INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the companion paper (Goossens and Van Opstal 2000), we reported that blinking affects various aspects of saccadic behavior in monkey. It appeared that air-puff-evoked blinks had a considerable influence on both the kinematics and the spatial trajectories of saccadic eye movements. When elicited before saccade onset, these reflex blinks also reduced the saccade latencies substantially. Despite the strong disruptive nature of the evoked blinks, visually elicited saccades remained quite accurate in the absence of visual feedback. These behavioral data support the idea that blinking interferes with the saccadic premotor system and that an active control system may compensate for the blink-related perturbations. The neural mechanisms underlying these saccade-blink interactions, however, have so far received little attention in the literature and are difficult to assess on the basis of behavioral data alone.

Up to now, neurophysiological experiments have shown that the tonic activity of brain stem omnipause neurons (OPNs) pauses during blinks as well as saccades (Cohen and Henn 1972; Fuchs et al. 1991; Mays and Morrisse 1994), thereby disinhibiting saccadic burst neurons in the pontomedullary brain stem. These findings suggest that OPNs mediate, at least partly, the tight latency coupling between saccades and blinks. However, as discussed in the companion paper (Goossens and Van Opstal 2000) and in view of current models of saccade generation, the changes in both the trajectory and the kinematics of blink-perturbed saccades cannot be simply ascribed to an interaction at this premotor level. Moreover, a linear superposition of saccade- and blink-related signals at the extraocular motoneurons cannot readily account for the observed saccade behavior either. It is conceivable therefore that blinking affects other premotor stages of the saccadic system as well.

Recent studies indeed hint at the possibility that also the midbrain superior colliculus (SC) could be an important site of saccade-blink interactions. For example, both in rats (Basso et al. 1996) and in monkeys (Gnadt et al. 1997), it has been observed that electrical microstimulation of the SC transiently reduces the magnitude of air-puff-evoked blinks. As in rats (Basso and Evinger 1996), the monkey SC presumably excites a nonsaccadic pathway that inhibits the reflex blink circuitry (Gnadt et al. 1997). However, it is still unknown whether blinking affects the discharge patterns of saccade-related collicular neurons and how this could in turn affect saccade generation. In the present paper, we therefore studied single-unit activity in the intermediate and deep layers of the SC with the use of the blink-perturbation paradigm.

It is well established that the SC has extensive projections to saccadic burst cells in the pontomedullary brain stem and that the SC is critically involved in the generation of normal saccadic eye movements (see e.g., Moschovakis and Highstein 1994; Sparks and Hartwich-Young 1989 for reviews). Many neurons in its intermediate and deep layers generate a burst of activity just before and during saccades directed to a particular region of the visual field, referred to as the movement field of the neuron. Together, these saccade-related burst neurons (SRBNs) form a topographically organized motor map in which anatomically nearby cells have overlapping movement fields (Lee et al. 1988; Mays and Sparks 1980; McIlwain 1982; Ottes et al. 1986; Robinson 1972; Schiller and Stryker 1972).

Until recently, it was generally held that the amplitude (R) and direction () of an impending saccade is specified by the location of the active cell population in the SC motor map rather than by the temporal discharge patterns of the recruited neurons. Several studies have suggested, however, that the collicular output may also determine the kinematics and trajectory of the saccade (Berthoz et al. 1986; Lee et al. 1988; Munoz et al. 1991; Van Opstal et al. 1990; Waitzman et al. 1991; Wurtz and Optican 1994). These findings have led to new quantitative models that place the SC inside the so-called local feedback loop. This internal feedback circuit is thought to control the saccade trajectory by a continuous comparison of the desired eye displacement signal with an internal representation (efference copy) of the actual eye displacement during the saccade (e.g., Arai et al. 1994; Droulez and Berthoz 1991; Lefèvre and Galiana 1992; Optican 1995; Van Opstal and Kappen 1993). Previously most models assumed that the local feedback loop is closed through assemblies of cells in the pontomedullary brain stem (Jürgens et al. 1981; Scudder 1988; Van Gisbergen et al. 1981).

A difficulty in studying the role of the SC in the control of saccades is the stereotyped relationship between the amplitude of normal saccades and their duration and peak velocity (referred to as the "main sequence") (Bahill et al. 1975; Fuchs 1967). In an attempt to overcome this problem, previous saccade-interruption paradigms have used intrasaccadic microstimulation of either the OPNs (Keller and Edelman 1994) or of the rostral SC (Munoz et al. 1996). In these experiments, it was found that the microstimulation not only stopped the eye in saccade mid-flight, but it also induced a brief pause in the discharge of collicular SRBNs. Shortly after the stimulation ended, the saccade resumed its course, and the same population of SC cells that was active before the stimulation was reactivated even though the resumed movement did not belong to the movement field of these cells. By contrast no, or only minimal, activity was found in cells whose movement field optimum matched the metrics of the resumed saccades. By what mechanism the same neurons are reactivated is still unclear. One possible explanation is local feedback, which indeed predicts a resumed discharge (e.g., Arai et al. 1994). The finding that the SRBN discharge rate during the resumed saccades still showed the same monotonic relation with the remaining motor error than the one obtained for uninterrupted saccades further supported this hypothesis (e.g., Das et al. 1995).

However, a problem that still hampers the interpretation of the saccade-interruption data are the stereotyped kinematics of the resumed movements and the lack of a change in eye-movement direction. Moreover, the potential danger of stimulating adjacent oculomotor pathways, both upstream and downstream from the SC, makes the interpretation of stimulation data less obvious then at first glance. For example, OPN and rostral SC stimulation not only stopped the saccade in mid-flight, but it also interrupted the SRBN discharge. Hence it appeared that the SRBNs did not represent the remaining motor error during the interruption period. It is unclear, however, whether the cessation of SRBN discharge could be a side effect of the electric stimulation. It is ambiguous therefore whether the SRBNs detected the perturbation (through feedback) or whether they instead caused the interruption of the saccade. By imposing natural perturbations on the saccadic system that affect both the kinematics and spatial trajectories of saccadic eye movements, the blink-perturbation paradigm may offer an opportunity to circumvent these difficulties.

In view of current models of the SC and the notion of local feedback control, the present paper investigates how blink-related spatial-temporal perturbations of the saccade trajectory are reflected in the SC activity patterns. A preliminary account of these experiments has been presented in abstract form (Goossens et al. 1996; Van Opstal and Goossens 1999).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

The neurophysiological data presented in this paper were obtained from three rhesus monkeys (Macaca mulatta) and were collected during the behavioral experiments described in the companion paper (Goossens and Van Opstal 2000). Details about the setup, surgical procedures, and methods used to measure eye and eyelid position as well as the applied behavioral paradigms, are described in that paper. Here we provide only a brief summary, and additional methods that were used to record and analyze the single-cell activity. All experiments were conducted in accordance with the European Communities Council Directive of November 24, 1986 (86/609/EEC) and were approved by the local university ethics committee.

Apparatus

The head-restrained monkeys were seated in a primate chair facing an array of 85 light-emitting diodes (LEDs) in an otherwise completely dark room. The horizontal and vertical components of the left eye position were measured with the double-magnetic-induction technique (Bour et al. 1984). Movements of the contralateral right eyelid were measured with the magnetic search-coil induction technique (Collewijn et al. 1975) by taping a small coil on the eyelid. Air puffs (20 ms, 1.4-1.8 Bar) were generated by a pressure unit and presented on the recording eye to elicit trigeminal blink reflexes.

The activity of single units was recorded with glass-coated tungsten microelectrodes (0.2-1.2 M) that were positioned in the SC using a hydraulic stepping motor (Trent Wells). The latter was mounted on the stainless steel recording chamber that was placed above a trephine hole at stereotaxic coordinates [AP, RL] = [0,0] and aimed at the SC. The electrode signal was amplified (Bak Electronics, Model A-1), low-pass filtered (15 kHz), and monitored on an oscilloscope. Action potentials were detected by a level detector and discriminated on the basis of their waveforms using a real-time decomposition of the first four principal components (e.g., Epping and Eggermont 1987). The accepted spike events were subsequently fed into a four-bit spike counter, which produced a stepwise DC output that was sampled at a rate of 1 kHz. This ensured that no spikes were missed, irrespective of the instantaneous discharge rate during vigorous bursts of action potentials.

SC localization and histology

The in vivo localization of the SC was based on the following criteria: 1) the stereotaxic coordinates of the recording sites corresponded closely to the coordinates of the SC as given by Snider and Lee (1961). The size of the area with visual- and saccade-related activity was also in accordance with this atlas. 2) A region without action potentials, corresponding to the superior cistern, was often passed before the electrode entered the superficial layers of the SC. Usually the upper boundary of the SC was readily recognized by the activity of visual neurons with limited contralateral receptive fields and response latencies of ~70 ms. 3) At deeper locations, typically between ~0.8 and 3 mm below the SC surface, clear saccade-related activity was encountered. In these layers, electrical stimulation with pulse trains (25 negative pulses of 0.5-ms duration at 500 Hz) at low current strengths (20-50 µA) evoked reproducible saccades that corresponded well with the movement fields of nearby saccade-related neurons. The latencies of stimulation-evoked saccades were ~20 ms, and thresholds for evoking saccades were typically <10 µA. 4) The topographic motor map of the SC (Robinson 1972) could reliably account for the optimal saccade vector of cells encountered in subsequent penetrations. And 5) after further lowering of the electrode, sustained auditory-evoked responses with short latencies, indicative for the inferior colliculus, were often encountered (Goossens et al. 1997).

After completing the experiments, two of the monkeys (SA and ER) were deeply anesthetized with an overdose of pentobarbital and perfused directly through the internal carotid artery with 2 l phosphate-buffered saline, pH 7.4, at 37°C, followed by 2 l of fixative containing phosphate-buffered 2% paraformaldehyde and 2.5% glutaraldehyde, pH 7.4. Immediately after perfusion, the midbrain was dissected out. Serial cryostat sections were made and prepared for standard histology (Nissl staining). Examination of the midbrain sections at low and high magnification showed that the penetrations were made through the SC. Most of the penetrations reached the intermediate and deep layers.

Experimental protocol

After isolating an SRBN, saccades were evoked toward all LEDs of the target array. These data were used for accurate calibration of the eye-position signals (see Goossens and Van Opstal 2000) and to estimate the location and extent of the cell's response field. Subsequently, the movement field of the neuron was characterized in detail by eliciting saccades to a series of targets presented inside and neighboring the cell's response field.

A typical movement field scan consisted of 85 different fixation/target configurations: 5 different target positions and 17 different initial fixation positions (usually within 5° from the center LED), yielding a spatial resolution down to 0.5°. In each trial, the monkey first looked a fixation spot that was presented for 800-1,600 ms (randomized). Then as soon as the fixation spot disappeared, a peripheral target was presented for another 900 ms at a pseudo-randomly selected location that the animal refixated with a saccade. When a better dissociation between visual- and saccade-related activity was required, the saccade latencies were increased by presenting the target 100 ms prior to the offset of the fixation point (overlap paradigm) (Fischer and Weber 1993).

After this standard procedure, which was usually repeated near the end of a recording session, the neuron's activity was tested with the use of the blink-perturbation paradigm that has been described in the companion paper (Goossens and Van Opstal 2000). In short, saccades were made in complete darkness from a fixation point to either one of five randomly selected and briefly flashed (50 ms) peripheral targets, typically into the central region of the cell's movement field. In 30% of the trials, air puffs were presented at a fixed moment after target onset (between 120 and 180 ms) to elicit a reflex blink (mean latency ~20 ms) near the onset of the saccade. Control and perturbation trials were randomly interleaved with catch trials. In the latter trials, target positions were chosen such that the evoked saccades could be matched to the successive eye-movement components of perturbed responses (see RESULTS, Fig. 11). Cells were also tested with the use of the fixation-blink paradigm, in which air-puff-evoked blinks were elicited while the animal attempted to fixate a straight-ahead fixation spot (see Goossens and Van Opstal 2000).

Data analysis

For details regarding saccade and blink detection procedures as well as statistic criteria that were used to discriminate between compensatory and noncompensatory responses, the reader is referred to METHODS and RESULTS of the companion paper (Goossens and Van Opstal 2000).

The raw single-cell activity was displayed in spike rasters aligned on specific events such as target onset, air puff onset, the onset or offset of a saccade, or the onset of a blink. All neurons that showed a sharp increase in their activity slightly preceding (~20 ms) and tightly linked to the onset of saccades directed into their movement field were considered saccade-related and were therefore subjected to further analysis. The location and extent of a cell's movement field was determined from the movement field scan data by plotting the number of spikes, counted from 20 ms before saccade onset to saccade offset, as function of saccade amplitude and direction. Quantitative descriptions of the movement fields were obtained by fitting two-dimensional Gaussian activation profiles to the cells' activity as function of saccade vectors in collicular motor map coordinates (see Ottes et al. 1986 for extensive details) (goodness of fit: r typically between 0.85 and 0.98).

The raw spike trains were converted into smoothed representations of a cell's instantaneous firing rate by constructing spike density functions (D). To that end, all spike events in a trial were substituted by Gaussian pulses of width  = 4 ms and height 1/() and summed to produce a continuous function of time (MacPherson and Aldridge 1979; Richmond and Optican 1987). Large values of the spike density function represent a high probability of spike occurrence, and the peak of the function represents the peak discharge of the cell (in spikes/s). To estimate the duration of the saccade-related burst during individual saccades more robustly, the width of the Gaussian was increased to  = 10 ms (see RESULTS).

Dynamic motor error (ME) was defined as the difference between saccade endpoint and instantaneous eye position. The average phase relations between spike density and dynamic motor error shown in Figs. 12, C-E, and 13, A-F, were obtained by averaging the spike density and dynamic motor error signals as function of time for series of matched eye movements. It was not always possible, however, to find a sufficient number of perturbed responses with closely matched movement profiles. To circumvent this problem, the spike density function of each trial was resampled as function of the declining radial motor error using spline interpolation. In this way, all responses with matched saccade amplitudes and directions could be used to compute the average phase relations for each neuron in Fig. 13, G and H, despite the large variability in movement kinematics.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Single-unit activity was recorded from a series of SRBNs encountered in the intermediate and deep layers of the SC of three monkeys. Twenty-nine of these neurons (PJ: n = 17; SA: n = 4; ER: n = 8) were studied with the use of the blink-perturbation paradigm in which visually evoked saccades were disturbed by air-puff-evoked reflex blinks. Of the 29 SRBNs tested, 25 neurons remained isolated long enough to obtain a full 2-D scan of their movement field as well as a sufficient series of perturbed saccades for the quantitative analysis described in the following text. For the remaining four cells, we gathered sufficient data in at least the blink-perturbation paradigm to verify that they showed qualitatively similar features to the more thoroughly studied set of 25 SRBNs.

Isolated collicular neurons were classified as SRBNs if they showed a sharp increase in their firing rate ~20 ms before and tightly linked to the onset of saccades directed into the cell's movement field, irrespective of other pre- and postsaccadic discharge properties. In saccade trials, most of the cells (25/29) also showed a brief visual response ~70 ms after the visual stimulus appeared inside their receptive field. A prelude of presaccadic activity that could start several tens of milliseconds before the high-frequency, saccade-locked burst was also frequently observed. Cells that were endowed with a considerable amount of long-lead (>100 ms) presaccadic activity also showed a considerable amount of postsaccadic activity over a period of 100 ms after the saccade and were typically encountered at deeper locations from the SC surface. Saccade amplitudes for which the recorded SRBNs were maximally recruited ranged between 8 and 50°, except for four cells in the caudal SC which had no optimum for saccade amplitudes <70°. When tested in fixation trials, SRBNs were silent, and they were not recruited when air puff stimuli were presented in fixation-blink trials.

Activity during compensatory saccades

When air puffs were used to elicit a reflex blink near the onset of saccades, not only were the stereotyped spatial-temporal properties of saccades disturbed (Goossens and Van Opstal 2000) but the discharge patterns of SRBNs in the SC also were clearly modified. This is illustrated in Fig. 1, which compares the discharge of a saccade-related neuron (sa6703) in the caudal SC during a control saccade (~60° to the left; Fig. 1A) and during a blink-perturbed saccade that landed close to the extinguished target (Fig. 1B). Figure 1B shows, in a qualitative way, the consistent features of blink-perturbed responses that were obtained in all three monkeys. First, at the onset of the blink, the eye rapidly deviated from its normal, straight trajectory (typical latencies ~20 ms relative to the air puff onset at the eye), and no sharp increase in the cell's discharge was observed around the eye-movement onset. Second, as the cell continued its low-level discharge, the eye movement continued, and the perturbation in both direction and velocity was compensated in complete darkness. Finally, the increase in movement duration to reach the target was matched by a comparable increase in the duration of the cell's discharge.

View larger version (16K): [in this window] [in a new window]   Fig. 1. Typical example of the persistent burst discharge of a collicular saccade-related burst neuron (SRBN) for a saccade that was perturbed by an air-puff-evoked reflex blink. Activity of this neuron (sa6703) was recorded during a control saccade (A) and during a blink-perturbed saccade (B) of comparable amplitude and direction ([R, ] = [60, 0] deg). Both movements were evoked from the fixation point (F) toward the briefly (50 ms) flashed target (T). Spatial trajectories of the 2 saccades are shown at the top. Subsequent traces show vertical eyelid position (L; in arbitrary units), eye position (H, horizontal and V, vertical; in °), and eye velocity (H, horizontal and V, vertical; in °/s). As in all subsequent figures, eyelid movements were recorded from the contralateral, unstimulated eye. Bottom traces show spike density (D;  = 4 ms; in spikes/s; see METHODS), as well as the individual action potentials (vertical lines). The neuron's optimum was at, or beyond 60° to the left.

To document the reproducibility of these findings among our sample of collicular neurons, Figs. 2 and 3 show the activity of two other representative SRBNs during a series of control trials and a series of perturbation trials. The data plotted in Fig. 2 were recorded from a neuron (pj5203) that was found in the central region of the SC motor map. Saccade vectors were matched in amplitude and direction, and data are aligned on different events, from top to bottom: saccade onset (Fig. 2, A-D), saccade offset (Fig. 2, E and F), and air puff onset (Fig. 2G). As shown in Fig. 2, right, the cell showed a brisk burst discharge for control saccades of optimal amplitude and direction ([R, ] = [20, 60] deg). The onset of the saccade-related burst ("motor burst" for short) preceded the saccade onset by ~20 ms (Fig. 2D) and peaked shortly before saccade onset. Subsequently the discharge declined sharply during the saccade until it stopped when the saccade ended (Fig. 2F). Neurons with such behavior have also been referred to as clipped burst neurons (Munoz and Wurtz 1995; Waitzman et al. 1991). During perturbation trials (Fig. 2, left), the cell's burst discharge was clearly disturbed along with the saccade kinematics. As may be observed in Fig. 2C, the neuron typically showed an irregular discharge around movement onset, except in two trials where the cell showed a near-normal burst (top rows of the spike rasters). The irregular, lower-frequency discharge was associated with the low-velocity saccades, whereas the two clear bursts were associated with the two high-velocity saccades (clearly identifiable in Fig. 2A). Figure 2G shows the same responses aligned with air puff onset. Note that the cell's (upcoming) burst activity was transiently suppressed shortly after the air puff arrived at the eye. About 30-40 ms after the air puff onset, the cell resumed its activity, and it continued its discharge until the perturbed saccade ended. The latter can be readily observed in Fig. 2E, which shows the data aligned with saccade offset. Figure 2H compares the two conditions by showing the difference between the average spike-density function for control and perturbed trials (data aligned with eye-movement onset). As was typically observed, the resulting difference waveform was endowed with an initial negative component, followed by a positive phase that ended with the average end of perturbed saccades. The biphasic profile illustrates both the initial suppression as well as the prolongation of the cell's burst discharge in the perturbed condition. Note that the negative component leads the eye-movement onset because the air puff typically preceded the onset of the impending saccade, thereby interfering already with the cell's presaccadic discharge.

View larger version (30K): [in this window] [in a new window]   Fig. 2. Clipped activity of an SRBN (pj5203) for a series of perturbed and control saccades made in darkness toward the position of targets flashed at [R, ] = [20, 60] deg. Amplitude and direction of the saccade vectors were matched and corresponded closely to the cell's optimum. A and B: radial eye position (E = ) and vectorial eye velocity ( = ) traces. Subsequent panels show spike rasters and averaged spike density aligned with saccade onset (C and D) with saccade offset (E and F) and with air puff onset in the perturbation trials (G), respectively. H: comparison of the cell's discharge during perturbed saccades and control saccades by showing the difference in the average spike-density waveforms when data were aligned with saccade onset. · · · , standard deviation. The cell's burst discharge during perturbed saccades is consistently prolonged until the movement ends, and a transient suppression of activity, shortly after the air puff onset, is apparent. Note the 2 trials where the perturbation occurred late in the saccade (top rows in the spike rasters).

View larger version (35K): [in this window] [in a new window]   Fig. 3. Unclipped activity of an SRBN (er1101) in the blink-perturbation paradigm. Same display format as in Fig. 2. Note the buildup of presaccadic activity and the gradual decline of the postsaccadic activity. The neuron's optimum saccade vector was at [R, ] = [14, 150] deg. Apart from a substantial amount of intrinsic noise, the cell's response showed comparable features as the neurons with a clipped discharge presented in Figs. 1 and 2. Note the 1 trial with a late perturbation (top row in the spike rasters).

Figure 3 shows comparable data obtained from an SRBN (er1101) that was endowed with long-lead presaccadic activity. As may be observed in Fig. 3D, this prelude activity was followed by a more intense burst just before and during control saccades of optimal amplitude and direction ([R, ] = [14, 150] deg). The discharge peaked at saccade onset and then declined rapidly toward the end of the saccade. After this high-frequency burst, the neuron showed a gradual decline of postsaccadic activity over a period of ~250 ms (i.e., unclipped discharge; see Fig. 3F). Collicular neurons that exhibit such temporal discharge characteristics have also been named buildup neurons (Munoz and Wurtz 1995). During perturbation trials (Fig. 3, left), the cell's activity was clearly suppressed within ~10-20 ms after the air-puff onset (Fig. 3G). As the air puffs were presented prior to the impending saccade, the onset of the suppression preceded the actual saccade onset (Fig. 3C), except in the one trial (top row of the spike raster) where the perturbation occurred toward the end of the saccade (see Fig. 3A). Following a near complete cessation of activity, the neuron resumed its discharge ~30-40 ms after the air-puff onset (Fig. 3G). Although the discharge of this neuron did not end at saccade offset, it may be inferred from the spike rasters in Fig. 3E that it showed an increased firing rate until the end of the perturbed eye movements. This prolongation of the cell's saccade-related activity is also evident from the difference between the average control and perturbed response (Fig. 3H), which drops to zero around the mean offset of the perturbed saccades.

Suppression and resumption of SRBN discharge

Since the air puffs often interfered already with the cells' pre-saccadic discharge, the impression was obtained that the changes in SRBN activity were the underlying cause rather than the consequence of modified saccade kinematics. To gain further insight into the possible mechanism that could underlie the suppression of SRBN discharge, we also examined responses in which the saccade was accompanied by a gaze-evoked blink. As shown in the companion paper (Goossens and Van Opstal 2000), the spatial-temporal perturbation resulting from gaze-evoked blinks were qualitatively similar to those obtained with air-puff-evoked blinks.

Figure 4 compares the responses of two different SRBNs that were recorded under both air-puff- and gaze-evoked blinking conditions. The two cells, er0902 (Fig. 4, A and B) and pj6802 (Fig. 4, C and D), were isolated in the left and right SC, respectively. Note that the burst activity of both cells was strongly suppressed following the onset of reflex blinks that were evoked by air puff stimulation of the left eye (Fig. 4, A and C). By contrast, no transient suppression was observed following the onset of gaze-evoked blinks (Fig. 4, B and D; thick traces), which had a comparatively small influence on the discharge of the SRBNs. Note that this difference in SRBN discharge is also reflected in the saccade kinematics, which were less dramatically disturbed by gaze-evoked blinks. The latter may be inferred from the control data that are superimposed in Fig. 4, B and D (thin traces). On average, also the air-puff- and gaze-evoked blinks were different. However, by selecting a subset of responses with comparable eyelid traces, it could be excluded that the differences in SRBN discharge were merely due to differences in blink magnitude (data not shown).

View larger version (32K): [in this window] [in a new window]   Fig. 4. Discharge of 2 different SRBNs under air-puff and gaze-evoked blinking conditions. A and B: SRBN found in the left superior colliculus (SC; er0902). C and D: SRBN found in the right SC (pj6801). Data were aligned to blink onset and averaged (thick traces). Average control responses are superimposed in B and D (thin traces). A strong transient suppression of activity following the onset air-puff-evoked blinks is apparent for both cells (A and C). By contrast, no transient suppression is observed in the case of gaze-evoked blinks (B and D). Note that the SRBN discharge as well as the saccade kinematics were only mildly affected in the case of gaze-evoked blinks.

As illustrated by the preceding examples, SRBNs showed a transient decrease in their discharge following the onset of an air puff. Although present in most cells tested, this suppression was often far from complete and variable from trial to trial. Moreover, at the time of the air puff (fixed re. target onset, see METHODS), the amount of presaccadic activity was variable due to considerable scatter in the onset latencies of the motor burst with respect to target onset (see Fig. 4, A and C). It was not possible therefore to quantify the onset latency and the duration of this phenomenon for each neuron with the same accuracy.

To obtain at least a crude estimate of the time course of the suppression and subsequent resumption of SRBN activity, we analyzed the discharge of those SRBNs (n = 7) that exhibited a strong suppression in a large number of trials. To that end, the spike-density functions from all perturbation trials in a recording session, except the ones in which the perturbations occurred very early or very late in the burst, were averaged (n between 20 and 80;  = 4 ms) and subsequently normalized with respect to the mean. Figure 5, A and B, shows the results of this procedure for each neuron (thin curves) when data were aligned to air puff onset and blink onset, respectively. Averaged data of the seven neurons (thick curves) are superimposed. As may be inferred from these data, the suppression starts, on average, ~10 ms after the air puff onset and leads the blink onset by an approximately similar amount. About 10-30 ms after the blink onset, the SRBNs resumed their discharge. Some caution is called for with regard to the interpretation of the data in Fig. 5B because movements of the contralateral, unstimulated eyelid were used to detect blink onsets. However, as indicated in the companion paper (Goossens and Van Opstal 2000), it is reasonable to assume that there is no significant latency difference between ipsi- and contralateral eyelid movements as the short-latency, uncrossed R1 component of the primate blink reflex hardly contributes to movements of the eyelid (Bour et al. 2000).

View larger version (27K): [in this window] [in a new window]   Fig. 5. Time course of the blink-related suppression and resumption of SRBN discharge. A: averaged and normalized spike density waveforms from 7 SRBNs (thin curves), aligned with air puff onset. Trials in which the perturbation occurred very early or very late in the burst were removed from the data sets. Each curve was normalized with respect to its mean. Although the precise onset of the suppression is ill defined for each neuron (see text), a crude estimate from the averaged cell data (thick curve) indicates a latency of ~10 ms relative to the air puff onset. B: same data as in A, but now the individual responses (n between 20 and 80) were aligned to the onset of the (consensual) blink. Note that the suppression starts, on average, ~10 ms before the blink onset. Approximately 10-30 ms after the blink onset, the neurons resumed their discharge.

Quantitative analysis of the saccade-related discharge

Unlike in previous saccade interruption paradigms, the evoked perturbations in the present experiments typically affected the entire saccade. Nevertheless the data presented so far indicate that the response properties of SRBNs during blink-perturbed saccades showed qualitatively similar features as have been observed during interrupt saccades (Keller and Edelman 1994; Munoz et al. 1996). Therefore to facilitate a comparison between the interruption data and the present data set, we started out with a comparable quantitative analysis.

BURST DURATION. The raw data in Figs. 1-3 indicate that the SRBNs showed a prolonged burst discharge until the perturbed eye movement ended. To quantify this feature, we determined the duration of the motor burst for each individual saccade into the cell's movement field. For neurons showing clipped or partially clipped activity in the control condition (n = 13), the burst duration was measured from 20 ms before saccade onset to the time when the spike density fell below 1% of the peak value. To determine, in a comparable way, the duration of the high-frequency motor bursts in cells with an unclipped discharge pattern (n = 12), we first estimated the end of their high-frequency burst in the averaged control data by looking at the transition point between the high- and low-frequency discharge (see Fig. 6 for an illustration). The ratio between the peak spike-density and the spike density at the end of the burst, thus derived from the averaged control data, was subsequently used to detect the duration of each individual burst (for the different neurons, actual cutoff values ranged between 10 and 30% of the peak discharge rate). The latter was done by a computer algorithm that computed spike density functions with a  of 10 ms (see METHODS). Because of the reduced firing rates during perturbed saccades (see e.g., Figs. 2 and 3; see also the quantitative data in following text), this relatively wide kernel was needed to obtain a more robust detection in both the perturbed and unperturbed condition (same  used in both conditions and for all neurons). Although this procedure tends to overestimate the absolute duration of individual bursts (depending on the actual cutoff value), this overestimate is virtually fixed (within ~5 ms) across trials. It therefore has a negligible influence on the slopes and correlations of the linear regression analysis reported in the following text.

View larger version (23K): [in this window] [in a new window]   Fig. 6. Duration of the motor burst vs. saccade duration for 2 different SRBNs. A and B: burst duration of an SRBN (pj6701) that showed a clipped discharge. Data points in A scatter closely around a regression line with a slope of 0.90, indicating that the end of the motor burst in control trials (open circle) and perturbation trials (filled circle) was tightly coupled to saccade offset. As indicated in B, burst duration was measured from 20 ms before the saccade onset (vertical dotted line) to 1% of the peak discharge (horizontal dotted line; "1% level"). Plotted are the averaged radial eye position (E) and the spike-density waveform (D;  = 10 ms) of a series of control saccades corresponding to the neuron's optimum vector ([R, ] = [50, 0] deg). C and D: comparable data for the high-frequency burst of an SRBN (er0904) that showed long-lead presaccadic activity. Note the buildup of prelude activity and the gradual decline of postsaccadic activity after the burst. Burst offset was taken at 30% of the peak discharge ("30% level"), where the mean spike-density waveform showed an inflection. Also for this neuron (optimum at [R, ] = [12, 180] deg), a significant relation between saccade duration and burst duration was obtained. This was most evident from the cell's responses in the perturbed condition (filled circle; correlation coefficient r = 0.87; slope = 1.13). Slopes and correlations indicated in A and C are based on pooled data from the 2 conditions.

View larger version (30K): [in this window] [in a new window]   Fig. 7. Quantitative summary of the relationship between saccade and burst duration in the sample of 25 collicular neurons recorded in 3 monkeys. Slopes and correlations are based on pooled data from control trials and perturbation trials (typically, n > 50) and are pooled for the different SRBN subtypes. A: intermediate to high correlations were obtained for the majority of neurons (n = 20; r > 0.4; P < 0.001). No significant correlation was obtained for 5 cells (r < 0.2; P > 0.1). B: range of slopes obtained by fitting linear regression lines to the data (see Fig. 6). Slopes tend to cluster around 0.7-0.8.

BURST MAGNITUDE. The raw data in Figs. 1-3 illustrate that the discharge rates were substantially lower during blink-perturbed saccades. However, the impression was obtained that the number of spikes in the motor burst was approximately similar for perturbed and control saccades. To better quantify this property, we counted the number of spikes in each burst for saccades of optimal amplitude and direction using a time window that ranged from 20 ms before saccade onset to saccade offset. For comparison, the total number of spikes before, during, and after the saccade also were quantified by counting all spikes over a fixed 800-ms window starting 100 ms after target onset (which excluded visually evoked activity). Perturbation trials in which the saccade kinematics appeared to be hardly affected were excluded from this analysis (typically <5-10% of the trials).

View larger version (27K): [in this window] [in a new window]   Fig. 8. Similar number of spikes in the burst for control and perturbed saccades toward the movement field center illustrated for 2 different SRBNs. The burst was taken from 20 ms before saccade onset until saccade offset. The total number of spikes before, during, and after the saccade was quantified over a fixed period of 800 ms starting 100 ms after target onset. A-D: data obtained from a clipped SRBN showing little prelude activity (pj6701). Same neuron as in Fig. 6, A and B. Histograms A and B show the number of spikes in the burst for control and perturbed saccades of optimal amplitude and direction. Histograms C and D show the total number of spikes in the same trials (800-ms window). The average number of spikes in the burst was 17 ± 5 (mean ± SD; n = 18) and 23 ± 8 (n = 15) in A and B, respectively. The average total number of spikes in C and D was 19 ± 6 and 23 ± 6, respectively. E-H: similar data from an unclipped cell showing long-lead presaccadic activity (er0904). Same neuron as in Fig. 6, C and D. The average number of spikes in the burst was 26 ± 3 (n = 20) and 30 ± 3 (n = 10) in E and F, respectively. The average total number of spikes in G and H was 63 ± 7 and 62 ± 9, respectively. Note that the number of spikes in the burst for control saccades and perturbed saccades was very similar for both neurons. Note also that the total number of spikes appeared to be similar too and that the number of spikes in the burst was only a small fraction of the total number of spikes for the neuron shown in E-H.

View larger version (16K): [in this window] [in a new window]   Fig. 9. Number of spikes in the burst and mean firing rate in the burst for control and perturbed saccades of optimal amplitude and direction for all 25 fully tested cells. A: the average number of spikes in the burst showed relatively little change even though the spatial-temporal properties of the saccades were always substantially modified. In most neurons, the number of spikes in the burst increased only slightly (P < 0.01 for n = 7 cells; P > 0.1 for n = 15 cells), resulting in a slope of 1.14 of the regression line. Average spike counts were typically based on 10 trials, and in no case were <5 trials used. B: a significant reduction of the mean firing rate in the burst was obtained for most cells (n = 17; P < 0.01), resulting in a slope of only 0.48 of the overall regression line. This is readily understood from the fact that comparable spike counts were obtained for movements of much longer durations.

View larger version (38K): [in this window] [in a new window]   Fig. 10. Mean firing rate in the motor burst is correlated with mean eye velocity. A: results for individual responses in control () and perturbation () trials of cell pj6701 (burst type). B: same analysis for cell er0904 (buildup type). C: summary of the correlation coefficients obtained for all 25 fully tested cells. For the far majority of cells (22/25), the correlation was significant (P < 0.01). Population average was 0.67 (median: 0.72; range: 0.28-0.90). D: slopes of the regression lines for all 25 neurons vary substantially from cell to cell, although clustering occurs between 0.0 and 0.25 (median: 0.28 spikes/s per °/s; range: 0.04-1.68 spikes/s per °/s).

Intrasaccadic discharge in relation to the 2-D saccade trajectory

The observed linkage between saccade duration and burst duration (Figs. 6 and 7) could be expected if the SC is part of the local feedback loop that is thought to control the saccade trajectory (see INTRODUCTION). To further investigate this possibility, we therefore analyzed the intrasaccadic discharge of collicular neurons in relation to the 2-D saccade trajectory.

SPATIAL PROPERTIES. Figure 11 shows a typical example of the results of this analysis for an SRBN that was most active for upward control saccades (movement field center [R, ]  [35, 90] deg). Figure 11, A-C, shows the 2-D trajectories of a perturbed saccade and a control saccade toward the same target (Fig. 11A; T at [R, ] = [27, 60] deg) as well as simultaneous records of the cell's discharge during each of these movements (Fig. 11, B and C, respectively). Figure 11, D-F, shows data of two visually evoked control saccades that matched to the two subsequent eye-movement phases that were present in the perturbed saccade. Comparing the perturbed response in Fig. 11B to the control response in Fig. 11C, one may observe that the duration of the saccade-related discharge during the perturbed response exceeded the normal burst duration by far. The most important feature, however, is that the cell resumed its burst activity after the blink-related suppression even though the direction of the compensatory movement (rightward) did not match the direction of the cell's movement field (upward). The latter becomes evident by noting that the compensatory movement in Fig. 11B was almost equivalent (albeit slower) to the rightward saccade plotted in Fig. 11E (cf., Fig. 11, A and D) and by noting that the cell was not at all recruited for this control saccade. Because the neuron resumed its burst activity, we conclude that the compensatory movement was still part of the initial saccade program specifying an eye displacement up and to the right. If the compensatory movement had instead been generated by a new motor command specifying a rightward eye displacement, the discharge of the neuron should have stopped completely when the eye moved rightward (see Fig. 11E). Note that these results closely resemble the results obtained in the interruption paradigms used previously (see INTRODUCTION) except that in this case the direction of the compensatory movement has been altered too. The initial phase of the perturbed response is also quite interesting. In this part of the response, the eye movement was directed into the cell's movement field (upward). As may be observed in Fig. 11, A and D, the amplitude and direction of this initial movement were comparable with that of the upward control saccade. Note, however, that during the upward movement in Fig. 11B, the cell showed a near-complete cessation of activity (see first 50-60 ms), whereas the cell discharged vigorously for the upward control saccade plotted in Fig. 11F. Because the SRBN showed a near-complete cessation of activity even though the eye moved in the optimal direction of the cell, we conclude that the SRBN discharge did not mediate the upward trajectory perturbation. If the upward movement had instead been generated a (modified) collicular command specifying an upward eye displacement, the cell should have continued its vigorous discharge during the upward movement (see Fig. 11F). Summarizing these results, it appears that the actual 2-D trajectory of the perturbed saccade was not reflected in the SRBN discharge pattern.

View larger version (18K): [in this window] [in a new window]   Fig. 11. SRBNs resumed their burst activity irrespective of the instantaneous eye-movement direction. Example from neuron pj4204. A: 2-dimensional (2-D) trajectories of a control saccade and a perturbed saccade of equal amplitude and direction evoked from the fixation point (F) toward a briefly flashed visual target (T, presented at [R, ]] = [27, 60]). The movement field optimum was at [R, ]  [35, 90]. B and C: plots of eyelid (L, vertical) and eye (H, horizontal; V, vertical) position as function of time for the 2 movements shown in A. Simultaneous records of the neuron's activity are plotted underneath. Shown are the spike density functions (D;  = 4 ms) and individual action potentials (vertical lines). D-F: similar data for an upward and a rightward control saccade matched, respectively, to the initial upward and subsequent rightward eye-movement phase of the perturbed saccade shown in A and B. Note that the burst activity of the cell was transiently suppressed (see 1st 50-60 after saccade onset in B) even though the perturbed saccade consisted of an initial upward movement that was directed into the cell's movement field (see burst in F). Note also that the neuron subsequently resumed its burst discharge until the eye had reached the target (see B) despite the increase in movement duration with respect to the control saccade in C and despite the fact that the rightward, compensatory movement fell outside the cell's movement field (see absence of activity in E).

DISCHARGE DYNAMICS. For all neurons that could be tested in detail, phase plots were constructed of instantaneous spike density versus radial motor error (ME; difference between desired and current eye displacement) by following the same method utilized in previous studies (Keller and Edelman 1994; Munoz et al. 1996; Waitzman et al. 1991). The majority of these cells (n = 17) showed a monotonic decline of spike density as function of decreasing radial motor error for control saccades toward the center of their movement field and were therefore subjected to a further analysis. Since blinks not only modified the saccade kinematics but also changed the 2-D saccade trajectories, additional phase plots were made of spike density versus the respective horizontal and vertical motor-error components. Spike density functions were computed with a  of 4 ms and shifted backward in time by 4-8 ms to obtain the best linear decline as function of radial motor error for control responses (Waitzman et al. 1991).

View larger version (33K): [in this window] [in a new window]   Fig. 12. Typical discharge dynamics in relation to the 2-D trajectories of visually evoked saccades, illustrated for 1 of the recorded SRBNs (pj5501). A: 2-D trajectories of a series of control saccades and blink-perturbed saccades toward targets flashed at [R, ] = [14, 60]. Amplitude and direction of the saccade vectors closely matched the cells' optimum vector, but the compensatory movement during perturbed saccades was approximately orthogonal to the control trajectories. B: spike density functions (D;  = 4 ms) and horizontal (H) and vertical (V) eye position components as function of time. Control movements and spike density functions are cutoff for clarity. Note sustained discharge during perturbed saccades until eye-movement offset even though the compensatory movements were made outside the cell's movement field (right and upward; not indicated). C-E: phase plots of spike density versus radial, horizontal, and vertical dynamic motor error (ME) components, respectively. Dynamic motor error is the difference between saccade endpoint and instantaneous eye position (HME = Ht  Hend; VME = Vt  Hend; RME = ). The curves begin at the largest motor error values and proceed, in a counterclockwise loop, toward 0. Time is implicit in these plots. The spike density functions were shifted backward in time by 7 ms to obtain the best linear decline as function of radial motor error for control saccades. Note scale differences. Each panel shows the sampled records of individual perturbed responses (dots at a time resolution of 2 ms). Solid lines show the averaged control responses (thin lines) and perturbed responses (thick lines). Dashed lines in C-E denote standard deviations of the averaged control curves. Note the monotonic, nearly linear, relation with each motor error component for control saccades. Also note quite different and complex, nonmonotonic curves in the case of perturbed movements. The final portion of these curves tends to coincide with the control curves in C and D but not in E due to the highly curved nature of the saccade trajectories (see also Goossens and Van Opstal 2000). Hence, the neuron showed a comparable activity near the end of the perturbed and control movements, even though the instantaneous eye-movement direction was distinctly different (see A).

View larger version (51K): [in this window] [in a new window]   Fig. 13. Relation between instantaneous spike density and dynamic motor error was different under the 2 experimental conditions. A-F: phase plots of intrasaccadic spike density as function of dynamic motor error for 2 representative neurons endowed with different presaccadic and postsaccadic discharge properties. Same layout as Fig. 12, C-E. A-C: spike density of a clipped SRBN (pj5203) as function of dynamic motor error. The activity of this neuron as function of time is shown in Fig. 2 (same responses, except the 2 with a late perturbation). D-F: similar data for an unclipped SRBN (er1101) endowed with long-lead presaccadic activity. Figure 3 shows the temporal discharge pattern of this neuron during the same responses (except the 1 response with a late perturbation). Note clearly different phase curves for control saccades (thin curves) and perturbed saccades (thick curves) in all 6 panels but qualitatively similar results for the 2 SRBNs. G: normalized phase curves for the 17 neurons analyzed in this way. Dashed curves: mean (thick) and standard deviation (thin) of the pooled control data from all 17 neurons. Solid curves: averaged data from each individual neuron (thin) in the perturbed condition with the population mean (thick) superimposed. Spike density was resampled as function of radial motor error to account for variability in saccade kinematics (see METHODS). H: spike density during perturbed saccades vs. spike density during control saccades at the same radial ME for each individual neuron (thin curves) with the population mean (thick curve) superimposed. Radial motor error and time are both implicit. Note that virtually all curves are nonmonotonic and that there are considerable deviations from the identity line (dashed).

Activity during saccades showing no compensation

So far we have presented an analysis of the SRBN discharge properties during saccades that compensated for the blink-related perturbations. Typically the SC cells remained active throughout these movements, irrespective of the instantaneous eye-movement trajectory. It is therefore of interest to also consider what happened in cases where compensation was absent (see also Goossens and Van Opstal 2000). Although this latter response mode was quite