| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
The Journal of Neurophysiology Vol. 80 No. 2 August 1998, pp. 849-862
Copyright ©1998 by the American Physiological Society
Department of Medical Physics and Biophysics, University of Nijmegen, 6500 HB Nijmegen, The Netherlands
 |
ABSTRACT |
|---|
 Abstract
 Introduction
Methods
Results
Discussion
References
|
|---|
Chaturvedi, Vivek and Jan A. M. Van Gisbergen. Shared target selection for combined version-vergence eye movements. J. Neurophysiol. 80: 849-862, 1998. Primates frequently make rapid binocular eye movements to reorient gaze in both direction and depth. To explain the unequal movements made by the two eyes, it often is assumed that they result from the combined action of a conjugate saccadic system and a vergence contribution. Clearly such a scheme can only yield coordinated binocular movements if both systems are guided by a shared or coupled target selection mechanism. To investigate the degree of cooperation at this level, we studied binocular refixations to target-nontarget double-stimuli in three-dimensional (3-D) space. Binocular eye movements were recorded in seven subjects using the scleral coil technique. In the experiments, 20% of trials were composed of a green target and a red nontarget, presented at the same time, but at different locations in 3-D space. These were alternated randomly with single-target trials (80%) in which the green stimulus was presented randomly at one of eight possible positions in 3-D space. Instructions to the subject emphasized either the speed or accuracy of response. Our findings show that typical features of the saccadic response to double-stimuli (bistability, averaging, and a speed-accuracy trade-off), as found in earlier two-dimensional studies, are also prevalent for initial binocular refixations to double stimuli in 3-D space. When the first saccadic response is directed to one of the two stimuli, the vergence system almost invariably makes the same choice. Likewise, when the saccadic system makes a short-latency averaging response, the vergence system shows a similar compromise. Statistical analysis shows a high correlation between saccadic and vergence target selection, strongly suggesting that the amplitude computation process of both subsystems is due to a common target selection stage that has access to information about stimulus location in 3-D space.
This paper deals with the question of how the saccadic and vergence systems work together during coordinated binocular gaze-shifts in three-dimensional (3-D) visual space. There are often many potential targets in our immediate environment, and, accordingly, a choice must be made by each oculomotor subsystem to achieve proper binocular foveation. Logically, the question arises as to which neural mechanisms are involved in ensuring that the saccadic and the vergence system both select the same target. To investigate this problem, we presented subjects with a target/nontarget task whereby a number of combined saccade-vergence responses were elicited. By analyzing the extent to which correct and incorrect stimulus choices in the two subsystems were correlated, we could study the coupling of the target selection mechanisms. Earlier studies on saccade-vergence interactions, reviewed further on, have generally concentrated on the response dynamics, and this has led to various interpretations about how these systems coordinate the execution of binocular eye movements.
Saccade-vergence interactions
Gaze-shifts in a natural environment generally require binocular refixations that have combined directional and depth components. In such responses, both the saccadic and the vergence subsystems are invoked to ensure that the respective lines of sight of the two eyes ultimately will intersect on the target of interest. It has become evident that, when tested in isolation, these two systems exhibit markedly contrasting dynamic characteristics (see for example: Collewijn et al. 1995 Target-selection studies
Target selection has, until now, has been studied in detail for the saccadic system. A number of frontal-plane double-stimuli studies (Becker and Jurgens 1979 Eye movement recording
By placing subjects in the center of two perpendicular rapidly alternating magnetic fields, we could record binocular eye position using the magnetic induction search coil technique (Collewijn et al. 1975 Calibration of binocular signals
During calibration the subject was instructed to fixate light-emitting diodes (LEDs) on a frontoparallel plane, placed at a distance of 125 cm from the subject. Thirteen horizontal and 13 vertical LEDs were positioned at different eccentricities (every 5°), from the straight-ahead direction, spanning a range of ±30°. The procedure was performed separately for each eye, and the center LED was aligned with the eye that was being calibrated. Both sets of recorded raw data were subjected to an off-line sinusoidal regression fit procedure to calibrate these signals. The horizontal angles of version (direction of the cyclopean eye) and vergence (angle between the lines of sight of the 2 eyes) were defined as being, respectively, the mean and difference of the two azimuth components. Interocular separation was assumed to be 6.5 cm.
3-D stimulus array and experimental design
After the coil calibration, a 3-D (horizontal, vertical, and depth) LED array (see Fig. 1) was mounted in front of the subject. The LEDs (5 mm diam) could be activated to be either green, to indicate the target, or red, to represent the nontarget. The position of the LEDs in the 3-D array was such that they required a horizontal displacement of 20° (either leftward or rightward), a vertical displacement of 10° (either upward or downward) and a vergence change of 2.5° (to either a near or far surface), all with respect to the central fixation point (aligned with the cyclopean eye and having a target vergence angle of 7.5°). All the LEDs in the array occupied positions on toroidal iso-vergent surfaces (Collewijn et al. 1988
Data analysis
The velocity of the eye movements was calculated by differentiating the position signals. After filtering with a 33-point, 75-Hz Rabiner digital filter, a threshold level (30°/s) was applied to the resulting velocity signal to detect saccades automatically; the results were always checked by visual inspection. For reasons that will be explained later, we concentrated the analysis on the binocular eye movement that occurred between the onset of the first and the onset of the second saccade (1st movement for short). Thus first movement onset was taken as the start of the primary saccade, and the end of the first movement was defined to coincide with the onset of the second (i.e., correction) saccade. If there was no correction saccade, or if the correction was delayed by >500 ms, then the first movement offset point was taken at 200 ms (correction saccades for single and double-stimuli typically occurred 150-200 ms after the first saccade) after primary saccade offset. The gain of the initial eye movement was defined to be the ratio of actual eye displacement during the first movement to the total eye displacement required for a precise movement to the target. Accordingly, a correct movement had a gain of +1, whereas a response that was directed accurately to the nontarget had a gain of Subjects
Experiments were carried out with seven male volunteers (AB, BW, JGO, HM, JVG, PH, and VC). Their ages ranged from 23 to 53 yr, and none had any known neurological or oculomotor disorders. Four subjects (JGO, HM, PH, and VC) wore corrective lenses during the experiment. Two subjects (JVG and VC) were familiar with the purpose of this study, whereas all the other subjects were kept naive. Before application of the coils, the surface of the eye was anesthetized with two drops of a local anesthetic (0.4% Novesine). Subjects were seated in a comfortable upright position while their heads were immobilized using a dental impression bitebar. All stimuli were viewed binocularly except during calibration. Experimental sessions generally lasted Our experiments were designed to investigate the target selection response for short-latency binocular refixations in 3-D visual space, with each response consisting of horizontal and vertical saccade components as well as a vergence component. To properly assess target selection for each subsystem during the double-stimulus task, it was essential to first analyze control responses to a single target.
Control responses
In the control trials, each subject had to move the binocular point of fixation to one of eight possible single green targets, at the corners of a 3-D LED array (see METHODS). Each response trajectory started from the same fixation point in the center of this array. These eye movements always required a horizontal saccade (either rightward or leftward) with an up or down vertical component, which had to be combined with a divergence or a convergence movement depending on whether the target appeared at the far or the near surface, respectively.
Normalization of the data
Because we wished to pool the results of the various types of trials (i.e., to targets that were positioned left/right, up/down, near/far), we standardized the data by computing gain values for both saccadic and vergence components of each first movement. Gain values were obtained by taking the ratio of the amplitude of the actual eye displacement to the required amplitude. These values, for saccades, were based on the vertical component because the required horizontal component, which was the same for target and nontarget, would not discriminate correct from incorrect responses. An accurately directed response of correct amplitude would have a gain of unity, whereas a wholly incorrect response, directed toward the nontarget, would have a negative gain of one. Note that the latter property of the gain measure holds as a consequence of the fact that target and nontarget always were presented on opposite corners of the 3-D LED array (i.e., corresponding to opposite version and vergence components). Not surprisingly, the pooled control responses of the saccadic and the vergence component, at the end of first movement, had gain values relatively close to unity (see Fig. 6). To characterize this variability in the control responses, we took the mean of all the values and computed ±2 SD on each side. In evaluating the responses to double stimuli (see further), the resulting boundaries will be used to characterize responses as correct or incorrect. To provide an impression of the relative contribution of each portion of the total first movement, the intrasaccadic and end of first movement gain values (and respective standard deviations) are shown in Table 1. As can be seen, the gain of the vergence response increases considerably during the postsaccadic episode. For the version response, the postsaccadic contribution, probably partly due to postsaccadic drift, is almost negligible.
Target/nontarget responses
To provide an impression of the wide range of possible responses to the target/nontarget double-stimulus task in 3-D space, we have selected three different trial responses to a stimulus configuration where the target was located downward and nearby, whereas the nontarget stimulus was upward and far away. The saccadic and vergence time profiles of each response are shown in Fig. 7, A-C. Again, the interpretation of the saccadic components is quite straightforward. Response 1 is typical of a wholly incorrect initial response that is later followed by an opposite corrective saccade. In contrast, response 3 is directly on target. Response 2 is an example of saccade averaging. It should be noted that the horizontal component of these three responses was roughly comparable. As stated earlier, we have focused our efforts on the description of the vertical saccadic component because this is most indicative of the target-selection process. In comparison with the saccade response, the occurrence of transient divergence and the absence of clear-cut fixation periods tend to make the vergence response picture somewhat more complex to analyze. Nevertheless, it can be seen clearly, in Fig. 7C, that this subsystem has made stimulus choices that are very similar to those made by the saccadic system. For example, in response 3, the vergence system moves directly to the target too. In the trial where the saccade is wrongly directed (response 1), the vergence response also initially heads to the far nontarget before correcting itself. Interestingly enough, when the saccade makes an averaging response, then so does the vergence response (response 2). To study the suggestion emanating from these examples, that the target selections in the two subsystems are linked, we analyzed to what extent the gain values of saccadic and vergence responses were correlated.
Test for system independence
In an effort to obtain an impression of the degree of association that each subsystem has with the other during the target/nontarget responses, we compared the two sets of gain values at the end of the first movement (as plotted in Figs. 8 and 9), using the statistical
Corrective binocular eye movements after erroneous initial responses
As explained above, the analysis of binocular responses in this study leans heavily on what we have called the first movement. The time limits that were used, first saccade onset to second saccade onset, require justification. The onset criterion was based on the strong impression, obtained from the recorded data, that the saccade and the vergence movement start virtually simultaneously. Cases where the movement in depth starts slightly earlier, with a slow drift, were sometimes noted, but the amplitude of these prelude vergence movements was always small (Chaturvedi et al. 1997
Speed-accuracy relationship
For fast refixations, saccade and vergence target/nontarget latencies did not show an obvious relation to the gain value of the refixation observed at the end of the first movement. By including experiments where the instruction emphasized accurate rather than fast refixation, which provided responses with a much wider range of latencies, we found that refixations in 3-D visual space have a similar speed-accuracy trade-off as was previously found for the frontal plane. This was done for two subjects (JVG and VC). Figure 11 shows the saccade (A) and vergence (B) gain values for both short- and long-latency responses. Incorrect and compromised responses are generally only prevalent in a limited time frame for the "fast" refixations. When the instruction during the experiment urged an accurate refixation of the target without first going to the wrong stimulus, then the latencies became significantly longer, the responses became less error-prone, and the gain values of both saccade and vergence eye movements were predominantly around unity.
Target selection in 3-D visual space
This study was designed specifically to investigate how target selection mechanisms can ensure unified action when the oculomotor system has to move in both direction and depth. It is widely believed that executing these refixations involves distinct oculomotor subsystems. There have been a number of studies that have concerned themselves with target selection in either the saccadic or the vergence system, but the problem of how the oculomotor system might handle the issue of target selection in combined saccade-vergence tasks has not been studied before.
Relation to earlier target selection studies
Although the target selection process of the saccadic system has been investigated extensively, relatively little is known about how the vergence system chooses its targets, and the question of how the two subsystems cooperate in target selection when presented with multiple stimuli has not been studied systematically at all.
Neural mechanisms of target selection
A fundamental problem in sensorimotor control concerns the neural mechanisms that are involved in the processes of visual target selection and the ensuing guided motor actions. A number of electrophysiological studies have used multiple stimulus paradigms in the frontal plane to investigate target identification and movement response selection before the generation of saccadic eye movements and to elucidate which neural areas are involved.
Neural correlates of the averaging effect
An interesting result of our study is that the vergence system was seen to make averaging responses when the saccadic system does likewise. This new information puts the issue of how the averaging phenomenon arises in a different perspective, in as far as averaging previously was considered to be a purely saccadic phenomenon.
Modeling aspects
Models describing saccade-vergence interactions generally do not address the issue of target selection. They do suggest, however, that there may be coupling in the timing (WHEN system) of the two systems, through the activity of omnipause neurons (Mays and Gamlin 1995 Conclusions
This study has shown that the saccadic and vergence system work together closely in selecting targets and making appropriate responses when stimuli are presented in 3-D visual space. This clearly is reflected in the instances when the saccadic system makes incorrectly directed or compromised responses. In each case, the vergence system makes a similarly directed response. This strong coupling in target selection by the two oculomotor subsystems is most parsimoniously explained by assuming a common selection system operating at a level where 3-D information is jointly available.
INTRODUCTION
Abstract
Introduction
Methods
Results
Discussion
References
; Erkelens et al. 1989; Yarbus 1967); saccades, responsible for the control of rapid changes of gaze in direction (version), are observed to be fast and conjugate (the eyes move equally in the same direction), whereas vergence responses (which rotate the eyes by equal amounts in opposite directions), required as a result of changing target distance, are disconjugate and relatively slow. These differences have led to the suggestion that conjugate and disconjugate eye movements use different neurophysiological substrates (Leigh and Zee 1983). The hypothesis of separate neural control systems has gained support with the discovery of vergence-related neurons in the midbrain (Mays et al. 1986).
suggested that a binocular gaze shift could be described as a simple linear summation of two distinct processes. It has become apparent from later work that, during binocular refixations in 3-D space, there is actually a degree of interaction between the two subsystems and that the ensuing response is too complex to be resolved algebraically into a fast conjugate saccadic and a slow vergence component (Kenyon et al. 1980; Ono et al. 1978). In this situation, the disconjugate response is observed to be substantially faster than during a pure vergence response (Enright 1984, 1986; Erkelens et al. 1989; Maxwell and King 1992; Oohira 1993). Two hypotheses have been put forward, thus far, to explain this.
). This hypothesis has received extra attention since a recent neurophysiological study (Zhou and King 1997) has shown that binocular eye movements may be under partial monocular and partial binocular control. It should be noted that the saccadic-depth contribution, proposed by this scheme, is incomplete in the majority of responses, so that an additional contribution of the slow vergence system is still needed. Furthermore, it cannot be excluded that this vergence contribution, clearly in evidence after the saccade, also may be present during the saccade. Another interpretation is that the vergence system is facilitated when a combined saccade-vergence response is elicited. Zee et al. (1992) recently proposed a model to account for this vergence facilitation. The model is composed of two parallel systems that are coupled through a shared initiation system. To explain the nonlinear interaction, it has been suggested that vergence burst-neuron activity is modulated by omnipause neurons during saccadic activity (Mays and Gamlin 1995, 1996; Zee et al. 1992).
; Coren and Hoenig 1972; Findlay 1980, 1982; Ottes et al. 1984, 1985) have shown that the initial response of the saccadic system, depending on stimuli separation, can be to choose one of the two stimuli (bistability) or to direct the first saccade to a point between two targets (averaging). These experiments also showed that a speed-accuracy trade-off occurs whereby short-latency saccadic responses become increasingly error prone. Comparatively little work has been done regarding vergence target selection, but it is known that the vergence system can choose a target from several alternatives (Erkelens and Collewijn 1991).
METHODS
Abstract
Introduction
Methods
Results
Discussion
References
; Robinson 1963). We used phase-sensitive detection techniques to demodulate the coil signals. The eye position data then were filtered, sampled, and stored on disk for further analysis (for details, see Chaturvedi and Van Gisbergen 1997).
; Koenderink 1992).
View larger version (44K):
[in a new window]
FIG. 1.
View of light-emitting diode (LED) array used for 3-dimensional (3-D) target-selection task. Each trial began with the subject attending the green fixation point F, which was aligned with the cyclopean eye and had a target vergence of 7.5°. In all trials, the green target (T) could appear at any 1 of the 8 indicated corner locations: requiring a horizontal displacement of 20°, a vertical displacement of 10°, and a vergence change of 2.5°, with respect to the fixation point. In double-stimulus trials (20%), the red nontarget (NT) appeared, simultaneously with target T but at a location that was diametrically opposite in vertical direction and depth; the horizontal component was always kept the same. In the double-stimulus trial shown, T is right, up, and far away, whereas NT is right, down, and nearby. The eyes are denoted by L and R. Interocular distance (L-R) always was assumed to be 6.5 cm.
View larger version (14K):
[in a new window]
FIG. 2.
Some theoretically possible saccade and vergence responses (seen from behind and from the side) for the double-stimulus configuration, depicted in Fig. 1. Extremes in potential behavior are shown to illustrate the concepts of concordant and discordant response modes. Concordant responses of both subsystems are shown in bold: the bold arrow from F to T denotes a correct response to the target location, whereas the dashed bold arrow is an incorrect response to the nontarget. Discordant responses (thin arrows) would occur if the saccadic and the vergence systems were to choose different stimuli (1 to T the other to NT). Because compromise responses are not shown, the set of examples is not exhaustive.
1. By using this normalized gain measure, it was possible to pool responses to stimuli at all eight locations. In the analysis, we concentrated on the vertical component of the saccadic response (the horizontal component was typically an accurate leftward/rightward movement) and the vergence response. Trials with a wrong fixation at the start of a trial (with a discrepancy of 
5° in the version angle or 1° in the vergence angle) and those having a predictive response (latency <75 ms) were excluded from the analysis.
40 min. Subjects were instructed to fixate the green target as quickly as possible (unless specified otherwise) but to refrain from making anticipatory responses.
RESULTS
Abstract
Introduction
Methods
Results
Discussion
References
, 1997; Oohira 1993; Zee et al. 1992). For this reason, this phase of the response was not considered to be a true reflection of the target-selection system, and therefore we did not analyze it as such. Its presence, however, had clear consequences on how the eyes completed their depth movement. The corresponding trajectories in oculomotor space of the typical trials shown as distinct (vertical) saccade and vergence components in Fig. 3, B and C, respectively, are depicted in side view in Fig. 4. These figures plainly illustrate that, because of the effect of transient divergence, the first part of the vergence response may look similar for both converging and diverging responses. As a consequence of this phenomenon, convergence movements needed a more sizeable postsaccadic vergence contribution than divergent responses. In the latter case, the eyes were already often close to target by the end of the first saccade.
View larger version (23K):
[in a new window]
FIG. 3.
Binocular eye movement responses of subject JVG to each of the 4 rightward targets in the 3-D LED array shown in Fig. 1. Shown are the horizontal version component (A), the vertical version component (B), and the vergence component (C). Response 1 was to the up-far target, response 2 was made to the down-far target, response 3 was directed to the up-near target, and response 4 was the eye movement to the target at the down-near corner. Rightward, upward, and convergence movements are depicted as positive deflections here and in all following figures. Note the occurrence of facilitated vergence during the saccade, and of a slow vergence phase after the saccade is completed. Transient divergence causes the eyes to initially move in the wrong direction for responses requiring convergence.
View larger version (11K):
[in a new window]
FIG. 4.
Vertical version component and the vergence responses, from Fig. 3, have been plotted against each other to give an impression of the single-target control trajectories, in 3-D oculomotor space. Fixation point (F) was always the same. Responses 1 (up) and 2 (down) are directed to the far targets, whereas responses 3 (up) and 4 (down) are directed toward the near targets. Slight quantitative discrepancies observed in vergence angle fixations and vertical displacements are due to small calibration errors, differences in (subject) interocular separation, and slight head misalignments (with respect to the plane of regard). Sample interval was 2 ms. 
, target positions recorded when the subject was required to fixate the LEDs during calibration.
View larger version (11K):
[in a new window]
FIG. 5.
Control response endpoints taken at the end of the 1st movement (i.e., onset of 2nd saccade). In this sideview plot (cf. Fig. 4), all controls on the right and the left side have been pooled. Subject JVG.
View larger version (13K):
[in a new window]
FIG. 6.
Gain values (ratio of actual eye displacement to desired amplitude) of controls at the end of 1st movement. Correct and accurate responses should have a gain of unity. Note that both saccade and vergence responses have gains scattering around this value. Limits, taken to be twice the standard deviation values for both sets of gain components, give an indication of the variability of both the saccade and vergence responses for single-target control responses. These boundaries have been used to categorise double-stimulus responses as wholly correct or incorrect. An example of a vergence gain larger than +1 can be seen in response 1 of Fig. 4. Subject JVG.
View this table:
TABLE 1.
Gain values of the vergence and version control responses
View larger version (22K):
[in a new window]
FIG. 7.
Selected examples of responses from subject JVG to a double stimulus pair where the target was located nearby, downward, and to the left while the nontarget was presented faraway, upward, and also on the left. Same conventions have been used as in Fig. 3 (i.e., A: horizontal version; B: vertical version; C: vergence). Interpretation of the responses can be most easily made on the basis of the vertical saccade component. Response 1 is initially wrongly directed and is followed by a large secondary saccade to the green target stimulus. Response 3 is an example of a correct response. Finally, response 2 shows averaging in the vertical saccade component and subsequently makes a corrective saccade. Vergence component shows a striking resemblance to the saccade counterpart. Response 1 is similarly wrong and makes a diverging response to the far nontarget. This error is not corrected straightaway but is delayed until the later corrective saccade occurs. Response 3 heads to the correct near target. Response 2 shows an averaging vergence response that coincides with the compromised saccade response.
1, respectively, this type of data presentation enables one to see, at once, whether the saccadic and vergence systems can show discordant behavior by directing their initial responses to different targets. This is, subsequently, seen not to be the case. Note that responses exhibiting opposite gain values (e.g., 1 and +1) for each subsystem were not observed. This hypothetical combination (where the systems would make diametrically opposing responses) would have meant that the binocular point of fixation would end up at a location where there was no stimulus at all (see Fig. 2).
View larger version (20K):
[in a new window]
FIG. 8.
Gain plot with all the data points from the target/nontarget paradigm showing a significant number of incorrectly directed and compromising responses for both saccades and vergence. Note that the saccade response is coupled closely to the vergence response. There are no discordant responses that exhibit near-unity gain values with opposite sign for saccade and vergence components. Limits (twice standard deviation), as defined in the controls of Fig. 6, have been drawn around both the correct (+1) and incorrect ( 1) response values. Two oblique lines joining these limits denote the variability of the control refixations onto that of the double-stimulus responses. Most of the latter data points fall within these limits of variability, indicating that similar deviations from perfectly concordant behavior (equal gains in both subsystems) would occur in single target controls along the line from target to nontarget. Numbers in the plot refer to the three responses depicted previously in Fig. 7. Converging and diverging responses have a fairly even distribution of correct, incorrect and averaging responses. Horizontal component of these data points is always directed correctly and constant in size, with an approximate gain value around + 1. Subject JVG.
), shows that there were some trials where the vergence refixation was correct but the saccade was clearly misdirected. Although such examples of discordant behavior were atypical and rare (we saw a total of only 7 such examples among all the subjects that were studied), they are interesting as illustrations that independent target selection is in fact possible.
View larger version (32K):
[in a new window]
FIG. 9.
Gain plots showing target/nontarget paradigm results in four other subjects. A: gain response plot of subject VC shows predominantly correct refixations along with a number of incorrect and compromised responses. B: subject PH, similarly, makes mostly target-directed responses. C: target selection behavior of subject JGO is bistable in nature. Responses are either entirely correct or entirely incorrect. There are no averaging responses at all. D: subject AB has a wide variability in responses. Note also a few exceptional discordant responses where the vergence gain is close to unity and the version gain is almost 1.
2 test of independence. This test provides a measure of the discrepancy between the observed (i.e., recorded) values and the values expected on the basis of independence. By grouping the two variables under consideration, version and vergence gain, for each individual subject, we computed a 2 value that indicated how closely the two subsystem responses were related to each other during a double-stimulus task.
1 are bounded, on either side, by confidence limits (described earlier). Because the majority of the responses occur within these criteria, the occasional aberrant datapoint that lies beyond these limits is neglected. This enables us to define three distinct rows and three distinct columns of data points, in a contingency table for version and vergence gain, where the different types of responses are paired together. For example, in the top right cell, one can find 13 responses that are considered to have approximately unity gain (+1) for both version and vergence; in the middle right cell, there are only four responses where the vergence is unity and the version has gain values between +1 and 1; in the bottom right cell, there are no responses where version has a negative gain with values around 1 while vergence is still at unity gain. Using the marginal totals of these grouped values, we computed the cell frequencies that one should expect to obtain if the two variables are independent of each other. For example, the value calculated for the top right cell denotes the expected number of responses that will have both a version and vergence component close to unity gain. The difference between each observed and expected number is squared and divided by the expected value. These values are summed over all the cells to obtain the 2 value. The 2 value was significant (P < 0.01) for all subjects, indicating that the null hypothesis, that no difference exists between the observed and expected values, could be rejected (see Table 2). Thus the version and vergence gain responses were not deemed to be independent of each other during a target selection task in 3-D space.
View this table:
TABLE 2.
Statistical evaluation of the target/nontarget responses
). The offset criterion obviously had to be applied somewhere between first saccade offset and second saccade onset. We chose the latter temporal boundary to be sure that the postsaccadic vergence contribution was included.
0.5) and plotted their time courses as gain values. Accordingly, their final endpoints would be expected ultimately to attain a positive gain of +1, the starting points would have a gain of zero, and the initial incorrect responses would betray themselves by having negative gains. The incorrect responses to the far nontarget, in the case of near target presentation (Fig. 10, A and B), show how the corrective vergence response, after the initial wrong movement, is linked closely to the corrective saccade and does not start its corrective procedure prior to the saccadic component. Thus it is as if the system needs time to select a new target, which then is used to launch both systems into a new coordinated movement toward the new goal just as was previously observed in the first movement. This is seen more strikingly for incorrect responses to a near nontarget, when a far target was presented (Fig. 10, C and D). In this case, the responses have a markedly large initial transient divergence response, which thereby ensures that the entire wrong vergence movement is not fully completed by the time the corrective saccade is started. It is interesting to see that, in these cases, the correct vergence response clearly is initiated along with the corrective saccadic movement. Because similar results were seen for all subjects, we believe that the end of the first movement is a justifiable temporal measure when analyzing these combined saccade-vergence refixations.
View larger version (21K):
[in a new window]
View larger version (26K):
[in a new window]
FIG. 10.
Time course of corrective vergence movements to a near target (A and B) and a far target (C and D). Subject JVG. A: correction of initially wholly incorrect movements to a near target in the target/nontarget paradigm. Gain time course of the vertical version component showing erroneous 1st saccades to the far nontarget followed, after a fixation period, by a large correction saccade to the near target. B: gain time course of the vergence component. As in the saccade, the subject's initial response is to the nontarget. Note how the vergence correction is closely linked to the saccade correction, showing no consistent signs of movement from the nontarget in the fixation period before the corrective saccade. C: correction of initially incorrect movements (directed to a near nontarget 1st) to a far target in a target/nontarget paradigm. Gain time course traces for the vertical saccade responses. D: gain time course of the vergence component. In all traces, the transient divergence is very prominent initially but is then overtaken by vergence in the opposite direction to the near nontarget. Note especially that the incorrect vergence response persists throughout the entire 1st movement. Correction of the vergence response coincides with the 2nd saccade, directed toward the far target.
View larger version (18K):
[in a new window]
FIG. 11.
Gain responses of 2 double-stimuli experiments, with different instructions, for subject JVG, are plotted here with respect to their latencies. A: saccadic onset was detected automatically by using a threshold value of 30°/s. This then was further checked by visual inspection. , taken from the experiment with the "fast" instruction, show that short-latency responses can have correct, averaging and incorrect endpoints of the first movement. , taken from an experiment with an "accurate" instruction, show that longer-latency responses are less error-prone and almost always have unity gain. B: vergence gain responses for the same trials.
DISCUSSION
Abstract
Introduction
Methods
Results
Discussion
References
; Coren and Hoenig 1972; Findlay 1980, 1982; Ottes et al. 1984, 1985). For large target separations, the saccadic system exhibits a bistable response, with saccades to either one stimulus or the other. When this separation is reduced, averaging saccades (directed to a location between the 2 stimuli) typically are found to be interspersed with responses near either stimulus. In other words, the initial saccadic response is not necessarily either correctly or wrongly directed but often lies in between the two stimuli.
). Erkelens and Collewijn (1991) have demonstrated that subjects can choose a small peripheral target, fixate it binocularly, and perceive it in binocular fusion, while disregarding other larger, more foveally located targets. Recently, Mallot et al. (1996) obtained indirect evidence (eye movements were not actually measured) that disparity-evoked vergence can be directed toward an average depth location. Their experiments, using stereograms with small disparity differences, were different from our experiments where the vergence response exhibited averaging-type behavior during large refixations to real targets.
had the novel idea to look at the averaging phenomenon from a binocular point of view. To this end, they studied binocular responses using a setup where refixations to a single target were elicited by changes in disparity. Their dichoptic study, which differs from our situation with real targets, specifically asked how the saccade was related to motor error in each eye. If fast eye movements had to bring the eyes directly on target (placed in the same hemifield of each eye), then unequal saccades in the two eyes would be necessary. Rather than seeing unequal saccades, Findlay and Harris observed that both eyes made the same saccade, aligning the cyclopean eye with the direction of the new target such that a correct refixation and, subsequently, binocular fusion of the target could be achieved through a follow-up pure vergence eye movement. This effectively means that the saccade coincided with the average of the two motor errors. This cyclopean saccade, therefore, did not contain any depth component and was purely conjugate. In our experiments, we have evoked comparable eye movements to real single targets at near and far locations. These responses, certainly those to distant targets, typically were composed of disconjugate saccades (see Fig. 3), which traversed a large part of the required depth during their trajectories, followed by final pure vergence movements. Similar findings on saccadic disconjugacy under these conditions have been reported extensively in the literature (Enright 1984, 1986; Erkelens et al. 1989; Maxwell and King 1992; Oohira 1993). It must be concluded that the disparity-driven "dichoptic" averaging effect, described by Findlay and Harris (1993), does not occur under our experimental conditions and pertains to a different issue than the averaging phenomenon observed in our study. The latter belongs to a different behavioral category that requires the use of multiple stimuli. Our main experiment, using the target/nontarget discrimination task, shows that binocular averaging in 3-D is just a generalization of saccadic averaging in the frontal plane. In both cases, the binocular point of fixation ends up at an intermediate position in 3-D space in between target and nontarget.
), that short-latency responses were unavoidably error prone. When the instruction to the subject was varied, by emphasizing either speed or accuracy of response, it was seen that errors could only be avoided if the saccadic response was delayed. Although it is reasonable to expect that this also will hold qualitatively for the present combined saccade-vergence refixations, it is conceivable, in theory, that the processing of depth and direction information might have different temporal characteristics causing the saccadic and vergence speed-accuracy trade-offs to be different. To explore this, we extended our experiments to incorporate different instructions to test for this possible latency dependence. The results show that the notion of a speed-accuracy trade-off is equally valid for 3-D binocular refixations; this explains the occurrence of wrongly directed responses at short latencies. Quantitatively, the improvement in accuracy with latency seems quite comparable for the saccade and vergence components. It can be noted that the shortest latencies of the eye movements shown in our speed-accuracy trade-off plot (see Fig. 11) are longer than those in similar experiments done by Ottes and coworkers (1984, 1985). This may be related to the findings of Honda and Findlay (1992), who noted that during binocular viewing, saccades to targets in a different depth plane showed increased latencies.
; Schall and Hanes 1993; Schall et al. 1995) has demonstrated that the frontal eye fields (FEF), known to be responsible for the generation of goal-directed eye movements, may be highly involved in the process of saccade target selection. The FEF are known to have reciprocal connections to parietal areas in the so-called "dorsal" pathway, which is associated with identifying the spatial location (WHERE) of items in the visual scene. Other electrophysiological studies have investigated target selection in the superior colliculus (Basso and Wurtz 1997; Glimcher and Sparks 1992; Ottes et al. 1987). It appears that early neural activity occurs shortly after information regarding the metrics of the target becomes available and that this activity is modulated by target uncertainty. It cannot be excluded that the observed activity is a reflection of the response selection process already observed in FEF.
has provided evidence that saccade targets and visual distractors are represented by different levels of neural activity in the lateral intraparietal area (LIP) of the monkey. Area LIP would be an extremely interesting area to search for neural correlates of combined saccade and vergence target selection by virtue of the fact that this area was found to contain neurons with broad response fields that cover a 3-D volume of space with tuning curves at different preferred depths (Gnadt and Mays 1995). This type of neuron may code signals that are related to eye movements in direction (frontoparallel plane) and depth (distance from plane of fixation).
; Van Opstal and Van Gisbergen 1990) have shown that the superior colliculus is involved in a saccadic averaging response. On the basis of their findings, both studies proposed that the neural correlate for the averaging process can be found either at the level of the superior colliculus or at a location further upstream. The data presented in the aforementioned studies, along with the results obtained in our experiments, are compatible with the idea of a linked saccade-vergence target selection process that will now be discussed.
, 1996; Zee et al. 1992).
, who found that oculomotor signals at the premotor level encode monocular saccadic eye velocity, indicate that the idea of a binocular version and vergence system probably is oversimplified. Although the study by Zhou and King (1997) provides a new view on saccade-vergence cooperation (see INTRODUCTION), refixations in 3-D still would require the cooperation of two systems: a fast system responsible for the change in direction and for part of the movement in depth and a slow postsaccadic vergence movement. As far as we know, there does not appear to be any studies that suggest that the slow postsaccadic and the fast intrasaccadic eye movement are generated by the same system. This, therefore, leads to a coordination problem, where both systems need to track the same target. Accordingly, one needs to explain how the slow vergence responses, which follow the initial saccade and appear to be continuations of the fast movement (see Fig. 10), arise. If the two systems always operate sequentially, the faster system theoretically could pass on its target selection information to the slower one, but it appears that a slow vergence movement actually may precede the saccade (Chaturvedi et al. 1997; Collewijn et al. 1997).
| |
ACKNOWLEDGEMENTS |
|---|
We thank A. van Beuzekom, S. Gielen, P. Hofman, P. Medendorp, and H. Misslisch for useful comments. We gratefully acknowledge the experimental and technical assistance of K. Hol, T. van Dreumel, H. Kleijnen, G. van Lingen, and W. Haerkens.
This study was supported by the Foundation for Life Sciences (SLW) and the Netherlands Organization for Scientific Research (NWO).
| |
FOOTNOTES |
|---|
Address for reprint requests: V. Chaturvedi, Dept. of Medical Physics and Biophysics, University of Nijmegen, P.O. Box 9101, 6500 HB Nijmegen, The Netherlands.
Received 30 December 1997; accepted in final form 3 April 1998.
| |
REFERENCES |
|---|
|
Abstract
Introduction
Methods
Results
Discussion
References
|
|---|
This article has been cited by other articles:
|
|
 |
|
  Q. Yang, M. P. Bucci, and Z. Kapoula The Latency of Saccades, Vergence, and Combined Eye Movements in Children and in Adults Invest. Ophthalmol. Vis. Sci., September 1, 2002; 43(9): 2939 - 2949. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Ferraina, M. Pare, and R. H. Wurtz Comparison of Cortico-Cortical and Cortico-Collicular Signals for the Generation of Saccadic Eye Movements J Neurophysiol, February 1, 2002; 87(2): 845 - 858. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Ferraina, M. Pare, and R. H. Wurtz Disparity Sensitivity of Frontal Eye Field Neurons J Neurophysiol, January 1, 2000; 83(1): 625 - 629. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. Chaturvedi and J. A. M. van Gisbergen Perturbation of Combined Saccade-Vergence Movements by Microstimulation in Monkey Superior Colliculus J Neurophysiol, May 1, 1999; 81(5): 2279 - 2296. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
