PROCESSING OF DIRECTION AND MAGNITUDE BY THE SACCADIC EYE-MOVEMENT SYSTEM’ R. L. Hoc’ and D. H. FENDER California Institute of Technology. Pasadena. California 9113. U.S.A. (Receired
25 Sepremhcr
1978)
Abreact-The function of the saccadic programming system has been studied using a doublestep target movement in two dimensions. The results of these experiments suggest the following hypotheses. The information processing in the programming of a saccade consists of direction computation and magnitude computation. If the new target-step arrives before the direction computation of the previous saccade is complete, this saccade will be cancelled. The partial program concerning the direction of the saccade is kept in a buffer memory: if the direction of the new saccade is not in a direction similar to the old one. this partial program has to be erased. which takes an extra 40-80 msec of processing time. There is a stage in which the direction of the saccade cannot be reprogrammed but the magnitude can still be reduced. In other words. the magnitude computation seems to finish after the direction computation. If the new target-step arrives after the computation of both direction and magnitude are complete. two full saccades will be observed.
INTRODUCTION In 1954, Westheimer pointed out that the response of the saccadic system to a pair of closely spaced targetsteps could reveal certain aspects of the programming processes underlying this activity; but despite a large amount of later work in this field (Wheeles et a[., 1966: Becker and Fuchs, 1969; Feinstein and Williams. 1972: Komoda PI u/., 1973; Lisberger er a/., 1975: Carlow et al., 1975; Becker and Jugens. 1975). the whole picture is still not clear. Some results from different laboratories conflict with each other; this may be explained by different experimental conditions, since most previous experiments have been designed to test part of the control system only, Minor invasion of activity for the supposedly unstimulated parts of the system can cause considerable variations in response. It is generally observed that saccadic eye movements have a reaction time of about 2OOmsec. but it is still not clear what kinds of information processing go on during this period. Although neurophysiological studies have gained some knowledge about how a saccadic eye movement is controlled by. premotor neural circuitry and the extraocular muscles, very little is known about how a saccadic eye movement is programmed. In most previous studies, only horizontal targetsteps have been used. In these cases the saccadic programming system may learn and take advantage of the fact that only two recti need be programmed for each eye. In our studies. the double target-step experimental paradigm was expanded to include twodimensional target movements and the direction of target-steps occurred in quasi-random order.
’ This research was supported in part by grants NS 03627 and RR07003 from the’ National Institutes of Health. z Present address: Bell Telephones Research Laboratories. Holmdel. NJ 07733. U.S.A.
These experiments exhibit some new results and some new interpretations are offered on the programming strategy of the saccadic eye movements. In this paper, the directional effect of the target movements on the reprogramming processes of the saccade is discussed.
METHODS Experirtte~trul Apparatus The apparatus used for this study is a flexible real-time computer system with data acquisition and graphic pattern generation capability. A pkdSIIIa screen was used to display the stimulus. The screen has 60 dots to the inch. and was placed 57 inches in front of the subject, thus the distance between each pair of dots corresponded to I min arc. Eye mocemenf
meusuremetl
t
A contact lens method as described by St-Cvr and Fender (1969) was employed throughout the experiments to measure eye movements. Duru anul_wis
The output voltage from the eye-position transducers corresponds to eye position. This was amplified and then digitized at I-msec intervals. The digitized voltage was read by the computer program for on-line data analysis. Twodimensional saccadic eye movements were first detected by the following algorithm. The eye position was examined every msec; a saccade was deemed to have occurred only if there were five consecutive increases or decreases above a threshold in either the horizontal or the vertical component. The saccadic onset time was set at the time at which the first increase or decrease occurred. In these experiments, the saccades did not last more than 3Omsec. hence the end position of a saccade was taken to be the position of gaze at SOmsec after the onset of the saccade. Target
moriotl
The stimulus was an approximately circular annulus of outside diameter substending 21 min arc: the width was 2 min arc. The target was viewed monocularly. AI the beginning of each trial. the subject was instructed to fixate
1421
R. L. Hoc and D H. FENDER Clclssl$ctrrlo/l 2 3 P
lb
0
0
Fig. I. Target motions in (a) single-step presentation: (b) adjacent presentation: (c) opposite presentation. Only the heavily outlined circles are illuminated sequentially in each presentation. on a target presented at the center of the screen. The computer was programmed to determine that the observer was fixating the target. The computer would then move the target two degrees to one of the eight positions shown in Fig. I. The target might them be moved again after various time intervals. Three paradigms for target presentaiions were used:
(I) Single-step presentation. The target remained starionary after the first jump (Fig. la). This served as the control experiment. (2) Adjacent presentation. After the first step the target jumped again to one of the two adjacent vertices of the octagon and stayed there until the next trial (F’ig. I b). (3) Opposite presentation. After the first step the target jumped again to the vertex opposite to the position of the first target-step and stayed there until the next trial (Fig. Ic). For stimuli 2 and 3 the time interval between the first and the second target jumps could be set to 50, 100 or 150msec. All target positions. stimulus modes. and time randomly. Subjects were were presented intervals instructed to follow the target-steps as quickly as possible. Two subjects were used in this study. RESULTS
ASD
of.rclccdtlrc
rrspon~rs
Wheeles er Al. (19661 classified the response of saccadic eye movements to this tuo-step stimulus as “one-saccade” or V‘tvvo-saccade” depending on the nature of the motion of the visual axis. In one-saccade responses. the saccade moved the subject‘s fixation from the original gaze position to the final target position: the first target step was ignored. In two-saccads responses. the first saccade brought the subject’s gaze to the first target position then the second saccade moved the fixation to the final target position. However. in two-dimensional e>s movements the above dichotomy is not so clear cut. For example. in the adjacent presentation (stimulus 1). there were cases in which subjects responded ivith two saccades but both in the direction of the final position of the target. The second saccade in this case could be a correction saccade: but if only horizontal eye movements are studied and both target steps are in the same direction. such a response mould be mistakenly classified as a two-saccade response rvhereas it should be classified as a one-saccade response with a correction saccade. This error might become serious for those studies involving large saccades in which a high percentage of correction saccades are found. We therefore use the term “reprogrammed saccade response” to replace the “one-saccade response” for those responses in ahich the first target-position was ignored. Examples of eye movements in these various categories are shown in Fig. 3. In the cases when the saccadic s!stem responded to both target-steps, the magnitude of the first saccade \vas sometimes much smaller than the magnitude of the first target-step. This short-fall could be due to reprogramming efforts of the saccsdic programming system in response to new target information; for example, the first saccade might not be totall! reprogrammed, but partial reprogramming was still possible. It could also be argued that the reduced magnitude was not due to reprogramming but was origi-
Double-step
Target Movement
/
_____I !
DlSCt_SSlOS
Several commonly used terms in the following cussion are defined here (Fig. 2):
dis-
ISI (inter-stimuli interval) is the time interval between two target-steps. RT is the average reaction time of saccadic eye movements in response to a single step target jump. m is the average reaction time of a reprogrammed saccade (see below) measured from the beginning of the-- first target-step. RT, and RT, are average reaction times of the first and the second saccades in two-saccade responses (see below) measured from the beginning of the first target step. Latencies measured from the begi&ng of the second target jump are (m - ISI) and (RT, - ISI).
h----RTr------! Fig. 2. Definitions
of RT,. RT.. RT;. ISI.
:. a”
Processing of direction and magnitude
20
I
1423
3
20
;. .s”?
20
3
2
.?’ 3
I :I P
Fig. 3. Typical eye movement responses; left, adjacent presentations: right. opposite presentation. (a) Reprogrammed saccadic response. (b) Intermediate two-saccade response. (c) Full two-saccade response.
nally programmed that way, and normally would be followed by correction saccade if there were no second target-step. If this argument holds. one would expect about the same percentage of short saccades in the control experiments (responses lo single targetsteps). two-saccade responses and reprogrammed saccade responses. Figure 4 shows the distribution of the ratio of the magnitude of the first or the only saccade to the target-step magnitude for one subject in the three response categories. and Table I shows the per-
Table I. Percentage of first saccades of a two-saccade response having a magnitude less than 3,4 of the first target-step
Single target-step Reprogrammed saccade Two-saccade response
26.7 20.0 44.5
centage of the first saccades falling short by 25% or more. There is a significant increase in short first sac-
(c) Ftrsl
_I
(
Saccade
I” Two-saccode
Response
J-J-T1525m 0.25
0.50 0.75 MAGNITUDE
I.00
.
I.50
RATIO
Fig. 4. Distribution of the first-saccade magnitude ratio for various stimuli and responses.
cades in the two-saccade response, hence these can probably be explained as partially reprogrammed saccades. Figure 5 shows the frequency distribution of the differences between the direction of the first or the only saccade and the eccentricity of the target. All three plots show a similar distribution. thus it seems appropriate to classify the saccadic responses to twostep target movements into three classes instead of two. (I) Reprogrammed saccade response. The initial saccade is completely reprogrammed and the first target-step is ignored. The saccadic eye movements are directed toward the final target-step only (Fig. 3a). (2) Intermediate two-saccade response. The subject responds to both target-steps. but the magnitude of the initial saccade is less than 3,‘4 of the first target step. The second saccade brings his gaze to the final target position (Fig. 3b). (3) Full two-saccade response. Same as intermediate saccade response except that the magnitude of the first saccade is greater than 3/4 of the initial target step (Fig. 3~). The reprogrammed saccade response and the intermediate saccade response. represent two levels of
1121
R. L. HOL and
r
(a i Smqie-step Respoo$e I
FESDER
Qrammed rrsponses and of intermediate two-saccadr responses is the same for both classes of adjacent presentations. The probabilities of the three types of saccadic response for ditftrent time intervals between target-steps are given in Table 2.
P
0.4
0. H.
1
0.3
0.2 Reaction
TI\lE
OF REPROCRASI>lED
SACCIDES
0.1 ~ Oi 0.4
-48
~1
( b) ReDragrommed
-40
-32 -24
-16
DIRECTIONAL
Fig.
5.
-8
0
8
ERROR
16
24
32
40
48
(deqrees)
Distribution of the directional error of the various stimuli and responses.
first
saccade for
reprogr~~mmab~lity of the saccadic system. It appears that there is a state in the programming process after which the direction of the saccade cannot be changed but the magnitude can still be shortened in response to new target information. This strengthens the suggestion by Komoda YKul. (1973) that computation of direction and computation of magnitude are separate processes. The intermediate two-saccade response further suggests that. in some circumstances. the computation of direction is completed earlier than magnitude computation. A priori. there seem to be two different classes of adjacent presentations. One class comprises the changes from horizontal eccentricities to any other position. in which case we have a onemuscle-pair motion reprogrammed into a motion involving more than one muscle pair. The other class of adjacent presetitations would be the converse of this. However, we find that the probabilities of reproTable
2. Variation
of the probabifity
ISI (msec)
50
Reprogrammed saccade Intermediate two-saccade Full two-saccade
0.83 0.085 0.085
I5
Wheelrs or u/. (1966) used a target step combination similar to our opposite presentation but only in the horizontal direction and found that the average reaction time of reprogrammed saccades measured from the second target-step (q - ISI) was about 40 msec longer than the normal reaction time of saccades. They explained this extra 40 msec as the time required to cancel the programming of previous saccades. However. Komoda et ui. (1973) found no extra cancellation time. and if both horizontal target-steps were in the same direction. less time than the normal latent! was needed to reprogram a saccade. In our opposite presentation. the two target-steps were in completely opposite directions. To reprogram a saccade in this case. agonist and antagonist oculomotor muscles had to switch their roles. In the adjacent presentation the two target-steps were at a 45 angle, hence to reprogram a saccade. the new saccadic eye movements may require activity in additional sets of oculomotor muscles. The average reaction times of reprogrammed saccade for opposite and adjacent presentations are shown in Fig. 6 which shows (m - ISI) as a function of ISI. For the adjacent presentation. the average reprogramming time were about the same as normal saccadic reaction times for both subjects. For opposite presentations extra time was observed for both subjects: it was about 4Omsec more for one subject and 80 msec more for the other. The distributions of the reaction times for one subject are shown in Fig. 7. This curve is multimodal, strengthening the idea that reprogramming is the factor determining the reaction time. The fact that more than the normal programming time of a saccade is needed to reprogram a saccade in the opposite direction may have important implications in our understanding of the mechanisms of saccadic programming. One possible explanation for this extra time is that it is required to switch the role of agonist and antagonist in the central program. In other words. in order to program LLnew saccade in the opposite direction. the previous programming for the antagonist must be erased. This explanation implies that in these early stages of saccadic programming, the control programs for both agonistic and antagonistic oculomotor neurons already exist. It is
of each t)pe interval Subject MB loo O.-t5 0.35 0.30 204 msec
of saccadic
response
150
50
O.IZS 0.38 0.492
0.81 0.10 0.09
with
inter-stimuli
Subject RW 100 0.42 0.15 0.43 I96 m5ec
I50 0.05 0.1-t 0.71
112.5
Processing of direction and magnitude 400-
cn Adjacent'
0
0
200-
,;
ET
I
loo
I
I
1
150
100
50
ISX fmsec)
Fig. 6. Average reaction times of the reprogrammed saccade measured from the onset of the second target step. possible that central processes do not program both agonistic and antagonistic motorneurons individually, but the inhibitive signals to the antagonistic motorneurons individually, but the inhibitive signals to the antagonistic motomeurons are driven from the signals sent to agonistic motorneurons through “sign conversion” processes of interneurons. If this is true, there is no need to cancel the previous programming of the antagonist. in this event the extra time could be explained as the cancellation time for the previous program for the agonistic motorneurons. There are two plausible theories to explain the extra pro_framming time for opposite presentation. For simphcity, let us call them the “antogonist cancellation theory” and the “agonist cancellation theory”. To differentiate between these two theories, the following modified two-target-step experiment was performed. Only the top, bottom. left and right vertices of the octagon were used (Fig. 8). The first target-step moved to any one of the four vertices and the second target jump moved to an adjacent vertex. This target pattern will be referred to as “orthogonal presentation”. If the saccade is reprogrammed, a completely different pair of agonistic and antagonistic motorneurons will be involved. If the agonist motorneurons which are required for a saccade to the first target positions are not needed for the reprogrammed saccadie eye movement, the agonist theory would require the previous program to these motorneurons to be cancelled. and extra cancellation time should be observed. Conversely, for the new reprogrammed saccadic eye movements. there are no antagonistic
motomeutrons which have been previously programmed. therefore no cancellation is needed. In summary, if an increased reaction time is observed for the reprogrammed saccade, this would argue against the 3
9 0
0-l I
0
2
3
~l
0.3
the orthogonal presentation.
Reprogrammed
0.2
0.1 100
200
0
Fig. 8. Two examples of target movement sequences for
st Sacade (N=?O)
0.4
2
300 400 Time (msec)
500
Fig. 7. Distribution of reaction times for various target presentations.
I426
R. L. Hoc
and D. H. FESDER
antagonist cancellation theor). while an unchanged reaction time would argue against cancellation theorq. In this experiment the orthogonal presentation i+as randomly intermixed with opposite presentations. The single step presentations of the previous experiment were used as the control. The results showed about the same percentage of reprogrammed saccade responses as the previous experiment. The aberage reaction times of the reprogrammed saccades measured from the second target step (v - ISI) for both opposite and orthogonal target presentations were the same, about 230msec. The first point that can be made from these data is that the reaction time for orthogonal presentation is increased. thus the antagonist cancellation theory is not upheld. Secondly. the change in the reaction time for the opposite presentation is the same as that for the orthogonal presentation. hence we might assume that the same mechanism is operative in both cases. The extra time seems to be used to erase the previous direction program even though that particular muscle will not be activated in the new reprogrammed saccade. This may hint at the existence of a buffer which will hold the partial results of the directional computation. If the direction of the reprogrammed saccade is not similar to the previous one. the results of the unfinished computation in the buffer may have to be erased before the new results can be put in. If the direction of the reprogrammed saccade IS m a direction similar to that of the cancelled saccade, the results of the direction computation require onlj some modification. Hence if the reprogrammed sacsade is in exactly the same direction as the cancelled saccade, the computation time for the reprogrammed
saccade should be shorter. This hhpathesis seems to explain our results ,nd also the discrepant> betueen Wheeles t’r (il. (1966) and Komoda ;‘r ,tl. I 19731.
REFEREICES Becker W. and Fuchs 4. 119691 Furth-r propertIc 01’ the human succadic skstsm: sqe mowmsnts and correction saccades wifh and \*tthou! rwai ti\ltion points. L’isiotl Ret 9. 11-!7-!:jg. Becker W. and Jugens R. (1975) Saszsdic reactions to double-step stimuli: evidence for model feedback and continuous information uptake. In B.ii(c. .Mrclr~oti.wrs U/ IEdited Owlnr .Moriliry ml Tlrrir C/imd I q~i~uriou~ by Lennerstrand and Bach-y-Rital. p. 519. Pergamon Press. Oxford. Carlou T.. Del’Osso L. F.. Troost B. T. DarotT R. B. and Birkett J. E. (1975) Saccadic eye moicment latencies to multimodal stimuli. C’isrort Rrs. 15. l?i’-1262. Fsinstein R. and Williams W. J. (19711 Interactions of the horizontal and vertical human oculomotor systems: the saccadic systems. I’isiwr Rrs. 12. 33-4-4. Komoda M. K.. Festinger L., Philips L. J Duckman R. H. and Young R. A. (19731 Some obser\ations concerning saccadic eye movements. C’isicw Rrs. 11. 1009-1020. Lisberger S. G.. Fuchs A. F.. King W. \I. and Evinger L. C. I 1975) Effect of mean reaction time on saccadic resoonse to two-step stimuli uith horizontal and vertical components. b’i.sim Rrs. II. 10~1-1025. St Cyr G. J. and Fender D. H. (19691 Nonlinearrties oi human oculomotor system: gain. c’isiou Rfs. 9. 1135-1216. Westheimer G. (19511 Mechanism of tic;adic e)e movements. ;Irclzs OplttM. 52. 7 10-721. Wheeles L. L.. Jr.. Bo>nton R. &I. and Cohen G. H. 1966. Eye movements responses to step and pulse-step stlmult.
J. opt. SOC. .hr. 56. 9X-960.
25 Sepremhcr
1978)
Abreact-The function of the saccadic programming system has been studied using a doublestep target movement in two dimensions. The results of these experiments suggest the following hypotheses. The information processing in the programming of a saccade consists of direction computation and magnitude computation. If the new target-step arrives before the direction computation of the previous saccade is complete, this saccade will be cancelled. The partial program concerning the direction of the saccade is kept in a buffer memory: if the direction of the new saccade is not in a direction similar to the old one. this partial program has to be erased. which takes an extra 40-80 msec of processing time. There is a stage in which the direction of the saccade cannot be reprogrammed but the magnitude can still be reduced. In other words. the magnitude computation seems to finish after the direction computation. If the new target-step arrives after the computation of both direction and magnitude are complete. two full saccades will be observed.
INTRODUCTION In 1954, Westheimer pointed out that the response of the saccadic system to a pair of closely spaced targetsteps could reveal certain aspects of the programming processes underlying this activity; but despite a large amount of later work in this field (Wheeles et a[., 1966: Becker and Fuchs, 1969; Feinstein and Williams. 1972: Komoda PI u/., 1973; Lisberger er a/., 1975: Carlow et al., 1975; Becker and Jugens. 1975). the whole picture is still not clear. Some results from different laboratories conflict with each other; this may be explained by different experimental conditions, since most previous experiments have been designed to test part of the control system only, Minor invasion of activity for the supposedly unstimulated parts of the system can cause considerable variations in response. It is generally observed that saccadic eye movements have a reaction time of about 2OOmsec. but it is still not clear what kinds of information processing go on during this period. Although neurophysiological studies have gained some knowledge about how a saccadic eye movement is controlled by. premotor neural circuitry and the extraocular muscles, very little is known about how a saccadic eye movement is programmed. In most previous studies, only horizontal targetsteps have been used. In these cases the saccadic programming system may learn and take advantage of the fact that only two recti need be programmed for each eye. In our studies. the double target-step experimental paradigm was expanded to include twodimensional target movements and the direction of target-steps occurred in quasi-random order.
’ This research was supported in part by grants NS 03627 and RR07003 from the’ National Institutes of Health. z Present address: Bell Telephones Research Laboratories. Holmdel. NJ 07733. U.S.A.
These experiments exhibit some new results and some new interpretations are offered on the programming strategy of the saccadic eye movements. In this paper, the directional effect of the target movements on the reprogramming processes of the saccade is discussed.
METHODS Experirtte~trul Apparatus The apparatus used for this study is a flexible real-time computer system with data acquisition and graphic pattern generation capability. A pkdSIIIa screen was used to display the stimulus. The screen has 60 dots to the inch. and was placed 57 inches in front of the subject, thus the distance between each pair of dots corresponded to I min arc. Eye mocemenf
meusuremetl
t
A contact lens method as described by St-Cvr and Fender (1969) was employed throughout the experiments to measure eye movements. Duru anul_wis
The output voltage from the eye-position transducers corresponds to eye position. This was amplified and then digitized at I-msec intervals. The digitized voltage was read by the computer program for on-line data analysis. Twodimensional saccadic eye movements were first detected by the following algorithm. The eye position was examined every msec; a saccade was deemed to have occurred only if there were five consecutive increases or decreases above a threshold in either the horizontal or the vertical component. The saccadic onset time was set at the time at which the first increase or decrease occurred. In these experiments, the saccades did not last more than 3Omsec. hence the end position of a saccade was taken to be the position of gaze at SOmsec after the onset of the saccade. Target
moriotl
The stimulus was an approximately circular annulus of outside diameter substending 21 min arc: the width was 2 min arc. The target was viewed monocularly. AI the beginning of each trial. the subject was instructed to fixate
1421
R. L. Hoc and D H. FENDER Clclssl$ctrrlo/l 2 3 P
lb
0
0
Fig. I. Target motions in (a) single-step presentation: (b) adjacent presentation: (c) opposite presentation. Only the heavily outlined circles are illuminated sequentially in each presentation. on a target presented at the center of the screen. The computer was programmed to determine that the observer was fixating the target. The computer would then move the target two degrees to one of the eight positions shown in Fig. I. The target might them be moved again after various time intervals. Three paradigms for target presentaiions were used:
(I) Single-step presentation. The target remained starionary after the first jump (Fig. la). This served as the control experiment. (2) Adjacent presentation. After the first step the target jumped again to one of the two adjacent vertices of the octagon and stayed there until the next trial (F’ig. I b). (3) Opposite presentation. After the first step the target jumped again to the vertex opposite to the position of the first target-step and stayed there until the next trial (Fig. Ic). For stimuli 2 and 3 the time interval between the first and the second target jumps could be set to 50, 100 or 150msec. All target positions. stimulus modes. and time randomly. Subjects were were presented intervals instructed to follow the target-steps as quickly as possible. Two subjects were used in this study. RESULTS
ASD
of.rclccdtlrc
rrspon~rs
Wheeles er Al. (19661 classified the response of saccadic eye movements to this tuo-step stimulus as “one-saccade” or V‘tvvo-saccade” depending on the nature of the motion of the visual axis. In one-saccade responses. the saccade moved the subject‘s fixation from the original gaze position to the final target position: the first target step was ignored. In two-saccads responses. the first saccade brought the subject’s gaze to the first target position then the second saccade moved the fixation to the final target position. However. in two-dimensional e>s movements the above dichotomy is not so clear cut. For example. in the adjacent presentation (stimulus 1). there were cases in which subjects responded ivith two saccades but both in the direction of the final position of the target. The second saccade in this case could be a correction saccade: but if only horizontal eye movements are studied and both target steps are in the same direction. such a response mould be mistakenly classified as a two-saccade response rvhereas it should be classified as a one-saccade response with a correction saccade. This error might become serious for those studies involving large saccades in which a high percentage of correction saccades are found. We therefore use the term “reprogrammed saccade response” to replace the “one-saccade response” for those responses in ahich the first target-position was ignored. Examples of eye movements in these various categories are shown in Fig. 3. In the cases when the saccadic s!stem responded to both target-steps, the magnitude of the first saccade \vas sometimes much smaller than the magnitude of the first target-step. This short-fall could be due to reprogramming efforts of the saccsdic programming system in response to new target information; for example, the first saccade might not be totall! reprogrammed, but partial reprogramming was still possible. It could also be argued that the reduced magnitude was not due to reprogramming but was origi-
Double-step
Target Movement
/
_____I !
DlSCt_SSlOS
Several commonly used terms in the following cussion are defined here (Fig. 2):
dis-
ISI (inter-stimuli interval) is the time interval between two target-steps. RT is the average reaction time of saccadic eye movements in response to a single step target jump. m is the average reaction time of a reprogrammed saccade (see below) measured from the beginning of the-- first target-step. RT, and RT, are average reaction times of the first and the second saccades in two-saccade responses (see below) measured from the beginning of the first target step. Latencies measured from the begi&ng of the second target jump are (m - ISI) and (RT, - ISI).
h----RTr------! Fig. 2. Definitions
of RT,. RT.. RT;. ISI.
:. a”
Processing of direction and magnitude
20
I
1423
3
20
;. .s”?
20
3
2
.?’ 3
I :I P
Fig. 3. Typical eye movement responses; left, adjacent presentations: right. opposite presentation. (a) Reprogrammed saccadic response. (b) Intermediate two-saccade response. (c) Full two-saccade response.
nally programmed that way, and normally would be followed by correction saccade if there were no second target-step. If this argument holds. one would expect about the same percentage of short saccades in the control experiments (responses lo single targetsteps). two-saccade responses and reprogrammed saccade responses. Figure 4 shows the distribution of the ratio of the magnitude of the first or the only saccade to the target-step magnitude for one subject in the three response categories. and Table I shows the per-
Table I. Percentage of first saccades of a two-saccade response having a magnitude less than 3,4 of the first target-step
Single target-step Reprogrammed saccade Two-saccade response
26.7 20.0 44.5
centage of the first saccades falling short by 25% or more. There is a significant increase in short first sac-
(c) Ftrsl
_I
(
Saccade
I” Two-saccode
Response
J-J-T1525m 0.25
0.50 0.75 MAGNITUDE
I.00
.
I.50
RATIO
Fig. 4. Distribution of the first-saccade magnitude ratio for various stimuli and responses.
cades in the two-saccade response, hence these can probably be explained as partially reprogrammed saccades. Figure 5 shows the frequency distribution of the differences between the direction of the first or the only saccade and the eccentricity of the target. All three plots show a similar distribution. thus it seems appropriate to classify the saccadic responses to twostep target movements into three classes instead of two. (I) Reprogrammed saccade response. The initial saccade is completely reprogrammed and the first target-step is ignored. The saccadic eye movements are directed toward the final target-step only (Fig. 3a). (2) Intermediate two-saccade response. The subject responds to both target-steps. but the magnitude of the initial saccade is less than 3,‘4 of the first target step. The second saccade brings his gaze to the final target position (Fig. 3b). (3) Full two-saccade response. Same as intermediate saccade response except that the magnitude of the first saccade is greater than 3/4 of the initial target step (Fig. 3~). The reprogrammed saccade response and the intermediate saccade response. represent two levels of
1121
R. L. HOL and
r
(a i Smqie-step Respoo$e I
FESDER
Qrammed rrsponses and of intermediate two-saccadr responses is the same for both classes of adjacent presentations. The probabilities of the three types of saccadic response for ditftrent time intervals between target-steps are given in Table 2.
P
0.4
0. H.
1
0.3
0.2 Reaction
TI\lE
OF REPROCRASI>lED
SACCIDES
0.1 ~ Oi 0.4
-48
~1
( b) ReDragrommed
-40
-32 -24
-16
DIRECTIONAL
Fig.
5.
-8
0
8
ERROR
16
24
32
40
48
(deqrees)
Distribution of the directional error of the various stimuli and responses.
first
saccade for
reprogr~~mmab~lity of the saccadic system. It appears that there is a state in the programming process after which the direction of the saccade cannot be changed but the magnitude can still be shortened in response to new target information. This strengthens the suggestion by Komoda YKul. (1973) that computation of direction and computation of magnitude are separate processes. The intermediate two-saccade response further suggests that. in some circumstances. the computation of direction is completed earlier than magnitude computation. A priori. there seem to be two different classes of adjacent presentations. One class comprises the changes from horizontal eccentricities to any other position. in which case we have a onemuscle-pair motion reprogrammed into a motion involving more than one muscle pair. The other class of adjacent presetitations would be the converse of this. However, we find that the probabilities of reproTable
2. Variation
of the probabifity
ISI (msec)
50
Reprogrammed saccade Intermediate two-saccade Full two-saccade
0.83 0.085 0.085
I5
Wheelrs or u/. (1966) used a target step combination similar to our opposite presentation but only in the horizontal direction and found that the average reaction time of reprogrammed saccades measured from the second target-step (q - ISI) was about 40 msec longer than the normal reaction time of saccades. They explained this extra 40 msec as the time required to cancel the programming of previous saccades. However. Komoda et ui. (1973) found no extra cancellation time. and if both horizontal target-steps were in the same direction. less time than the normal latent! was needed to reprogram a saccade. In our opposite presentation. the two target-steps were in completely opposite directions. To reprogram a saccade in this case. agonist and antagonist oculomotor muscles had to switch their roles. In the adjacent presentation the two target-steps were at a 45 angle, hence to reprogram a saccade. the new saccadic eye movements may require activity in additional sets of oculomotor muscles. The average reaction times of reprogrammed saccade for opposite and adjacent presentations are shown in Fig. 6 which shows (m - ISI) as a function of ISI. For the adjacent presentation. the average reprogramming time were about the same as normal saccadic reaction times for both subjects. For opposite presentations extra time was observed for both subjects: it was about 4Omsec more for one subject and 80 msec more for the other. The distributions of the reaction times for one subject are shown in Fig. 7. This curve is multimodal, strengthening the idea that reprogramming is the factor determining the reaction time. The fact that more than the normal programming time of a saccade is needed to reprogram a saccade in the opposite direction may have important implications in our understanding of the mechanisms of saccadic programming. One possible explanation for this extra time is that it is required to switch the role of agonist and antagonist in the central program. In other words. in order to program LLnew saccade in the opposite direction. the previous programming for the antagonist must be erased. This explanation implies that in these early stages of saccadic programming, the control programs for both agonistic and antagonistic oculomotor neurons already exist. It is
of each t)pe interval Subject MB loo O.-t5 0.35 0.30 204 msec
of saccadic
response
150
50
O.IZS 0.38 0.492
0.81 0.10 0.09
with
inter-stimuli
Subject RW 100 0.42 0.15 0.43 I96 m5ec
I50 0.05 0.1-t 0.71
112.5
Processing of direction and magnitude 400-
cn Adjacent'
0
0
200-
,;
ET
I
loo
I
I
1
150
100
50
ISX fmsec)
Fig. 6. Average reaction times of the reprogrammed saccade measured from the onset of the second target step. possible that central processes do not program both agonistic and antagonistic motorneurons individually, but the inhibitive signals to the antagonistic motorneurons individually, but the inhibitive signals to the antagonistic motomeurons are driven from the signals sent to agonistic motorneurons through “sign conversion” processes of interneurons. If this is true, there is no need to cancel the previous programming of the antagonist. in this event the extra time could be explained as the cancellation time for the previous program for the agonistic motorneurons. There are two plausible theories to explain the extra pro_framming time for opposite presentation. For simphcity, let us call them the “antogonist cancellation theory” and the “agonist cancellation theory”. To differentiate between these two theories, the following modified two-target-step experiment was performed. Only the top, bottom. left and right vertices of the octagon were used (Fig. 8). The first target-step moved to any one of the four vertices and the second target jump moved to an adjacent vertex. This target pattern will be referred to as “orthogonal presentation”. If the saccade is reprogrammed, a completely different pair of agonistic and antagonistic motorneurons will be involved. If the agonist motorneurons which are required for a saccade to the first target positions are not needed for the reprogrammed saccadie eye movement, the agonist theory would require the previous program to these motorneurons to be cancelled. and extra cancellation time should be observed. Conversely, for the new reprogrammed saccadic eye movements. there are no antagonistic
motomeutrons which have been previously programmed. therefore no cancellation is needed. In summary, if an increased reaction time is observed for the reprogrammed saccade, this would argue against the 3
9 0
0-l I
0
2
3
~l
0.3
the orthogonal presentation.
Reprogrammed
0.2
0.1 100
200
0
Fig. 8. Two examples of target movement sequences for
st Sacade (N=?O)
0.4
2
300 400 Time (msec)
500
Fig. 7. Distribution of reaction times for various target presentations.
I426
R. L. Hoc
and D. H. FESDER
antagonist cancellation theor). while an unchanged reaction time would argue against cancellation theorq. In this experiment the orthogonal presentation i+as randomly intermixed with opposite presentations. The single step presentations of the previous experiment were used as the control. The results showed about the same percentage of reprogrammed saccade responses as the previous experiment. The aberage reaction times of the reprogrammed saccades measured from the second target step (v - ISI) for both opposite and orthogonal target presentations were the same, about 230msec. The first point that can be made from these data is that the reaction time for orthogonal presentation is increased. thus the antagonist cancellation theory is not upheld. Secondly. the change in the reaction time for the opposite presentation is the same as that for the orthogonal presentation. hence we might assume that the same mechanism is operative in both cases. The extra time seems to be used to erase the previous direction program even though that particular muscle will not be activated in the new reprogrammed saccade. This may hint at the existence of a buffer which will hold the partial results of the directional computation. If the direction of the reprogrammed saccade is not similar to the previous one. the results of the unfinished computation in the buffer may have to be erased before the new results can be put in. If the direction of the reprogrammed saccade IS m a direction similar to that of the cancelled saccade, the results of the direction computation require onlj some modification. Hence if the reprogrammed sacsade is in exactly the same direction as the cancelled saccade, the computation time for the reprogrammed
saccade should be shorter. This hhpathesis seems to explain our results ,nd also the discrepant> betueen Wheeles t’r (il. (1966) and Komoda ;‘r ,tl. I 19731.
REFEREICES Becker W. and Fuchs 4. 119691 Furth-r propertIc 01’ the human succadic skstsm: sqe mowmsnts and correction saccades wifh and \*tthou! rwai ti\ltion points. L’isiotl Ret 9. 11-!7-!:jg. Becker W. and Jugens R. (1975) Saszsdic reactions to double-step stimuli: evidence for model feedback and continuous information uptake. In B.ii(c. .Mrclr~oti.wrs U/ IEdited Owlnr .Moriliry ml Tlrrir C/imd I q~i~uriou~ by Lennerstrand and Bach-y-Rital. p. 519. Pergamon Press. Oxford. Carlou T.. Del’Osso L. F.. Troost B. T. DarotT R. B. and Birkett J. E. (1975) Saccadic eye moicment latencies to multimodal stimuli. C’isrort Rrs. 15. l?i’-1262. Fsinstein R. and Williams W. J. (19711 Interactions of the horizontal and vertical human oculomotor systems: the saccadic systems. I’isiwr Rrs. 12. 33-4-4. Komoda M. K.. Festinger L., Philips L. J Duckman R. H. and Young R. A. (19731 Some obser\ations concerning saccadic eye movements. C’isicw Rrs. 11. 1009-1020. Lisberger S. G.. Fuchs A. F.. King W. \I. and Evinger L. C. I 1975) Effect of mean reaction time on saccadic resoonse to two-step stimuli uith horizontal and vertical components. b’i.sim Rrs. II. 10~1-1025. St Cyr G. J. and Fender D. H. (19691 Nonlinearrties oi human oculomotor system: gain. c’isiou Rfs. 9. 1135-1216. Westheimer G. (19511 Mechanism of tic;adic e)e movements. ;Irclzs OplttM. 52. 7 10-721. Wheeles L. L.. Jr.. Bo>nton R. &I. and Cohen G. H. 1966. Eye movements responses to step and pulse-step stlmult.
J. opt. SOC. .hr. 56. 9X-960.
