JN Fuel your research with LabChart
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
 QUICK SEARCH:   [advanced]


     


J Neurophysiol 90: 1011-1026, 2003; doi:10.1152/jn.00193.2002
0022-3077/03 $5.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (16)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Scudder, C. A.
Right arrow Articles by McGee, D. M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Scudder, C. A.
Right arrow Articles by McGee, D. M.

Adaptive Modification of Saccade Size Produces Correlated Changes in the Discharges of Fastigial Nucleus Neurons

Charles A. Scudder1,2 and David M. McGee2

1 Department of Otolaryngology, University of Pittsburgh, Pittsburgh, Pennsylvania 15213; 2 Department of Neurobiology, University of Pittsburgh, Pittsburgh, Pennsylvania 15213

Submitted 15 March 2002; accepted in final form 14 April 2003


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 DISCLOSURES
 ACKNOWLEDGMENTS
 REFERENCES
 
Saccade accuracy is known to be maintained by adaptive mechanisms that progressively reduce any visual error that consistently exists when the saccade ends. We used an experimental paradigm known to induce adaptation of saccade size while monitoring the neural correlates of this adaptation. In rhesus monkeys where the medial and lateral recti of one eye were surgically weakened, patching the unoperated eye and forcing the monkey to use the weakened eye induced a gradual increase in saccade size in both eyes until the viewing, weak eye almost acquired the target in one step. Subsequent patching of the weakened eye gradually reversed the situation, so that the saccades in the viewing, normal eye decreased from an initial overshooting to normal. In the caudal fastigial nuclei of unadapted monkeys, neurons typically exhibit an early burst of spikes that is correlated with the onset of contraversive saccades and a later burst of spikes that is correlated with the termination of ipsiversive saccades. Comparing the discharges of the same fastigial neurons recorded before and during adaptation, this basic pattern did not change, but some parameters of the discharges did. The most consistent changes were in the latency of the burst for ipsiversive saccades, which was positively correlated with saccade size (1.28 ms/deg), and in the number of spikes associated with contraversive saccades, which was also positively correlated (0.55 spikes/deg). The former was more important when saccade size was decreasing, and the latter was more important when saccade size was increasing. Based on current knowledge of the anatomical connections of fastigial neurons, as well as on the effects of cerebellar lesions and on recordings in other structures, we argue that these changes are appropriate for causing the associated changes in saccade size.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 DISCLOSURES
 ACKNOWLEDGMENTS
 REFERENCES
 
Because saccades have too short a duration for vision to guide the eye to its target, saccade accuracy must be maintained over the long term by mechanisms that assess saccade error at the end of each saccade and modify subsequent movements. The existence of such adaptive mechanisms has been demonstrated in patients with paresis of the extraocular muscles in one eye (Abel et al. 1978Go; Kommerell et al. 1976Go; Optican et al. 1985Go) and in monkeys with surgically induced paresis (Optican and Robinson 1980Go; Scudder et al. 1998Go). Because of the conjugate oculomotor command conveyed to both eyes, eye movements are necessarily smaller in the paretic, "weak" eye than in the normal eye. When the normal eye is monocularly viewing, the subject makes saccades to visible targets accurately, but when the subject is forced to use the weakened eye by patching the normal eye, saccades initially undershoot the target. However, saccades gradually increase in size until they are nearly accurate. When the patch is switched and the normal eye is used again, saccades in this eye initially overshoot the target, but adaptive mechanisms gradually restore normal accuracy. The ability to adapt saccade size has also been experimentally shown by displacing the target during the time that the human or monkey subject is generating a saccade that would have acquired the target ("intrasaccadic target steps"; Deubel et al. 1986Go; McLaughlin 1967Go; Miller et al. 1981Go; Scudder et al. 1998Go; Straube et al. 1997Go). Initially, saccades land near the original target and miss the displaced target, but adaptation of saccade size and/or direction gradually produces saccades that land near the displaced target.

A variety of evidence has led to the belief that the cerebellum participates in the control of saccade size and direction. Lesions of the posterior vermis that include lobule VII, especially unilateral lesions, result in saccades that are markedly dysmetric (Aschoff and Cohen 1971Go, 1972Go; Barash et al. 1999Go; Optican and Robinson 1980Go; Ritchie 1976Go; Selhorst et al. 1976aGo,bGo; Takagi et al. 1998Go). Saccade size may differ from target displacement by a factor of 2, indicating that the cerebellar signals have a potent effect on the saccadic burst generator located in the brain stem. Similar, transient deficits are produced in monkeys and cats by local inactivation of the target area of lobule VII axons, that is, the caudal fastigial nucleus. Inactivation of this part of the nucleus by cooling (Vilis and Hore 1981Go) or by injection of the GABA agonist, muscimol, results in saccades to the contralesional side that are hypometric and saccades to the ipsilesional side that are hypermetric (Goffart and Pelisson 1994Go, 1998Go; Robinson et al. 1993Go). Finally, microstimulation of lobule VII slightly before or during contralaterally directed saccades has the ability to foreshorten these saccades (Keller et al. 1983Go; Ohtsuka and Noda 1991bGo).

Anatomical and physiological data support the hypothesis that saccade size might be controlled by the cerebellum and, moreover, have suggested ways that this might occur. Retrograde and anterograde tract tracing has shown that fibers from the caudal fastigial nucleus project mainly contralaterally to the pontine and mesencephalic reticular formations as well as to the superior colliculus (Batton et al. 1977Go; Gonzalo-Ruiz et al. 1988Go, 1990Go; Kurimoto et al. 1995Go; May et al. 1990Go; Sato and Noda 1991Go). Efferent terminals seem to target the specific areas where saccade-related neurons are located (Noda et al. 1990Go; Scudder et al. 2000Go). These neurons include the premotor "saccadic burst neurons," that is, excitatory burst neurons (EBNs) and inhibitory burst neurons (IBNs), as well as interneurons, that is, long-lead burst neurons (LLBNs) and midline omnipause neurons (OPNs). EBNs, IBNs, and LLBNs exhibit a vigorous burst of spikes preceding and during ipsiversive saccades, whereas OPNs exhibit a cessation (pause) in their tonic firing at the same time (cf. Fuchs et al. 1985Go; Scudder et al. 2002Go). As discussed in more detail later, the OPNs make inhibitory connections with EBNs and IBNs (Curthoys et al. 1984Go; Nakao et al. 1980Go) and are thought to permit the discharge of these premotor burst neurons by the cessation of tonic firing. Therefore OPNs probably control both the onset and termination of premotor bursts. Given that saccade size is determined by the number of spikes in these burst neurons (Kaneko et al. 1981Go; Keller 1974Go; Scudder et al. 1988Go; Strassman et al. 1986aGo,bGo; van Gisbergen et al. 1981Go), control of either the burst frequency and/or the burst duration would bring about control of saccade size. The brain-stem connections of caudal fastigial neurons are therefore well suited for this function.

Extracellular recording of cells in the caudal fastigial nucleus have revealed further clues about how the cerebellum might control saccade size. Saccade-related cells are found in a circumscribed area, which Noda and colleagues have called the fastigial oculomotor region (FOR), and exhibit a burst of spikes that precedes the onset of contraversive saccades by an average 10–19 ms (Fuchs et al. 1993Go; Ohtsuka and Noda 1990Go, 1991aGo). The range of burst leads is large but, nonetheless, the discharges seem well timed to affect the discharges of the premotor burst neurons (EBNs and IBNs) on the contralateral side while these latter neurons exhibit their maximal bursts (in their "on-direction"; cf. Fuchs et al. 1985Go). Because the vast majority of cerebellar nuclear projections to the pons appear to be excitatory (Angaut and Sotello 1989; Kitai et al. 1976Go; Schwarz and Schmitz 1997Go; Verveer et al. 1997Go), one hypothesis is that the burst of FOR neurons during contraversive saccades contributes to the on-direction burst in contralateral EBNs and IBNs, and therefore to the production of force in agonist motoneurons and the reduction of force in antagonist motoneurons, respectively. An effect on the onset and/or offset of the OPN pause is also possible. FOR neurons also exhibit a more delayed burst during ipsiversive saccades that seems to be best correlated with the end of the saccade. Ohtsuka and Noda (1991aGo) report that the burst for individual neurons begins at a fairly fixed interval before saccade end that averaged 31 ms, regardless of saccade size. By the contralateral projection, this late burst could augment the late off-direction discharges of EBNs and IBNs (Scudder et al. 1985; van Gisbergen et al. 1981Go). Augmenting the off-direction discharges would have the effect of "braking" the saccade (Ohtsuka and Noda 1991aGo). However, the tight relation of the FOR-neuron burst to saccade end would seem to require a more secure mechanism, that is, indirect control of all burst neurons by affecting the resumption of OPN firing.

Fastigial neurons also project to the superior colliculus and could affect saccade size by this pathway. However, there are conflicting data regarding the strength of this projection (reviewed by Scudder et al. 2002Go) and whether the projection is just to the "fixation neurons" (Munoz and Wurtz 1993Go) in the rostral colliculus (May et al. 1990Go) or is more widespread (Batton et al. 1977Go; Gonzalo-Ruiz et al. 1988Go, 1990Go; Kurimoto et al. 1995Go).

Given that mechanisms are available for cerebellar control of saccade size and that these mechanisms do exert powerful effects on saccade size, it stands to reason that the cerebellum could mediate the adaptive processes described above. Nonetheless, evidence for this hypothesis is meager. The fact that the saccadic dysmetrias produced by chronic cerebellar lesions are enduring and are not corrected by adaptation argues for this hypothesis. Goldberg et al. (1993Go) extended this conclusion by showing that monkeys with large electrolytic lesions including the fastigial nuclei had no capacity to change saccade size when subjected to intrasaccadic target displacements. Optican and Robinson (1980Go) combined large lesions of the monkey posterior cerebellum with muscle weakening in one eye. Monocular viewing with the normal eye resulted in enduring hypermetria, as expected from other binocular studies, but monocular viewing using the weakened eye did not produce sufficient dysmetria for a critical test.

The current study attempted a direct test of the hypothesized role of the cerebellum in saccade adaptation by searching for the neural correlates of that adaptation. The discharges of neurons in the caudal fastigial nucleus of our monkeys were recorded extracellularly during ongoing adaptation produced by alternately patching a surgically weakened or normal unoperated eye. Our approach was to correlate a variety of FOR-neuron discharge parameters with saccade size as it changed to reveal those discharge parameters that were modified in a repeatable manner. We applied this approach consistently because the hypothesis is that adaptation functions to reduce the amplitude error (i.e., change saccade size) regardless of the precise neural mechanisms that mediate the change. As noted above, possible mechanisms include those that modulate EBN and IBN burst frequency and those that modulate burst duration, which also has the effect of modulating saccade duration. In fact, our results show that FOR-neuron discharges change during adaptation in multiple ways that can modify EBN and IBN burst frequency and duration, albeit not always in ways that maintain normal saccade amplitude–duration relationships.

A preliminary report based on a smaller sample of neurons has appeared as an abstract (Scudder 1998Go).


    METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 DISCLOSURES
 ACKNOWLEDGMENTS
 REFERENCES
 
Three juvenile rhesus monkeys were used in these experiments. All procedures were approved by the Institute Animal Care and Use Committee and conform to guidelines issued by the National Institutes of Health.

Surgery

Monkeys were surgically prepared for eye movement recording during two aseptic procedures in which they were deeply anesthetized with halothane (1.5% in oxygen/nitrous oxide) after induction with ketamine. In the first surgery, a magnetic search coil was sutured to the sclera of the right eye and three acrylic lugs for the stabilization of the head were implanted as described previously (Scudder et al. 1998Go). The front lug also served as the receptacle for a metal stalk that held the eye patch. After training (see following text), monkeys underwent a second surgery to weaken the eye muscles of the left eye, to implant a search coil in the left eye, and to implant a recording chamber directed at the fastigial nuclei. The medial and lateral recti were detached from the globe and cut to about two-thirds their normal length. A length of absorbable suture bridged the gap between the cut end of each muscle and the original attachment of its tendon to help control the place where the cut muscle would reattach to the globe. The second search coil was implanted in the same manner as the first. The 20-mm-diameter recording chamber was placed in a craniotomy over the midsagittal plane. It was aimed 5–7 mm posterior to ear-bar zero and angled 20° posteriorly from stereotaxic vertical. The recording chamber was attached to the bone with self-tapping screws and dental acrylic. The chamber was filled with antibiotic gel and covered with a cap. For the duration of the experiments, the antibiotic gel was replaced daily and the dura cleaned with cotton swabs.

Behavioral apparatus and experimental tasks

Experiments began 3–4 wk after the second surgery. They were conducted in a dimly illuminated soundproof booth with the monkeys seated in a primate chair and their heads restrained. Surrounding their heads were two orthogonal pairs of magnetic induction coils that were used in conjunction with the implanted search coil for the measurement of eye movements (CNC Engineering). Eye movement signals were calibrated by having the monkey alternately fixate stationary targets at ±15° horizontal and vertical eccentricity.

Monkeys viewed a red, 0.3° target spot projected onto a textureless white screen formed into a vertically oriented quarter cylinder and situated 75 cm in front of the monkey. The target spot was generated by a laser diode and deflected by an X-Y mirror galvanometer system (General Scanning) under computer control (PDP 11/73). The computer determined the target location, monitored eye movements, and dispensed applesauce reward when on-target conditions were met. For training, food-deprived monkeys were required to fixate the target spot within an error "window" of 6° in all directions to receive a dollop (about 0.1 ml) of applesauce. As performance improved, the error window was reduced to 2° and monkeys were required to fixate >=1.5 s after an on-target saccade.

One pattern of 16 possible targets located along 8 equally spaced directions on 8°- and 15°-radius circles and with a central fixation point was used to collect information about the relation between FOR-neuron discharges and the size and direction of saccades. The fixation point and any of the pseudorandomly chosen peripheral target locations were alternately illuminated so that the monkey made saccades to and from the fixation point in each of 8 directions. A 1.3- to 2.0-s period of fixation triggered the illumination of the next target. A second pattern consisting of two possible target locations separated horizontally by 20° was used to collect all data relating to adaptation of saccade size and the correlated changes in FOR neuron discharges. This pattern was used because it produced the most rapid adaptation (Scudder et al. 1998Go) and thereby maximized the probability of obtaining significant changes in a short period of time.

Long-term and short-term adaptation of saccade size was produced by forcing the monkey to view the world through the eye having surgically weakened eye muscles. Vision in the normal unoperated eye was blocked using an opaque rigid patch that mated with a receptacle attached to the front acrylic lug on the monkey's head. Reverse adaptation (recovery) was produced by patching the weak eye. Monkeys were never allowed to experience binocular vision, either during the experiments or when housed in the animal facility. By long term, we mean that the monkeys would wear the patch on the same eye every day for 1 wk, beginning on a Friday. When recordings began the following Monday, the monkey had had sufficient time that saccades in the viewing eye could reach an asymptotic value of saccadic gain (saccade amplitude/target-step amplitude; Scudder et al. 1998Go), which we will call the initial state. When a suitable neuron was isolated (see following text), data from about 200 saccades were collected using first the circular target pattern and then the two-target linear pattern. The eye patch was then switched so the monkey viewed through the unadapted eye and data were again collected using the two-target linear pattern. Further data from 100 to 200 saccades were collected 20 min after switching the patch, and every 10–20 min thereafter as the saccadic system adapted to the new viewing condition. A few neurons were isolated long enough that adaptation slowed as saccade gain approached its asymptotic value for the particular viewing eye (0.8–0.95; Scudder et al. 1998Go). When this occurred, the target was then stepped intrasaccadically to produce additional visual error. If the monkey was viewing with the weakened eye, the target was stepped in the forward direction 3–4° (15–20% of the initial target step); if the monkey was viewing with the normal eye, the target was stepped in the backward direction 3–4°.

If the neuron was lost after only a brief period of isolation (<20 min of adaptation), the patch was switched back to cover the original eye and we attempted to isolate a new neuron. If isolation was maintained for a longer time and then lost, the patch was switched to the original eye and the remainder of the experiment was used to return saccade size to its initial state. Regardless, the monkey was returned to its cage with the patch on the original eye so that saccade gain would return to the same initial state for the experiment the next day. After 1 or 2 wk of such sessions, the patch would be placed over the original viewing eye on Friday afternoon so that the monkey would enter the lab adapted to viewing with the opposite eye during the next set of experiments.

Neuronal recording

Neuronal action potentials were recorded with commercially available 0.005-in. tungsten microelectrodes with 15-µm exposed tips (Micro Probe, Potomac, MD). Electrodes were loaded into a 24/21-gauge concentric cannula assembly that protected the electrode as both were lowered into the brain and through the tentorium cerebelli. The outer 21-gauge cannula was 8 mm shorter than the inner 24-gauge cannula so that only the latter penetrated the tentorium. Electrodes were driven the last 10-mm to the fastigial nucleus with a Trent-Wells motorized hydraulic microdrive.

Neurons exhibiting discharges related to saccadic eye movements were encountered 0.7–2.0 mm lateral to the midline in the most caudal 1–2 mm of the fastigial nucleus. The majority of neurons exhibited a burst of spikes during saccades, and some had very pronounced bursts (peak rates >500 spikes/s), which served as an unambiguous marker for the area. Preliminary data were collected for all isolated saccade-related neurons using the circular target pattern. Units selected for recording during adaptation had to have had excellent, stable isolation throughout the preliminary data collection. Neurons were also required to have a saccade-related response of >=20 spike/s, and to have a stronger response during horizontal saccades than during vertical saccades. Neurons failing the latter criteria were rejected under the supposition that they might function to control vertical and not horizontal saccades. Neurons meeting these criteria were tested during short-term adaptation using the procedures described above.

Data collection and analysis

Neuronal signals were conventionally amplified and filtered (0.2–10 kHz). Most data were digitized on-line. Spikes were converted to TTL pulses with a simple Schmitt trigger, and were given a time stamp to the nearest 0.1 ms. The 4 eye channels (horizontal and vertical eye position for both eyes) as well as the 2 target channels were sampled every 4 ms. Data were also binary-encoded for storage on VCR tape (Vetter 4000A). Two cells were digitized off-line using the same program, but using the taped backup data. Seven other cells were digitized off-line, and spikes were detected in software using DataWave Experimenters Workbench. Temporal resolution for spike occurrence was again 0.1 ms, and the sampling rate for the eye and target channels was 4 ms.

Neuronal discharge parameters and saccade metrics were computed from the digitized data using an interactive program similar to that used previously (Scudder et al. 1988Go). The user manually scrolled the data and placed a cursor near desired saccades. The computer (Pentium-based) found peak velocity and independently searched backward and forward in time for the first point where eye position reached that occurring during fixation before and after the saccade, respectively. Temporal resolution was improved by fitting the 4 adjacent persaccadic samples with a second-order polynomial with the vertex forced through the steady-state eye position. The time coordinate of the vertex was taken as the start or end of the saccade. The computer marked these points and the user accepted them or manually marked those that were in error (errors were usually in the vertical channel where eye velocity could change direction in the middle of a horizontal saccade). The user manually marked the approximate start and end of bursts and/or pauses in the neuronal discharges. The computer used the spike succeeding (preceding) the first mark as the onset of a burst (pause), and the spike preceding (succeeding) the second mark as the end of the burst (pause). From these marks and the eye positions at these marks, the computer determined the amplitudes of the horizontal and vertical saccadic components, and horizontal and vertical component durations. The computer also determined latency from saccade onset to burst onset, burst duration, and other discharge parameters. Because the firing rate of FOR neurons was frequently above the resting level at saccade termination, and because subsequent spikes obviously cannot affect the size of the terminated saccade, only spikes occurring between burst onset and saccade termination were included in the computation of number of spikes, average frequency (number of spikes/duration), and peak frequency. The last was defined as the average frequency of the three consecutive spikes spanning the shortest time interval.

As noted by Fuchs et al. (1993Go), burst onset could not always be marked with high reliability. In these cases, burst onset was determined from averaged instantaneous frequency data. The interval between successive spikes was assigned a frequency value equivalent to the reciprocal of the interval. Frequency was averaged across saccades by aligning all data on saccade onset. Burst onset was defined as the time when averaged frequency crossed the criterion of one third the difference between resting rate and peak burst frequency. For a given unit, the frequency criterion was determined from the data at the start of adaptation, and this same absolute value was applied to data acquired at all subsequent times. Measurements obtained with this method usually differed by a few milliseconds from the averaged discharge latency, but determinations at different times during adaptation were almost always in the same rank order for the two methods. Among neurons with well-defined burst onsets, the changes in burst onset during adaptation were nearly identical for the two methods. Similarly, burst end was poorly defined for some units and a graphical method was again used. Burst end was defined as the time when firing rate had dropped to 40% of the difference between peak rate and resting rate.

Saccades were excluded from the statistical analysis if the reaction times were too short (<70 ms) or too long (>600 ms), if the monkey was not initially on target (±1.5°), or if the saccade was clearly not directed at the target. Saccades were also excluded if their durations exceeded 3.5 times the average value reported by Fuchs (1967Go) for a given size, or if the saccade size was 3.5 SDs away from the mean for the particular data set (about 0.3% of the data). Averaged saccade metrics (e.g., amplitude) and discharge parameters were always computed separately for leftward and rightward saccades. Measurements were statistically compared using Student's t-test, and 2-tailed P values reported.

Data on saccade metrics after long-term adaptation were collected in association with the 28 neurons included in this study as well as with neurons that were not isolated long enough to be included. Means for each neuron were computed, and the average and SD of these means are reported. Preadaptation means and SDs of the discharge parameters (first section of RESULTS section) were obtained using the 20° target steps. Therefore some data were collected with the monkey adapted long term to using the normal eye and other data were collected with the monkey adapted long term to using the weak eye. The average saccade size for the combined data was 23.1 ± 4.4° (SD). Linear correlations were obtained between discharge and saccade parameters under two different circumstances, and the difference is important for understanding our findings. In the first circumstance, data were collected before short-term adaptation using the circle task, and the correlations describe how the discharges respond as saccade size and duration vary both naturally and to different size target steps. In the second, data were obtained during short-term adaptation, and the correlations describe how the discharges respond as saccade size changes because of adaptation. The process of obtaining the latter correlations are described more fully in the RESULTS.

Plots of average eye position versus time and averaged discharge rate were generated for illustrations. Instantaneous frequency was computed as described above, and eye position and frequency traces were aligned on saccade onset and averaged. Each data point corresponds to one 4-ms sample interval.

Anatomy and histology

At the completion of all experiments, monkeys were deeply anesthetized and perfused through the heart with buffered saline followed by buffered 4% paraformaldehyde. The fixed brain was sectioned coronally and sections were processed for neuroanatomical tracers (to be reported separately) and counterstained with Neutral Red. We confirmed that electrode tracks passed through the caudal pole of the fastigial nucleus, but we did not attempt to plot the location of recorded cells on individual tracks.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 DISCLOSURES
 ACKNOWLEDGMENTS
 REFERENCES
 
A total of 127 saccade-related neurons encountered in the caudal fastigial nucleus met the criteria for prolonged recording during adaptation of saccade size (see METHODS). Of these, only 28 met the additional criteria of being isolated long enough for the saccade size to change >=5%. This report focuses on these 28 neurons; 18 from monkey M, 7 from monkey S, and 3 from monkey G. Eighteen neurons were recorded during size increases, whereas 10 were recorded during size decreases.

General features of included neurons: preadaptation data

The neurons used in this study were generally similar to those reported elsewhere (Fuchs et al. 1993Go; Ohtsuka and Noda 1990Go, 1991aGo). Averaged data from a typical neuron are illustrated in Fig. 1 for 20° horizontal target steps. All the neurons exhibited a burst of spikes during contraversive saccades that began on average 4.2 ± 8.1 ms (SD, n = 28) before the onset of saccades averaging 23° (see METHODS). Although other investigators have emphasized the early lead of this burst (18.5 ms, Ohtsuka and Noda 1991aGo; 10.5 ms, Fuchs et al. 1993Go), the burst began after saccade onset in 25% of our sample. In fact, one monkey (S) had a predominance of such neurons, with a mean lead of –3.7 ms. The peak discharge, which averaged 229 spikes/s for the full sample, nearly always occurred during the saccade. Some neurons appeared to have two bursts: one near saccade onset, and a second smaller burst before saccade end. For 7 neurons, the burst nearly always (>85% of saccades) ended before the saccade ended, whereas for 7 other neurons, including that in Fig. 1, the burst nearly always (>85%) ended later than the saccade. The remaining 14 units exhibited intermediate behavior, with some bursts ending before and some later than the saccade. For the whole sample, the average end of the burst occurred at about the end of the saccade. The burst was preceded by a 20–50% diminution in firing rate in 8 units and was followed by a diminution in another 5, as revealed in the averaged firing rate plots.



View larger version (22K):
[in this window]
[in a new window]
 
FIG. 1. Averaged discharges of typical saccade-related fastigial nucleus neuron illustrated for rightward saccades (top) and leftward saccades (bottom). Neuron was recorded in left nucleus, so rightward movement (illustrated as upward) is contraversive and leftward movement is ipsiversive. For both top and bottom panels, averaged horizontal position of viewing weak eye (H.Eye), spike rasters for first 12 saccades, and averaged firing rate (FR) are illustrated. For averaging, saccades were aligned on saccade onset (thin vertical line). Neuron has burst onset times and peak frequencies typical of sample, but burst durations are longer than average.

 

All neurons also exhibited a burst of spikes during ipsiversive saccades. For saccades averaging 23°, this burst began after saccade onset in all but 4 neurons. The average latency, which is defined hereafter as ipsiversive burst lag, was 23.9 ± 20.5 ms (SD, n = 28). Within neurons, the variability (SD) of ipsiversive burst lag averaged 11.1 ms (range 5.1–27.4, n = 28). Expressed relative to the end of the saccade, the average burst began 47.3 ± 18.8 ms (SD, n = 28) before saccade end. Based on their finding that the average interval between burst onset and saccade end was invariant regardless of saccade size, Ohtsuka and Noda (1991aGo) argued that the function of this burst is to terminate the saccade. We examined this possibility by correlating ipsiversive burst lag against saccade duration as the monkey made a spectrum of saccade sizes to the 8° and 15° target steps. In our sample, there were only 4 neurons where the interval between burst onset and saccade end was nearly invariant, but there were 2 other neurons where the interval between saccade start and burst onset (burst lag) was nearly invariant. In the majority of neurons, behavior was in between. Overall, burst lag was significantly correlated with saccade duration in 24 neurons, but on average, increased 0.55 ± 0.27 ms (SD, n = 28) for every 1-ms increase in saccade duration (average r = 0.68, range 0.07–0.98). This slope is considerably less than the 1.0 ms/ms slope stipulated by Ohtsuka and Noda. The variability about this relation (square-root of the error variance) was 9.0 ms (range 2.4–22.9).

Ipsiversive burst lag was also correlated with saccade size (1.22 ± 0.71 ms/deg, n = 28), but less well than with saccade duration (average r = 0.56, range 0.04–0.94). Thus the dependency on size could be a byproduct of the correlation between saccade size and saccade duration.

The ipsiversive burst was preceded by a full cessation of firing in 13 units, and a significant reduction (e.g., 50%) in another 7. The firing rate at the end of the saccade remained elevated above the resting rate in most of the neurons (22 of 28), and this was followed by a diminution or outright pause in firing in 6 units. Overall, these discharge patterns fit within the "burst," "pause-before-burst," and "burst-before-pause" types described by Fuchs et al. (1993Go). A small number of neurons that exclusively paused were also encountered, but are not included in this report.

Adaptation of saccade size

The process of adapting saccade size using the appropriate patching of the paretic or normal eye is illustrated for monkey "M" in Fig. 2, A–D. Initially (Fig. 2A), monkeys wore the patch over the weak eye and viewed the target with the normal eye. Target steps elicited saccades in the normal eye (top traces) that accurately acquired the target, but saccades measured in the nonviewing weak eye were approximately half the size of the target step (bottom traces). After recording initial saccade and neuronal-discharge data, the patch was switched to cover the normal eye. Figure 2B illustrates typical eye movements recorded immediately thereafter. Saccades in the (now viewing) weak eye were still severely hypometric, thereby requiring multiple corrective saccades to finally acquire the target. Simultaneously, the initial saccade measured in the (nonviewing) normal eye was still roughly equal to target displacement (dotted line). Saccade size measured in both eyes began increasing immediately (Scudder et al. 1998Go) and changed measurably over 20 min. Saccade size in the weak eye ultimately reached an asymptotic value that undershot the target (Scudder et al. 1998Go), but this value was not achieved over the period that FOR neurons remained isolated. Long-term asymptotic horizontal gains measured in the viewing weak eye in monkeys M, G, and S were 0.85, 0.95, and 0.90, respectively. The saccade in Fig. 2C exhibits this long-term adapted gain. Horizontal gain measured in the nonviewing normal eye was about 1.8. For 10 of the neurons, this was the initial state of adaptation. Switching the patch to cover the weak eye produced the eye movements illustrated in Fig. 2D. Saccade size was initially the same as that measured in the previous panel (Fig. 2C), so that saccades measured in the viewing normal eye severely overshot the target. Saccade size decreased more rapidly than saccade size previously increased. Long-term patching of the weak eye resulted in eye movements comparable to those illustrated in Fig. 2A. Although this readaptation to use of the normal eye is operationally a "recovery," we caution that this word carries mechanistic implications that we believe do not apply to short-term adaptation of saccade size. The vertical component of the saccades changed negligibly during the patch-switching paradigm.



View larger version (14K):
[in this window]
[in a new window]
 
FIG. 2. Adaptation of saccade size produced by patching of either normal or weak eye. Four stages of adaptation (A–D) are shown for typical single trials in clockwise order. Horizontal position is illustrated for visible target (Targ), normal eye (N.Eye), weak eye (W.Eye), and unseen target relative to patched eye (dotted lines). Vertical target and vertical eye positions were unchanged during adaptation. Initially, patch covered weak eye and normal eye viewed target (A). Adaptation was initiated by switching patch to cover normal eye so that saccades in viewing weak eye were severely hypometric (B). After viewing with weak eye for a day or more, saccade size increased in both eyes (C). Finally, switching patch back to cover weak eye resulted in severely hypermetric saccades in viewing normal eye (D). Long-term adaptation to this situation produced decreases in saccade size and return to normal-sized saccades in normal eye (A).

 

In the long term, increases in saccade size produced by viewing with the weak eye were accomplished by increases in saccade duration and increases in saccade velocity. Measurements were made in the normal eye, and data from rightward and leftward saccades, which were somewhat but not significantly different, were combined. Long-term increases in saccade size for 20° target steps averaged 11.5° and were accompanied by an increased duration of 29.3 ms (67.7 ± 8.0 to 97.0 ± 13.7) and an increased peak velocity of 43°/s (548 ± 45 to 591 ± 68). These differences were significantly different (P < 0.02, n = 51). The change in duration was the dominant factor in increasing saccade size; the 62% change in saccade size was accomplished with a 43% increase in duration and an 8% increase in velocity. Adapted and unadapted "main sequence" parameters (Bahill 1975) were compared for 10° and for 15° saccades in the normal eye (values of maximal overlap for adapted and unadapted saccades using the circular target pattern). Adapted saccades had slightly, but significantly, smaller durations and velocities. For instance, unadapted and adapted durations for 15° saccades averaged 57.7 ± 5.5 (n = 45) and 50.9 ± 6.7 (n = 55) ms, respectively, whereas unadapted and adapted peak velocities were 519 ± 41 and 491 ± 42 deg/s. The smaller saccade durations and velocities associated with saccades of the same size imply that the basic shape of the velocity profiles must have changed.

Short-term changes in saccade size (those occurring during recording sessions) were smaller than the long-term changes. For the 18 gain-increasing experiments, saccade gain (weak eye) began at around 0.61 (range 0.43–0.76) and increased to 0.69. Saccade size (normal eye) began at an average 22.5° and increased by 1.3–10.6°, with a mean increase of 4.3°. Saccade size began at a value >20° because the muscle weakening produced nonlinearities in the orbit (Scudder et al. 1998Go) and because adaptation had been attempted before the successful recording in some electrode penetrations. For the 10 gain-decreasing experiments, saccade size began at an average 25.5° and decreased by 1.3–13.2°, with a mean decrease of 6.3°. All changes were statistically significant at P < 0.01 or better.

As in the long term, short-term changes in saccade size were usually accomplished by changes in saccade duration and peak velocity, although there was considerable variability. In some cases, for instance, saccade duration changed negligibly and/or saccade velocity did not change or changed in the "wrong" direction (velocity decreased as size increased). In these cases, the shape of the velocity profiles must have changed during adaptation. To illustrate this, we plotted the averaged velocity profiles of rightward saccades recorded during a session in which saccade size, but not duration or peak velocity, had increased after 40 min of adaptation (Fig. 3). The 11% increase in saccade size was accomplished by a relative increase in velocity during the deceleration phase of the saccade. Similar results were obtained in two more conditions from two neurons where saccade duration and peak velocity changed negligibly (leftward and rightward saccades were separately averaged for a total of 56 conditions). For the remaining conditions, we computed the product of averaged peak velocity times averaged duration and compared the changes in this product during adaptation to the averaged changes in saccade size. If waveshape did not change, this product would scale directly with saccade size so that the percentage change in the product and the measured size would be equal. There was a gross inequality in 6 additional conditions; the percentage changes in this product were 2.5 to 4 times smaller than the percentage changes in saccade size. All these instances of unusually small increases in duration and peak velocity were obtained during gain-increasing experiments, and most (7/9) were obtained from one monkey (S). Both the long- and short-term data show that adapted saccade duration and peak velocity need not fall on the main sequence, and that these two parameters are not sufficient for characterizing all the changes that occur during adaptation.



View larger version (10K):
[in this window]
[in a new window]
 
FIG. 3. Comparison of saccade trajectory from an adaptation experiment in which saccade duration and velocity changed negligibly despite increase in saccade size. Traces plot average horizontal angular position of viewing weak eye (H.Eye) and corresponding velocity (H.Vel.) for 34 saccades collected during start of adaptation (Start) and 20 saccades collected 40 min later (40 min). All saccades were to 20° target steps and were aligned on saccade onset (vertical dotted line) for averaging. Saccades were selected to have durations within ±4 ms of average duration for all saccades during recording epoch (90.8 ms at Start, 90.6 ms at 40 min). This selection resulted in saccade durations, peak velocities, and sizes within 1% of full sample for each time epoch. Comparing Start and 40-min epochs, average peak velocities were nearly identical (360 vs. 357°/s, respectively), but saccades recorded at 40 min maintained higher velocity until their near simultaneous end (91.7 vs. 90.4 ms for Start and 40 min, respectively).

 

Changes in discharge patterns during adaptation

Based on the possible mechanisms mentioned in the INTRODUCTION by which FOR neurons might affect saccade size, the discharge onset time, discharge offset time, peak frequency, average frequency, and number of spikes were all considered important parameters to monitor during adaptation. One parameter exhibiting changes during adaptation was the burst lag for ipsiversive saccades. Because the ipsiversive burst was postulated to "brake" or truncate the saccade, earlier discharges should be associated with smaller saccades, just as they are in unadapted animals. Figure 4 illustrates this association for one neuron. Plots are averages of 54 and 44 saccades aligned on saccade onset. Before the experiment, the monkey had been wearing the patch on the normal eye so that switching the patch to the weak eye caused hypermetric saccades in the now viewing normal eye (Fig. 4, top). Saccade amplitude averaged 26 ± 1.6° in response to the 20°-target step. However, after 2 h of performing the saccadic tracking task, average saccade amplitude in the normal eye declined to 19.4 ± 1.7°, and there was a corresponding significant decrease in the average burst lag from 73.0 ± 14.5 to 60.1 ± 11.4 ms (P < 0.0001, n = 98).



View larger version (20K):
[in this window]
[in a new window]
 
FIG. 4. Changes in discharges of fastigial oculomotor region (FOR) neuron produced by adaptation of saccade size. For both top and bottom panels, averaged horizontal position of viewing normal eye (H.Eye), spike rasters for first 12 saccades, and averaged firing rate (FR) are illustrated. Data are averaged across ipsiversive (leftward) saccades aligned on saccade onset (left thin vertical line). Preadaptation data (top) were obtained with monkey previously adapted to viewing with weakened eye, but now viewing with normal eye. Saccade size averaged 26 ± 1.6° (SD, n = 54) and overshot 20° target step (not illustrated). Onset of neuronal discharge is marked with right thin vertical line. Late increase in firing rate is associated with corrective saccades (not illustrated). Postadaptation data (bottom) illustrate smaller saccades (19.4° on average, n = 44) produced by 2 h of continued tracking with normal eye. Onset of neuronal discharge has advanced 19 ms relative to control discharge, whose onset is marked by right thin vertical line. Average onset, as marked by computer user on individual discharges, advanced 13 ms.

 

To quantify the results in this and in the other neurons, burst lag was plotted as a function of saccade size for all time points where data were collected. Figure 5 illustrates 8 such plots for 8 neurons selected to show the range of the data. For neurons plotted on the right-hand side, the monkey had previously been viewing with the weak eye, and data were collected after the patch was switched so that the normal eye became the viewing eye (Gain– condition). This is like Fig. 4, where initially overshooting saccades (size >20°) decreased in amplitude over time. Note that the size sometimes went below 19° (a normal saccade size to a 20° target) because of the use of backward intrasaccadic target steps (see METHODS). For neurons plotted on the left-hand side, the monkey had previously been viewing with the normal eye, and data were collected after the weak eye became the viewing eye (Gain+ condition). Gain increased during adaptation in these cases. Figure 5 illustrates that for some neurons, increasing saccade sizes were associated with increasing burst lags, although there were others where lag decreased. Among these is a neuron whose ipsiversive burst began before saccade onset (burst lag <0).



View larger version (15K):
[in this window]
[in a new window]
 
FIG. 5. Burst lag (time from saccade onset to burst onset) associated with ipsiversive saccades is plotted as function of saccade size in viewing eye for 8 representative neurons. Nonoverlapping curves were selected to show range of data. Symbols plot average lag against average size for 38–70 saccades collected at discrete intervals during adaptation. Traces to left of 20°-size plot data for experiments in which gain started low and increased as monkeys tracked target with their weakened eye (Gain+ experiments). For traces to right, gain started high and decreased as monkeys tracked target with their normal eye (Gain– experiments). Ipsiversive burst lag decreased as saccade size decreased during Gain– experiments, but changed more variably during Gain+ experiments. Error bars: ±1SE of burst lag or saccade size. Most standard errors were smaller than symbol size (±1.6 ms lag, ±0.28°). SDs for 8 neurons averaged ±10.6 ms for lag, and ±1.34° for size. Neuron in Fig. 4 corresponds to uppermost plot (solid squares).

 

To compare data from neurons with different magnitudes and directions of adaptation on an equal basis, discharge data were normalized to +1° change in saccade size. This is equivalent to computing the slope of the relationship between burst lag and saccade size. Neurons that behave according to the hypothesis that the ipsiversive burst acts to brake or terminate the saccade would exhibit increasing lags during gain-increasing experiments and decreasing lags during gain-decreasing experiments, but the same positive slopes in both cases. Therefore slope is the most appropriate measure for testing whether changes in burst parameters are consistent with the hypotheses outlined in the INTRODUCTION. Slopes of burst lag versus saccade size (measured using data from the normal eye regardless of which eye was viewing) were estimated using linear regression through averaged data points, as in Fig. 5. Correlation coefficients ranged from 0.08 to 0.99 (absolute value, and excluding neurons with 2 data points) and averaged 0.75. We include data yielding both statistically significant and nonsignificant (low or zero) slopes because the latter (e.g., no change) is a perfectly plausible outcome of these experiments. Figure 6 illustrates the results of this analysis, and confirms the conclusions drawn from Fig. 5. Sixty-four percent (18/28) of the neurons had positive slopes, and the mean slope (1.28 ± 3.0 ms/deg, n = 28) was significantly different from zero. Figure 6 also shows that neurons responded quite differently during gain-decreasing and gain-increasing experiments. For the gain-decreasing experiments, most neurons exhibited positive slopes, and the mean slope was significantly greater than zero (Table 1). For gain-increasing experiments, slopes were more uniformly distributed about zero, and the small positive average slope was not significantly different from zero.



View larger version (34K):
[in this window]
[in a new window]
 
FIG. 6. Histogram of slope of regression of ipsiversive burst lag against saccade size for 28 neurons. Slopes recorded during gain-increasing (Gain+) experiments clustered around zero with slight positive mean (thick arrow). Slopes recorded during gain-decreasing (Gain–) experiments were mostly positive with significant positive mean (thin arrow).

 

View this table:
[in this window]
[in a new window]
 
TABLE 1. Slope of saccade and discharge parameters over saccade size during adaptation

 

If ipsilateral burst lag exerts control over saccade duration, there should be a correlation between changes in these variables as adaptation progresses. Figure 7 shows this correlation for the sample of 28 neurons. Changes in burst lag relative to that at the start of adaptation are plotted against changes in saccade duration for all time points where data were collected. The slope of this relation was 0.64 and the correlation coefficient was 0.71 (P < 0.0001, n = 76). Some scatter is expected because ipsiversive burst lag is surely not the only determinant of saccade duration. In addition, 4 atypical "untuned neurons" (squares), to be described more fully later, almost align along a slope opposite to that of the rest of the sample. Elimination of these 4 neurons and elimination of data points where burst lag did not actually change (<2 ms) improved the correlation coefficient to 0.84 (P < 0.0001, n = 50) and changed the slope somewhat to 0.80.



View larger version (23K):
[in this window]
[in a new window]
 
FIG. 7. Scatter plot illustrating that changes in ipsiversive burst lag are correlated with changes in saccade duration during adaptation. Each point represents difference between lag or saccade duration measured at initiation of adaptation and that measured at subsequent time periods (e.g., 20, 40, 60 min). Data from gain-decreasing experiments are shown in light symbols and data from gain-increasing experiments are shown in dark symbols. Light and dark squares plot data from 4 "untuned" neurons that had atypical discharges (see RESULTS, last section). Change in burst lag and saccade duration are well correlated (r = 0.71, n = 76, P < 0.0001) with slope of 0.64 ms/ms.

 

Analyses like those of Figs. 5 and 6 were conducted for other ipsiversive burst parameters. For the whole sample, number of spikes, average discharge frequency, and peak discharge frequency during ipsiversive saccades did not change significantly with adaptation (Table 1). However, this was due to opposing changes in gain-increasing and gain-decreasing experiments. Number of spikes, average frequency, and peak frequency were mostly positively correlated with saccade size in the former case, and negatively correlated in the latter case. Differences between these parameters for gain-increasing and gain-decreasing conditions were all statistically significant.

Changes in burst parameters during contraversive saccades are illustrated in Fig. 8 for one neuron recorded during a gain-increasing experiment. Preadaptation saccade size (top panel) recorded from the weak eye increased from 10.4 ± 1.4 to 12.4 ± 1.2° (postadaptation, bottom panel), a highly significant change (P < 0.0001, n = 101). Associated with that increase in saccade size was a significant increase in peak burst frequency from 254 ± 42 to 277 ± 47 spikes/s (P < 0.02), and a significant increase in burst duration from 54 ± 13 to 69 ± 15 ms (P < 0.0001). The increase in burst duration was mostly produced by a later termination of the burst. Changes in burst frequency and duration collectively produced a significant increase in number of spikes from 10.5 ± 2.5 to 13.1 ± 2.4 spikes (P < 0.0001).



View larger version (21K):
[in this window]
[in a new window]
 
FIG. 8. Changes in discharges of FOR neuron produced by adaptative increase in saccade size. For both top and bottom panels, averaged horizontal eye position (H.Eye), spike rasters for first 12 saccades, and averaged FR are illustrated. Data are averaged across contraversive (leftward) saccades aligned on saccade onset (thin vertical line). Preadaptation data (top) were obtained with monkey previously adapted to viewing with normal eye, but now viewing with weak eye. Saccade size measured in weak eye averaged 10.4 ± 1.4° (SD, n = 59), and undershot the 20° target step (not illustrated). Postadaptation data (bottom) illustrate larger saccades [12.4 ± 1.2° (SD, n = 42)] produced by 30 min of continuous tracking with weak eye. Magnitude of peak frequency in top panel is indicated by arrow in bottom panel, showing that burst frequency increased as result of adaptation. Burst duration also increased, which together with burst frequency, produced increase in number of spikes.

 

This pattern of changes was reflected in the sample as a whole. Burst frequencies, burst onset- and end times, and number of spikes were regressed against saccade size during adaptation, and the resulting slopes were tabulated. Figure 9A shows the most robust finding of this study: the number of spikes in the burst was positively correlated with saccade size in all but one neuron, and the mean slope (0.55 ± 0.44 spikes/deg, n = 28; Table 1) was highly significantly different from zero. Slopes associated with gain-increasing and gain-decreasing experiments were similar but not identical (Table 1). As in Fig. 8, changes in the number of spikes were produced by changes in burst duration and discharge frequency. Figure 9B shows that peak discharge frequency was positively correlated with saccade size in most (21/28) neurons, and the mean value of the slope was significantly greater than zero (Table 1). The mean slope for gain-increasing experiments was greater than that for gain-decreasing experiments, but not significantly (P {approx} 0.25). Burst duration was also strongly and positively correlated with saccade size for most neurons. Figure 9C shows that the end of the burst, measured graphically or by marking individual bursts (see METHODS), was positively correlated with saccade size for 24 of 28 neurons and had a significant positive mean. A caveat is that this change in burst end did not fully contribute to the change in number of spikes because burst end occurred after the end of the saccade for about half of all bursts (the spike count was truncated at saccade end; see METHODS). So the changes in number of spikes can loosely be attributed to three factors: 1) changes in peak frequency, 2) changes in burst end for those neurons with shorter bursts, and 3) changes in saccade duration acting on those neurons with longer bursts. Finally, a fourth factor contributing to a change in number of spikes was a moderate change in burst lead (0.54 ms/deg), which was about 1/4th of the change in burst end, and in the complementary direction (increased size was associated with earlier burst onset and later burst end).



View larger version (30K):
[in this window]
[in a new window]
 
FIG. 9. Histograms of slope of regression of number of spikes (A), peak frequency (B), and burst-end (C) against size of contraversive saccades during adaptation. Mean values of slope for number of spikes and burst-end are similar for Gain+ and Gain– experiments, so average of full sample is marked with single arrow. Mean values of slopes for peak frequency are not similar (Table 1), and thus are marked separately for Gain– experiments (thin arrow) and Gain+ experiments (thick arrow).

 

Can all neurons contribute to size changes?

This study seeks to determine whether changes that occur in the discharges of FOR neurons are appropriate for causing the changes in saccade size observed during saccade adaptation. However, it is not necessarily true that all saccade-related neurons are candidates for this function. For instance, Sato and Noda (1991Go) raised the possibility that some neurons preferentially or exclusively control vertical saccades by showing that some caudal fastigial neurons are retrogradely labeled from the mesencephalic reticular formation (the site of the vertical burst generator) and not the pontine reticular formation. Similarly, Noda et al. (1992Go) found a separation of sites in the fastigial efferent pathways of the uncinate fasciculus where microstimulation elicited vertical rather than horizontal saccades. We attempted to obviate this possible problem by preselecting neurons for study. Among neurons that do project to the pontine reticular formation, not all may serve the same function. Scudder et al. (2000Go) found that only a fraction of fastigial neurons retrogradely labeled from the EBN and from the IBN regions were double labeled, and there was a partial separation of neurons projecting to each area. Consequently, we searched for criteria that might isolate neurons that are poor candidates for participation in the adaptation of horizontal saccades.

The discharges recorded during the circle task before adaptation were analyzed for directional characteristics using the model described earlier. That is, prototypical horizontal FOR neurons have an early burst containing more spikes for contraversive saccades and a later burst with fewer spikes for ipsiversive saccades. Two neurons had their longest lead and maximum number of spikes for downward saccades, whereas their longest lag and fewest spikes occurred during upward saccades. These two were classified as vertical neurons, and two others were classified as vertical-oblique. During adaptation, the number of spikes, burst duration, and peak frequency in the contraversive burst were all positively correlated with saccade size for all 4 neurons, and the slopes were within 1SD of the mean for the full sample. Similarly, for 3 of these neurons, burst lag for ipsiversive saccades was strongly positively correlated with saccade size during adaptation, whereas one vertical neuron exhibited negative correlations. In short, these neurons behaved much like the other 24 neurons with regard to encoding saccade size during adaptation.

Analysis based on the circle task also revealed 4 neurons that had minimal or no directional tuning based on burst latency and number of spikes. In addition, these cells had very poor (nonsignificant) preadaptation correlations between ipsiversive burst lag and saccade size or saccade duration; and finally, ipsiversive burst lag itself was close to zero or negative (the burst actually led saccade onset). These neurons would not appear to be good candidates for controlling saccade size according to the hypotheses stated in the INTRODUCTION, and in fact they were not. During adaptation, 2 of the neurons exhibited very small or negative correlations between number of spikes and contraversive saccade size, and 3 exhibited negative correlations between ipsiversive burst lag and ipsiversive saccade size. This latter point is illustrated in Fig. 10, which also illustrates the more general point that the best predictor of the change in ipsiversive burst lag during adaptation ("Lag Sensitivity") was the absolute burst lag that was recorded at the initiation of adaptation. These variables were related with a correlation coefficient of 0.57, which is significant at P < 0.005. Other variables were not significantly correlated with change in burst lag during adaptation. In fact, the preadaptation slope of the correlation between ipsiversive burst lag and saccade size was a poor predictor of the preadaptation slope (r = 0.19; P > 0.3).



View larger version (21K):
[in this window]
[in a new window]
 
FIG. 10. Scatter plot showing that neurons with longest ipsiversive burst lags changed the most during adaptation. Abscissa is absolute burst lag recorded at initiation of adaptation, and ordinate is slope of regression of ipsiversive burst lag against saccade size obtained during adaptation (as in Figs. 5 and 6). The 28 neurons were divided into 3 types. Most neurons exhibited classical discharge pattern with strong, early burst for contraversive horizontal saccades and weaker, later burst for ipsiversive saccades (circles). Vertical neurons exhibited strong, early burst for down saccades and weaker, later burst for up saccades, whereas oblique neurons exhibited discharge patterns intermediate between vertical and horizontal neurons (triangles). Untuned neurons exhibited approximately equal bursts in all directions and tended to have earliest bursts among all neurons (squares). Regression between slope and ipsiversive burst lag was significant at P < 0.005 (r = 0.57, slope = 0.081 ms deg1 ms1, n = 28).

 

Elimination of the 4 untuned neurons from the sample averages produced minor (<10%) changes for most variables. However, the sensitivity of ipsiversive burst lag to saccade size increased 38% to 1.77 ms/deg. Sensitivity increased for both the gain-decreasing and the gain-increasing experiments, so that the mean for the latter approached significance (P = 0.10). The other significant change was a 44% reduction (to 1.88 spikes s1 deg1) in the sensitivity of contraversive peak discharge frequency to saccade size during gain-decreasing experiments. For gain-increasing experiments, sensitivity of peak frequency increased moderately (9%), so that the difference between gain-increasing and gain-decreasing experiments approached significance (P = 0.07). Elimination of the 4 vertical and oblique neurons produced little additional change in the average sensitivity of discharge parameters to adaptation of saccade size.


    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 METHODS
 RESULTS
 DISCUSSION
 DISCLOSURES
 ACKNOWLEDGMENTS
 REFERENCES
 
This study has shown that there are several parameters of FOR-neuron discharges that change significantly as saccade size is adaptively modified. The number of spikes in the burst for contraversive saccades positively covaried with saccade size in all neurons but one, and this was in turn produced by changes in peak burst frequency and in burst duration. Ipsiversive burst lag positively covaried with saccade size in the majority of neurons, but neurons changed differently during the gain-increasing and gain-decreasing experiments. Overall, gain decreases were associated with larger changes in ipsiversive burst lag, but smaller changes in contraversive peak frequency, when compared with gain increases. This dichotomy was more evident after removal of the directionally untuned units having ipsiversive bursts that led saccade onset. Assuming that the changes in FOR discharges that we recorded are partly responsible for the changes in saccade size, as we argue later, our data suggest that short-term gain decreases are not accomplished by simply reversing the mechanism of short-term gain increases. However, a full reversal presumably occurs over the long term.

An asymmetry in the mechanisms mediating gain increases and decreases is consistent with the properties of synaptic plasticity. In the cerebellum, for instance, where plasticity exists throughout the sequence of synapses from the mossy fibers to the deep cerebellar nuclei (Aizenman et al. 1998Go; Armano et al. 2000Go; Crepel and Jaillard 1991Go; D'Angelo et al. 1999Go; Hansel et al. 2001Go. Morishita and Sastry 1996Go; Ouardouz and Sastry 2000Go; Salin et al. 1996Go), long-term potentiation and long-term depression do not always coexist at the same synapses or are more readily induced at some synapses than others. Long-term potentiation and depression at the same synapse often occur at different rates and are produced by distinctly different stimuli. Consequently, it is possible that the changes in one set of synapses that mediate gain increases are not as readily reversed, allowing a different set of synapses with more auspicious properties to initially mediate the gain decreases.

Given the dichotomy in neuronal changes during gain-increasing versus gain-decreasing experiments, we had expected to see a dichotomy in the eye movement responses, with larger changes in saccade peak velocity and smaller changes in duration for gain increases compared with gain decreases. However, no significant differences were observed. A partial explanation might be found in the change in saccade waveshape, whereby the increase in velocity found during gain-increasing experiments occurred in the epoch after peak velocity. A second explanation is that changes in peak frequency probably also affect saccade duration using mechanisms that are described later.

Apparently, not all FOR neurons participate in the modification of saccade size. The untuned neurons, whose bursts resembled the contraversive burst of other neurons in all directions, were poor candidates for controlling saccade size according to the mechanisms summarized in the INTRODUCTION and described more fully later. In some ways, they behaved oppositely to the other neurons during adaptation. On the other hand, one vertical and two oblique neurons, which at first might seem to be poor candidates for controlling the size of horizontal saccades, did participate in saccade adaptation. The paradox can be resolved by noting that there are not discrete populations of horizontal and vertical FOR neurons, but rather a continuum of best directions (Fuchs et al. 1993Go; Ohtsuka and Noda 1991aGo). Consequently, many of the FOR neurons with off-vertical preferences might nonetheless project to the horizontal burst neurons in the pons in addition to projecting to vertical burst neurons in the mesencephalon. Such neurons could have adaptively altered discharges during horizontal saccades and not have altered discharges during vertical saccades because there can be separately adapted horizontal and vertical inputs from the cerebellar cortex (Deubel 1987Go; Noda and Fujikado 1987Go).

In the remaining discussion, we consider the question of whether the changes in neuronal firing we observed during adaptation might cause the changes in saccade size. The question has two parts: 1) are there plastic changes occurring in the cerebellum that produce the changes in FOR-neuron firing, or do the cerebellar signals change as a byproduct of plastic changes occurring in structures that provide input to the cerebellum, and 2) are the changes in FOR-neuron discharges appropriate and sufficient to produce changes in saccade size.

The location of plastic synapses

To address the first question, we begin by considering the possible sites that both provide input to the cerebellum and might undergo plastic changes. Figure 11 shows a simplified block diagram of the saccadic system that is based on known connections. We have ignored cortical inputs from the frontal eye fields and parietal cortex, which principally impinge on the superior colliculus and project more weakly to the brain stem (Leichnetz and Gonzalo-Ruiz 1987Go; Lynch et al. 1985Go; Stanton et al. 1988Go). The superior colliculus is therefore regarded as a nodal point for collecting inputs from the cortex and basal ganglia and, in turn, issues the immediate saccadic command to the brain-stem saccadic burst generator. The superior colliculus also projects heavily to the cerebellum by way of nucleus reticularis tegmenti pontis (NRTP) and parts of the pontine nuclei in both cats and monkeys (Batini et al. 1978Go; Gerrits and Voogd 1987Go; Gonzalo-Ruiz and Leichnetz 1990Go; Graham 1977Go; Harting 1977Go; Huerta and Harting 1982Go; Kawamura 1974; Scudder et al. 1996a; Thielert and Thier 1993Go). The neurons of the saccadic burst generator (EBNs, IBNs, and OPNs) do not project to the cerebellum (Strassman et al. 1986aGo,bGo), and their signals may reach the cerebellum indirectly only through the smaller pathways originating in the pontine reticular formation and in raphe pontis (Fig. 11, thin lines).



View larger version (23K):
[in this window]
[in a new window]
 
FIG. 11. Schematized diagram of relationship between cerebellum and saccadic burst generator. Comma