| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Physiology, W. M. Keck Foundation Center for Integrative Neuroscience, and Neuroscience Graduate Program, University of California at San Francisco, California, 94143
 |
ABSTRACT |
|---|
 Abstract
 Introduction
Methods
Results
Discussion
References
|
|---|
Ferrera, Vincent P. and Stephen G. Lisberger. Neuronal responses in visual areas MT and MST during smooth pursuit target selection. J. Neurophysiol. 78: 1433-1446, 1997. We recorded the activity of single neurons in the middle temporal (MT) and middle superior temporal (MST) visual areas in two macaque monkeys while the animals performed a smooth pursuit target selection task. The monkeys were presented with two moving stimuli of different colors and were trained to initiate smooth pursuit to the stimulus that matched the color of a previously given cue. We designed these experiments so that we could separate the component of the neuronal response that was driven by the visual stimulus from an extraretinal component that predicted the color or direction of the selected target. We found that for all cells in MT and MST the response was primarily determined by the visual stimulus. However, 14% (8 of 58) of MT neurons and 26% (22 of 84) of MST neurons had a small predictive component that was significant at the P 
0.05 level. In some cells, the predictive component was clearly related to the color of the intended target, but more often it was correlated with the direction of the target. We have previously documented a systematic shift in the latency of smooth pursuit that depends on the relative direction of motion of the two stimuli. We found that neither the latency nor the amplitude of neuronal responses in MT or MST was correlated with behavioral latency. These results are consistent with a model for target selection in which a weak selection bias for the intended target is amplified by a competitive network that suppresses motion signals related to the nonintended stimulus. It is possible that the predictive component of neuronal responses in MT and MST contributes to the selection bias. However, the strength of the selection bias in MT and MST is not sufficient to account for the high degree of selectivity shown by pursuit behavior.
A fundamental issue for both visual attention and voluntary movement control is the problem of selection. How do the systems that initiate voluntary movements or covert shifts of attention choose a particular target from an array of candidates? We have studied this problem with the use of voluntary smooth pursuit eye movements in monkeys as a model system. The smooth pursuit system in primates is a sensorimotor pathway that serves to stabilize the retinal image of moving targets. The pursuit system shows a high degree of selectivity in the way it transforms sensory input into motor output. One aspect of selectivity in the pursuit system is the ability to select one of several moving stimuli and to faithfully track the motion of the selected target while filtering out the motion of distractors (Ferrera and Lisberger 1995 Experiments were conducted on two male rhesus monkeys (Macaca mulatta) weighing ~6 and 7 kg. The same monkeys were used in earlier behavioral studies (Ferrera and Lisberger 1995 Behavioral tasks
Monkeys were trained to track moving targets presented on a cathode ray tube monitor. We used a step-ramp target motion paradigm to minimize the occurrence of saccades during pursuit initiation (Rashbass 1961
Visual stimulation
Pursuit targets were generated and controlled by a Univision Piranha video framebuffer with an on-board microprocessor (Texas Instruments TMS 34020). The output from the video board was displayed on a calibrated 20-in. color monitor (Barco) with a 60-Hz noninterlaced refresh rate. The monitor stood at a viewing distance of 30 in. so that the display area subtended roughly 30° horizontally by 20° vertically. The spatial resolution of the display was 1,280 pixels × 1,024 lines, and the depth was 8 bits per pixel. Pursuit targets were small (0.9°) colored squares presented on a uniform gray background. The luminance of the fixation mark, cue, and targets was 15.0 cd/m2, whereas the background was 5.0 cd/m2.
Eye movement recording
Eye position was monitored with the use of a monocular scleral search coil system (CNC Engineering). Separate horizontal and vertical eye position signals were fed through an analog differentiator (low-pass, Neuronal recording and data collection
When behavioral training was complete, a recording chamber (20 mm diam) was implanted on the skull overlying the superior temporal sulcus (STS) or the operculum. We used two approaches to MT and MST: a horizontal approach through V1 with the recording chamber angled 15° from the horizontal plane (monkey B) and a vertical approach through the inferior parietal lobule (monkey F). In both cases, the center of the recording chamber was positioned 15 mm lateral to the midline. Each day a hydraulic microdrive was mounted on the recording chamber, filled with sterile mineral oil, and sealed. Transdural recordings were made with the use of Pt/Ir or tungsten extracellular electrodes with impedances of ~1-2 M Verification of recording sites
For each monkey, we initially sampled over a wide area of the STS, effectively "mapping" the RF topography of the posterior and anterior banks. These RFs were used to guide electrode penetrations and to help with the assignment of individual units to MT or MST. However, not every RF corresponds to a single unit used in the study. Many RF plots correspond to multiunit recording sites and some correspond to sites that were not used because their RFs were too eccentric or they did not respond to small moving targets. It should be noted that many sites were excluded because the RFs were too large to plot within the confines of our tangent screen, which was limited to the central 25° of the visual field (as measured along the horizontal meridian). Figure 2 shows the relationship between RF size (square root of RF area) and eccentricity for single units and multiunit sites that were assigned to MT (n = 76) or MST (n = 36). We calculated linear regession lines for RF size versus eccentricity, and the slopes of these lines are in general agreement with other work (Komatsu and Wurtz 1988a
Responses to single targets
Before assessing the role of MT and MST in target selection, it is useful to determine whether responses were appropriate for identifying the direction, color, and position of the moving stimuli. We did this by extracting responses to single targets that were presented on random trials during the course of the target selection task. We constructed peristimulus time histograms for all cells that provided sufficient data (28 in MT, 83 in MST). These histograms are plotted in Fig. 3. Data for MT and MST were combined. Figure 3A shows responses at the preferred location, whereas Fig. 3, B-D, shows responses at nonpreferred locations. The black histograms are the responses to the preferred direction or color and the gray histograms are responses to the nonpreferred direction or color. For each cell, the trial-by-trial responses were first summed and then normalized to the number of trials. Then the responses for all cells were summed and normalized by the number of cells. The Y-axis therefore represents the firing rate per cell per trial. The X-axis represents time in milliseconds. Histograms were smoothed with the use of an exponential low-pass filter with a time constant of 10.0 ms. The bar on the X-axis demarcates the first 100 ms of target motion. Smooth pursuit eye movements were generally initiated after the end of this 100-ms interval.
Selection bias in MT and MST
We recorded from MT and MST neurons with the use of a task designed to dissociate the sensory and cognitive components of the neuronal response. Figure 4 shows the responses of a single neuron recorded in MST to a subset of trials for this task. Each column shows responses to the visual stimulus configuration depicted at the top of the column. Filled squares represent the green stimulus and open squares represent the red stimulus. The cell had a large RF that covered both stimuli. Each row corresponds to a particular instruction about which stimulus to track (green or red). Each of the eight subplots shows the spike histogram (unit), stimulus onset (stim), and horizontal eye velocity (E
Relationship between neuronal and behavioral latency
We have previously shown that the presence of a moving distractor shifts the latency of pursuit (Ferrera and Lisberger 1995
Relationship between neuronal response amplitude and behavioral latency
It is possible that behavioral latency might be correlated not with neuronal latency but neuronal response amplitude. This would be the case if there were a firing rate threshold that needed to be reached before an eye movement was initiated (Hanes and Schall 1996
Estimation of the "net" directional signal
As a final measure, we were interested in determining the net directional signal that is available to stages of the pursuit pathway downstream from MT and MST. In other words, what does the activity in MT and MST look like to areas that are "reading out" the distributed representation of image motion? In Fig. 10 we show the summed activity of all MT and MST neurons (the histograms were constructed as in Figs. 3 and 6) when the target moved in either the preferred (black) or null (gray) direction. In Fig. 10, A and B, we included responses to stimuli comprising a target and distractor where the target was presented at the preferred location for each cell. In this case, there is a robust directional signal regardless of whether the target and distractor move in the same (Fig. 10A) or opposite (Fig. 10B) direction. This confirms the results of Fig. 9, which also show that there is little net influence of distractors at nonpreferred locations.
The main goal of this study was to determine how multiple discrete moving targets are represented in MT and MST and whether that representation is biased in a way that predicts which target will be used to guide smooth pursuit. For MT, it is known that there is a crude topographic map of retinal position (Maunsell and Van Essen 1983a
What accounts for the behavioral selectivity of pursuit?
The strength of the selection bias we measured raises some interesting practical issues in relating neuronal selectivity to behavioral selectivity. One can imagine two extremes of selectivity, which we will refer to as "strong selection" and "weak selection." These correspond respectively to the "winner-take-all" and "vector-averaging" dichotomy suggested by others (Groh et al. 1995 What accounts for the behavioral latency of pursuit?
Behavioral latencies for smooth pursuit show a characteristic dependence on distractor motion that can be seen in Fig. 8 and Table 1. However, we found no systematic relationship between behavioral latency and either neuronal response latency or amplitude. The dissociation between responses in MT and MST and the latency of pursuit is interesting because others have found that manipulations of the visual stimulus can lead to correlated changes in neuronal and behavioral latency. For example, Krauzlis (1991) Theoretical considerations
The current results are surprising because a number of theoretical arguments and experimental observations gave reason to expect much stronger response modulation in MT and MST during target selection. Recently, Treue and Maunsell (1996)
INTRODUCTION
Abstract
Introduction
Methods
Results
Discussion
References
, 1997). The goal of this study is to understand how cortical motion processing in the middle temporal (MT) and middle superior temporal (MST) visual areas contributes to this selectivity.
; Lisberger et al. 1987). There is a great deal of evidence suggesting that these signals are provided by the cortical "motion pathway," which includes MT and MST. These areas are part of a pathway through parietal visual cortex that is primarily concerned with spatial and motion information (Albright 1984; Maunsell and Van Essen 1983b; Zeki 1978) and is thought to be involved in visual orienting and guidance (Andersen 1987; Goodale and Milner 1992; Ungerleider and Mishkin 1982). One of the major subcortical targets of MT and MST is the dorsolateral pontine nucleus (Boussaoud et al. 1992; Brodal 1978; Fries 1981; Glickstein et al. 1980, 1985; Maunsell and Van Essen 1983a; Ungerleider et al. 1984), which projects to the cerebellum and is thought to play a role in the generation of smooth eye movements (Brodal 1979, 1982; Gerrits and Voogd 1989; Mustari et al. 1988; Suzuki and Keller 1988; Suzuki et al. 1988; Their et al. 1988). Microstimulation of MT or MST during pursuit can affect smooth eye velocity (Groh et al. 1995; Komatsu and Wurtz 1989), whereas lesions of these areas lead to characteristic deficits in pursuit (Dursteler and Wurtz 1988; Dursteler et al. 1987).
,b). These cells constituted ~30% of neurons in MT and MST and were distributed preferentially in regions where receptive fields (RFs) are near to or include the fovea (Desimone and Ungerleider 1986; Komatsu and Wurtz 1988a); pursuit cells were found in roughly equal proportions in foveal MT (MTf), ventrolateral MST (MSTl), and the dorsal pole of dorsomedial MST (MSTd). Some pursuit cells had a maintained response even when the target was stabilized on the retina or turned off briefly, thereby eliminating retinal image motion (Newsome et al. 1988). This "extraretinal" signal was strongest in MSTd, weak (but not absent) in MTf, and of variable strength in MSTl. At present it is not known to what extent the responses of pursuit cells represent signals related to eye position, eye velocity, or an abstract representation of the pursuit target in a nonretinocentric coordinate frame. Almost nothing is known about the role of these neurons in target selection. Do neuronal responses in MT and MST form an equally weighted representation of all moving stimuli or are they biased in a way that favors those stimuli that are to become the target of an eye movement? A more precise way to state this is to ask whether the response of MT or MST neurons is determined entirely by the retinal stimulus or whether it predicts the identity or movement of the intended target.
, 1997).
METHODS
Abstract
Introduction
Methods
Results
Discussion
References
, 1996). Our methods were approved by the UCSF Institutional Animal Care and Use Committee. Monkeys were trained to move voluntarily from their home cage to a primate chair. A method modified from Wurtz (1969) was used to train each monkey to attend a stationary target. Surgery was then performed under sterile conditions to implant a coil of wire on one eye (Judge et al. 1980) and to secure a platform to the skull for head restraint (Miles and Eighmy 1980). For all subsequent training and experiments, the monkey's head was secured to the ceiling of the primate chair and a set of field coils was lowered over the chair so that we could use a magnetic search coil to monitor horizontal and vertical eye position. The eye coil was calibrated by having the monkey attend to targets at different positions, and the monkey was subsequently required to keep the direction of gaze within 2-3° of target position. Correct performance of the task was rewarded with drops of fruit juice or water.
). Trials were initiated by requiring the monkey to look at a stationary central fixation light. After a short interval, a moving perifoveal target appeared. At the same time, the central fixation light was turned off and the monkey was required to track the target by initiating a smooth pursuit eye movement. The monkey was given a liquid reward provided gaze was kept directed toward the desired target for the duration of the trial. The monkey's performance was monitored by tracking eye position relative to a ±3° fixation window centered around the target. Several steps were taken to ensure that the monkeys did not make anticipatory eye movements. First, the initial target location relative to the fixation mark was randomized from trial to trial. Second, the direction of target motion (toward or away from the vertical meridian) was also randomized. Third, trials were aborted if the monkey initiated an eye movement before the fixation light went off.
) (Fig. 1). The monkey initially fixated on a small (0.4°) white square in the center of the screen. After a few hundred milliseconds, the white fixation mark was replaced by a colored square (red or green) of the same size and luminance. This was the cue. The color cue lasted for 250 ms and then changed back to the white (Fig. 1A). After a second time interval of 200 ms, the fixation target disappeared and two moving stimuli appeared (1 red, 1 green). This second interval was used so that neuronal responses to the onset of the pursuit targets would not be confused with any color-dependent response to the offset of the cue. The monkey's task was to pursue the stimulus that matched the color of the cue (we call this stimulus the target and the nonmatching stimulus the distractor). After the stimuli appeared, the monkey initiated horizontal smooth pursuit with a latency generally in the range of 80-160 ms. On some randomly interleaved trials, only a single stimulus appeared so that we could measure the animal's "normal" pursuit under similar visual conditions. The color, initial position, and direction of the target were randomized trial-to-trial so that the monkey could not anticipate the direction of the required eye movement.
View larger version (14K):
[in a new window]
FIG. 1.
Illustration of the basic color-cue task that the monkeys were trained to perform. A: time course of trial events. At the start of each trial, the animal foveated a white central fixation target. Shortly thereafter, the fixation target changed to 1 of 2 colors, red or green ("cue"). The cue was presented for 250 ms, then it disappeared, and 200 ms later 2 colored moving targets appeared. The animal's task was to look at the target that matched the color of the cue, which was accomplished by initiating a smooth eye movement in the appropriate direction. The other target is referred to as the "distractor." B: spatial representation of the task. All stimuli were presented on a 20-in. color video monitor. In the standard configuration, a target of either color could appear in any of 4 locations (dashed squares) and move either toward or away from the vertical meridian. The distractor appeared in 1 of the other 2 locations that did not have the same vertical position as the target. Target and distractor were separated by 3° vertically. The distractor was always a different color than the target and could move in either the same or the opposite direction. Given these constraints, a block of trials consisted of all random permutations of target color, direction, and position and 3 distractor conditions (no distractor, same direction, opposite direction) above or below the initial fixation position. To tailor the display for each individual cell, we could rotate and scale the stimulus positions in polar coordinates. We could independently rotate and scale the velocities (speed and direction) of the targets. The target and distractor axes of motion remained parallel.
3 dB at 25 Hz) to yield horizontal and vertical eye velocity. The eye position and eye velocity signals were then digitally sampled by computer at 1 kHz per channel and stored on disk for further analysis.
at 1 kHz (Wolbarsht et al. 1960). Signals from the microelectrode were amplified, filtered, and monitored on an oscilloscope and audio monitor.
2 times the peak of the background noise. Action potentials were converted to digital pulses with a time-amplitude window discriminator. The time of occurrence of action potentials was recorded with a precision of 0.01 ms on the same time base used for recording vertical sync pulses from the video board. When a unit was isolated, stimulus parameters such as the axis and speed of target motion were adjusted to optimize its response. Data were then collected while the animal performed the target selection task. When data collection ended for the target selection tasks, we attempted to map the RF of the unit by hand while the monkey fixated a stationary spot.
). We subsequently limited our recordings to regions of MT and MST representing the central 10°, which is where pursuit initiation is optimal. We found two distinct representations of the central visual field in the STS. One was located more dorsally on the anterior bank and corresponded to MSTd. The other was located more ventrally on the posterior bank and corresponded to MTf or MSTl. We were not able to distinguish reliably between MTf and MSTl. In the remainder of the paper, the designation "MST" refers to cells that were clearly in MSTd, whereas "MT" refers mainly to cells that were in MTf but includes some cells that might have been in MSTl.
View larger version (15K):
[in a new window]
FIG. 2.
Receptive field (RF) size vs. eccentricity. For a given eccentricity, RFs in the middle superior temporal (MST) visual area tend to be larger than those in the middle temporal (MT) visual area, although there is substantial overlap. Solid line: linear regression for the MT data (slope = 0.69, Y-intercept = 1.0, r = 0.83). Dashed line: linear regression for the MST data (slope = 1.16, Y-intercept = 3.0, r = 0.70).
). The borders of MT were located on the basis of its distinctive myeloarchitectonics (Van Essen et al. 1981). The locations of the cortical recording sites were estimated on the basis of microdrive readings and distance relative to cortical borders.
RESULTS
Abstract
Introduction
Methods
Results
Discussion
References
View larger version (36K):
[in a new window]
FIG. 3.
Responses in MT and MST to single moving pursuit targets sorted by direction, color, and location. A: responses at the preferred location for each cell. Black histograms: responses to the preferred direction or color. Gray histograms: responses to the nonpreferred direction or color. B-D: responses at nonpreferred locations. Black bars underneath each histogram: 1st 100 ms of stimulus motion or the "open-loop" interval for pursuit. X-axis: time (ms). Y-axis: per-cell, per-trial response (spikes/s). Histograms have been smoothed with the use of a low-pass exponential filter with a time constant of 10 ms.
, who recorded mainly from neurons that started to respond well after the initiation of pursuit. This could be due to different stimulus conditions in the two studies or to the possibility that we isolated different subpopulations of neurons. Direction selectivity is apparent at all stimulus locations. From the position sensitivity, one can infer that two discrete moving stimuli will activate topographically distinct but overlapping populations of neurons. Color selectivity was poor, although it should be noted that we made no attempt to measure cell's true color preferences, which would entail the use of a much broader spectrum of target colors. When we compared responses to the green and red stimuli on a cell-by-cell basis, we found no systematic preference for one color or the other (MT: green average = 20.9 spikes/s, red average = 21.6 spikes/s, paired t-test, green vs. red P = 0.49; MST: green average = 30.5 spikes/s, red average = 29.1 spikes/s, paired t-test, P = 0.13). This indicates that the small color preference shown in Fig. 3 probably was not contaminated by a luminance artifact that systematically biased responses for one color or the other.
), respectively. The eye velocity traces have been shifted vertically to fit in the space provided. The beginning of each trace represents zero eye velocity. Positive (upward) deflections correspond to rightward movements and negative (downward) deflections correspond to leftward movements. This cell had both an early and a late response. The vertical dashed lines demarcate the time of the early response. The late response was highly selective for leftward image motion and was strongest during rightward eye movements. The early response (indicated by the vertical dashed lines) was triggered by the retinal stimulus but was significantly stronger when the animal was instructed to track the green target (unpaired t-test, P < 0.05). The early response averaged 46.3 spikes/s on "track green" trials, whereas the response on "track red" trials averaged 29.5 spikes/s. This cell was typical of those that showed significant extraretinal response modulation.
View larger version (22K):
[in a new window]
FIG. 4.
Responses of a dorsomedial MST (MSTd) neuron recorded while the animal performed a target selection task. This unit has an early response (- - -) when the animal is cued to track the green target (top), but a much weaker response when the animal is cued to track the red target (bottom). Early response is correlated with the color of the target, but not the direction of the eye movement. unit: neuronal response histogram. stim: target motion onset pulse. E : eye velocity trace; traces have been shifted vertically to fit in the space allowed. In each case, the early part of the trace represents 0 eye velocity. Upward deflections: eye movements to the right. Downward deflections: eye movements to the left. Histogram binwidth:8 ms.
).
0.05). Cells that passed both tests are plotted as filled circles. Cells that passed only the consistency test are plotted as open circles and cells that passed neither test are plotted as small squares. Although the selection bias is generally rather weak, both areas have a substantial number of cells that passed both tests (14% in MT and 26% in MST).
View larger version (14K):
[in a new window]
FIG. 5.
Selection bias in MT and MST. Responses to the same retinal stimulus were sorted according to the color of the cue. Filled circles: differences that were significant at the P 0.05 level. A: selection bias in MT. Fraction in the bottom right corner: proportion of cells with significant differences. B: selection bias in MST.
0.05). In MT, for the distractor-same condition, we found that 5 of 45 (11%) cells were significantly modulated versus 8 of 58 (14%) cells in the distractor-opposite condition. For MST, 22 of 84 (26%) cells were significantly modulated for distractor-same conditions and 20 of 84 (24%) for distractor-opposite conditions.
where Rg is the response following a green cue and Rr is the response following a red cue. We computed an index for the overall response and for each distractor condition (same/opposite) separately. The average modulation index was0.2 (MT: 0.2 ± 0.02, median = 0.14; MST: 0.2 ± 0.02,median = 0.14). The average modulation index for units with significant effects was in the range of 0.25-0.35 (MT: 0.34 ± 0.07, median = 0.35; MST: 0.32 ± 0.05, median = 0.26). Results for the same and opposite distractor conditions were similar. The cell in Fig. 4 had a modulation index of 0.22.
(1)
View larger version (40K):
[in a new window]
FIG. 6.
Time course of selection bias in MT and MST revealed by peristimulus time histograms. A: cells that showed a significant bias for "track green." B: cells that showed a significant bias for "track red." In both cases, the black histogram is the response to the preferred instruction and the gray histogram is the reponse to the nonpreferred instruction. Y-axis is scaled to the average response per cell per trial (spikes/s). X-axis: time (ms). Bar on the X-axis: 1st 100 ms of target motion. Histograms were smoothed with the use of a low-pass exponential filter with a time constant of 10 ms.
0.05) effect for color and 1 had a significant effect for direction. For 83 cells in MST, 4 were significant for color and 6 were significant for direction. The statistical significance of the effects is diluted by the fact that responses were averaged over optimal and nonoptimal stimuli, thus increasing the response variance.
View larger version (13K):
[in a new window]
FIG. 7.
Ability of cells to predict the color or direction of the intended target. Responses to all stimuli with target and distractor moving in opposite directions were sorted according to the color and direction of the target. Differential response to color is plotted on the X-axis and differential response to direction is plotted on the Y-axis. Data for MT ( 
) and MST (
) were combined. Axes are scaled in spikes/s.
, 1997). Distractors moving in the same direction as the target reduce the latency of pursuit, whereas distractors moving in the opposite direction increase latency. The latency shift can be interpreted as evidence of a competitive mechanism for target selection. We were therefore motivated to examine the relationship between neuronal response latencies and the latency of smooth pursuit. We did this on a cell-by-cell and trial-type-by-trial-type basis by taking all trials of a given type and constructing average spike histograms and eye velocity traces. Only trials in which the target moved toward the fovea were included. The neuronal and behavioral latencies were determined by visual inspection of the averaged responses: the action potential peristimulus time histograms for neuronal latency and the horizontal and vertical eye velocity traces for behavioral latency. Most cells had fairly crisp and reliable responses, so that there was no difficulty in determining the onset of the response. Cells that did not have clear onsets for targets moving toward the fovea were excluded from this analysis. The analysis was performed on 38 MT cells and 41 MST cells. For each individual cell that was used, neuronal latencies were taken only from trial types that had clear response onsets; thus responses to targets moving in the null direction were often excluded. Behavioral latencies were recorded for every trial type. In addition, neuronal responses were excluded if they occurred later than 200 ms or were clearly related to retinal image motion subsequent to the initiation of pursuit.
View larger version (26K):
[in a new window]
FIG. 8.
Neuronal and behavioral response latencies. A: MT neuronal vs. behavioral latency sorted by distractor condition (+, single targets; , target and distractor same direction; 
, target and distractor opposite directions). Scatterplot shows only data for trial types that provided both a neuronal and a behavioral latency measurement. Distributions below scatterplot: neuronal latencies sorted by distractor condition, with latencies for single targets in gray. Distributions at right of scatterplot: behavioral latencies for all trial types, again sorted by distractor condition with single targets in gray. B: data for MST plotted in the same format. Horizontal and vertical scales for both the scatterplots and the distributions are identical. All latencies were taken from averaged histograms with 4-ms time bins. For display purposes only, the data points in the scatterplots were randomly jittered by ±2 ms in both dimensions to reduce overlap among points falling in the same bin.
View this table:
TABLE 1.
Relationship between neuronal and
behavioral latency
). If counterdirectional movement suppresses neuronal responses in MT and MST, as suggested by others (Qian and Andersen 1994; Snowden et al. 1991), this could account for the longer behavioral latencies in the distractor-opposite condition. We tested this by comparing neuronal responses to different distractor conditions (single, same, and opposite). For each cell, we first found the single-target condition (direction and position) that gave the strongest response. We then computed the response to that target when it was paired with a distractor moving in the same direction and when paired with a distractor moving in the opposite direction. We performed this analysis for the same set of cells shown in Fig. 4. In Fig. 9 the response to paired targets (, same condition; , opposite condition) is plotted versus the response to single targets. There were no systematic differences between single and paired targets (i.e., target + distractor) or between same and opposite direction distractors. For the two distractor directions, we computed the difference in response (same opposite). The distribution of these differences is show in Fig. 9, insets. The mean differences are not significantly different from zero (MT: 0.04 ± 0.91, median = 0.71; MST: 0.15 ± 0.75, median = 0.63). If pursuit latency is related to response amplitude, then the presence of a distractor should have no behavioral effect, which is contrary to what is observed.
View larger version (16K):
[in a new window]
FIG. 9.
Response amplitude sorted by distractor condition. For each cell, the reponse to a single target moving in the preferred direction is plotted against the response to the same target paired with a distractor moving in the same direction ( ) and paired with a distractor moving in the opposite direction (). Inset: distribution of differences for the same and opposite distractor conditions (same opposite). A: results for MT. B: results for MST.
View larger version (23K):
[in a new window]
FIG. 10.
"Net" direction selectivity in MT and MST for 2 moving stimuli. A: average response to stimulus configurations where the target was at the preferred location for each cell. Black histogram: response when the target moved in the preferred direction. Gray histogram: reponse when the target moved in the null direction. Difference between the preferred and null responses is plotted below. Distractor moved in the same direction as target. B: same as A, except that the distractor moved in the opposite direction to the target. C: average response when either the target or distractor was at the preferred location. Black histogram: target moving in the preferred direction. Gray histogram: target moving in the null direction. Distractor moved in the same direction as the target. D: same as C, but distractor moved in the opposite direction to the target. Y-axis is scaled to the average response per cell per trial (spikes/s). X-axis: time (ms). Bar on X-axis: 1st 100 ms of target motion. Histograms were smoothed with the use of a low-pass exponential filter with a time constant of 10 ms.
DISCUSSION
Abstract
Introduction
Methods
Results
Discussion
References
,b) combined with an orderly representation of direction (Albright 1984). One might therefore imagine that two moving targets should be represented by two peaks of activity in MT corresponding to direction columns at different topographic locations. Ideally, one would like to compare the amplitude of these two peaks to find out whether the peak corresponding to the distractor is smaller than that corresponding to the target. However, because we could only measure the activity at a single site, we instead measured the response to the same retinal stimulus while varying the instruction (cue) given to the animal regarding which target to track. We were thus able to dissociate the response of a single neuron into one component that was attributable to sensory (retinal) signals and another that was attributable to cognitive (extraretinal) factors. We refer to the latter as the selection bias or predictive component of the response.
). Therefore it is reasonable to expect that selection effects should be found in areas that represent either the color or the motion of the target (or both). Some of the selection bias found in MT and MST seemed to be related to color of the target rather than its direction (e.g., Fig. 4). The color-dependent component of the selection bias might reflect attentional modulation in cortical areas that are more sensitive to color and that have connections with MT and MST, such as V4. Our single-unit results with the use of the color task are in general agreement with another study in which monkeys were cued for the shape or location of the target (Recanzone et al. 1993). Our results are also consistent with a report that color can enhance neuronal sensitivity to motion in MT (Croner and Albright 1996).
). The notion of attention as an outcome of competition among sensory signals goes back at least as far as Broadbent (1958) and has been employed in various computational models of attention and eye movements (Ferrera and Lisberger 1995, 1996; Koch and Ullman 1985; Scheinberg and Zelinsky 1993). One version of such a model is shown in Fig. 11. In this model, target and distractor velocity signals are processed by a competitive network and the output of this network is the desired eye velocity command used by the motor system to generate the eye movement. A selection bias that favors the direction of the target determines the output of the network. The current results indicate that the predictive component of responses in MT and MST might be a correlate of the selection bias in the model.
View larger version (17K):
[in a new window]
FIG. 11.
Conceptual model for target selection. Target and distractor signals are processed by a bistable competitive network. Desired target is determined by a selection bias that influences the outcome of the competition.
; Salzman and Newsome 1994). Under conditions similar to those of the present experiments, pursuit behavior shows strong selection. The direction of pursuit is identical to the direction of the selected target and is not influenced by the motion of the distractor (Ferrera and Lisberger 1997). [It should be noted that when there is no attentional bias, the direction of pursuit is intermediate between that of the target and distractor (see Groh et al. 1995; Lisberger and Ferrera 1996).] However, the effects of attention that have been documented in visual cortex, including those of the present study, clearly indicate weak selection. These effects are almost always limited to less than half of the cells examined, typically ~25-40%. In the cells that show significant attentional modulation, the mean effect is usually something like 30% enhancement or suppression of the response. This leaves a considerable gap between cortical physiology and eye movement behavior. To bridge this gap, we have proposed that there is a competitive network that serves to amplify the weak selection bias, converting it into an all-or-none output. It has been shown mathematically that some types of competitive networks can generate winner-take-all output from an arbitrarily small selection bias (Yuille and Grzywacz 1989), so that weak cortical selection is not necessarily an obstacle for the generation of accurate movements.
; Moran and Desimone 1985; Treue and Maunsell 1996). Although we did not explore this systematically, we kept the separation between the target and distractor in the range of 3-12°. Figure 3 shows that there are substantial responses to stimuli presented at nonpreferred locations. From this it may be inferred that most stimulus locations were inside the RF of a given cell. In addition, behavioral measurements indicate that distractor motion affects the latency of pursuit as long as the distractor is within 16° of the target (Ferrera and Lisberger 1996) and that it makes no difference if the target and distractor are in the same or opposite hemifields (Ferrera and Lisberger 1995). These considerations suggest that the neural site of target selection should have RFs or horizontal connections that are bilateral and cover at least the central 16° of the visual field, like those in MSTd or dorsolateral pontine nucleus (Suzuki et al. 1990). We have also observed that pursuit performance tends to break down at small separations. When the target and distractor are separated by <3°, animals make substantially more mistakes and latencies are exceptionally long, suggesting that the pursuit system has difficulty resolving motion signals within small RFs, such as those in MTf.
; Preddie et al. 1995). We considered manipulating the task in a way that would extend pursuit latencies beyond their normal range, but we felt that this would have distorted our results. Effects of attention that occur on longer time scales might not be relevant to pursuit initiation under natural conditions, where response latencies are on the order of 100 ms.
found that varying the "motion onset delay" of a visual target (i.e., the amount of time the target remains stationary before it begins to move in the step-ramp paradigm) affects pursuit latency such that increasing the delay leads to shorter latencies. Movshon et al. (1990) subsequently showed that increasing motion onset delay also reduces neuronal latencies in MT in anesthetized monkeys. Kawano et al. (1994) showed a similar correlation in alert monkeys. They found that increasing the speed of a target reduces the latency of ocular following (a smooth eye movement evoked by movements of a large-field stimulus) and also reduces the latency of neuronal responses in MST. In fact, the slope of the function relating neuronal to behavioral latency was nearly 1.0, indicating a very close temporal correspondence between responses in MST and the generation of ocular following. For visuomotor behaviors such as smooth pursuit, visual processing delays necessarily result in delayed motor responses. By showing a dissociation between visual responses in cortex and behavioral latency shifts induced by competing stimuli, we can argue that these behavioral effects are not due to sensory processing delays.
). One candidate for such a mechanism is the competitive network mentioned earlier. In previous work, we have shown that distractor-induced shifts in pursuit latency are consistent with the internal dynamics of a competitive network (Ferrera and Lisberger 1995, 1997). The lack of a correlation between neuronal response latency or amplitude and behavioral latency indicates that neither MT nor MST is the site of this competitive bottleneck.
; Shadlen et al. 1996). The "evidence" in this case is the directional bias remaining after one averages out the confounding effects of distractor motion. If one assumes that the pursuit system uses a weighted sum of activity in MT and MST (each neuron's response weighted by its preferred direction), then the direction bias available to downstream stages of the pathway is roughly equivalent to the differential response to motion shown in Fig. 7 and Fig. 10, C and D. Figure 10C shows that when the target and distractor move in the same direction, there is a strong directional signal. But when target and distractor move in opposite directions (Fig. 7 and Fig. 10D), the net directional signal is many times weaker. The directional signal in Fig. 10C also has much greater statistical reliability than that in Fig. 10D. The greater magnitude and reliability of the direction signal should translate into faster responses when the target and distractor move in the same direction.
reported robust modulation of responses in MT and MST when monkeys performed a discrimination task requiring covert attention to moving targets. If there is a necessary link between covert shifts of attention and the programming of eye movements, as postulated by Rizzolati et al. (1987), then one would predict robust attentional effects during eye movement tasks in cortical areas that are modulated by covert attention. Another theory, known as the "what/where" or "action/perception" hypothesis, predicts that attentional effects in the parietal pathway, including MT and MST, should be stronger for visual guidance tasks than for covert attention tasks (Goodale and Milner 1992; Milner and Goodale 1995; Ungerleider and Mishkin 1982). The observation of weak modulation during smooth pursuit target selection contradicts aspects of both theories and indicates that there might be a dissociation between the effects of covert attention and eye movements in MT and MST. This complements the observation that there is selective enhancement of responses in the frontal eye field preceding saccades (Goldberg and Bushnell 1981) but not during covert attentional shifts. Bushnell et al. (1980) also found both attention-related and saccade-related enhancement in area 7a of parietal cortex. The overall pattern of results favors the idea that there are distinct but overlapping cortical substrates for overt and covert orienting, with prefrontal cortex being more involved in overt shifts and parietal cortex more involved in covert shifts of attention. This is consistent with the view that there are distinct mechanisms for visual attention and motor response selection (Pashler 1991).
). On the other hand, cells in MST with RFs away from the fovea have visual responses that are suppressed during maintained pursuit (Erickson and Thier 1991). It now seems likely that MST (and to a lesser extent, MT) is involved in distinguishing the "true" motion of the target from image motion of the stationary background that is induced by the eye movement, a distinction that is important for perceptual stability. However, even though MT and MST respond vigorously to stimulus motion before the initiation of pursuit, they appear not to have a prominent role in the decision that determines which of several moving stimuli will ultimately become the target of a smooth eye movement. Cognitive signals in MT and MST therefore appear to be more closely related to pursuit maintenance than to pursuit initiation.
| |
ACKNOWLEDGEMENTS |
|---|
We thank Dr. Karl Gegenfurtner for sharing graphics software and Dr. John Maunsell for helpful comments.
This work was supported by the McDonnell-Pew Program in Cognitive Neuroscience Fellowship JSMF 92-38 to V. P. Ferrera and National Eye Institute Grant EY-03878 to S. G. Lisberger.
| |
FOOTNOTES |
|---|
Present address and address for reprint requests: V. P. Ferrera, Dept. of Psychiatry and Center for Neurobiology and Behavior, Columbia University, 722 West 168th St., New York, NY 10032.
Received 13 December 1996; accepted in final form 21 May 1997.
| |
REFERENCES |
|---|
|
Abstract
Introduction
Methods
Results
Discussion
References
|
|---|
This article has been cited by other articles:
|
|
 |
|
  W. K. Page and C. J. Duffy Cortical Neuronal Responses to Optic Flow Are Shaped by Visual Strategies for Steering Cereb Cortex, April 1, 2008; 18(4): 727 - 739. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Spering and K. R. Gegenfurtner Contrast and Assimilation in Motion Perception and Smooth Pursuit Eye Movements J Neurophysiol, September 1, 2007; 98(3): 1355 - 1363. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Bakola, G. G. Gregoriou, A. K. Moschovakis, V. Raos, and H. E. Savaki Saccade-Related Information in the Superior Temporal Motion Complex: Quantitative Functional Mapping in the Monkey J. Neurosci., February 28, 2007; 27(9): 2224 - 2229. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. J. Majaj, M. Carandini, and J. A. Movshon Motion Integration by Neurons in Macaque MT Is Local, Not Global J. Neurosci., January 10, 2007; 27(2): 366 - 370. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
X. Li and M. A. Basso Competitive Stimulus Interactions within Single Response Fields of Superior Colliculus Neurons J. Neurosci., December 7, 2005; 25(49): 11357 - 11373. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. J. Krauzlis The Control of Voluntary Eye Movements: New Perspectives Neuroscientist, April 1, 2005; 11(2): 124 - 137. [Abstract] [PDF]
|
|
|
|
|
|
J. A. Perge, B. G. Borghuis, R. J. E. Bours, M. J. M. Lankheet, and R. J. A. van Wezel Temporal Dynamics of Direction Tuning in Motion-Sensitive Macaque Area MT J Neurophysiol, April 1, 2005; 93(4): 2104 - 2116. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Tanimoto, M. Takagi, T. Bando, H. Abe, S. Hasegawa, T. Usui, A. Miki, and D. S. Zee Central and Peripheral Visual Interactions in Disparity-Induced Vergence Eye Movements: I. Spatial Interaction Invest. Ophthalmol. Vis. Sci., April 1, 2004; 45(4): 1132 - 1138. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. J. Krauzlis Recasting the Smooth Pursuit Eye Movement System J Neurophysiol, February 1, 2004; 91(2): 591 - 603. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. W. Bisley, D. Zaksas, J. A. Droll, and T. Pasternak Activity of Neurons in Cortical Area MT During a Memory for Motion Task J Neurophysiol, January 1, 2004; 91(1): 286 - 300. [Abstract] [Full Text]
|
|
|
|
|
|
D. Liston and R. J. Krauzlis Shared Response Preparation for Pursuit and Saccadic Eye Movements J. Neurosci., December 10, 2003; 23(36): 11305 - 11314. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. L. Port and R. H. Wurtz Sequential Activity of Simultaneously Recorded Neurons in the Superior Colliculus During Curved Saccades J Neurophysiol, September 1, 2003; 90(3): 1887 - 1903. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. W. Heuer and K. H. Britten Contrast Dependence of Response Normalization in Area MT of the Rhesus Macaque J Neurophysiol, December 1, 2002; 88(6): 3398 - 3408. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Tanaka and S. G. Lisberger Role of Arcuate Frontal Cortex of Monkeys in Smooth Pursuit Eye Movements. II. Relation to Vector Averaging Pursuit J Neurophysiol, June 1, 2002; 87(6): 2700 - 2714. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I-h. Chou and S. G. Lisberger Spatial Generalization of Learning in Smooth Pursuit Eye Movements: Implications for the Coordinate Frame and Sites of Learning J. Neurosci., June 1, 2002; 22(11): 4728 - 4739. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. L. Gardner and S. G. Lisberger Linked Target Selection for Saccadic and Smooth Pursuit Eye Movements J. Neurosci., March 15, 2001; 21(6): 2075 - 2084. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. P. Ferrera Task-Dependent Modulation of the Sensorimotor Transformation for Smooth Pursuit Eye Movements J Neurophysiol, December 1, 2000; 84(6): 2725 - 2738. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Kahlon and S. G. Lisberger Changes in the Responses of Purkinje Cells in the Floccular Complex of Monkeys After Motor Learning in Smooth Pursuit Eye Movements J Neurophysiol, December 1, 2000; 84(6): 2945 - 2960. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Xing and R. A. Andersen Memory Activity of LIP Neurons for Sequential Eye Movements Simulated With Neural Networks J Neurophysiol, August 1, 2000; 84(2): 651 - 665. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. H. Recanzone and R. H. Wurtz Effects of Attention on MT and MST Neuronal Activity During Pursuit Initiation J Neurophysiol, February 1, 2000; 83(2): 777 - 790. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Kahlon and S. G. Lisberger Vector Averaging Occurs Downstream from Learning in Smooth Pursuit Eye Movements of Monkeys J. Neurosci., October 15, 1999; 19(20): 9039 - 9053. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. H. Britten and H. W. Heuer Spatial Summation in the Receptive Fields of MT Neurons J. Neurosci., June 15, 1999; 19(12): 5074 - 5084. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Seidemann and W. T. Newsome Effect of Spatial Attention on the Responses of Area MT Neurons J Neurophysiol, April 1, 1999; 81(4): 1783 - 1794. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. H. Reynolds, L. Chelazzi, and R. Desimone Competitive Mechanisms Subserve Attention in Macaque Areas V2 and V4 J. Neurosci., March 1, 1999; 19(5): 1736 - 1753. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
