| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Laboratory of Sensorimotor Research, National Eye Institute, National Institutes of Health, Bethesda, Maryland 20892-4435
Submitted 31 July 2003; accepted in final form 15 October 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
The principle of corollary discharge (Sperry 1950
) is that when a brain region sends a motor command downstream, it simultaneously sends a copy upstream to be used as information about the movement (Fig. 1A). As a way to monitor movements, corollary discharge (or efference copy) (von Holst and Mittelstaedt 1950) has two main advantages over sensory cues: it provides information even before movement onset and it does not rely on the integrity of sensory receptors. The concept of corollary discharge has helped to answer questions as diverse as how crickets chirp without damaging their hearing (Poulet and Hedwig 2002) and why we can tickle others but not ourselves (Blakemore et al. 2000). Impaired corollary discharge may contribute to pathologies including schizophrenia (Feinberg and Guazzelli 1999; Ford et al. 2001) and may underlie many of the sensory and motor deficits that follow brain damage (Angel 1980; Baizer et al. 1999; Duhamel et al. 1992b; Gaymard et al. 1994; Haarmeier et al. 1997; Heide et al. 1995; Rafal 1994; Versino et al. 2000). Neurons in several parts of the primate brain seem to receive corollary discharge because their visual responses change just prior to movements (Duhamel et al. 1992a; Reppas et al. 2002; Richmond and Wurtz 1980; Thiele et al. 2002; Tolias et al. 2001; Umeno and Goldberg 1997).
|
), we described a pathway from superior colliculus (SC) to frontal eye field (FEF; Fig. 1B) that transmits a variety of signals including bursts of activity just prior to contraversive saccades (i.e., saccades made into the contralateral visual field). This activity provides information about when a saccade will occur and where it will go. Because it travels away from the eye muscles and up to the FEF, a cortical region involved in visual analysis and saccadic planning (Schall 1997), we hypothesized that the activity was a corollary discharge.
The goal of the present study was to test this hypothesis by examining the consequences of inactivating the pathway. Signals sent from the SC to the FEF are relayed by neurons in the mediodorsal thalamus (MD), something, from an experimental point of view, that is highly fortuitous; shutting off those neurons should interrupt signal flow in the pathway without directly affecting either the pathway's source, the SC, or its target, the FEF (Fig. 1B). We injected the GABAA agonist muscimol at the sites of previously recorded MD relay neurons (Fig. 1C) (Lomber 1999). This should temporarily inactivate the neurons because GABAA receptors are found throughout MD (Steriade et al. 1997).
We reasoned that if the pathway conveys corollary discharge signals, then during MD inactivation, a monkey should be impaired at keeping track of its contraversive saccades. One task that requires the internal monitoring of saccades is the double-step task (Becker and Jürgens 1979; Hallett and Lightstone 1976; Mays and Sparks 1980), which has been used extensively in prior studies to test for corollary discharge deficits (Duhamel et al. 1992b; Gaymard et al. 1994; Heide et al. 1995). In this task, subjects make saccades sequentially to the locations of two flashed targets (Fig. 2A). To make the second saccade correctly, a subject must keep track of where its first saccade goes; thus deficits in monitoring first saccades should disrupt second saccades. Visual feedback is unavailable because the targets disappear before the eyes move, and information from eye muscle proprioception seems to be negligible (Bridgeman 1995; Guthrie et al. 1983; Lewis et al. 2001; Steinbach 1987). Keeping track of the first saccade should depend primarily on monitoring the corollary discharge of that saccade.
|
These corollary discharge deficits, although not directly observable, should be detectable in the double-step task through their influence on second saccades. To illustrate this, Fig. 2C depicts a monkey's internal estimate of its behavior (top) and its actual behavior (bottom). The monkey's goal is to make a second saccade that reaches the second target location, and to do this, it must monitor its corollary discharge representation of the first saccade. From trial to trial, the monkey will monitor its corollary discharge signals and make adjustments so as to keep its second saccade end-points constant (Fig. 2C, top), but because the first saccade does not really change, these adjustments will actually introduce errors into the second saccade end points (Fig. 2C, bottom). Compared with the normal behavior (Fig. 2C, bottom left), if there is a corollary discharge accuracy deficit, the second saccade end points will shift systematically (bottom middle), and if there is a corollary discharge precision deficit, these end points will become more scattered (bottom right).
Finally, a critical expectation is that these changes in second saccades should be lateralized, occurring only in trials involving contraversive first saccades. This is because the putative corollary discharge signal (presaccadic activity in the SC-MD-FEF pathway) is related almost exclusively to contraversive saccades (Sommer and Wurtz 2004b).
We found that our hypothesis was supported: MD inactivation caused contralateralized deficits in both the accuracy and the precision of corollary discharge. This provides direct evidence that the presaccadic signals sent from SC to FEF represent corollary discharge information about saccades. A brief report regarding the accuracy deficit was published previously (Sommer and Wurtz 2002).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
We used monkeys B and C from our study of MD relay neurons (Sommer and Wurtz 2004b). Both monkeys had been implanted with scleral search coils for measuring eye position, a post for holding the head, and chambers for accessing the brain. All procedures were approved by the Institute Animal Care and Use Committee and complied with Public Health Service Policy on the humane care and use of laboratory animals.
Injection sites
We injected at a site only if it met three criteria: past recordings at the site must have yielded identified MD relay neurons (i.e., only the sites shown in Fig. 1C were considered), recordings the day before injecting at the site had to demonstrate strong multiunit activity, indicating that the site was still healthy, and electrical stimulation at the site had to evoke saccades because we used changes in stimulation-evoked saccades to assess whether muscimol was successfully injected (see following text). At each site, we injected either muscimol dissolved in saline (5 µg/µl) or saline alone as a control for incidental effects (e.g., pressure or dilution) that might be caused by liquid infusion. Table 1 lists all the experiments (1st column) and the agent injected in each (2nd column). One site in monkey C (Fig. 1C, left, 
) was extraordinarily well suited for inactivation, as it met all three criteria and, moreover, was where we had found a particularly rich concentration of MD relay neurons [cf. Fig. 2A, left, hatched circle, in Sommer and Wurtz (2004b)]; the surrounding sites were less attractive for injecting (less robust neuronal activity, no saccades evoked, or both). We restricted our injections in monkey C to this one site to minimize inactivation of MD areas that did not contain relay neurons. As shown in Table 1, we performed different tests during each inactivation at this site. In monkey B, we injected at three sites that met our criteria (
, 
, and 
in Fig. 1C, right). We terminated the study once we ran out of sites that met our criteria for inactivation. Although the number of inactivation sites was limited, it was nonetheless sufficient to demonstrate significant, lateralized corollary discharge deficits in both the pooled data and in each monkey individually as reported in the RESULTS.
|
We inserted a 30-gauge cannula through a 23-gauge guide tube into the brain until the cannula tip was at the depth of previously recorded MD relay neurons. The other end of the cannula was attached to a 10-µl Hamilton syringe, and we injected muscimol manually in boluses of 0.2 µl every 30 s. A major technical issue was ensuring that muscimol was ejected from the cannula. We aborted the experiment if resistance was felt that might indicate a plugged cannula tip or if muscimol ejection was not verified by changes in stimulation-evoked saccades. Four injections (2 in each monkey) were aborted for these reasons and are not included in this report. An insulated wire threaded through the cannula, with its exposed tip extending 500 µ beyond the cannula tip, allowed us to electrically stimulate at each injection site (Crist et al. 1988). We found that stimulating at relay neuron sites (biphasic cathodal-anodal pulses, 0.25 ms/phase, 350-Hz, 150-ms trains) often evoked saccades having scattered vectors (not stereo-typed as when stimulating SC or FEF) after long latencies (mean: 96 ms after stimulation onset) and at moderate current thresholds (mean: 36 µA); other stimulation results are beyond the scope of this report. Our indicator of successful muscimol ejection from the cannula was a marked increase in current threshold (typically >2 times the initial level) around 20-30 min after the injection, when muscimol typically begins having strong effects (e.g., Aizawa and Wurtz 1998). This rise in threshold may have been due in part to tissue displacement; we did not systematically compare the effects of muscimol versus saline injection to test this. Regardless of the exact mechanism(s) causing it, the increased threshold confirmed that liquid had been ejected into the brain. The increase in current threshold marked the start of the "during" inactivation period discussed throughout this report. The current threshold increase always persisted throughout the session, paralleling the multiple hour time course of muscimol (e.g., see Lomber 1999).
General protocol
The day before an injection we recorded at the targeted site to ensure that neurons were active there and collected "before" behavioral data. The day of the injection we collected baseline stimulation data, injected muscimol and collected stimulation data to verify its delivery, and collected during behavioral data. On a later day (usually the next day) we collected "after" behavioral data to measure recovery.
Experimental apparatus
During each experiment a monkey faced a tangent screen onto which visual stimuli were projected. Ambient illumination was dim light or darkness. In dim-light experiments, we used an LCD projector that even at its darkest setting illuminated the tangent screen with a diffuse gray background (0.1 cd/m2). Visual stimuli were a blue fixation spot and red targets (0.6 cd/m2). In darkness experiments, we used lasers to produce red dots for the fixation spot and targets. There was no other source of light in the room during each trial, and between trials a lamp was illuminated to prevent dark-adaptation. We sampled eye position and recorded task events every 1 ms.
Double-step task
The purpose of this task was to test whether MD inactivation impaired corollary discharge (see INTRODUCTION). A monkey fixated a spot, the spot disappeared, and two targets briefly appeared in sequence (Fig. 2A). To receive reward (a drop of water), the monkey had to make two sequential saccades to the locations of the targets. The timings of target presentation (specifically, the 1st target duration, the gap between offset of the 1st target and onset of the 2nd, and the 2nd target duration) varied depending on the target pair and the monkey because we always fine-tuned the timings until errors were minimized. For example, if for a certain target pair a monkey often erred by making its first saccade to target 2 instead of target 1, we lengthened the target 1 duration, shortened the target 2 duration, and changed the gap between targets 1 and 2 until the errors were minimized. The only constraint was that the overall sum of the target 1 duration, the gap between targets 1 and 2, and the target 2 duration be less than 
180 ms, approximately the minimum reaction time of first saccades relative to target 1 onset. This constraint was imposed because we required all visual stimuli to disappear before the first saccade began so that there would be no visual feedback from the targets that might inform the monkey whether the first saccade was made or where it landed. In practice, the typical target 1 durations were 130 ms, the gap was 20 ms, and the target 2 duration was 30 ms. On-line, we used a real-time saccadic velocity detector to abort trials in which the first saccade began too early; off-line, we inspected every trial to make sure that the more precise saccade detection algorithm in our analysis program agreed with this on-line analysis, and we eliminated further trials as needed. Spatially, we randomized a variety of target pairs as will be explained in RESULTS.
Single-step task
The purpose of this task was to see if inactivation affected the generation of saccades. A monkey fixated a spot, the spot disappeared, a gap of 150 ms ensued, and then a single target appeared. The monkey had to make one saccade directly to the target. In a "target-present" version of this task, once the peripheral target appeared, it stayed lit until after the monkey made a saccade to it. In contrast, a "target-absent" version involved carefully chosen parameters so that the target presentation closely matched the presentation of target 2 in the double-step task. We did this to see how monkeys responded to a flashed target when there was no intervening first saccade, i.e., when corollary discharge was not a factor. The 150-ms gap approximated the target 1 duration plus the gap between targets 1 and 2 in the double-step task, and when the target appeared, it stayed lit for the same length of time (typically 30 ms) that target 2 would have appeared in the double-step task. Because of the brief target flash, the saccade did not begin until 120 ms after the target disappeared (saccadic reaction times were 150 ms), and thus the saccade went to the remembered location of the target. This memory requirement was nearly identical to that required for making saccades to target 2 in the double-step task. Spatially, we randomized a variety of target locations (see RESULTS).
Analysis
On-line, in the double-step task we used rather large virtual windows around the targets for judging whether saccades were correct, because we did not want monkeys to receive feedback (due to loss of reward) that their behavior had changed if their second saccades became drastically different due to corollary discharge impairment. Typically the windows were 10° horizontally by 10° vertically around target 1 and 20 x 20° around target 2. Initial fixation windows were typically 5 x 5° and sometimes a bit larger; this was somewhat liberal but necessary because our monkeys seemed to find it difficult to sustain their fixation on the laser spots in total darkness, particularly when fixation was required at eccentric locations. In the single-step task, we used smaller windows, typically 5 x 5°, around the target.
Off-line, we detected saccades in the data with software that used a template-matching algorithm. We manually reviewed every trial to verify the saccade detection and to remove error trials. Correct trials were those in which the first and second saccades landed in well-defined clusters of end points within a few degrees (exact distance depended on the target configuration) of the first or second target location, respectively. Most error trials (see Fig. 13C) were obvious by inspection and included 20-35% of trials in which the first saccade went to the second target, 5-10% of trials in which the first saccade was correct but the second saccade did not occur or went to the wrong location, 10% of trials in which the first saccade went nowhere near the first or second target, and very rare saccades that were curved or otherwise atypical (e.g., the strongly curved upward 2nd saccade in Fig. 3A, bottom). The only ambiguity in identifying error trials came from first saccades that landed between the first and second targets ("averaging" saccades); we decided to accept such saccades as correct if they landed closer to the first than the second target.
|
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Figure 3 shows data from an example inactivation. The monkey had to make a rightward saccade and then an upward saccade in response to two flashed targets. Figure 3A, top, shows all trials performed by the animal before MD was inactivated. Most saccades were made in the appropriate sequence, going toward the first target and then the second. During MD inactivation (Fig. 3A, bottom), the monkey's behavior was nearly the same except that the second, upward saccades seemed to land, on average, at more contralateral locations. We selected from these raw data only trials in which the monkey made correct sequences (Fig. 3B; see METHODS), and in Fig. 3C, we summarize these correct trials by showing the means and SDs of the initial fixations, first saccade end points, and second saccade end points before and during inactivation. The most obvious effect of the inactivation seemed to be a roughly horizontal shift in second saccade end points. We examined this quantitatively by testing whether the end points were significantly different during versus before inactivation along the horizontal dimension and, for comparison, along the vertical dimension (criterion level P < 0.025, Bonferroni corrected from P < 0.05 due to testing in 2 directions). Figure 3C shows that in this example case, there was indeed a significant contraversive shift in second saccade end points. In contrast, neither the initial fixations nor the first saccade end points shifted in either the horizontal or vertical direction.
We note that in Fig. 3 and in subsequent examples it is obvious that the saccadic sequences often were somewhat inaccurate relative to the actual target locations. This was because, first, upward shifts sometimes occurred as a normal behavior of the monkeys. Such shifts are common when saccades are made to remembered target locations in darkness (Gnadt et al. 1991), and in our experience, this happens even in dim light for some monkeys (e.g., monkey C). Second, we purposely avoided over-training the monkeys. If they were to establish and execute programmed sequences of saccades as a rote response to a particular target pair stimulus (e.g., Kowler 1990), they could then perform the task without continuously monitoring their first saccades. We therefore changed target configurations frequently and allowed monkeys to practice on them for only a few days, at most, before each injection experiment.
Figure 4A shows the three target configurations we used (left) and example saccadic sequences generated by the monkeys in response to each (right). Each configuration involved horizontal first saccades followed by vertical or oblique second saccades. In each injection experiment, we used a single configuration with all the target pairs of that configuration randomized by trial. The first two configurations, A and B (Fig. 4A, top and middle), involved a central fixation point and centrifugal first saccades. The advantage of these two configurations was that the sequence was unpredictable at the start of fixation; the disadvantage, however, was that second saccades began at an eccentric location. Second saccades were the most crucial movements for us to measure, so it would be preferable if they started at the center so that they could be measured optimally (maximum linearity of the eye coil system was in the center) and so that their trajectories would be unaffected by mechanical limits at the orbital extremes. Thus in a third configuration, C (Fig. 4A, bottom), we used centripetal first saccades and second ones starting at the center of the screen. The disadvantage of this configuration was that it introduced predictability into the paradigm (the initial fixation location provided information as to whether the sequence would be rightward or leftward), so to counter this problem, we introduced extra randomness by adding many more second target locations. Each configuration, A, B, or C, resulted in a different pattern of saccadic sequences that approximately matched the arrangement of targets in the configuration (Fig. 4A, right), and thus it appeared that the monkeys did try to perform the sequences correctly even though hampered by vertical upshifts and only limited training, as noted in the preceding text. Second saccades clearly were attempts to reach the second target locations and were not just default saccades.
|
|
Our first observation as shown in the preceding examples was that MD inactivation often caused a systematic shift in second saccade end points indicative of a corollary discharge accuracy deficit. Next we will document this effect quantitatively.
TRIALS INVOLVING CONTRAVERSIVE FIRST SACCADES. Figure 6 summarizes the results by pooling data from all injections and both monkeys. In each experiment, the data from a single target pair represented one "case" for analysis; e.g., one case was shown in Fig. 3C and six more in Fig. 5, A-F. Combining all the inactivation experiments, there were 22 cases involving contraversive first saccades. In nearly every case (82%, 18/22), there was a contraversive shift in second saccade end points during MD inactivation (Fig. 6A, left). Shifts were individually significant in half of the cases (11/22) and the average shift was significantly greater than zero. The shifts were approximately the same in each monkey (Fig. 6A, left, inset graphs) and they occurred regardless of whether the task was performed in total darkness or in dim light (not shown; mean shifts were 1.12° in n = 16 dark cases and 1.13° in n = 6 dim light cases).
|
As noted in the preceding text, the contraversive direction of the shifts implied that the underlying corollary discharge vectors shrank (Fig. 2C, middle). It follows that the maximum possible contraversive shift should equal the first saccade amplitude (10° in target configuration A, 20° in configurations B and C). To estimate the severity of the deficit, we compared the observed shifts to these maximal shifts. As reported previously (Sommer and Wurtz 2002), in the 11 cases showing a significant shift (Fig. 6A, left), there was a 19% average deficit. Considering all 22 cases, there was a 10% average deficit.
IPSIVERSIVE SACCADES AND SALINE CONTROLS. We analyzed two other types of trials for comparison with the preceding results. First, because MD relay neurons represent contraversive saccades almost exclusively (Sommer and Wurtz 2004b), we expected MD inactivation to have little or no effect on corollary discharge of ipsiversive saccades. Second saccades following ipsiversive first saccades, therefore, should be unperturbed during MD inactivation, and that is what we found. Figure 6D, left, shows the distribution of horizontal shifts in second saccade end points for all 22 ipsiversive control cases; here, a corollary discharge deficit would be indicated by an ipsiversive shift. There were a few more individually significant ipsiversive shifts than contraversive ones (5 vs. 2) and the overall mean trended toward the ipsiversive direction, but this was not significant (P > 0.025). As a second point of comparison we expected that trials involving contraversive first saccades should be unaffected if only saline, sans muscimol, was injected; this too was found to be true (Fig. 6D, right).
ANGLE OF THE SHIFT IN SECOND SACCADE END POINTS. The average shift during inactivation, in trials involving contraversive first saccades, was significant horizontally but not vertically (Fig. 6A, left), but that does not necessarily mean that the exact angle of the shift was horizontal. If the shift angle were appreciably different from horizontal, it would imply that the underlying corollary discharge signal not only shrank on average but also rotated. To quantify the exact shift angle, in each case we connected the means of the second saccade end points before and during inactivation with a line. The angle of the line relative to horizontal (defined as 
) was the shift angle while the length of the line (defined as D) was the shift distance. For the case shown in Fig. 3C, for example, the line had an angle = +25.9° and a length D = 2.3°. We treated each line as a vector and plotted all of them in Fig. 7. Vectors from inactivation experiments in which the first saccade was contraversive are shown in Fig. 7A (the vector corresponding to the Fig. 3C example is shown with a dashed arrow). The overall mean vector (larger, white arrow) was highly significant (P < 0.01), as zero fell outside the 99% confidence ellipse regarding its tip location. This mean vector had = +5.0° and D = 1.1°. Because the average shift was nearly horizontal, the underlying corollary discharge signal did seem to be shortened but not appreciably rotated. In contrast, for the cases involving ipsiversive first saccades (Fig. 7B) and for the saline controls (Fig. 7C), the mean vectors were not significant (the 95% confidence ellipses included 0), which confirms that these cases exhibited no overall shift.
|
We expected that a precision deficit in corollary discharge would increase the scatter of second saccade end points (Fig. 2C, right), but this did not occur (Fig. 8A); in trials involving contraversive first saccades, the SD of second saccade end points did not change during inactivation (paired t-tests, P = 0.36 for horizontal SD, 0.62 for vertical). If there was increased scatter, it was not detectable against the background scatter. A subtle deficit was suggested, however, by the fact that in Fig. 8A more data points fell above the unity line than below it (25 vs. 19). This prompted us to try a different technique for identifying a precision deficit.
|
To measure how well corollary discharge matches the first saccade on a trial-by-trial basis, once again we used second saccades as our behavioral indicators. Figure 8B shows one case of saccadic sequences during and before inactivation, and C shows each individual second saccade vector. To see if the monkey detected where each second saccade started (i.e., where each 1st saccade landed), we replotted all the second saccades in Fig. 8D by spreading them out along the vertical dimension while maintaining their relative horizontal locations (they are arranged so that the saccade starting most to the left is at top and that starting most to the right is at bottom—not chronologically by trial number). At least before inactivation (Fig. 8D, left), the monkey did act as if it knew where each second saccade began; moving from left to right, the vectors tend to swing counterclockwise as if compensating for the different initial positions. In contrast, during inactivation (Fig. 8D, right), the monkey showed little evidence of knowing where each first saccade landed; all the angles of the second saccades were about the same regardless of where the saccades began.
To quantify this effect, we compared the observed second saccade vectors (Fig. 8C) with the ideal vectors that would have been expected from perfect compensation. Ideal vectors were modeled (Fig. 9A) by connecting the initial position of each observed second saccade to the mean end-point location of the second saccades (we assume this mean location was where the monkey intended its saccades to land). For quantification, we measured saccadic direction (
) and amplitude (
; see Fig. 8D). First we evaluated how well the monkey adjusted its second saccade directions. Figure 9B plots observed versus ideal for every trial. Before inactivation (bold circles and line), the values were highly correlated with a linear regression slope near 1.0. Thus the monkey was quite adept at adjusting its second saccade directions, implying that it knew where each first saccade landed; corollary discharge was precise. During inactivation (thin triangles and line), however, precision seemed impaired because the correlation became insignificant. Next we evaluated how well the monkey adjusted its second saccade amplitudes. Figure 9C shows that the monkey was poor at this both before and during inactivation (observed was not correlated with ideal ). The criterion level for a significant correlation was P < 0.025 because we looked for changes in both direction and amplitude of the saccades.
|
(Table 2, 1st row from top) and also in the average slope of the regression between observed and ideal (Table 2, 2nd row). Figure 9D (left) graphs this change in slope and, for comparison, the insignificant changes for ipsiversive first saccade trials (middle) and saline control trials (right). No other significant effects were found (Table 2).
|
As an aside, these results also demonstrate that the monkeys did not preprogram their saccadic sequences. The monkeys normally tailored the direction of their second saccades to compensate for changes in their first saccade end points, meaning that they only made a second saccade after learning about the first one. Our strategies of not over-training the monkeys and frequently changing the target configurations therefore seemed to successfully prevent preprogramming.
In sum, the precision analysis revealed both a positive and negative result. The positive finding was twofold: monkeys normally adjusted their second saccade directions from trial to trial to compensate for slight fluctuations in first saccades, and this was disrupted by inactivation. This shows that corollary discharge is normally precise and that MD inactivation impairs this precision. The negative finding was that monkeys showed little ability to adjust their second saccade amplitudes. This may explain why we found no precision deficit in our initial SD analysis (Fig. 8A). Second saccade amplitudes both before and during inactivation were nonideal— essentially "noisy"—and this may have contributed so much scatter to the end points that SD increases caused by an impaired ability to adjust second saccade directions could not be detected. Only by analyzing second saccade directions in isolation could we uncover the precision deficit.
Generation of single saccades
In the preceding sections, we considered whether corollary discharge, i.e., information about saccades, was impaired by MD inactivation; next we consider whether saccade generation itself was impaired. In brief, it was not. As shown in the preceding text, inactivation had little or no effect on the generation of contraversive first saccades in the double-step task (Fig. 6B), and previously (Sommer and Wurtz 2002) we showed one example of unaffected saccades made in the single-step task, demonstrated that the dynamics of single saccades were unaffected, and reported that there were no overall accuracy or latency deficits. Here we illustrate saccades made in a different version of the single-step task and provide complete accuracy, precision, and latency results.
Figure 10A shows an example experiment that was notable because it contained the largest changes in single saccades during MD inactivation that we ever found. In contrast to the previously published example that showed saccades from the target-present version of the task (Sommer and Wurtz 2002), this example (Fig. 10A) shows saccades from the target-absent version. The target-absent version was more challenging in that the target was flashed only briefly and its location had to be remembered during the reaction time (see METHODS). In each trial, the monkey made a single saccade from the center to a target appearing at one of the locations shown in Fig. 4B (configuration B), and shown are the means and SDs of the saccadic end points before and during MD inactivation along with the radial distance (D value) of each shift. To see if the shifts were significant, we tested them along the two cardinal axes, as we did in the double-step task, and superscripts following the D values indicate end points that were significantly shifted horizontally (h) or vertically (v). As can be seen, shifts were small in magnitude, primarily vertical in direction, and about as likely to affect contraversive as ipsiversive saccades. Figure 10B shows the distributions of shift sizes of contraversive saccades (top) and ipsiversive saccades (bottom) from all the single-step cases. The average shift size was not significantly different between contraversive and ipsiversive saccades, and a few individually significant horizontal or vertical shifts were seen for both contraversive and ipsiversive saccades. The precision of single saccades was not affected by inactivation either (Fig. 10C; all paired t-tests of during-before changes yielded P > 0.025). Unlike in the double-step task, we saw no alternative way of testing for a precision deficit; it is possible there was a small deficit in the precision of saccade generation that was undetectable. Finally, reaction times were not affected (Fig. 10D; all paired t-tests yielded P > 0.05).
|
; Hikosaka and Wurtz 1985) and in the FEF (Dias and Segraves 1999; Sommer and Tehovnik 1997), presumably due to silencing these structures' descending output neurons (Everling and Munoz 2000; Munoz et al. 1991; Segraves 1992; Segraves and Goldberg 1987; Sommer and Wurtz 2000, 2001). Alternative explanations
Returning to the double-step task results, we think that corollary discharge deficits provide the simplest explanation for the second saccade disruptions in trials involving contraversive first saccades. Nevertheless, alternative explanations should be considered.
WERE SECOND SACCADE END-POINT SHIFTS DUE TO FIRST SACCADE END-POINT SHIFTS? Figure 6B (left) showed that in one case there was a significant contraversive shift in first saccade end points during MD inactivation, and Fig. 11A (left) shows the average saccades from this case. The slight (0.8°) but significant shift in first saccade end points may have contributed to the significant shift in second saccade end points of 1.8°. To see if there was an accumulation of horizontal error from first to second saccades in general, we plotted the horizontal shift in second saccade end points as a function of the horizontal shift in first saccade end points (Fig. 11A, right). If the errors accumulated, then larger first saccade shifts should have tended to produce larger second saccade shifts; the data should be directly correlated. There was no correlation, however (P = 0.888). First saccade end-point shifts did not seem to cause second saccade end-point shifts.
|
and ideal for these cases decreased from an average of 0.82 before inactivation to 0.28 during inactivation, but after inactivation the average slope rose to 0.81, representing almost perfect recovery (Fig. 11B, right). WERE SECOND SACCADE END-POINT SHIFTS DUE TO GENERAL ROTATIONS OF ALL SACCADES? It could also be that the second saccade end-point shifts were an artifact of a direction deficit in saccade generation; maybe all saccades rotated contraversively. For example, we assumed that the second saccade end points in Fig. 12A (same as in Fig. 5C) shifted rightward because of inaccurate corollary discharge. It could be, however, that any saccade made in an obliquely upward direction would have rotated to the right. To test this, in two of the single-saccade control tasks discussed in the preceding text, we had targets arranged (configuration B in Fig. 4B) so as to match the double-step target arrangement (configuration C in Fig. 4A) during the same injections, and we asked whether single, obliquely upward saccades also were rotated. They were not. Figure 12B shows the single saccade control that matched the trial type of A. The single saccades were much more accurate than the second saccades of the double-step task, of course, as would be expected. During inactivation, there was no rotation and consequently no contraversive shift (Fig. 12B, top); this was always the case (bottom).
|
WERE PRECISION DEFICITS DUE TO IMPRECISE VISION OR MEMORY? An alternative explanation for the precision deficits also posits a visual or memory problem. Perhaps corollary discharge precision was normal during inactivation, but the precision of detecting or remembering the target location was impaired; this also could have caused second saccade directions to deviate from the ideal during inactivation as in Fig. 9B. We found no evidence for this, however. Such impairments should have increased the scatter of second saccade end points, but this did not occur [Fig. 8A; testing for such a visual or memory deficit was in fact the original motivation for analyzing these data in Sommer and Wurtz (2002)]. As a follow-up, we looked at precision in the target-absent version of the single-step task. Detecting and remembering the target in this task should be of similar difficulty as detecting and remembering the second target in the double-step task. Precision in the single-step task, however, was unaffected by inactivation (Fig. 10C; nearly all the data are from the target-absent version). From what we can tell, therefore, the precision of detecting and remembering the second target remained normal. The best explanation for the deficit in compensation (Fig. 9B and D, left) is that the precision of corollary discharge was impaired.
Bilateral impairments
All the deficits in the double-step task discussed thus far affected only trials in which the first saccade was contraversive, but other, bilateral, impairments also occurred.
INCREASED REACTION TIMES. Reaction times increased almost completely bilaterally during inactivation. Figure 13A (left) shows that for all double-step trials pooled together (correct + wrong), mean reaction times of first saccades increased significantly regardless of whether they were contra- or ipsiversive. Just considering correct trials, reaction times of first saccades increased slightly contraversively but not ipsiversively (Fig. 13A, 2nd graph from left) while inter-saccadic intervals (2nd graph from right) and reaction times of second saccades (right) both increased bilaterally.
DECREASED PERCENT CORRECT. Recall that correct trials in the double-step task were those in which two saccades were made sequentially toward the target locations. The percentage of such trials relative to all trials was the percent correct. We found that percent correct decreased bilaterally during inactivation (Fig. 13B). To better understand this, we analyzed which kinds of error trials increased. A major, bilateral increase occurred in error trials in which the first saccade went to target 2 and stayed there until the trial ended (Fig. 13C, top and bottom, leftmost bars). We think this was related to the bilateral increase in first saccade reaction time (Fig. 13A, left) because longer reaction times in double-step trials are associated with errors in which first saccades go to the second targets (Sommer and Tehovnik 1999). Two other increases in errors also occurred in trials involving ipsiversive first saccades (Fig. 13C, bottom): errors in which first saccades were correct but then no second saccades appeared (middle bars) and errors in which the first saccade headed to neither the first nor the second target (rightmost bars).
GENERAL LETHARGY. The monkeys occasionally seemed to become lethargic during inactivation. We were very familiar with the monkeys' behaviors, so when the following effects occurred less than an hour after muscimol injection (once with monkey B, twice with monkey C), we thought they were striking: the monkeys shut their eyes, ceased to work even for greater reward, and slept. When this occurred, we aborted the experiment and returned the monkeys to their cages. The next day the monkeys were completely recovered. None of the data in this report were collected during such episodes.
We do not know the underlying cause of the reaction time increases, the performance deficits, or the lethargy. Importantly, however, because all these impairments were bilateral, they could not have been related to the highly lateralized second saccade impairments that we considered markers of corollary discharge deficits.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
) to a target (
).
