INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

This study was designed to investigate the involvement of the monkey superior colliculus (SC) in binocular refixations to targets in three-dimensional (3D) visual (V)-space requiring combined saccade-vergence responses. It is well known from previous electrophysiological studies that the SC is involved in the control of saccades (for review see Sparks and Mays 1980; Wurtz 1996) through its connections with saccadic burst cells (see Munoz and Wurtz 1995; Sparks 1978) and its association with the omnipause neurons (OPNs) in the brain stem (Buttner-Ennever and Horn 1994; Gandhi and Keller 1997; Raybourn and Keller 1977). In contrast relatively little is known about the neuronal circuitry controlling vergence movements at the premotor and higher levels. This has made it rather difficult to understand the neural mechanisms controlling eye movements that are composed of both directional and depth components. An early suggestion by Yarbus (1967) was that the saccadic and vergence systems function rather independently of each other. Investigators long had the tendency to study each oculomotor subsystem in isolation from the other and the finding that saccade and vergence responses have strongly contrasting dynamics lent some credibility to the Yarbus concept. Early modeling studies (for overview see Collewijn et al. 1995), by emphasizing these distinctions, helped reinforce the conviction that the two systems are quite different in nature. It is commonly accepted that the saccadic system (classically defined as the system that causes the eyes to move equally in the same direction to fixate a target of interest in the frontal plane) does not process its inputs continuously. The saccadic system is normally suppressed by tonic pause cell activity that must be inhibited before a saccade can be made (gating). Saccadic eye movements are much too fast to allow sensory feedback during the movement. In contrast to this intermittent behavior, however, the vergence system (which moves the eyes in opposite directions in the horizontal plane to maintain binocular alignment when gaze is shifted between targets at different distances) was often portrayed as a continuous system relying on V-feedback.

As a sign that insights are changing a more recent suggestion regarding the control of 3D gaze shifts proposed that saccades and fast vergence have a shared gating system for movement initiation (Mays and Gamlin 1995, 1996; Zee et al. 1992). Common gating would be reflected in a strong temporal (WHEN) coupling between the latencies of saccade and fast vergence responses. The finding that pure vergence movements can be slowed down by electrical (E)-stimulation in the OPN region suggests that pause neurons can indeed affect the vergence system, directly or indirectly (Mays and Gamlin 1996). These studies suggest that there is a greater degree of interaction between the two subsystems than was previously perceived.

Arguing further against thinking in terms of independent subsystems, it has been shown that binocular refixations, made up of combined version-vergence eye movements, are not simply linear summations of the required components executed in isolation (Collewijn et al. 1995, 1997; Enright 1984, 1986; Erkelens et al. 1989; Maxwell and King 1992; Oohira 1993). The vergence component, when combined with a saccade, is substantially faster than a normal pure vergence response, which implies that sensory feedback would be unlikely in this case. A recent study provided strong evidence for a shared saccade-vergence target selection system, where the metrics (WHERE) of the saccadic and vergence components of planned movements are jointly determined (Chaturvedi and Van Gisbergen 1998). In line with this idea, this study explores to what extent the SC, generally held to be a saccadic control center, may also be involved in the control of vergence movements.

Possible role of SC in neural control of vergence eye movements

For the sake of clarity, three different possible scenarios (see Fig. 1) regarding the involvement of the SC in vergence control will now be developed. One extreme point of view (the "noninvolvement" hypothesis), shown in Fig. 1A, entertains the view that saccades and vergence are implemented by entirely separate control systems. This implies dedicated WHERE and WHEN mechanisms for each distinct oculomotor subsystem. Because the SC is obviously heavily involved in saccadic control, simple logic implies that this view precludes its involvement in vergence movements. Any artificial intervention in the colliculus would only affect the conjugate response and have no vergence effect whatsoever.

View larger version (17K): [in this window] [in a new window]   Fig. 1. Schematic of 3 models that depict different possible degrees of collicular involvement in saccade-vergence coordination. Each has a common 3-dimensional (3D) input, based on a higher-level common target-selection mechanism (see Chaturvedi and Van Gisbergen 1998). A: noninvolvement hypothesis exhibits the notion that saccade and vergence control is carried out by 2 distinct and separate oculomotor systems. A distinguishing characteristic of this model is the individual gating (WHEN) mechanism for both subsystems. The common 3D input is first separated into directional and depth components. The horizontal and vertical signals are fed to the 2-dimensional (2D) superior colliculus (SC) that controls the conjugate saccadic response. The SC sends a displacement signal to purely saccadic burst cells (Bs) downstream. When the bursters are enabled by the saccadic gate, embodied by the omnipause neurons (OPNs) in the brain stem, they send a velocity signal to the motoneurons. A very similar scenario is depicted for the vergence system. This presupposes pure vergence-command neurons in a separate unknown neural region (?), a dedicated timing mechanism, and pure vergence bursters (Bv). B: indirect involvement model differs only from the previous scheme in that it has a common gating mechanism. Thus the timing of both subsystems is identical, but each controls its own metrics. C: direct involvement model follows logically as another possibility from the previous 2. It assumes not only a common timing (WHEN) mechanism but also a common metric-specification (WHERE) system. Here the SC is assumed to code both directional and depth information for binocular responses in 3D visual space. OPNs control the gating of bursters (Bsv) that are involved in coding both the direction and depth coordinates of rapid eye movements in each eye separately (see Zhou and King 1998).

An alternative idea (the "indirect involvement" hypothesis), shown in Fig. 1B, is that saccades and vergence have a shared WHEN system (Mays and Gamlin 1995, 1996; Zee et al. 1992). As for the dedicated WHERE system, the scenario of indirect involvement entails that the SC specifies only the metric of saccadic eye movements, leaving vergence-dedicated WHERE control to some other neural center (Judge and Cumming 1986; Mays 1984; Mays et al. 1986). The postulated WHEN coupling in both systems may be implemented through gating by omnipause cells further downstream. Accordingly, the indirect involvement hypothesis implies that collicular microstimulation may indirectly have an effect on vergence initiation but not on its metrics.

A third possibility to be considered (the "direct involvement" hypothesis) is that the SC is involved in the WHEN and WHERE mechanisms of both saccade and vergence subsystems (see Fig. 1C). A number of electrophysiological studies (Bacon et al. 1998; Billitz and Mays 1997; Dias et al. 1991; Gnadt and Beyer 1998; Jiang et al. 1996; Mays 1996) proposed that the SC may indeed have some involvement with eye movements in depth. If this view is correct, stimulating in the SC in principle should affect not only the onset of saccade and vergence responses but also bias their metrics.

It occurred to us that a suitable and effective way of testing the 3D colliculus hypothesis would be to intervene artificially in the SC by E-stimulation. We carried out microstimulation perturbation experiments in the caudal SC with the intention of studying the initiation (WHEN) and metric-specification (WHERE) mechanisms of 3D eye movements. Before we explain the logic behind this approach, we briefly review earlier SC stimulation studies.

Earlier E-stimulation studies in SC

Neurons in the deeper layers of the caudal SC burst for saccadic eye movements in a limited range of directions and amplitude (Sparks and Mays 1980; Sparks et al. 1976; Wurtz and Goldberg 1972). High-frequency stimulation at caudal collicular sites elicits short-latency fixed-vector saccades (Schiller and Stryker 1972; Robinson 1972) by recruiting the brain stem burst neurons that are normally kept silent by omnipause cell inhibition (Raybourn and Keller 1977). Glimcher and Sparks (1993) have shown that low-frequency E-stimulation in the caudal region of the monkey SC, before the initiation of a V-guided saccade, can influence the metrics (direction and amplitude) of the ensuing target-directed response. They suggest, in agreement with results from previous studies (Becker and Jurgens 1979; Glimcher and Sparks 1992; Sparks and Mays 1983), that signals specifying the metrics of saccades develop gradually after target presentation. Sparks and Mays (1983) used high-frequency stimulation in caudal collicular sites to elicit an E-induced saccade before an impending V-guided saccade and found that the metrics of the E-saccade could be altered by target location. The degree to which the saccade was dominated by the E- or the V-stimulus depended on the time interval between target onset and stimulation onset. They suggested that stimulation of the SC caused the V-induced motor error signal, which was being specified at the time, to be expressed.

Possible outcomes of combined EV-stimulation

To explore the three hypotheses outlined previously (see Fig. 1) we applied a similar paradigm in a 3D V-task to see if the V-specified vergence signal that is presumably building up would manifest itself in the E-triggered response in a similar way to the saccadic signal described previously. In our experiments we perturbed the system by applying E-stimulation at a time when the monkey was preparing or just initiated a 3D refixation to a newly presented target. Our goal was to see whether any effects on the vergence component along with the expected change in the saccade response could be demonstrated.

On the basis of results found previously in the frontal plane (Sparks and Mays 1983), the saccadic vector can be expected to exhibit a gradual change in its metrics from purely E- to more V-dominated when the E-stimulus is applied later and later in the trial. However, one would still expect an E-effect in late-stimulation trials where the monkey just initiated the V-guided response itself so that V-dominance will not reach the 100% level (Schlag-Rey et al. 1989). In other words, the expected effect of the perturbation on the saccadic system involves both a WHEN and WHERE aspect; saccadic responses can be triggered prematurely, and even when they are not (in the case of self-initiated movements) their metrical properties are affected by the E-stimulus. In the case of vergence, predictions are different for each model. First, if one takes the view that the SC is a saccade-related area (noninvolvement), one would not expect any vergence component as a result of premature stimulation. Alternatively, the indirect involvement hypothesis, which rests on the assumption of common gating, suggests the possibility that a premature E-triggered vergence response toward the V-stimulus may emerge. Finally, if the SC is involved in the control of the WHEN and WHERE systems of both saccades and vergence (hypothesis 3), one would expect not only a vergence response resulting from premature stimulation but also when E-stimulation is applied to ongoing self-initiated movements.

This study illustrates the actual effect of the combined E- and V-stimulation on the initiation and metric specification of binocular refixations. The results show that the WHEN and WHERE control, for both saccadic and vergence systems, is closely related and can be manipulated by artificial intervention in a stereotyped fashion.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Animal preparation and neurophysiological procedures

SUBJECTS. The experiments were performed in two adult male rhesus monkeys (monkeys I and II), weighing 8-9 kg, that were trained to accurately fixate and follow V-targets presented in 3D V-space. The animals were water deprived and received a liquid reward after each correct trial. All surgical and experimental procedures were approved by the university committee for the use of experimental animals.

RECORDING OF NEURONAL ACTIVITY. Extracellular activity in the SC was recorded with glass-coated tungsten microelectrodes (impedance 0.5-1.2 M) inside a transdural guide tube. After amplification and filtering the electrode signal was fed into a level detector for spike detection.

Experimental procedures and setup

BINOCULAR EYE POSITION RECORDING AND CALIBRATION. Two-dimensional (2D) binocular eye positions were recorded with the double magnetic induction technique (Bour et al. 1984). Two alternating perpendicular magnetic fields (horizontal: 30 kHz; vertical: 50 kHz) induced eye position-dependent currents in the implanted eye rings, which in turn induced secondary currents in sensitive pickup coils that were mounted directly in front of each eye (interocular distance: 3.5 cm). A nulling coil, placed some distance away from each eye on a rigid manipulator, electronically canceled the primary (eye position independent) signal component induced by the magnetic fields. After amplification and demodulation by lock-in amplifiers (PAR 128A), the raw horizontal and vertical eye position signals were low-pass filtered (3 dB at 200 Hz; 4th-order Bessel filter), sampled with 12-bit resolution at 500 Hz per channel, and stored for off-line analysis (details are described subsequently). The resolution of the recording technique was ~0.2° for eccentricities up to 25°. Careful inspection revealed that cross talk between the signals from the two eyes was negligible (<1° error for large horizontal angles) for our purposes. This was confirmed by other investigators who were using the same technique in cat studies (Malpeli 1998).

STIMULUS PRESENTATION AND DATA ACQUISITION. Experiments were performed while the monkey sat, head fixed, in a primate chair. Light emitting diodes (LEDs; diameter 5 mm), positioned on a large circular board at a distance of 125 cm from the monkey, were used as V-targets to guide eye movements. The central LED was aligned with the cyclopean eye of the monkey. A number of LEDs were positioned on an adjustable perspex rod, roughly at eye level and closer to the monkey, thereby enabling the presentation of V-targets at various locations in the depth plane. Target presentation and stimulus timing were controlled by a PC. All experimental paradigms required the monkey to fixate a V-target before other targets were presented. For each trial, data recording always started 300 ms before the offset of this initial fixation point (FP) and continued for 2 s.

Perturbation paradigm

The objective of the perturbation paradigm was to elicit a saccade by E-microstimulation in the caudal region of the SC while the monkey was preparing a saccade to a V-target. The paradigm exploited the fact that, when a combination of E-stimulation and V-evoked activity occurs at different sites in the collicular map, the system may generate a saccade whose timing is related to E-stimulation onset, whereas its metrics may also reflect a V-guided influence, depending on relative timing of the E- and the V-stimulus. We tested whether 3D refixations were similarly affected, especially with regard to vergence metrics. To gauge the relative effects of the two stimuli in various temporal combinations, we recorded two control responses, i.e., an E-elicited eye movement (see Fig. 2A) and a purely V-evoked eye movement (see Fig. 2B). To determine the compromising character of the saccadic response in the forthcoming experimental paradigm we ensured that these controls clearly differed from each other in direction. The location of the FP was chosen freely in 3D space in these experiments, but once chosen it remained unchanged throughout the session.

View larger version (13K): [in this window] [in a new window]   Fig. 2. Schematic representation of control responses in the perturbation paradigm. A: short-latency electrical (E) saccade is elicited (upward in this scheme) by stimulating suprathreshold in the SC when fixation spot is extinguished. This saccade vector corresponds with the chosen stimulation site in SC and appears to have no vergence component. B: V-evoked eye movement in 3D space to the light-emitting diode target presented downward from the fixation point (FP) and closer to the monkey. C: basic idea behind EV-trial in the perturbation paradigm where E-stimulation is applied at various E-delays () after the V-target was presented nearby. In this schematic example the stimulation pulse is applied fairly early, and after a short latency a saccade is elicited (EV-response). This is followed by a V-guided saccade that redirects the eyes to the V-target. Although it is not entirely clear what the disconjugate response will do, there was no a priori reason to expect that the initial EV-response would show vergence. The E-latency is the time delay between stimulation onset and the occurrence of the EV-saccade. V-latency of the saccade is the time between target onset and saccade onset.

To record the E-induced binocular eye movement (E-control), an E-stimulation pulse train was applied in the deep layers of the caudal SC while the monkey was fixating an LED. The FP was extinguished on stimulation, and after the E-saccade the monkey was rewarded. The pulse train (duration 50 ms; frequency 500 Hz) consisted of constant current biphasic pulses (BAK, Model BPG-1) with a duration of 0.2 ms. Initially, the threshold value for evoking saccades was determined by finding the current intensity where 90% of all stimulations led to a saccadic response. Experiments were then conducted with a suprathreshold current level of up to twice this value (between 20 and 70 µA) such that most saccades had a fixed amplitude. The E-control paradigm was repeated at the end of the experiment.

To obtain the purely V-guided control response, the monkey first had to fixate the FP. Directly after FP offset, a V-target was presented at one of a number of selected positions in 3D space, and the monkey made a V-guided binocular refixation (V-control) to the target. Care was taken to position the V-targets with respect to the fixation plane in either a far plane (requiring diverging eye movements) or a near plane (requiring convergence responses) such that there was a clear 3D directional difference between the E- and the V-response in both direction (Fig. 3A) and depth (Fig. 3B). An attempt was made to align both near and far targets such that refixations from the FP to these target LEDs required V-responses with saccades of roughly the same direction and amplitude. After choosing the target positions, a third set of experiments was started where fixation and depth-plane (both near and far) targets were randomly interleaved and where 30% of the trials required refixations with a vergence component.

View larger version (15K): [in this window] [in a new window]   Fig. 3. The same schematic control responses from Fig. 2 projected on 2 different planes to show the spatial relations. Perturbation paradigm is set up such that there is a clear directional and depth difference in the E-elicited control and the V-evoked control. A: frontal plane view of FP (fix), target (tgt), E-control (E), and V-control (V). E-control saccade vector is directed obliquely upward. V-target was aligned in amplitude with the E-saccade and was directed to the lower hemifield. B: in a depth plane view of the controls we show that the E-control was not expected to have any vergence component, whereas V-response was directed nearby.

To explore the effect of combined E- and V-stimulation (see Fig. 2C), in some of these 3D trials (20%), denoted as EV-trials, a stimulation pulse train was applied at various short delays () after the presentation of the V-target to induce a premature binocular eye movement. The onset time of the stimulation pulse train was varied from immediately after target onset to after V-saccade onset. E-triggered EV-responses were usually followed by a later V-guided response to the V-target, which corrected for the E-perturbation. E-stimulation generally only occurred when either the near or far target was presented. To minimize the possibility of predictive behavior during the V-guided refixations, a number of V-catch trials (where the FP was extinguished and no target was presented), stimulation catch trials (where no target was presented and stimulation was applied at FP offset), and eccentric targets, located along the four perpendicular axes in the far plane, were incorporated in each experimental session and presented at random times. The monkey was only rewarded at the end of the trial if correct fixation of the presented target was achieved. The monkey was rewarded in the catch trials and in the control E-trials if the fixation target was correctly attended to. Whenever the monkey did not fulfill a requirement set in our paradigm, the FP or target disappeared immediately and no reward was given. All experiments were done under dim lighting conditions.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

E- and V-control responses

The perturbation experiments in this study were designed to investigate the involvement of the monkey SC in binocular refixations to targets in 3D V-space requiring combined saccade-vergence responses. The basic idea was to evaluate the effect of E-stimulation at one SC site on the 3D eye movement elicited by the V-target at another location in the SC map. As described previously (see Fig. 3) the E- and V-controls differed from each other in direction and depth, similar to the examples from a typical site shown in Fig. 4. The two panels on the left show both control responses projected together in the frontal and depth planes, respectively, whereas the three right-hand panels illustrate their corresponding time-courses. We aligned the LED target (tgt) in 3D space such that there was a clearly discernible difference in the vertical component of the two controls, keeping the magnitude of their horizontal components roughly similar. The V-response shown here, directed toward a near target, has a vergence component that can be separated into a fast intrasaccadic vergence phase and a subsequent slower movement. The eye movement generated on E-stimulation (E-response) is purely saccadic, and no evident vergence response is elicited. The slight divergence displacement during the E-saccade can be attributed to transient divergence (Collewijn et al. 1988, 1995; Maxwell and King 1992; Oohira 1993; Zee et al. 1992).

View larger version (26K): [in this window] [in a new window]   Fig. 4. Trajectories of actual control responses are presented in the format of Fig. 3 for both frontal and depth plane (horizontal plane of regard) views. The E-response exhibits no vergence component. The V-response is directed to a near target. Here, the FP (fix) is somewhat above eye level and slightly to the left of the cyclopean eye. Target location (tgt) was chosen such that, although vertical and vergence components differ considerably for both E- and V-controls, horizontal amplitudes are almost identical. In the time-course profiles the control responses are plotted with respect to V-latency. Data points occur every 2 ms. The E-response onset was temporally aligned with the V-response for sake of comparison. Onset and offset of the stimulation pulse train is indicated by the dotted lines. Data are from site I-3.

Qualitative observations

TEMPORAL ASPECTS. To introduce the analysis of the data, examples of EV-responses are depicted in Fig. 5, where the onset of the stimulation pulse train occurred at early, intermediate, and late stages after V-target presentation. The saccadic time courses are presented in Fig. 5 in the top plots, and the vergence components are presented in the bottom panels. By inspecting the time course of the vertical component of the combined EV-trials relative to the E- and V-responses (controls) we can evaluate how saccadic displacement changes as a consequence of E-stimulation. As will become clear, saccade behavior showed no surprises and conformed to the expectations formulated previously.

View larger version (30K): [in this window] [in a new window]   Fig. 5. Time courses of typical EV-responses, with E-stimulation (between dotted lines) applied at different points in time, alongside E- and V-controls (earlier shown in Fig. 4). Saccadic profiles are shown at the top, and the corresponding vergence components are shown in the bottom panels. In the first case (left panel), when stimulation is early, the E-response for both sets of components looks remarkably similar to the E-control. In other words, an E-saccade is elicited with no accompanying vergence response. The near target is subsequently fixated by an ensuing V-directed, combined version-vergence eye movement. In the next case (middle panel), when stimulation onset occurs a little later after target presentation, the E-triggered response is seen to be composed of a saccade and a small but unmistakable target-directed, fast vergence component. This is directly followed by a slow vergence eye movement in the direction of the target. After a short delay, a V-directed saccade and a corresponding fast vergence component occur to fixate the target. This same pattern is followed in the right panel. In this case, however, it appears that the later the stimulation occurs the larger the initial E-saccade and accompanying vergence components are. Whereas the saccadic time courses in the top plots can be clearly marked as having 2 distinct fast movements, the vergence profile in contrast has a more complex character. Data are from same site as in Fig. 4.

METRICAL ASPECTS. To make this distinction we now reinspect the component displacements of the binocular refixations to compare them with the predictions put forward by each hypothesis. The notion of indirect involvement (see Fig. 1B) suggests that the E-stimulus causes a train of events that allows a V-induced vergence signal to be expressed prematurely but without adding any vergence WHERE contribution. Accordingly, if this view is correct, one should find averaging of E- and V-induced effects in the saccadic component but not in the vergence signal. As a corollary, V-induced movements that were E-perturbed in midflight in their saccadic component should not show any vergence effects.

View larger version (24K): [in this window] [in a new window]   Fig. 6. Spatial trajectories of the trials and controls shown in Fig. 5 in frontal and depth planes. It is clear that the EV-responses show an averaging effect for both the vertical saccade and the vergence component, where the degree of compromise appears to be related to stimulation onset. Even for late stimulation, full-blown vergence is not seen. There is an unexplained increase in horizontal amplitude, despite the 2 controls being closely aligned.

MIDFLIGHT PERTURBATION EFFECTS. In Fig. 7 we present three examples of V-initiated responses where E-stimulation is applied in midflight after onset of the binocular refixation. It is a crucial finding that such a late E-stimulus perturbs not only the saccadic but, remarkably enough, also the vergence component. In each case the saccade time course closely follows the V-control until stimulation onset, after which the response deviates from its initial V-guided path toward the E-control. Interestingly, there is also a corresponding disturbance to the vergence response, temporally linked to the saccadic perturbation, where the V-guided vergence component breaks off its initial target-directed trajectory and diverges temporarily. As soon as the saccade picks up where it left off then so does the vergence. This is seen clearly in the spatial trajectories of the trials in Fig. 7, A and C, which are shown in Fig. 8.

View larger version (27K): [in this window] [in a new window]   Fig. 7. Time courses of 3 EV-responses (A-C) where stimulation (between dotted lines) was applied during midflight, displayed in a format similar to Fig. 5. E-stimulation briefly perturbs the target-directed response toward the E-control for both the saccade and vergence response, although the eye movement was already initiated voluntarily. Whereas the saccade stops briefly after E-modification, the vergence continues toward the target, albeit slowly. Data are from site I-3.

View larger version (23K): [in this window] [in a new window]   Fig. 8. Trajectories in both planes of trials A and C from Fig. 7 are shown together with control responses. The deviation and averaging effects for saccade and vergence components are very clear.

Quantitative analysis of metrical perturbation effect

In this analysis, it is important to keep track of how the binocular refixations in the E-triggered EV-trials, the V-stimulus and the E-stimulus, are temporally related to one another. Because E-stimulation was applied at a number of discrete delays relative to the V-stimulus, we obtained E-triggered responses over an extensive time period of target signal buildup, ranging from zero (E-stimulation at V-target onset) to nearly complete (V-latency range). When the E-stimulation was applied at a late point in time, the V-guided response could begin before stimulation onset so that in these cases E-latency (saccade onset with respect to E-stimulation onset) was negative. When E-stimulation occurred at a relatively early stage, all EV-responses were E-initiated, having rather constant E-latency values of ~20-30 ms. As one would predict, when E-delays (time between V-stimulus onset and E-stimulation onset, see Fig. 2C) were made long enough to enter the domain of V-latencies, the probability of getting a V-triggered EV-response increased sharply. To understand the additive effects of E- and V-stimulation, it is useful to consider how the characteristics of EV-responses develop both as a function of their timing relative to the V-stimulus (V-latency) but also as a function of E-latency. We will now proceed to present data, considered from either perspective.

METRICAL DEPENDENCE ON V-LATENCY. We present the entire range of EV-responses in Fig. 9, where the component displacements are now plotted as a function of V-saccade latency. Saccade amplitudes are depicted in Fig. 9A, and intrasaccadic vergence is shown in Fig. 9B. The mean control values for E-stimulation and V-guided refixations are also shown. If we first inspect the saccadic component in Fig. 9A we can see that E-triggered EV-responses () only attain roughly up to halfway completion. These EV-responses are fully E-dominated when their V-latencies are short (e.g., ~100 ms). The V-influence increases when the E-stimulus is delayed such that V-latency is prolonged. However, for long V-latencies we do not see complete V-dominance, at best a compromise between the two influences. This averaging effect continues to be present in the V-triggered EV-responses (), which show a large degree of scatter in saccadic displacement. Similar conclusions can be drawn for the intrasaccadic vergence as a function of V-latency, shown in Fig. 9B. Here a compromising phenomenon is also seen for both E- and V-triggered EV-responses, and it is apparent that metrically the latter set of data points can range almost anywhere between E- and V-domination. E- and V-triggered EV-responses show a marked degree of overlap. A much clearer picture emerges if one analyzes how the interplay of competing E-stimulus and V-stimulus effects depends on the time shift between the E-stimulus and response onset (E-latency).

View larger version (17K): [in this window] [in a new window]   Fig. 9. Component displacements of EV-responses are presented as a function of V-latency (taken as primary saccade onset with respect to target onset). The dotted lines are the mean of all V (cntrlV)- and E (cntrlE)- control responses. A: responses show a slight trend from heavy E-influence to increasing V-influence in (vertical) saccadic displacement as latency increases. In the realm of actual V-latencies (>200 ms) both E- and V-triggered responses scatter between halfway completion and complete V-influence. B: similar trend is shown for vergence displacement. The E-influence is large for early V-latencies, and the V-influence increases for the V-latency range.

METRICAL DEPENDENCE ON E-LATENCY. This point is illustrated in Fig. 10 where each rapid component of the binocular refixation is plotted as a function of E-latency, linking the displacement in each component with its latency relative to the E-stimulus onset. A great deal of information is contained in this plot, and we start by considering the saccadic behavior, shown in Fig. 10A. Notice that EV-responses with the longest positive E-latencies, obtained by very early E-stimulation, are indistinguishable from the E-controls (top dotted line). The other extreme, responses that are fully V-dominated (indistinguishable from V-controls, indicated by the bottom dotted line), are those V-triggered movements where the E-stimulus came too late to affect the movement at all, that is, after saccade offset. In between these two extremes there is a gradual transition of data points, which are observed to be at an intermediate stage of displacement when E-latency is zero. Because it seems reasonable to assume that the V-buildup signals reached the V-control level in all V-triggered EV-responses (i.e., negative E-latencies), the fact that the intercept value does not equal the V-control level indicates that the E-stimulus must be exerting a metrical effect on ongoing V-initiated movements. As previously noted, this is the critical distinction between hypotheses 2 and 3. Figure 10 shows that the E-triggered saccades become more and more V-affected as the E-latency diminishes toward zero. This clear change from E- to V-dominance continues further as E-latencies become increasingly negative. The picture that emerges therefore is one of an averaging process that blends the evoked E- and V-signals, causing a gradual trend of data points that link the E-control response to the V-control response.

View larger version (20K): [in this window] [in a new window]   Fig. 10. Component displacements of EV-responses are presented as a function of E-latency (defined to be time between saccade onset and stimulation onset). The linear regression line fits the data points in the time window of ±40 ms. A: clear change in saccadic displacement occurs as positive latencies become smaller, i.e., as stimulation onset is delayed and E-latency decreases. Responses with negative latencies (V-triggered responses) are still heavily influenced when the E-stimulus occurs in midflight. A full saccade only occurs after ~40 ms when stimulation comes after saccade offset. B: similar trend is seen for the vergence component. Early stimulation (large positive latencies) does not lead to any vergence. When E-latency is 0, only halfway completion is attained. V-triggered responses () still do not exhibit full-blown vergence responses. This is also only seen after ~40 ms.

QUANTITATIVE EVALUATION OF HYPOTHESES. To gain a quantitative measure for each site that characterizes the important relationship between the metrics of the responses and the E-latencies, we applied a linear regression analysis to the data points within a selected temporal window. By choosing only the responses that make the transition from the E- to the V-control lines (roughly between ±40 and 40ms) we were able to gauge the intercept of the EV-responses on the dashed line in Fig. 10 representing zero E-latency. For the site shown in Fig. 10A, the intersection point is roughly midway between the E-control and V-control lines. This provides strong evidence that the two regions of activity (E- and V-activated) still compete strongly with each other after the V-guided refixation was initiated. To ascertain to what extent the picture emerging from Fig. 10, favoring hypothesis 3, is representative we provide a quantitative overview of all sites (see Table 1).

Similarities in saccade and vergence perturbation effects

CORRELATED PERTURBATION EFFECTS. If the basic idea behind direct involvement is correct, the possibility has to be considered that the proposed involvement of the SC in both saccades and vergence may cause functional linkages between the WHERE signals of the two oculomotor subsystems by the introduction of a common noise source. To assess whether such couplings can in fact be demonstrated, we further examined the perturbation data from Table 1. What causes the variability in P values is not known (see DISCUSSION for a suggestion), but it is clear that if the scatter in saccadic P values strongly correlates to the variation in vergence P values, this would support the idea that these causes work equally on both subsystems. We compare P indices for saccades and vergence for all the sites, from both monkeys, in Fig. 11. The best-fit line has nearly a unity slope and a correlation coefficient that is highly significant (P < 0.01), showing that the perturbation effects in the saccade and vergence components of the binocular response covary and, interestingly, that they are about equally strong.

View larger version (12K): [in this window] [in a new window]   Fig. 11. Perturbation values for both sets of components, for all sites in both monkeys, are plotted against each other. The linear regression line is shown and has a correlation coefficient of r = 0.86 (n = 22). Sites from monkey 1 () are perturbed somewhat more, in general, than sites from monkey II (). Both sets lie fairly close to the regression line.

ANALYSIS OF TRIAL-TO-TRIAL VARIABILITY. The plots in Fig. 10 show that there is a linear relationship between the degree of perturbation and E-latency both for saccades and vergence. This strong relation with E-latency was observed for almost all sites, as shown by the data presented in Table 2 for both saccadic and vergence components. There also appears to be a considerable degree of scatter in displacement for the V-triggered responses that are perturbed in midflight. Given that a trial-to-trial variation exists in the relative effects of the V- and E-activity, one wonders whether this scatter in the saccadic and the vergence component is also correlated. In other words, if the V-effect in a particular trial is especially small or large in the saccadic component, relative to the fit curve, would the corresponding vergence component then show concordant variations? To explore the possibility that a common noise source is affecting the metrical effects in both components on a trial-by-trial basis we plotted the various component displacements against each other in Fig. 12. Of course, if the trial-by-trial scatter were independent in both subsystems, one would still expect the two components to be correlated in this plot because of their shared dependency on E-latency (shown in Fig. 10). The impression at first glance is that the very strong relationship between the conjugate and disconjugate responses of each binocular refixation is unlikely to be merely an epiphenomenon of the shared latency dependence. As Table 2 shows, a high degree of metrical correlation between saccades and vergence was seen for all sites. In more than two-thirds of the sites, 85% or more of the variability in vergence can be accounted for on this basis. To determine the contribution of trial-to-trial variability, it is necessary to filter out the latency contribution with partial correlation analysis. After this was done (see Table 2), we see that a substantial correlation remains in the majority of sites. This means that the trial-to-trial variability, causing scatter about the metric-latency relationship (see Fig. 10), is not independent for the saccadic and vergence components. To illustrate, in one-half of the sites 50% of this variability can be ascribed to this common noise factor (see Table 2).

View larger version (13K): [in this window] [in a new window]   Fig. 12. Component displacements plotted against each other for the site shown in Fig. 10. The EV-data points, spanning the range between the 2 sets of controls (*), marked by E and V, exhibit a strong correlation (R2 = 0.86; n = 51). As saccadic displacement increases, intrasaccadic vergence increases accordingly. This is true not only for the E-triggered () responses but also for the V-triggered responses ().

Vergence changes are not on epiphenomenon of varying saccade duration

In our analysis, we concentrated on intrasaccadic vergence as a clearly defined signal for comparison with the saccade behavior. However, this means that the following caveat has to be considered. One might argue that modification of the saccade metrics by the E-stimulus might, by virtue of the amplitude-duration relationship of saccades, influence the time-window in which rapid vergence can express itself. Although our experimental setup was expressly designed to sidestep this potential problem, by attempting to keep the amplitude of the saccade vector nearly constant (see METHODS), it was inevitable that saccades would vary in amplitude and therefore also in duration. The question to be answered now is to what extent these unintended duration effects may have determined intrasaccadic vergence displacement indirectly. To illustrate the problem, we examined this issue for the responses shown in Fig. 12. The data points depicted in Fig. 13 show the intrasaccadic vergence as a function of saccade duration. It appears that there is only a weak relationship between the two parameters, which in this site would only account for some 22% in the variance of the vergence component (r² = 0.22). To gauge how significant a role saccade duration plays in causing the metrical correlation with the vergence signal, we performed a partial correlation analysis for all sites by controlling for the influence of saccade duration. If the strong correlation in Fig. 12 were an epiphenomenon of variations in saccade duration, it would be lost after performing this further statistical analysis. What becomes clear from Table 2 is that the strong correlation between the saccade and vergence displacements, as seen for all sites, cannot be unduly considered to be an epiphenomenon of varying saccade duration. The r² values in the column on the righthand side of Table 2 reflect the genuine relation between the metrics of saccades and vergence after controlling for saccade duration effects. The relationship between the saccade and vergence metrics clearly remains very strong.

View larger version (14K): [in this window] [in a new window]   Fig. 13. Intrasaccadic vergence displacement presented as a function of saccade duration for the entire set of EV data points shown in Fig. 12. The low correlation indicates a weak relationship that does not account for the strong correlation seen in Fig. 12.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

We undertook neurophysiological experiments in the SC to learn more about the neural substrate of combined saccade-vergence eye movements. The oculomotor system was perturbed by E-stimulation in the SC at a time when the monkey was preparing, or just initiated, a 3D refixation to a newly presented target. Our goal was to see whether any effects on vergence, besides the expected changes in saccade responses, could be demonstrated. Results from our perturbation paradigm have shown not only the occurrence of premature E-triggered (and V-directed) vergence responses but also the weighted averaging of the two sets of stimuli (E and V) as a function of stimulation onset for both sets of components. Furthermore, data from V-triggered EV-responses show that the vergence response reveals a highly similar degree of perturbation as the corresponding saccadic portion.

Absence of vergence in E-induced refixations

To differentiate between the E- and V-triggered portions of EV-responses in the perturbation paradigm, the experiment was designed such that separate E- and V-stimulation yielded control responses with clearly distinct saccade and vergence directions (see Fig. 4). The results indicate that the greatest part of the required convergence, in the V-controls, takes place during the fast intrasaccadic phase. The E-control trials consistently yielded the saccade expected on the basis of the stimulated site in the SC motor map but did not exhibit any vergence response. This data are in line with findings recently presented by Billitz and Mays (1997), who observed that stimulation of caudal and rostral sites, while a monkey fixated a distant target, did not lead to any vergence effects. If interpreted in a straightforward manner as a sign that the SC is only involved with saccade control, this fact would lead one to believe that E-stimulation, applied in conjunction with a V-target in 3D EV-trials, will only interfere with the required saccadic response, leaving vergence unaffected. It is evident from our results (see Figs. 10 and 11 and Table 1) that this explanation has become untenable.

Rejection of noninvolvement hypothesis

Yarbus' (1967) suggestion that refixations in 3D space are a result of distinct and independent oculomotor subsystems and neurophysiological work indicating that saccades and vergence responses may be coded in different areas of the brain (Judge and Cumming 1986; Mays 1984; Mays et al. 1986) provides a basis for the noninvolvement hypothesis, outlined in the INTRODUCTION. If, as in this proposal, each subsystem is considered to be an independent entity, then no room is left for a common WHEN and/or WHERE control mechanism for the two components. On the basis of the results from previous studies (Glimcher and Sparks 1993; Sparks and Mays 1983) one would expect, in the perturbation paradigm, that the saccadic response shows a gradual change in displacement between the two sets of controls as the onset time of the E-stimulus is delayed. A specific prediction from the noninvolvement model is a zero-vergence response (as seen in the E-control) for all positive E-latencies (i.e., cases where the E-stimulus is applied before any V-triggered response occurs). In addition, a full convergence response, unmitigated by the E-stimulus, should occur when a V-guided response is initiated toward the near target, in other words, when E-latencies are negative.

Because both vergence predictions are in conflict with the data, the model must be discarded. Next we will evaluate the second model that can account for the subset of our experimental results showing that (V-directed) vergence is elicited prematurely as a result of the E-stimulation in the SC. This novel result (see Fig. 5) shows that, by intervening in the SC during the time period of V-signal buildup, one can affect not only the saccadic response but also the vergence response.

Evaluation of indirect-involvement model

PREMATURE E-TRIGGERED VERGENCE RESPONSES. E-stimulation, in the indirect involvement model, opens the gate for both sets of components (in whichever state of preparation they may be). If we assume that the vergence signal is built up gradually, this model can explain why the V-directed vergence signal is able to manifest itself prematurely after application of the E-stimulus in the SC.

E-PERTURBATION OF V-TRIGGERED 3D RESPONSES. In the case of saccades, it was well established that E-stimulation in midflight causes a perturbing effect on the V-guided response (Schlag-Rey et al. 1989). In other words, a full-blown saccade response will only be observed if the E-stimulus is delayed until after saccade completion. This is also evident from our data presentation in Fig. 10A. If we stimulated earlier, we observed that the saccade only reached roughly halfway completion by the time the V-latency range was reached. Not only does the saccade response intersect the zero latency axis far below V-control level (see Fig. 10A) but it continues to be influenced by the E-stimulus when the response is V-guided.