation trials• Finally again, the amount of induced adaptation was tested with 90 Overlap test saccades. Again, six subjects participated in this experiment. Figure 3 presents saccadic gain as a function of the temporal delay D between peripheral target appearance and saccade onset• The O denote the data from the pre-adaptive test phase, the • those from the post-adaptive test phase. For the unadapted subjects the fastest reactions have latencies below 150 msec; these saccades tend to show slightly reduced amplitudes. Otherwise, saccadic gain of the unadapted saccades is largely independent of the delay D, yielding a mean value of 90 4- 1.6% (SE). This picture changes drastically after double step adaptation. Obviously, the amount of gain change induced by the double step stimuli now depends strongly on the amount of available processing time. Saccades reacting immediately to the target onset (i.e. with latencies

3534

HEINER DEUBEL 100

105 r--n

1O0

ae

95 .E o

90

._.q

85

I

._c ._o 13

8

o

80

o o u3

L

7O 0

100

L

I

I

I

I

u o

100

I

,ae,

90

85

.E

85

80

I

i

i

I

I

._o

70

70

105

I

I

I

I

I

60

100 200 300 400 500 600

0

80

I

I

I

I

I

90

._t2

85

75

I

0

I

I

I

i

I

I

I

I

I

[msee]

85 80 75

o

70 65

70

i

I

100 200 300 400 500 600

90 .E

75

65

I

95

80

70

I

100

I

100

o

I

Saccade delay D

95 ._=

85

I

65 I

Saccode deloy D [reset]

t-m

I

80 75

0

I

I

o

75

100 200 300 400 500 600

95

60

I

90

6O

90 "6 m 32

I

95

Saccode delay D [msec]

c

I

65

75

r-~

I

95

I

I

I

I

l

I

60

I

100 200 300 400 500 600

100 200 300 400 500 600

Saccade delay D [rnsec]

Saccade delay D [msec]

I

I

I

I

100 200 300 400 500 60(

Saccade delay D

[reset]

FIGURE 3. Experiment 2. Effect of double step adaptation on saccadic gain in an Overlap test paradigm. The data are given for six subjects. Each graph shows saccadic gain as a function of the temporal delay D between peripheral target appearance and saccade onset. The O denote the data from the pre-adaptive test phase, the • those from the post-adaptive test phase.

below 200 msec) show, consistently for all six subjects, almost complete adaptation to the 25% backward step. Mean post-adaptive gain for these fastest responses is 7 3 _ 1.9%. Later saccades, however, are considerably less affected by the previous conditioning; in three of the six subjects, the gain change is no longer present for the latest responses occurring between 500 and 600 msec after target onset. Averaged over the six subjects, gain of these late saccades is 87.4 _+ 1.7%. Unfortunately, this experiment did not include longer overlap durations. Results from pilot studies however suggest that gain changes induced by double steps disappear completely for overlap durations longer than 1000msec (Deubel, 1993b).

110

I

I

t

110

I

t--m

The next experiment investigated whether double step adaptation transfers to saccades directed to memorized targets. The experimental sequence [Fig. I(E)] was very similar to the previously described Overlap paradigm except that the peripheral target was initially presented for only 100 msec. Randomly selected overlap durations between 0 and 1100 msec were applied. The experimental sessions started with a pre-adaptive test phase in which blocks with 64 of these memory-guided saccades and 60 saccades to single target steps [Fig. I(C)] were interleaved. Then the subject went through a long adaptation phase

I

I

I

be

be

be

IO0

100

100 E 0

O

o~

g0 . __q "D O O

r

Experiment 3: step adaptation does not transfer to memory-guided saccades

--~

0

0

~5

-o

0

u

80

90

Cn

90

80

o 0 0

80

O G~

70 70 0

I

I

300

600

I

I

900 1200 1500

Saccode delay D [msec]

I

7O 0

300

IJ

600

900 1200 1500

Saccade delay

D

[msec]

0

300

I

600

I

I

900 1200 1500

Saccade delay D

[rnsec]

FIGURE 4. Experiment 3. Double step adaptation and memory-guided saccades. Experimental results from the pre-adaptive (open symbols) and post-adaptive (solid symbols) test phases for the three subjects. The squares denote mean saccadic latency and gain from the single step test trials. The circles provide mean saccadic gain in the Memorized target control conditions, as a function of the delay (D) between target flash and saccade onset.

SELECTIVE ADAPTATION

OF REACTIVE

including 400 Double step trials [Fig. 1(B)]. In the final post-adaptive test phase, 64 "Memorized target" conditions and 60 Single step conditions were applied again. Three subjects participated in the experiment. Figure 4 compares the experimental results from the pre-adaptive (open symbols) and post-adaptive (solid symbols) test phases for the three subjects. The squares denote mean saccadic latency and gain from the Single step trials. Obviously saccadic conditioning was successful, reducing mean gain of these saccades from 92.2% before adaptation to 79.0% after adaptation. The circles provide mean saccadic gain in the Memorized target control conditions, as a function of the delay (D) between target flash and saccade onset. Both pre- and post-adaptive controls show a pronounced tendency for larger undershooting with longer delays. However, the amount of transfer of adaptation is small for the faster responses and disappears completely for the saccades with longer onset delays. It can be concluded that step adaptation does not transfer to saccades directed to target position that have to be memorized for a second or longer.

The first experiment demonstrated that when stimulustriggered saccades are adapted, no transfer of adaptation occurs to scanning saccades. The question now arises of

I

i

i

i

_

o

5.5

6.0

_1

5.5

-

I

SACCADES

3535

whether it is also possible to adapt scanning saccades without affecting the stimulus-driven responses. In order to investigate this question the Scanning paradigm [Fig. I(A)] was combined with intrasaccadic displacements of the stimulus. A session started with 80 control experiments consisting of blocks of scanning saccades and single step trials [Fig. I(A, C)]. Then 200 trials were performed in which scanning saccades received conditioning feedback by systematically displacing the whole display by 25% of the item separation (1.5 deg) during each scanning saccade. The shifts occurred into the opposite direction of the saccade. Two blocks, each of 30 Single step test trials [Fig. I(C)], were interleaved with these training stimuli. Six subjects participated in this experiment. Figure 5 presents the experimental results for each of the six subjects. As in Fig. 2, the • in each graph display the amplitudes of the individual saccades directed to the stepping targets, given as a function of subsequent trial number. The © denote the mean magnitude of the four horizontal saccades elicited in the Scanning situation. The solid and dashed curves represent the results of fitting the individual data with an exponential function. As in Expt 1, the subjects consistently revealed larger amplitudes in the Scanning situation (mean + SEM: 5.64 + 0.05 deg) as compared to the stepping target situation (5.15 + 0.14 deg). With the introduction of the conditioning stimulus, the magnitudes of the scanning

Experiment 4: adaptation of scanning saccades transfers only partially to reactive saccades

6.0

AND VOLITIONAL

i

i

n . 6 . 0 ~_1

I

i

5.5

°o

• *

i

i

... Io

• .oqo8

-

5.0

.c

5.0

-

_4~o~tf~o • •: -• " .

5.0

.. "

r, E o

4.5

_

~-~.,.:.~-.

~.:o

o,,



~

4.5

..

.o o

o o U')

a.

4.0

I

.

° "

0

I

1O0

~.~'l 200

[

0

400

I

6.0

i

I

I

200

300

6.0 - oo"

i

7

--,.t

5.5

5.5

5.0

5.o

0

°

:.~

o."

5.5

S" 8_-"~,,~%%.-. Oo°°o.,~_'o..o--~,~o-~.

5.0

o

4.5

"0 o

o~*.



?~

4.0

oo

.

o

o-

o

o ~

?

o o , o

"



"0 o

-'o

o o °

.

o ~l~'::

200

6.0 _r. ~.-,-:

~

o.

.

?:

300

400

Trial -

-] ..

~...o" o _~,~.~A'o°o



¢1. E 4.5 o

1 O0

#

:

• • ,

0

400

..



o

3.5

I

1 O0

Trial - #

8

o ~3

*°o°

Trial -



4.0

°

• • •

3.5

q.

300

I:

"~° o

-.o

-."



o CO

3.5 i

4.5

". o

4.0

~ . : ;

I

.

.'| -

;'-

o

~

.

,

. ,

:.:" •



,

"

:

"

0

."

#

4.5



o

~ 4.o

4.o

o

Q

N 3.5 0

I

I

I

1 O0

200

300

Trial - #

3.5 400

0

I

I

I

I

1 O0

200

300

400

Trial-

#

3.5 0

I

I

I

I

1 O0

200

300

400

Trial -

#

5. E x p e r i m e n t 4. Effect of the adaptation of scanning saeeades on reactive saccades. The • in each graph display the amplitudes of the individual saccades directed to the stepping targets, given as a function of subsequent trial number. The O denote the mean magnitude o f the four horizontal saecades elicited in the Scanning situation. The solid and the dashed curves represent the results of fitting the individual data with an exponential function. FIGURE

3536

HEINER DEUBEL

saccades decreased quickly and consistently in all subjects. The average final gain reduction was found to be 1.08 _ 0.15 deg. The mean of the time constants T of the adaptive process was 46.2 trials, which is considerably faster than for the Double step adaptation condition. It should be taken into account, however, that each of these points represents the average of four subsequent saccades that are trained in each trial. The magnitudes of the interleaved Single step test saccades also become reduced by the ongoing adaptation process, but the mean size reduction is only 0.4 + 0.11 deg. This means that there is indeed some transfer of adaptation of scanning saccades to the reactive saccades, however, the transfer is relatively small, amounting to only 37% of the gain reduction effect induced on the scanning saccades. ANOVA confirmed a highly significant interaction between the test conditions and the time of testing IF(l,5) = 84.5]. Post-hoc Newman-Keuls tests demonstrated significant amplitude changes for both test conditions (P < 0.01).

Experiment 5: adaptation of scanning saccades transfers to delayed responses in an Overlap paradigm The second experiment demonstrated that double step adaptation transfers to short-latency but not to long-latency saccades in an Overlap paradigm (Fig. 3). From the double dissociation indicated in Expt 4 it should be expected that adaptation of scanning saccades should, however, affect late saccades in an Overlap paradigm. This was tested in four subjects. In each session blocks of trials with the Single target step condition [Fig. I(C)] and the Overlap test condition with a temporal overlap of O = 600msec [Fig. I(D)] were interleaved with long phases of adaptation of scanning saccades. The experimental data for the three of the four subjects are shown in Fig. 6 (the results for the fourth subject are very similar). The O in each graph present the sizes of the scanning saccades as a function of trial number; the exponential fit is shown as the solid curve. As in the previous experiment, adaptation in all three subjects is fast and consistent. Again, transfer of adaptation to the reactive saccades in the Single step condition (e) is rather

6.5

I gk

-

I

I

~-~ 6.(?

.._ ~E

5.5-



.

~

[

"

A~

6.0

o"

"o

"

",. ~.. ~ ~ ............ ~ . . . . . .

4.0 3.5 / i " 0

p 1 O0

t 200 Trial-

I

I

6.5

i

I 300

#

%

AxA



~.

5.0

i 400



5.5

o -, ,v.

e..

~...-

6.0

-.

"~ 5.0

°

~2

5.5

N-N: --*

8

i

I

I

t

I

0

100

200

300

400

Trial -

f~

4.o 3.5

"f ,i,

4..5

(,'3

.

o

o

~ 4.5 ?, g 4.0 t/l

Oo

,.5

Finally, I tested the effect of adaptation of scanning saccades on saccadic eye movements to memorized targets. The procedure was identical to the previous Overlap experiment except that the Memorized target test condition [Fig. I(E)] was applied, using overlap times of 600 and 1100 msec. Three subjects participated in the experiment. The experimental results are shown in Fig. 7. The O in each graph represent the sizes of the scanning saccades as a function of trial number, demonstrating, again, fast and consistent adaptation to the conditioning stimulus. Again, transfer of adaptation to the fast, reactive saccades in the Single step condition (.) is rather small as indicated by the solid curves. The amplitudes of the saccades to the memorized targets, however, exhibit considerable gain reductions that are largely identical to the amplitudes of the scanning saccades. It can be concluded that adaptation of scanning saccades transfers completely to saccades directed to memorized targets. ANOVA revealed a significant interaction between the test conditions and the time of testing [F(2,4) = 17.6; P 0.05).

5.5

"

f--

5.0- ~ , ~ ~ , o

zxl

Experiment 6: adaptation of scanning saccades transfers to memory-guided saccades

.

i a~o%. --

6.5

I

I

•. " A ~

small, as indicated by the solid curves. The amplitudes of the saccades in the Overlap condition (A), however, reveal considerable gain reduction in the course of adaptation. So, as expected, fast saccades are largely unchanged, while late saccades in the Overlap condition are affected by a previous adaptation of scanning saccades. ANOVA shows a significant interaction between the test conditions and the time of testing [F(2,6) = 30.11; P < 0.001]. Newman-Keuls tests demonstrate significant amplitude changes for the Scanning and the Overlap conditions (P < 0.01), but not for the reactive saccades ( e > 0.05).

"4"s""

.....

1

I

I

I

0

100

200

300

Trial -

#

F I G U R E 6. E x p e r i m e n t 5. Effect o f a d a p t a t i o n o f s c a n n i n g saccades o n delayed saccades in a n overlap p a r a d i g m . The O in each g r a p h present the sizes o f the s c a n n i n g saccades as a f u n c t i o n o f trial n u m b e r ; the e x p o n e n t i a l fit is s h o w n as the solid curve. -• - - R e a c t i v e saccades in the single step c o n d i t i o n . • -A- - A m p l i t u d e s o f saccades in the overlap condition.

~'

I

4-00

SELECTIVE ADAPTATION OF REACTIVE AND VOLITIONAL SACCADES 6.5

I

.

L~ •

I

I

6.5

1 T

6.0

IA ~.

• • • •

A i ,ao •~

i

i

6.5

i



^ %%

5.5

#

t. •

'~

6.0

~-~ 6.0

'-"

5.5

5.5

5.0

5.0

(~.

4.5

~

4.5

g 4.o

g 4.0

o

o

m

3.5

3.0

Y~

A

0

I

i

I

I

1O0

200

300

400

-~.#.

~1

I

4

~' %'$ ~I

I

"G

I

". .

5.0

E



Oo

°

"0

AA

3537

o aq,M~lk

-

.":.

4~.

~.'~,

.:

~

,.

",

,

,.

¢1_

E o

4.5

4.0

3.s

o

3.0

3.5 3.0

100

Trial- #

200 Trial - #

300

400

0

I

I

i

I

100

200

300

400

Trial - /~

F I G U R E 7. Experiment 6. Effect of adaptation of scanning saccades on memory-guided saccades. The O in each graph present the sizes of the scanning saccades as a function of trial number; the exponential fit is shown as the solid curve. - • - - Reactive saccades in the single step condition. • .A. • Amplitudes of saccades to the memorized target.

DISCUSSION This study is the first to demonstrate that double step adaptation and adaptation during saccadic scanning leads to plastic gain changes that are specific for the type of saccadic response. Adaptation with the stepping target transfers neither to the saccades in the Scanning situation nor to delayed saccades in an Overlap paradigm, nor to memory-guided saccades. Conversely, when saccades are adapted in the Scanning situation, induced gain changes transfer to overlap and memory-guided saccades, but barely to saccades following steps of a single target. This indicates that there is not just a single gain parameter determining the amplitudes of saccadic responses (for saccades in a certain direction), but that at least two separate and largely independent mechanisms are involved in the generation of these two types of saccadic responses. What are the features that distinguish the two saccade modes? One of the characteristics of all of the former experiments with double-step adaptation was that the subject's eye movements were completely guided by the stimulus: all these saccades were "reactive" saccades in the sense that they were triggered by an external event, and that the location of where the eye had to go was determined by the location of the single stimulus in the visual field. Most of our everyday saccades are not determined by external events, however; we intentionally decide where to move the eyes next and when to trigger a saccade. The Scanning paradigm thus aimed at mimicking a situation where the saccades are not determined by the onset of a single target in an otherwise empty field, but where the subject intentionally selects a target from several alternatives, has ample time to prepare the sequence of eye movements and performs a self-paced timing of the scanning eye movements. These saccades are termed "volitional" here in order to emphasize their intentional and voluntary character. The results from this investigation show that the reactive and the volitional types of saccadic responses can be selectively adapted, suggesting the existence of separate adaptive mechanisms for the metrical control of these saccades. Moreover,

the paradigm may serve as a methodological means for categorizing different types of saccadic responses: obviously, delayed saccades and memo;y-guided saccades can be identified as belonging to tee category of volitional saccades. There is recent experimental evidence (Deubel, 1994) that anticipatory saccades to predictive target locations are also affected by scanning adaptation, but not by double step adaptation, allowing to categorize them as volitional. Also, it should be expected that anti-saccades (HaHett, 1978) should show the signs of volitional responses. The finding that, when internally guided saccades are adapted, this conditioning transfers only to a small degree to stimulus-elicited saccades is a confirmation of recent findings by Erkelens and Hulleman (1993) who also suggested from their data that internally triggered saccades can be selectively adapted, leaving stimulustriggered eye movements largely unaffected. However, for the adaptation of internally guided saccades these authors used very large (50% of initial step) backward displacements of a target the subjects were certainly aware of. Also, target positions were not varied during the session. Accordingly, oculomotor learning took place within a few saccades, arguing for a cognitive-strategic cause of saccade size change. Other behavioural studies also hint at a dissociation of endogenously and exogenously controlled saccades. So, Lemij and Collewijn (1990) found that saccades to stationary targets are more accurate than saccades to jumping targets. Collewijn, Erkelens and Steinman (1988) demonstrated that the amplitudes of endogenous saccades do not show the 10% undershooting that is normally seen in saccades to suddenly appearing targets. Further, it is known that volitional saccades, such as saccades directed to memorized targets, exhibit slightly lower maximum velocities than target-driven saccades (Smit, van Gisbergen & Cools, 1987). The question arises of whether other aspects of saccadic eye movements that are under adaptive control would also show a similar dissociation between reactive and volitional response modes. Indeed, this seems to be the case for the adaptation ofpost-saccadic eye drift (Deubel,

3538

HEINER DEUBEL

1991b, 1993a). In these experiments post-saccadic drift was induced by systematic exponential target movements at the end of each saccade. The results demonstrated that the amount of induced post-saccadic eye drift is indeed dependent on the way of testing: transfer of adaptation is significantly smaller to scanning saccades and to spontaneous saccades in the dark than to reactive saccades (Deubel, in preparation). Given the proposed classification of saccades into "reactive" and "volitional" responses is correct, of what type then are the saccades we perform in our normal visual environment, outside the laboratory? It is probably reasonable to assume that the normal oculomotor activity consists of a mixture of both, saccades scanning static objects and of saccades reacting to transient changes in our environment (such as the appearance of a car in the visual periphery). According to this hypothesis, both types of responses are continuously controlled by the respective adaptive mechanisms. This is in agreement with clinical reports and non-human lesion studies (e.g. Kommerell et al., 1976; Optican & Robinson, 1980) showing that plastic changes of saccade parameters developed during normal behaviour transfer, at least in part, to saccades to flashing lights in a laboratory environment. The data shown here are another demonstration of the amazing adaptability of certain aspects of the saccadic response to the requirements presented by the environment. Some of the adaptations found are extremely fast, even when the system is confronted with the additional demand for disconjugate changes (van der Steen, 1993; Eggert & Kapoula, 1993; Eggert, Kapoula & Bucci, 1995; van der Steen & Bruno, 1995). In considering the general validity of ultra-fast adaptations as truly adaptive processes, it is important to be aware of the possibility that these changes could reflect the switching to a specific strategy rather than the plastic modification of oculomotor parameters. A previous study (Deubel, 1995) demonstrated that visual stimulus features (e.g. colour and form of the target, background structure) cannot be used by the saccadic system to switch between different sets of response parameters suggesting that there is no specificity of saccadic gain control with respect to visual aspects of the target. The finding that visual target features p e r se cannot serve for selecting specific response modes seems to argue against the hypothesis that ultra-fast adaptations represent the cognitively controlled selection of response strategies. It should be mentioned that experimental indications for context-specific adaptive mechanisms are not limited to the saccadic system: recently, evidence has been accumulating for contextspecific gain switching even in the vestibulo-ocular reflex (Tan, Shelhamer & Zee, 1992; Shelhamer, Robinson & Tan, 1992). The findings of this study also have important implications for the understanding of the neural control of saccadic eye movements. So, the results provide strong arguments that reflexive, stimulus-guided saccades and intentionally elicited saccades are controlled by different neural mechanisms. Separate and largely independent

adaptive control systems seem to exist for the protection of both the visuo-motor reflex pathways and the pathways for intentionally induced eye movements. Dependent on the conditions under which adaptation occurs, either endogenous or exogenous saccades are affected. It is very tempting to speculate that the neural substrate of the reflexive system might be the rather direct retino-collicular pathways, maybe including the primary cortex, while the substrate for the intentionally controlled saccades may include the frontal eye fields (e.g. Schiller, 1985). Moreover, the experiments provide new insight concerning the neuronal site where the adaptation may take place. They make it unlikely that the adaptive gain modification occurs in lower oculomotor centres such as the brainstem, where the collicular and frontal eye field signals have presumably already converged. The hypothesis of a higher-level site of saccadic adaptive control has already been put forward by the author in the context of experiments demonstrating an amazing directional selectivity of adaptive changes of saccade gain (Deubel, 1987) and post-saccadic drift (Deubel, 1991b). Simulations seem to suggest that these findings are not compatible with adaptation occurring at the level of the peripheral oculomotor system being largely organized in cartesian (horizontal and vertical) components (Deubel, 1993a). Further evidence against low-level adaptive control comes from neurophysiological studies by FitzGibbon and Goldberg (1988). They found that saccades evoked by stimulation of the superior colliculus at the site corresponding to the position of the adapted saccade vector remain identical to those evoked before the adaptation. This finding is compatible with adaptive changes on afferent connections to the intermediate superior colliculus. Moreover, the present study raises new questions as to the role of the cerebellum in the selective adaptation of both response types. It is well-established that cerebellar structures, especially the cerebellar midline vermis and the fastigial nuclei are essential for the adaptive maintenance of saccadic accuracy (e.g. Ritchie, 1976; Optican & Robinson, 1980; Dean, Mayhew & Langdon, 1992). From the experimental findings presented above the question arises whether these cerebellar mechanisms are responsible for maintaining proper functioning of both of the postulated pathways. In order to evaluate possible differences of cerebellar involvement in both types of saccadic responses, investigations applying the above experimental paradigms to patients suffering from cerebellar diseases have been performed recently. Preliminary results from these studies indeed suggest that cerebellar lesions may specifically affect only one of both response types, namely the reflexive saccades. We studied a patient suffering from a bilateral lesion of the fastigial nucleus caused by a large cystic tumor extending into the medial cerebellum (Straube, Deubel, Spuler & Biittner, 1995). The accuracy of both reflexive saccades to stepping targets and volitional saccades was examined 6 days after surgery, in paradigms identical to the ones described here. For the saccades occurring in response to the stepping targets, the patient's responses showed the prominent

SELECTIVE ADAPTATION OF REACTIVE AND VOLITIONAL SACCADES

saccadic overshoots that form the characteristic clinical signs for a disease of the midline cerebellum. For the saccades that occur in the Scanning situation, however, no such signs were observable. In other words, while the patient's reflexive saccades were clearly hypermetric, the internally guided saccadic responses were perfectly normal. This indicates that the cerebellum, and possibly the cerebellar fastigial nucleus in particular, is involved in the generation of externally triggered saccades toward a jumping target but not in the generation of scanning saccades toward a stationary target. Otherwise, direct physiological evidence for a differential role of the cerebellum is still rare. Ohtsuka and Noda (1991) describe a difference between the discharge of the cerebellar fastigial nucleus neurons for memory-guided and externally triggered saccades as opposed to spontaneous saccades. Further indications for a basic dissociation between reflexive and volitional actions come from the area of motor control. So, Bizzi, Kalil and Morasso (1972) reported that the neuro-muscular activity underlying active eye-head coordination displays distinct characteristics depending upon whether these were stimulustriggered movements or the result of prediction. Frens and Erkelens (1991) find different interactions of hand and eye movements depending upon whether the target was actually visible or the movements were anticipatory. Bridgeman, Lewis, Heit and Nagle (1979) and Paillard (1987) propose, on the basis of a variety of experiments, a distinction between two separate mappings of spatial relationships that might independently contribute to the nervous organization of spatial behaviour. The first refers to a sensorimotor mode which is uniquely spatial, generally unconscious, and motor-oriented. The second, "representational" mode of processing spatial information refers to a system with a more symbolic, conscious content, forming the basis of perception. Finally, there may be some relations of the proposed dissociation to the field of visual attention. A number of workers (e.g. Jonides, 1981; Nakayama & Mackeben, 1989) report basically different properties of shifts of visual attention elicited by exogenous cues (such as the abrupt onset of a peripheral stimulus) and endogenous, volitional shifts of attention.

REFERENCES Abel, L. A., Schmidt, D., Dell'Osso, L. F. & Daroff, R. B. (1978). Saccadic system plasticity in humans. Annals of Neurology, 3, 313-318. Bizzi, E., Kalil, R. E. & Morasso, P. (1972). Two modes of active eye-head coordination in monkeys. Brain Research, 40, 4548. Bridgeman, B., Hendry, D. & Stark, L. (1975). Failure to detect displacement of visual world during saccadic eye movements. Vision Research, 15, 719-722. Bridgeman, B., Lewis, S, Heit, G. & Nagle, M. (1979). Relation between cognitive and motor-oriented systems of visual position perception. Journal of Experimental Psychology: Human Perception and Performance, 5, 692-700. Collewijn, H., Erkelens, C. J. & Steinman, R. M. (1988). Binocular co-ordination of human horizontal eye movements. Journal o f Physiology, 404, 157-182.

3539

Crane, H. D. & Steele, C. M. (1985). Generation-V dual-Purkinje-lmage eyetracker. Applied Optics, 24, 527-537. Dean, P., Mayhew, J. E. W. & Langdon, P. (1992). A neural net model for adaptive control of saccadic accuracy by primate cerebellum and brainstem. In Moody, J. E., Hanson, S. J. & Lippmann, R. P. (Eds), Advances in neural information processing systems 4 (pp. 595-602). San Mateo, Calif.: Morgan Kaufmann. Deubel, H. (1987). Adaptivity of gain and direction in oblique saccades. In O'Regan, J. K. & Lrvy-Schoen, A. (Eds), Eye movements: From physiology to cognition (pp. 181-190). Amsterdam: Elsevier. Deubel, H. (1991 a). Adaptive control of saccade metrics. In Obrecht, G. & Stark, L. W. (Eds), Presbyopia research (pp. 93-100). New York: Plenum Press. Deubel, H. (1991b). Plasticity of metrical and dynamical aspects of saccadic eye movements. In Requin, J. & Stelmach, G.E. (Eds), Tutorials in motor neuroscience (pp. 563-579). Dordrecht: Kluwer. Deubel, H. (1993a). Learning in the oculomotor system: Implications for models of saccadic control. In d'Ydewalle, G. & van Rensbergen, J. (Eds), Perception and cognition: Advances in eye movement research (pp. 377-389). Amsterdam: North Holland. Deubel, H. (1993b). Context specificity of saccadic adaptation. Investigative Ophthalmology and Visual Science (Suppl.), 34, 3947. Deubel, H. (1994). Selective adaptation of intentional and reactive saccades. In Fuchs, A. F., Brandt, Th., Biittner, U. & Zee, D. S. (Eds), Contemporary ocular motor and vestibular research: A tribute to David A. Robinson (pp. 206-208). Stuttgart: Thieme. Deubel, H. (1995). Is saccadic adaptation context-specific? In Findlay, J. M., Kentridge, R. W. & Walker, R. (Eds), Eye movement research." Mechanisms, processes and applications (pp. 177-187). Oxford: Elsevier. Deubel, H. & Bridgeman, B. (1995). Fourth Purkinje image signals reveal eye lens deviations and retinal image distortions during saccades. Vision Research, 35, 529-538. Deubel, H., Wolf, W. & Hauske, G. (1986). Adaptive gain-control of saccadic eye-movements. Human Neurobiology, 5,245 253. Eggert, T. & Kapoula, Z. (1993). Fast disconjugate adaptations to aniseikonia. Neuroscience Abstracts, 18, 102.7. Eggert, T., Kapoula, Z. & Bucci, M. P. (1995). Fast disconjugate adaptations of saccades: Dependence on stimulus characteristics. In Findlay, J. M., Kentridge, R. W. & Walker, R. (Eds), Eye movement research: Mechanisms, processes and applications. Oxford: Elsevier. Erkelens, C. J. & Hulleman, J. (1993). Selective adaptation of internally triggered saccades made to visual targets. Experimental Brain Research, 93, 157-164. FitzGibbon, E. J. & Goldberg, M. E. (1988). Long-.term saccadic adaptation does not affect the saccades evoked by electrical stimulation of the superior colliculus. Frens, M. A. & Erkelens, C. J. (1991). Coordination of hand movements and saccades: Evidence for a common and a separate pathway. Experimental Brain Research, 85,682-690. Hallett, P. E. (1978). Primary and secondary saccades defined by instructions. Vision Research, 18, 1279-1296. Henson, D. B. (1978). Corrective saccades: Effect of altering visual feedback. Vision Research, 18, 63-67. Jonides, J. ( 1981). Voluntary vs. automatic control over the mind's eye's movement. In Long, J. & Baddeley, A. (Eds), Attention and performance IX. Hillsdale, N.J.: Erlbaum. Kommerell, G., Olivier, D. & Theopold, H. (1976). Adaptive programming of phasic and tonic components in saccadic eye movements. Investigative Ophthalmology, 15,657 660. Lemij, H. & Collewijn, H. (1990). Differences in accuracy of human saccades between stationary and jumping targets. Vision Research, 29, 1737 1748. McLaughlin, S. C. (1967). Parametric adjustment in saccadic eye movements. Perception & Psychophysics, 2, 359-361. Miller, J. M., Anstis, T. & Templeton, W. B. (1981). Saccadic plasticity: parametric adaptive control by retinal feedback. Journal of Experimental Psychology: Human Perception and Performance, 7, 356-366.

3540

HEINER DEUBEL

Nakayama, K. & Mackeben, M. (1989). Sustained and transient components of focal visual attention. Vision Research, 29, 1631-1647. Ohtsuka, K. & Noda, H. (1992). Saccadic burst neurons in the oculomotor region of the fastigial nucleus of macaque monkeys. Journal of Neurophysiology, 65, 1422-1434. Optican, L. M. & Robinson, D. A. (1980). Cerebellar-dependent adaptive control of primate saccadic system. Journal of Neurophysiology, 44, 1058-1076. Paillard, J. (1987). Cognitive versus sensorimotor encoding of spatial information. In Ellen, P. & Thinus-Blanc, C. (Eds), Cognitive processes and spatial orientation in animal and man (pp. 43-77). Dordrecht: Martinus Nijhoff. Ritchie, L. (1976). Effects of cerebellar lesions on saccadic eye movements. Journal of Neurophysiology, 70, 1246-1256. Schiller, P. (1985). A model for the generation of visually guided eye movements. In Rose, D. & Dobson, V. G. (Eds), Models of the visual cortex (pp. 62-70). New York: Wiley. Shelhamer, M., Robinson, D. A. & Tan, H. S. (1992). Context-specific gain switching in the human vestibulo-ocular reflex. Annals of the New York Academy of Sciences, 656, 889-891. Smit, A. C., van Gisbergen, J. A. M. & Cools, A. R. (1987). A parametric analysis of human saccades in different experimental paradigms. Vision Research, 57, 1745-1762.

van der Steen, J. (1993). Nonconjugate adaptation of human saccades: Fast changes in binocular motor programming. Neuroscience Abstracts, 18, 102.3. van der Steen, J. & Bruno, P. (1995). Unequal amplitude saccades produced by aniseikonic patterns: Effects of viewing distance. Vision Research, 35, 3459-3471. Straube, A., Deubel, H., Spuler, A. & Biittner, U. (1995). Differential effect of a deep bilateral cerebellar nuclei lesion on externally and internally triggered saccades. Neuro-Ophthalmology, 15, 61-74. Tan, H. S., Shelhamer, M. & Zee, D. S. (1992). Effect of head orientation and position on vestibulo-ocular reflex adaptation. Annals of the New York Academy of Sciences, 656, 158-165. Wolf, W. & Deubel, H. (1993). Enhanced vision with saccadic eye movements: An artefact due to phosphor persistence of display screens? In Proceedings of the 15th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (pp. 1383-1384). New York: IEEE.

Acknowledgements--The author is indebted to Silvia Hieke for her never-ending patience and persuasive competence in organizing and running the experiments. The study was supported by the "Science" program of the European Economic Community (Contract No. SC1"-CT91-0747).