| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Program in Neuroscience, Wake Forest University School of Medicine, Winston Salem, North Carolina 27157; 2 Department of Neurobiology and Anatomy, Wake Forest University School of Medicine, Winston Salem, North Carolina 27157
Submitted 23 January 2003; accepted in final form 24 April 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
; Schlag-Rey and Schlag 1984). Despite some intriguing findings and provocative interpretation, these two reports remain the only full accounts of the visual- and oculomotor-related properties of neurons in a group of central thalamic regions that collectively might be considered to constitute an oculomotor thalamus (OcTh). Oculomotor regions of thalamus include several nuclei within and at the borders of the internal medullary lamina (IML), a band of fibers that roughly bisects the thalamus into medial and lateral halves. Included within Schlag-Rey and Schlag's original description of areas having gaze-related activity are the paracentral (Pc), centrolateral (CL), and central superior lateral (CSL) nuclei. These 3 nuclei are found within the IML and are considered members of the rostral intralaminar group (Groenwegen and Berendse 1994; Van der Werf et al. 2002). Also included as part of OcTh are regions that border the IML, including paralaminar regions of the mediodorsal (MD), ventral anterior (VA), and ventral lateral (VL) nuclei as well as the lateral dorsal nucleus (LD), which is found at the superior extreme of the IML.
In separate reports, Schlag-Rey and Schlag (1984) and Schlag and Schlag-Rey (1984) provided thorough qualitative accounts of neurons that were active in association with self-paced spontaneous saccades and neurons that were responsive to visual stimulation in the absence of any specific motor requirement. Although rudimentary by today's standards, the spontaneous saccade and fixation tasks used in these studies revealed neural activity related to visual stimuli, saccades, active fixation, and eye position.
At the time of Schlag and Schlag-Rey's reports, most published accounts of neurons with visual and/or oculomotor activity were from recordings of the midbrain's superior colliculus (SC; for review see Sparks 1986), with a handful of studies targeting regions downstream (e.g., pontine gaze centers, Hepp and Henn 1983; King and Fuchs 1979; motor nuclei, Fuchs and Luschei 1970, 1971; Robinson 1970; Schiller 1970) or upstream (e.g., frontal eye fields, Bizzi 1968; Goldberg and Bushnell 1981; Schiller 1980) to the SC. Although clearly visual-related or saccade-related, the properties of neurons recorded in central thalamus were in many ways distinct from those recorded in the SC and downstream structures more directly associated with motor outflow. Comparatively, thalamic response types were more diverse, less stereotyped in their relation to the physical parameters of saccades (e.g., amplitude, direction), and exhibited little or no systematic topography. This, along with the fact that the intralaminar nuclei (ILN) were known to have widespread cortical and subcortical connections, led Schlag and Schlag-Rey to propose a "higher-order" role for signals in oculomotor thalamus. Rather than carry precise information about "what" would occur, Schlag and Schlag-Rey proposed that central thalamic signals contributed to determining "when" and "how" a saccade occurred by coordinating the function of cortical visuomotor regions as a saccade was being planned and executed.
Conceptually, at least, Schlag and Schlag-Rey's postulate of central thalamus as a "controller" seems timely, given recent advances regarding the neural basis of cognitive functions such as attention, working memory, and perceptual-motor decision making (for reviews see Colby and Goldberg 1999; Glimcher 2001; Miller and Cohen 2001; Schall and Thompson 1999). Electrophysiological studies have greatly expanded knowledge of how frontal eye fields (FEF), supplementary eye fields (SEF), posterior parietal (PPC), prefrontal (PFC), and cingulate cortices function in these capacities. Paralleling studies in cortex, however, has been a growing appreciation for the importance of cortical–subcortical interactions in sensorimotor control. As such, concepts of basal ganglia (Mink 1996; Redgrave et al. 1999) and cerebellar function have been dramatically revised in recent years to accommodate their potential roles in mediating most, if not all, of the aforementioned cognitive functions (Albin et al. 1989, 1995; Graybiel et al. 1994; Hikosaka et al. 2000; Houk et al. 1996; Kim et al. 1994; Middleton and Strick 1994; Mink 1996; Thach 1996).
Although electrophysiological studies of central thalamus remain few, accumulating anatomical evidence continues to strengthen the link between these regions and visuomotor function by placing the ILN and paralaminar regions of MD, VA, and VL at the center of several putative computational loops. Thalamocortical loops consisting of reciprocal connections with FEF, SEF, PPC, and PFC have been established (see Groenewegen and Berendse 1994; Jones 1985; Macchi and Bentivoglio 1986 for reviews). Paralaminar VA is an obligatory synapse in the basal ganglia "oculomotor loop" (Alexander et al. 1986), which originates within the FEF and SEF (Lynch et al. 1996; Shook et al. 1991), flows through the caudate nucleus, the substantia nigra pars reticulata (SNr), and thalamus [ventral anterior nucleus (VAmc)] before returning to its point of cortical origin. Along with cortical terminations, the anterior intralaminar nuclei and paralaminar regions of VA and VL (McFarland and Haber 2000) project to regions in the caudate nucleus that are also targeted by FEF and SEF, a projection that may serve to modulate activity within the so-called oculomotor loop (see Groenewegen and Berendse 1994 for review). A projection from paralaminar MD to FEF is known to carry information from the SC (Sommer and Wurtz 1998, 2002) and may thus be the final link in a cortico-subcortical loop that originates and returns to FEF (FEF–SC–MD–FEF). OcTh also mediates the influence of deep cerebellar nuclei on cortical activity. The gaze-related cerebello-thalamic projection arises from the dentate nucleus and terminates in paralaminar VL (Area X of Olszewski) that in turn projects to both the FEF and the SEF (Lynch et al. 1994, 1996; Shook et al. 1991).
Despite compelling anatomical evidence of a central thalamic role in visuomotor control, little is known about the visuomotor properties of neurons in these regions. Only very recently have a few studies examined the activity of these neurons using controlled behavioral tasks. Although not in the context of a saccadic task, recordings from the centre median–parafascicular complex (Cm-Pf) of the caudal intralaminar group indicate that neurons in this region represent behaviorally relevant sensory stimuli (Matsumoto et al. 2001) and may participate in attentional orienting (Minamimoto and Kimura 2002). Most recently, Tanibuchi and Goldman-Rakic (2003) reported that neurons in MD show spatially selective activity in the context of visually guided and memory-guided saccade tasks, a finding that is consistent with known connections between MD and PFC.
In this study we systematically evaluated the capacity for central thalamic neurons to represent information relevant to the performance of goal-directed saccadic eye movements. Recordings were concentrated in areas previously investigated by Schlag and Schlag-Rey (1984a,b), including paralaminar regions of the VA and VL nuclei along with nuclei of the IML. In contrast to the original studies of these regions, we used a visually guided delayed saccade task. This task couples a specific sensory stimulus, by an instructed delay, to a specific saccadic response. This task readily distinguishes between the sensory-contingent and saccade-related activities of individual neurons and permits classification based on their conjunction. Determining whether a neuron is visual-related, motor-related, or visuomotor is relevant for understanding the processing stage or stages to which it contributes. Having identified sensory- and motor-related response components, we further characterized neurons by quantifying both the timing and spatial selectivity of these components.
The delayed saccade task also permitted evaluation of each neuron's capacity to maintain task-relevant information throughout an instructed delay. Spatially selective "delay-period" activity may be the hallmark of a neuron with the potential to participate in "higher-order" aspects of sensorimotor function. In numerous visuomotor regions, such activity has been associated with the neural correlate of factors such as movement selection (Glimcher and Sparks 1992), motor planning (Barash et al. 1991a,b; Bracewell et al. 1996; Mazzoni et al. 1996; see Andersen 1995 for review), spatial attention (see Colby and Goldberg 1999 for review), and perceptual judgment (see Glimcher 2001; Schall and Thompson 1999; Shadlen and Newsome 1996 for reviews). We determined whether— and if so, with what fidelity—thalamic neurons represented spatial information throughout the imposed delay.
In brief we found activity in central thalamus to represent each phase of the visually guided delayed-saccade task with many constituent neurons modulated in association with more than one task period. Across the sample, considerable variability in the timing and spatial selectivity of visual- and saccade-related responses suggested a range of possible functions. We also report that a significant number of neurons were found to carry spatial information throughout the instructed delay period, a finding that supports the idea that central thalamus participates in higher-order aspects of visuomotor control. Preliminary versions of these results have appeared in abstract form (Massoglia et al. 2001; Wyder and Stanford 2000; Wyder et al. 2001).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
All surgical and experimental protocols complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, USDA regulations, and the policies set forth by the Wake Forest University School of Medicine Animal Care and Use Committee (ACUC). Three rhesus monkeys (Macaca mulatta) were prepared for chronic single-unit recording. Each monkey underwent two sterile surgical procedures while under general isoflurane anesthesia. During the first surgery, an MRI-compatible titanium post was attached to the skull using titanium screws and orthopedic bone cement. Also at this time, a preformed loop of Teflon-coated stainless steel wire (eye coil) was implanted beneath the conjuctiva to circumscribe the cornea of one eye (Judge et al. 1980). During subsequent training/recording sessions, the post served to restrain the monkey's head, whereas the eye coil provided an analog signal of eye position (Fuchs and Robinson 1966; Robinson 1963). Recovery from the initial surgery required 2–4 wk, during which time analgesics and antibiotics were administered as required.
Fully recovered animals were trained on the behavioral task (see following text). Once trained to a criterion level of performance, a second surgery was performed to place an MRI-compatible recording cylinder (Crist Instrument) over the oculomotor thalamus (OcTh). A presurgical MRI was carried out to optimize the stereotaxic coordinates of the cylinder for individual monkeys. The recording cylinder was positioned over a small craniotomy (about 15 mm diameter) and secured with titanium screws and bone cement. Daily recording sessions began on full recovery (2–3 wk).
Recording procedures
Eye position was recorded using the search coil method (Fuchs and Robinson 1966; Robinson 1963). Briefly, the monkey sat in a primate chair with head restrained at the center of a pair of orthogonal (horizontal and vertical) magnetic fields. The magnetic fields induced current to flow within the surgically implanted eye coil. This current, when decomposed into horizontal and vertical components, yielded analog signals proportional to the angular relationships between the eye coil and the horizontal and vertical magnetic fields, respectively (i.e., eye position). Horizontal and vertical eye positions were sampled and stored at 500 Hz.
Neural activity was recorded using parylene-coated, tungsten microelectrodes (Micro Probe) having impedances between 1.0 and 1.5 M
at 1 kHz. Activity was amplified, filtered (300 Hz to 4 kHz), and monitored using an oscilloscope and an audio monitor. The action potentials of single neurons were isolated using a time/amplitude window discriminator and spike times stored at a resolution of 10 µs. Electrodes were advanced through a dura-piercing cannula and advanced to OcTh by hydraulic microdrive. Generally, the electrode was advanced rapidly to 15 mm below the surface of the dura and then more slowly to detect landmark changes in background activity. A "quiet" period as the electrode passed through the lateral ventricle was followed by a return of activity on entry into dorsal thalamus. Within the thalamus, oculomotor regions were identified by observing/listening for modulations that coincided with task events (see Experimental design, below). We recorded from any isolated neuron that appeared to be modulated in association with any phase of the task.
Experimental design
During training and subsequent recording sessions, monkeys were seated in a primate chair in a very dimly lit room. The stimulus display consisted of an array of light-emitting diodes (LEDs). The distance between adjacent LEDs was either 1 or 2 in., which, at a viewing distance of 57 in., corresponded to 1 or 2° of visual angle (Cartesian coordinates), respectively. Maximum horizontal and vertical stimulus eccentricities were 24 and 21°, respectively. Standard operant methods were used to train monkeys to look toward visual targets for liquid reward (drop of water or juice). Neural data presented in this report were collected primarily in association with performance of a visually guided delayed-saccade task, diagrammed in Fig. 1. Each trial began with the presentation of a central fixation stimulus (left), which the monkey had to acquire within 500 ms. After a variable interval (300–700 ms), a second stimulus was illuminated at an eccentric location (middle). The monkey was required to withhold eye movement to the eccentric stimulus until the fixation light was extinguished (GO signal, right). This interval, the delay period, ranged from 500 to 1000 ms. Once given the GO signal, the monkey was required to look to the target within 500 ms and maintain fixation on the target for an additional fixed interval (200 or 500 ms) to obtain a liquid reward.
|
A majority of OcTh neurons were tested for direction selectivity by choosing targets from a circular array of eight target locations. The 8 locations were spaced equally (at 45°) on the circumference of an imaginary circle with 0° (360°) corresponding to straight right. Because LEDs were organized in a Cartesian coordinate frame the polar coordinates of target LEDs were not exact, but chosen to best approximate the desired polar coordinates. Target direction was randomized across trials and presented within blocks of fixed eccentricity at either 6, 10, or 20° (i.e., the radius of the array). The eccentricity that yielded the highest level of activity was estimated on-line and chosen for the first block of trials. Time permitting, direction tuning was tested for a second or third eccentricity. A minority (30/162) of neurons was not tested with the circular target array. For these neurons, targets were presented within the receptive/movement field as estimated on-line by the experimenter.
Data analysis
In the context of the delayed saccade task (see Fig. 1), a prototypical visual-motor neuron generates an early transient burst of activity in response to the sensory stimulus and, after the imposed delay, a motor-related burst associated with the saccade (e.g., see Fig. 4A). In addition, sustained activation that intervenes between the sensory and motor-related bursts is not uncommon (e.g., see Fig. 4A). Analyses were tailored to estimate the direction selectivity and the time of occurrence of sensory and motor-related modulations.
|
|
|
Direction tuning
Figure 2A illustrates average instantaneous firing frequency histograms (bin width = 2 ms) for activity associated with saccades to each of 8 visual targets. The 8 targets were spaced equally at 45° on the circumference of an imaginary circle with 0° (360°) corresponding to straight right. Proceeding counterclockwise, 90, 180, and 270° corresponded to up, left, and down, respectively. Target eccentricity (i.e., radius) was 10° in this example. Each histogram represents the average of multiple trials (N) for movements to a particular target with activity for all trials aligned on saccade onset (t = 0) before averaging. For each target direction, a moving window was used to find the 100-ms epoch for which average firing frequency was highest (Step 1). Searching from 250 ms before to 400 ms after saccade onset (dashed vertical lines), the average firing rate was computed across successive groups of 50 bins (100 ms), each time incrementing the window by 1 bin (2 ms). For each direction, the epoch found to have the highest average firing rate is bracketed by solid vertical lines. Of the 8 epochs defined by this method, that with the highest average firing frequency was selected for the purpose of estimating the parameters of direction tuning (Step 2). In this example, the interval with the highest average firing (gray highlight) corresponded to the time period from 58 ms before saccade onset to 42 ms after saccade onset for movements to the 90° (up) target. The polar plot of Fig. 2B summarizes the relationship between mean firing rate and target direction for this interval that, as expected, shows a strong bias for the 90° target.
Direction tuning was assessed based on the interval identified as having the highest average firing frequency (see Step 2 above). The firing rate within this interval was determined for saccades to each of the 8 targets and plots of firing rate versus saccade direction were fit with a Gaussian function (Fig. 2C). Least-squares fit of the Gaussian function
 |
With the appropriate modifications, the same method was used to evaluate the tuning of decreases in activity. Accordingly, the initial search returned the 100-ms period with the lowest average firing rate (Step 1), and the Gaussian tuning function (Step 2) specified an inverse peak.
Timing
For each neuron, the timing of task-related modulations (Step 3) was estimated from average activity profiles that were based on activity for the 3 directions defining the peak of the Gaussian tuning function. Accordingly, for the example in Fig. 2, movements to the 90, 45, and 135° targets were included in the timing analysis (Fig. 2C, gray highlight). The resulting averaged histogram, synchronized on saccade onset (t = 0) is shown in Fig. 3A. With the appropriate average frequency histogram generated, the goal was to identify a statistically significant change in average firing rate. The first step was to choose an appropriate interval for estimating a baseline firing rate. For motor-related activity, we used a 100-ms interval beginning 350 ms before and ending 250 ms before saccade onset (dotted vertical lines). During this period, the monkey was engaged in stable fixation. This epoch was long after presentation of the target (minimum of 300 ms after target onset) and would therefore exclude visual stimulus–related onsets in activity.
We used a resampling method to determine the likelihood that any given firing rate could result by chance from random samples drawn from the baseline population. Assuming that average firing rate remains stationary over time, we can consider the 100-ms baseline interval to be composed of 50 independent samples (2-ms bins) of firing frequency. We sampled (with replacement) this baseline population, each time drawing N samples (n = number of trials contributing to the averaged histogram; n = 33 for this example), and each time calculating the mean of the N samples. The probability density function shown in Fig. 3B is the result of 1,000 iterations of this procedure and plots the probability of obtaining any given mean firing rate from the baseline distribution. The corresponding cumulative density function is shown (Fig. 3C). Beginning at the end of the baseline period (t = –250 ms) and proceeding until 400 ms after saccade onset, the spike rate of each successive bin was evaluated against the resampled baseline distribution (null hypothesis) until the null hypothesis was rejected (P < 0.01) for four consecutive bins. The onset of the burst was taken as the time corresponding to the first of the 4 bins to exceed the criterion level. For the neuron shown in this example, the average frequency had to exceed 69.0 spikes/s to meet the P < 0.01 criterion level. A solid line corresponding to the criterion frequency is shown on each plot. A solid vertical line on the average firing frequency histogram (Fig. 3A) indicates the time at which the activity exceeded this threshold, which in this case was –64 ms (i.e., 64 ms before saccade onset). A second vertical line corresponds to the first of 4 successive bins that failed to meet significance, thus defining the end of this transient activation (46 ms after saccade onset).
The same method was used to evaluate the timing of decreases in activity, except that the analysis sought 4 consecutive bins that were significantly below baseline.
For neurons that were not tuned, or for neurons not tested with the circular target array (limited number of target locations chosen based on the on-line estimate of best area), timing was based on the average response to all target positions at a given eccentricity.
Tuning and timing of visual activity
Figures 2 and 3 illustrate our methods for evaluating motor-related activity. The same 3-step method, with appropriate alterations, was used to quantify sensory-linked increases and decreases in activity. For this, rasters were aligned on target onset rather than saccade onset. The initial search was performed over a 300 ms interval beginning at target onset. Gaussian functions were fit to plots of average firing rate versus target direction, not saccade direction, and the baseline period chosen for the timing analysis corresponded to a period of stable fixation 100 ms before target onset. The bin by bin search for statistically significant modulations in activity proceeded from the time of target onset to the end of the delay period (GO signal).
Modulation index
To compare the strength of sensory- and motor-related modulations for neurons that responded in association with both events (n = 71), a modulation index (MI) was calculated as follows
 |
Histology
During a single, final recording session, electrolytic lesions were made by passing 10 µA for 20 s at several locations. Lesion sites were chosen to mark the locations and boundaries of the regions where neurons were recorded. One week postlesion, monkeys were sedated with ketamine, administered an overdose of sodium pentobarbitol, exsanguinated, and perfused with heparanized saline and 4% paraformaldehyde. The brain was blocked, equilibrated in 30% sucrose, and frozen sections were cut at 50-µm thickness. Every other section was mounted and stained for Nissl substance (cresyl violet).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
Figure 4 illustrates the task-related activity profile of a representative visual-motor neuron. In most respects, the activity profile of this OcTh neuron is qualitatively similar to that for visual-motor neurons found in areas such as the SC or FEF (see DISCUSSION). In Fig. 4A, rasters (top) and average frequency histograms (bottom) for the same set of trials are shown synchronized both to target onset (left) and saccade onset (right). This neuron displayed two transient bursts of activity, one stimulus-linked and one motor-linked. It is worth noting that, given typical reaction times in the range of 150–300 ms, these bursts could not have been resolved had they been recorded in the context of a simple reaction time task. Along with permitting distinctions between sensory- and motor-related modulations, the delayed saccade task reveals a neuron's capacity to carry information throughout an instructed delay (see INTRODUCTION). In this example, delay period activation is evident as sustained and increasing activity that effectively bridges the gap between the sensory and motor-related bursts.
A primary objective of this study was to provide a quantitative characterization of neurons in OcTh that would be comparable to those that exist for more commonly and more recently studied visuomotor regions. Toward this end, we focused primarily on the essential measures of timing and spatial tuning for visual- and motor-related modulations (see METHODS). For this example neuron, recorded in the left hemisphere, the sensory burst (Fig. 4A, left) began 68 ms after stimulus onset (t = 0) and endured for 60 ms (i.e., to t = 128 ms; second horizontal black bar). The motor-related burst (Fig. 4A, right) began 64 ms before saccade onset, reached a peak coincident with saccade onset, and declined with saccade execution, returning to baseline near the end of the saccade at a postsaccadic time of 46 ms (second horizontal black bar). Direction tuning was assessed separately for the sensory and motor bursts and the corresponding tuning functions are shown below for the visual-related (Fig. 4B, left; horizontal gray bar in A) and saccade-related periods (Fig. 4B, right; horizontal gray bar in A). Estimates of preferred direction were reasonably consistent at 68° for the visual- and 83° for the motor-related transient, in both cases indicating a preference for stimuli/saccades in the upper quadrant of the contralateral (right) space. Tuning for the visual epoch (35°) was somewhat sharper than that for the motor epoch (55°). Finally, motor-related activity was considerably more vigorous than sensory-related activity as quantified by the estimates of amplitude provided by the fitted Gaussians (visual = 45 spikes/s; motor = 126 spikes/s).
Figure 5 plots the distributions of timing and direction tuning estimates for the entire sample. In general we found the onset times of visual and motor-related modulations to vary over considerable ranges, but each distribution to be largely defined by a single mode. As for the example neuron just described, the majority of stimulus-related onset times cluster tightly near a poststimulus time of 100 ms (Fig. 5A). Considering only this primary mode (n = 72; onsets <300 ms) and excluding the minor mode consisting of 9 neurons with onset times exceeding 300 ms (see following text, Visual-related activity, for discussion), the mean poststimulus latency was 98 ± 58 ms. In contrast, however, the onset times of motor-related modulations were broadly distributed, ranging from 248 ms before saccade onset to 442 ms after saccade onset (mean = 26 ± 118 ms) with postsaccadic modulations (80/152: 53%) as common as presaccadic (72/152: 47%) modulations (Fig. 5D).
|
Overall, visual- and motor-related activities were broadly tuned for direction. Of the 81 neurons with visual-related responses, 71 were tested for direction tuning with the radial target array. Of these, 46 (65%) showed significant tuning as estimated by least-squares fit of the Gaussian function. Of the 152 neurons with motor-related modulations, 124 were tested with the radial target array with 83 (67%) well fit with a Gaussian. The corresponding estimates of tuning index (Fig. 5, B and E) and preferred direction (Fig. 5, C and F) for these subsets of neurons are also shown. Considering increases and decreases together, the mean tuning indices for significantly tuned stimulus-related and saccade-related modulations were 48 ± 36 and 55 ± 34°, respectively. Thus OcTh neurons were quite strongly influenced by task-related events over an entire quadrant of sensory/motor space. Preferred directions for sensory-contingent modulations were almost entirely (41/46 = 89%) contralateral (Fig. 5C), but a contralateral bias, although present (46/83 = 55%), was much less pronounced for motor-related activity (Fig. 5F). Table 1 summarizes these results separately for increases and decreases in sensory and motor-related activity. Increases and decreases were observed in similar proportion for sensory-related (77%/23%) and motor-related modulations (72%/28%). These results are considered in more detail below in the individual sections on sensory- and motor-related activity.
|
We note that lack of a significant Gaussian fit does not necessarily imply that a visual or motor-related response was omnidirectional (i.e., roughly equivalent responding for all target directions). For visual activity, 5/25 "untuned" neurons appeared to be truly omnidirectional, having strong stimulus-linked activity that was not selective for target direction. Omnidirectionality was somewhat more common for motor-related responses, with 18/41 cases suggesting no preference for direction. The remaining response versus direction functions were varied, consisting of some that approached but failed to meet our statistical significance criterion for a Gaussian fit (visual: 4/25; motor: 2/41); those with a clear direction preference, but one that was poorly described by a Gaussian function (visual: 5/25; motor: 13/41); and finally, those that were ambiguous by virtue of being composed of very weak (visual: 11/25; motor: 8/41) responses.
Visual-related activity
Other than response polarity, the summary histograms of Fig. 5, A–C do not provide a strong basis for categorical distinctions among OcTh neurons on the basis of visual-related modulations in activity. As stated in METHODS, the search for a statistically significant modulation in activity proceeded forward in time from the time of visual stimulus onset to the end of the delay period (a minimum of 500 ms). It can be reasonably argued that the few neurons (n = 9) with very long latencies (>300 ms) are not "visual-related" in any meaningful sense. Inspection of these few examples suggested that these modulations anticipated later saccade-contingent modulations. As noted above, the cluster of onset times near 100 ms is more consistent with that which has been reported for sensory-related activity in other visuomotor areas (see DISCUSSION). Examples of individual neurons drawn from this population are shown in Fig. 6 to illustrate some of the qualitative features of these neurons. For each neuron, Fig. 6, A–I illustrates rasters and average firing frequency histograms aligned on stimulus onset and composed of responses to preferred direction stimuli (see legend for details). As was true for the example of Fig. 4, many neurons showed a relatively high-frequency transient that was tightly linked to the stimulus onset (Fig. 6, A and B). For others, the transient was less pronounced and, as evident in the rasters, occurred less reliably (Fig. 6, C and D), and for still others, the transient was largely absent (Fig. 6, E and F).
|
The neurons depicted in Fig. 6, A–F show fairly substantial differences along several dimensions (e.g., onset time, transient duration, transient strength, etc.); however, attempts to define classes of neurons based on any single quantifiable metric were not successful. Rather, the variability observed in any given measure was more suggestive of a response continuum than a category boundary. As noted, one clear categorical distinction among sensory-contingent modulations was whether activity increased or decreased. Examples of three neurons with sensory-linked decreases in activity are shown in Fig. 6, G–I. As noted above, just under one-quarter (23%) of the visual-related modulations were decreases in firing rate. Like increases, decreases were varied with some clearly transient (Fig. 6I), others less obviously so (Fig. 6H), and still others clearly sustained (Fig. 6G). Although the timing and tuning distributions for decreases overlapped those of increases (see Fig. 5), decreases were somewhat less likely to be fit by a Gaussian function (decreases: 10/19; 53%; increases: 36/52; 70%) and, when fit, were more broadly tuned for direction (decreases: 83 ± 40°; increases: 38 ± 29° t-test, P < 0.001).
Although differences in onset latency did not suggest categories of visual response, neurons clearly differed with respect to the quality (i.e., transient/sustained) of the response. Figure 7 summarizes the time course of visual-related modulations for the 72 neurons for which poststimulus modulation occurred within 300 ms. Each neuron is represented as a horizontal line with the start of the line indicating the earliest time (i.e., onset) at which activity differed from baseline (above for increases, below for decreases). Line endings signify times at which activity returned to baseline levels unless occurring at 500 or 750 ms (vertical dashed lines show delay period end), in which case firing was significantly above (increases) or below (decreases) baseline throughout the entire delay period. When present, a dot signifies the time at which a higher-frequency transient gave way to a lower level of firing. The offset times of high-frequency transients were determined with the same procedure used for determining onset times (see METHODS) simply by choosing a posttransient baseline period and searching backward in time for a significant increase above baseline.
|
Figure 7 parses neurons into 3 groups. Within a given group, neurons are arranged from bottom to top in order of increasing latency. Forty-six of the 56 neurons that were excited by visual stimuli showed an initial transient response component (1–46 on plot). For many (n = 30), the relatively high-frequency transient gave way to a lower frequency of sustained activity that often persisted throughout the delay period (e.g., Fig. 6B) and that, in some cases, was preceded by a brief pause in firing (e.g., Fig. 6A). We found no consistent qualitative or quantitative differences between neurons with transient-only and transient-sustained responses. Ten neurons (47–56; shaded) clearly lacked a transient component but demonstrated low-frequency sustained activity throughout the delay period (e.g., Fig. 6F). Neurons 57–72 (unshaded) decreased their activity in association with visual stimuli. Decreases, like increases, could be either transient (e.g., Fig. 6I) or sustained (e.g., Fig. 6G).
Effect of target eccentricity on tuning for direction
As described in METHODS, our primary index of spatial selectivity was direction tuning and we attempted to examine direction selectivity at what was estimated on-line to be the most effective of 3 possible eccentricities, 6, 10, and 20° (a few neurons were examined at 4°). Nineteen visually responsive neurons were examined at multiple target eccentricities for all 8 directions with 8 neurons tested at 2, and 11 neurons tested at 3 eccentricities. Of these 19 cells, 13 were significantly tuned for direction at more than one eccentricity with 2 neurons tuned at 3 eccentricities, 11 at 2 eccentricities, 4 at a single eccentricity, and 2 not tuned. Comparison of Gaussian tuning functions, shown in Fig. 8 for 8 different neurons, illustrates several points. First, direction-selective OcTh neurons are also selective for eccentricity. Considering the peaks of the Gaussian tuning functions, it is apparent that firing rate varied with stimulus eccentricity for most of these example neurons and, furthermore, that different OcTh neurons had different preferred eccentricities. The neuron in Fig. 8A, for example, fired most strongly at an eccentricity of 6° (dotted line) with firing rate decreasing monotonically with increasing target eccentricity (solid line: 10°; dashed line: 20°). In contrast, the neuron depicted in Fig. 8B responded most strongly at 20°, the largest eccentricity that we could test with the circular stimulus array (LED board 24 x 21°). For these two neurons, the firing rate loosely approximated a multiplicative function of target direction and target eccentricity, the primary effect being on the amplitude of the Gaussian tuning function with other parameters (preferred direction, baseline, tuning index) relatively unchanged. For other neurons, changes in amplitude, tuning index, and baseline were present in varying combinations and to varying degrees, although preferred directions remained fairly congruent across eccentricity.
|
The examples shown in Fig. 8 suggest that the direction selectivity of OcTh neurons was often the product of both excitation and suppression. The baseline level of activity, obtained during steady fixation before stimulus onset, is indicated as a horizontal dashed line on each of the plots in Fig. 8. Note that, for several of these neurons, activity evoked by stimuli at nonpreferred directions was reduced compared with prestimulus levels (Figs. 8, A, B, D, and E). A somewhat analogous (opposite in sign) phenomenon was observed for neurons that were predominantly suppressed by visual stimulation. Thus for example, the tuning functions shown in Fig. 8, G and H are the product of suppression of background activity. In Fig. 8G, tuning at 10° eccentricity is the result of suppression at just 3 directions in contralateral space (unshaded), whereas at an eccentricity of 20°, suppression is much more powerful and extends to all target directions (i.e., 360°). A similar result is shown in Fig. 8H for a second neuron.
Motor-related activity
As shown in the summary histograms of Fig. 5, saccade-related modulations could be either increases or decreases and these could be either pre- or postsaccadic. In contrast to visual-related modulations, the timing of the activity with respect to the saccade can clearly be the basis for a functional distinction among neurons with motor-related activity. Examples of 8 neurons, each with a saccade-related transient, are shown in Fig. 9. For each neuron, rasters corresponding to saccades in the preferred direction are shown aligned on saccade onset (black vertical line; t = 0). Tick marks indicating the times of the "GO" signal (i.e., fixation offset, green tick) and saccade offset (red ticks) are shown for each trial. Proceeding from Fig. 9, A–H, the onset of the saccade-related burst becomes progressively later, ranging from 124 ms before saccade onset (Fig. 9A) to 68 ms after saccade onset (Fig. 9H). As suggested by the summary histograms (see Fig. 5), the intervening examples (Fig. 9, B–G) illustrate that pre- and postsaccadic modulation did not represent a simple dichotomy; rather, the temporal relationship between saccade onset and the burst onset varied continuously. The same could be said about other temporal parameters of the transient. Note that, as a group, neither the time of peak firing nor the time of burst offset bore any consistent relationship to the onset or offset of the saccade. Thus for example, the neurons in Fig. 9, A–C were primarily presaccadic, fired maximally near the time of saccade onset and ceased firing before or near the time of saccade offset. In contrast, those in Fig. 9, D–F began firing later (at, or just after saccade onset), attained peak rate near the time of saccade termination and declined thereafter. Finally, those depicted in Fig. 9, G and H did not begin to fire until approximately 20–40 ms after the saccade had ended.
|
The relatively high incidence of postsaccadic activity in OcTh is particularly interesting. Such late activity could serve a multitude of functions as putative feedback carrying information about the ongoing or recently completed saccade (see DISCUSSION). Alternatively, the possibility that this activity could be related to something other than the saccade itself had to be considered. Visually guided delayed saccades were made to an LED illuminated against a uniformly black background in a very dimly lit room. Thus one possibility is that activity was the consequence of a saccade that brought the visual stimulus into or through the receptive field. Given an average visual response latency near 100 ms and minimum latencies on the order of 50 ms (see Figs. 5 and 7), this is a valid possibility only for relatively late postsaccadic activity such as that shown in Fig. 9, G and H, which begins after saccade offset. Given these afferent delays, activity that began during a saccade (e.g., Fig. 9, E and F) occurred too early to have been visually evoked. For those with late postsaccadic activation, we compared responses for visually guided saccades and saccades to remembered locations in which the target was extinguished before saccade onset. Examples of two postsaccadic neurons that responded identically for visually guided (gray) and remembered saccades (black) are shown in Fig. 10, A and B (red ticks: saccade offset). Rather than a saccade-induced sensory response, these neurons appeared to carry a late saccade-related signal. The neuron depicted in Fig. 10B reliably fired after small corrective saccades as well and these bursts are evident between 200 to 300 ms after the primary saccade on visually guided trials (correctives rarely occurred on memory trials). In all, there were 23 neurons with a postsaccadic burst that was temporally coupled to the saccade. Of these, 20 were examined for both visual and memory trials with 12/20 apparently motor-related as evidenced by roughly equivalent activity for stimulus-present and stimulus-absent trials.
|
For 8 neurons, the postsaccadic burst was attenuated or completely absent for saccades to remembered target locations. Although it is not possible to rule out the "sensory" explanation for these neurons, we observed that neurons with pronounced differences in firing on target-present and target-absent trials were just as likely to be presaccadic (11/19) as postsaccadic (8/19). For example, the neurons depicted in Fig. 10, C and D showed a strongly attenuated response for saccades made in the absence of the target (black). In one case the activity is clearly presaccadic (Fig. 10C) and in the other (Fig. 10D) it begins near the time of saccade onset. On the basis of timing, both would appear to be saccade-related. Taken together, it seems most likely that neurons that differentiated between visual and memory trials, whether pre- or postsaccadic, carried motor-related activity specific for saccades to available visual goals.
Motor decreases
Of all saccade-related modulations, 43 (28%) were decreases. Mean onset time for decreases (68 ± 132 ms) was not significantly different from that for increases (11 ± 103 ms), with both slightly postsaccadic. However, as was true for visual-related activity, the firing rate versus direction functions were less likely to be fit with a Gaussian for decreases (14/37; 38%) than for increases in activity (69/87; 78%) and, when fit, decreases tended to be more broadly tuned (decreases: 87 ± 29°; increases: 48 ± 32°; t-test, P < 0.001). Moreover, compared with increases, a somewhat higher proportion of motor-related decreases (9/14; 64%) had ipsilateral preferred directions (28/69; 41%) (see Fig. 5; Table 1).
Contralateral versus ipsilateral
Presaccadic neurons in principally visuomotor areas like the SC and FEF have almost exclusively contralateral preferred directions (see DISCUSSION). As shown above, ipsilateral tuning was not uncommon for the motor-related modulations of OcTh neurons. For instance, 5 of the examples shown in Fig. 9 had ipsilateral preferred directions (Fig. 9, A, B, and F–H) and 3 had contralateral preferred directions (Fig. 9, C–E). Nothing among these examples would suggest any quantitative (e.g., timing) or qualitative differences between ipsilateral- and contralateral-tuned neurons. To verify this, we compared OcTh neurons with contralateral (n = 46) and ipsilateral (n = 37) preferred directions to determine whether they differed with respect to motor-related timing or tuning. Neither mean onset time (contra: 10.7 ± 100.4 ms; ipsi: –7.0 ± 87.5 ms) nor tuning index (contra: 52.6 ± 25.6°; ipsi: 57.6 ± 43°) was significantly different for neurons with contralateral and ipsilateral preferred directions.
Effect of target eccentricity on tuning for direction
Twenty-six neurons with direction-selective motor-related activity were examined for effects of target eccentricity. Nine neurons were tested at 2, and 17 neurons at 3 target eccentricities. Of these, 20 met the criterion for Gaussian direction tuning at more than one eccentricity with 6 neurons tuned for direction at 3 eccentricities, 14 at 2 eccentricities, and 6 at a single eccentricity. Figure 11 compares the motor-related direction tuning functions obtained at different eccentricities for 6 OcTh neurons. As was true for visual-related activity, saccade-related modulations showed selectivity for eccentricity with a primary effect on firing rate (e.g., Fig. 11A). When significant direction tuning was present at multiple eccentricities, the parameter estimates (e.g., preferred direction) were largely consistent. The example also illustrates that, as expected, tuning for direction breaks down for eccentricities that fail to evoke adequate levels of activity (Fig. 11, B, D, and F, unfitted data points).
|
Transient and sustained timing
The examples (see Figs. 9 and 10) shown to this point have emphasized saccade-related transient increases, and appropriately so. Most saccade-contingent modulations were increases (109/162; 67%) and the majority (77/109; 71%) of these were transient or had a pronounced transient component. Figure 12A summarizes the timing of neurons with saccade-related transients for both increases (1–77; unshaded) and decreases (78–96; shaded). The conventions are the same as described for Fig. 7, except that dots indicate the time of maximum average firing frequency for each transient. Responses were considered transient if the modulation in activity returned to baseline before the end of the trial. All but a handful of transient activations began within ±100 ms of saccade onset with most reaching their peak firing near or just after (within 75 ms) saccade onset. The longest lines in Fig. 12A indicate 2 unusual neurons with a pronounced transient followed by lower frequency sustained activation until the end of the trial. Figure 12B illustrates the same timing summary for neurons that lacked a transient component to their response. These responses were considered sustained because activity did not return to baseline before the end of the trial (lines ending at 250 ms indicate neurons recorded with a 200-ms target-fixation interval). Once activated, these neurons (32 increases; 24 decreases) often fired for the duration of the trial until reward was issued and the monkey was allowed to break fixation from the target (t = 250 or 500 ms). The distribution of onset times for sustained neurons clearly differed from that for transient neurons, with many sustained modulations beginning more than 100 ms after saccade onset. On average, transients began earlier and closer to saccade onset (5.62 ± 100 ms) than did sustained activations (64.5 ± 128 ms) and this difference was significant (t-test; P < 0.005).
|
Our operational definition of "motor-related" included any significant modulation in activity that occurred during the time period from 250 ms before saccade onset to the end of the trial. With rasters aligned on saccade onset, the tendency for impulses to occur at the same time across trials would suggest a temporal linkage to the motor event; however, rather than convey information about the saccade itself (e.g., timing, metrics), activity could reflect (e.g., change in orbital position) or anticipate (e.g., reward) some consequence of the saccade. This possibility seems remote for transients that are tightly linked to the saccade, but seems more likely for sustained activations.
Eye position-related activity
For one group of neurons, sustained postsaccadic activation was related to changes in eye position. These neurons fired at a steady rate during periods of stable fixation with firing rate a reflection of eye position in the orbit. Figure 13 shows an example of one such neuron. Figure 13A presents rasters for individual trials aligned on saccade onset (vertical line) for purely horizontal saccades (red tick marks, saccade offset) to contralateral (below) and ipsilateral (above) targets of 10° (gray) and 20° (black) of eccentricity. Before saccade onset, during fixation on central LED, the rate of sustained activity was intermediate to that for fixations to the left or right. The relationship between mean firing rate and eye position (eccentricity) is plotted in Fig. 13B. Each of the 4 sets of data points plots firing rate versus eccentricity for targets along one of the 4 possible axes defined by the 8 target circular target (i.e., horizontal, vertical, 2 diagonals). Note that the function is steepest for fixations along the horizontal axis (circles). Along this axis, the neuron was completely silent for ipsilateral fixation positions but the firing rate increased monotonically as a function of progressively more eccentric contralateral fixation positions. The rate–position functions for fixations along the 2 diagonals are similar but have slopes that are somewhat less steep, whereas there is no systematic trend apparent for fixations along the vertical axis (diamonds). This pattern of results indicates that it was not simply eccentricity, but eccentricity along a preferred axis (horizontal in this case) that determined the firing rate. Consistent with this, Fig. 13C shows good agreement between the individual curves when the data points are replotted as a function of the horizontal component of eccentricity.
|
250 ms. Top group (unshaded): all decreasing responses. Two vertical dotted lines mark 50 and 100 ms after target onset and are shown only for reference.
