HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) All Versions of this Article: 95/2/1078    most recent 00431.2005v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (11) Citing Articles via Google Scholar Google Scholar Articles by Sereno, A. B. Articles by Amador, S. C. Search for Related Content PubMed PubMed Citation Articles by Sereno, A. B. Articles by Amador, S. C.

Attention and Memory-Related Responses of Neurons in the Lateral Intraparietal Area During Spatial and Shape-Delayed Match-to-Sample Tasks

¹Neurobiology and Anatomy, University of Texas—Houston Medical School; and ²Division of Neuroscience, Baylor College of Medicine, Houston, Texas

Submitted 27 April 2005; accepted in final form 2 October 2005

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
When a monkey attends to, remembers, and looks toward targets, the activity of some neurons in the lateral intraparietal area (LIP) changes. We recorded from isolated neurons during both a spatial and a shape match-to-sample task to examine and characterize voluntary active processes in LIP. Many LIP neurons show spatially selective activity during the delay period that depends on the location of the sample, but for most cells, this activity does not differ between the two tasks. Although much past work in posterior parietal cortex has explained responses in this region in terms of active processes such as decision-making and motor planning, our findings suggest that much of that activity represents more passive processing. Nevertheless, we do see a significant minority of units that demonstrate instruction-dependent activity during the delay period, suggesting that these units could represent the neural correlates of voluntary or active processes. Separately, we found that during the presentation of the sample stimulus and test array, some units show stronger responses to the stimulus in the shape-matching task when the animal must attend to the shape of a stimulus. This elevated response to the sample during the shape task provides evidence for feature-based attention in LIP. Attention to shape is a property that has not previously been described in primate cortex.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
The discovery of specialized cortical visual areas that represent different stimulus properties (Cowey 1979; Felleman and Van Essen 1991; Maunsell and Newsome 1987) has helped establish the idea that many perceptual dimensions are processed independently. Many behavioral theories of attention (Cave and Wolfe 1990; Rossi and Paradiso 1995; Treisman and Gelade 1980) have also proposed functionally separate maps for different properties, and these ideas have been supported by physiological findings (Corbetta et al. 1991; Courtney et al. 1996; Goldman-Rakic 1988). The best-characterized segregation in sensory processing in cerebral cortex is that between a ventral stream that includes areas of temporal cortex and is involved in visual form analysis and object recognition (the "what" pathway) and a dorsal stream that includes areas of parietal cortex and is important for vision related to spatial relations and actions (the "where" or "how" pathway) (Baizer et al. 1991; Goodale and Milner 1992; Maunsell and Newsome 1987).

This distinction between object and spatial properties has also been proposed as an important factor in understanding the mechanisms of short-term memory and is the basis for the idea that object and spatial memory are also segregated (Funahashi et al. 1990; Rao et al. 1997; Wilson et al. 1993). Posterior parietal cortex, and the lateral intraparietal area (LIP) in particular, has been extensively studied over the past two decades, defining its important role in spatial attention and spatial short-term memory (Andersen 1987; Bushnell et al. 1981; Colby 1991; Goldberg et al. 1994; Robinson et al. 1978). However, recent work has shown that responses of neurons in LIP are also sensitive to nonspatial attributes of stimuli (Gifford and Cohen 2005; Toth and Assad 2002), including reports demonstrating appreciable selectivity for two-dimensional (2D) (Sereno and Maunsell 1998) and 3D shapes (Sereno et al. 2002). In addition, many units in LIP show enhanced activity during periods when the animal must remember simple geometric 2D shapes (Sereno and Maunsell 1998).

It is not known whether activity related to remembering a location or remembering a shape is selectively engaged when the animal must actively remember that information (e.g., working memory) or if the activity occurs in a more passive or automatic way when such stimuli are viewed (e.g., spatial and object priming). This distinction between reflexive (passive, bottom-up) and voluntary (active, top-down) cognitive processing has been most clearly defined with respect to spatial attention and saccadic eye movements (e.g., Gaymard et al. 1998; Klein et al. 1992; Munoz 2002; Pierrot-Deseilligny et al. 1991; Sereno 1996). Much work has documented differences in behavior, physiology, and anatomy between these reflexive and voluntary processes (e.g., Briand et al. 1999; Connolly et al. 2002; Gaymard et al. 1999; Larrison-Faucher et al. 2004; Matsuda et al. 2004; Mort et al. 2003; Munoz 2002; Seidlits et al. 2003; Sweeney et al. 1996). Specifically, with respect to spatial attention, both a nonpredictive peripheral cue and a centrally presented, predictive, symbolic arrow cue will result, for a certain interval of time, in subjects being able to better detect, identify, and discriminate stimuli at the cued location. The peripheral sudden onset cue reflexively draws attention, whereas the central symbolic cue causes a voluntary shift of attention and much work suggests that there are important differences in the behavior (e.g., time course), physiology, and structures that are critically involved in these two forms of spatially selective attention (e.g., Corbetta et al. 1993; Klein et al. 1992; Rafal et al. 1989; Rosen et al. 1999; Sereno et al. 2006). However, much previous work in neurophysiology has not distinguished between these processes. For example, in a memory-saccade paradigm, a peripheral cue indicates the location that is to be remembered. Hence, neural activity for this location could be related to automatic, reflexive processing of the cue (i.e., reflexive spatial attention, passive memory trace, automatic saccade programming) or to an active voluntary processing of the cue that indicates that 100% of the time the cued position is the position to be remembered (i.e., voluntary spatial attention, active working memory, voluntarily planned saccade programming). Despite the ambiguity, many studies suggest that the neural activity that results represents a voluntary process (e.g.,Chafee and Goldman-Rakic 1998; Funahashi et al. 1990). In fact many different active processes have been proposed to account for the selective activation during the delay period including a working memory interpretation of delay period activity (Colby et al. 1996; Goldberg et al. 2002), an intentional interpretation of delay period activity (Andersen 1995; Snyder et al. 2000), as well as a decision or gain theoretic account of delay period activity (Platt and Glimcher 1999; Shadlen and Newsome 1996).

To examine and disambiguate voluntary processes in LIP, we use a modified delayed match-to-sample task that keeps task conditions identical and varies only what the animal must actively attend to and remember (i.e., either the sample shape in the shape-matching task or the sample location in the location-matching task). The animal is cued as to which task he will perform by a briefly presented central symbolic cue at fixation that precedes the trial. It is important to note that we do not eliminate passive processes (given we are using a peripheral sample); however, these passive processes should be identical across the two tasks. Hence, in the shape-matching task, in addition to the active process of attending to and remembering the sample shape, there will be passive shape and spatial processes that reflexively occur after presentation of a peripheral sample shape. Likewise, in the location-matching task, in addition to the active process of attending to and remembering the sample location, there will be the same passive shape and spatial processes that reflexively occur after presentation of a peripheral sample shape. Thus in a carefully controlled design, we test here whether tasks requiring active attention and memory for shape versus spatial location differentially modulate the activity of individual cells in area LIP.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
Animals, animal care, and surgical procedures

Two rhesus monkeys (Macaca mulatta) were first implanted with a head post and scleral search coil in an aseptic surgery and then trained to perform two delayed match-to-sample tasks (DMTS) and a passive fixation task, obtaining a liquid reward for every correct trial. After completion of behavioral training, a recording chamber was implanted. The chambers were centered, 3–5 mm posterior and 10–12 mm lateral, over the right cerebral hemisphere. All experimental protocols were approved by the University of Texas and Baylor College of Medicine Animal Welfare Committees and were in accordance with the National Institutes of Health Guidelines.

Stimulus shapes

Sample and test stimuli were chosen from a subset of eight possible shapes (see Sereno and Maunsell 1998). Each shape was a simple 2D black-and-white geometric form. Each shape fit within the same-sized square region and had an equal number of white pixels.

Behavioral tasks

DMTS TASKS. The two animals performed shape and location DMTS tasks that presented trials with identical sample stimuli and test arrays (see Fig. 1). Shape and location task trials were either randomly interleaved on a trial-to-trial basis or presented in nine trial blocks. In both tasks, the animal fixated a 0.1° spot in the center of the video display throughout each trial. Eye position was monitored using the scleral search-coil method. After a minimum of 400 ms of fixation, a 0.5° instruction cue was presented surrounding the fixation spot for a minimum of 250 ms. For the shape task, the cue was a set of four small colored circles positioned at the vertices of an imaginary square centered on the fixation point. This cue signaled to the animal that it needed to remember the shape of the sample. In the location task, the cue was a white square outline that flickered on and off indicating to the animal that it needed to remember the position of the sample. After the cue offset, the animal maintained fixation for another 500 ms. Then a sample stimulus was presented (for 200–400 ms) in one of three locations. For most units, the sample was immediately followed by a high-contrast full-screen pattern mask (duration 27 ms) and then a variable delay period (ranging from 600 to 2,700 ms among the units recorded). Finally, three test shapes appeared together on the display, equidistant from the fixation spot. In the shape DMTS task, the animal was required to make an eye movement to the test shape that matched the sample shape for a juice reward. In the location DMTS task, the animal was required to make an eye movement to the test location that matched the sample location for a juice reward. The latency of the saccade was defined as the time after test onset at which point the eye left the fixation window (1° window, centered on the fixation spot.) During recording from each unit, sample and test stimuli were a subset of three of eight possible shapes. For most of the units recorded, stimulus eccentricity was between 3.5 and 6.0° and stimulus size was between 0.4 and 0.8°. In general, when eccentricity of the stimuli was increased, the stimulus size was increased to ensure that the stimuli remained easily discriminable. Catch trials comprised 20% of trials. In catch trials in the shape DMTS task, the test array did not include the sample shape, which was replaced instead by one of the five other shapes. In catch trials in the location DMTS task, the test array did not include any stimulus in the correct target location. For both shape and location task catch trials, the animal was rewarded for maintaining fixation on the central spot until it disappeared (after an additional 1.3–1.9 s).

View larger version (17K): [in this window] [in a new window]   FIG. 1. Schematic diagram of the location and shape-matching tasks. Each trial began when the animal fixated a spot in the center of the video display. A small cue centered on the fixation spot instructed the animal as to which task he was to perform. After the cue disappeared, a sample shape was presented and followed by a brief presentation of a full screen pattern mask. Finally, after a delay period, a test array of 3 shapes appeared. On trials in which the animal had been instructed to remember the location of the sample (location task), the animal was required to make an eye movement (indicated by an arrow during the test period) directly to the shape that appeared in the remembered sample location to obtain a juice reward. On trials in the shape-matching task, the animal was required to make an eye movement to the test shape that matched the sample shape to obtain a juice reward.

Recording procedures

Monkeys performed the matching tasks while the recording electrode was slowly advanced with a hydraulic micropositioner in search of neurons. We recorded any neuron that we could isolate well and that appeared stable. Once a neuron was well isolated, we presented a stimulus at varying locations (typically 8 locations equally spaced around the fixation spot) to determine the location in the receptive field that elicited the strongest response and then selected the three locations for the matching tasks. The three possible stimulus positions were configured so that at least one stimulus position fell within the receptive field in the location that elicited the strongest response at a given eccentricity. In addition, the three locations chosen were typically equally spaced (120° apart). We then presented all shape stimuli at the preferred location and determined the three shapes for the matching tasks. For most units, the two shapes that elicited the strongest response and the one shape that elicited the weakest response were selected. For most units, data were then collected with the two matching task trials intermixed (either randomly or alternated in 9 trial blocks). For a few units, data were collected for each matching task sequentially. As long as the isolation could be maintained, recordings were collected for 216 correct matching trials [or 12 trials per each of 9 sample conditions (3 shapes by 3 locations) for each of 2 matching tasks]. If the unit remained well isolated, we recorded responses during the passive fixation task for all eight shape stimuli (6 trials each).

We recorded at least five repetitions (median = 12) for each combination of sample shape and sample location from 122 isolated neurons in the lateral bank of the intraparietal sulcus in the right hemisphere of both animals. There were 85 units recorded from one animal and 37 recorded from the second animal. In addition, for 52 of these isolated neurons with spatially selective delay period acitivity, we recorded at least three trials (median = 6) for each sample shape in the passive fixation task. For the passive fixation task, there were 36 units recorded from one animal and 16 recorded from the second animal. Because there were no obvious differences between the two animals, the data have been combined.

Behavioral monitoring, eye position and spike sampling, and on-line data analysis were performed under computer control. Stimuli were presented on a 20-in computer monitor (75 Hz), positioned 57 cm from the animal. While animals performed the experimental tasks, action potentials were recorded extracellularly with either transdural Pt/Ir (1–2 M) or tungsten microelectrodes (1–2 M, Microprobe). Signals from the electrode were amplified, filtered, and transformed into pulses by a window discriminator. Spike times were recorded with 1-ms resolution.

Anatomical localization

We recorded from 122 isolated neurons in the lateral bank of the intraparietal sulcus of two animals. Histological reconstruction from one animal showed that 79 of the 85 units assigned to the lateral bank were within LIP defined as including a densely myelinated zone and a region extending 2 mm further dorsal (Andersen et al. 1990; Ungerleider and Mishkin 1982). The remaining six units were within another millimeter further dorsal, and because they were 4 mm below the surface and gave vigorous visual responses, we have included them in the current analysis. Histological reconstruction from the second animal is not available, but the physiological results from that animal were consistent.

Data analysis

For statistical analyses in the DMTS tasks, F tests were performed on the average rate of firing in four different trial intervals: for the sample period, starting 50 ms after the onset of the sample stimulus and ending with stimulus offset (200–400 ms durations across the population of recorded units); for the delay period, the last 600 ms of the delay period; for the test period, starting 50 ms after the onset of the test array and ending 150 ms after test array onset; and for the peri eye movement (EM) period, starting 100 ms before the onset of the EM and ending 100 ms after the EM onset. For analyses of the sample and delay periods, trials were sorted based on the task instruction, sample shape, and sample location (see Fig. 2A). For analyses of the test and EM periods, trials were sorted based on the task instruction, target shape, and target location (see Fig. 2B). For analyses of shape effects, the average firing rate for each shape was collapsed across matching tasks and locations (see highlighted histograms in Figs. 4 and 5, broken down by task). For analyses of spatial effects, the average firing rate for each location was collapsed across matching tasks and shapes (see highlighted histograms in Fig. 3, broken down by task). For analyses of task instruction effects, the average firing rate for each matching task was collapsed across shapes and locations (see highlighted histograms in Fig. 6). Similar to the analyses of task instruction effects, for analyses of check and change effects, the average firing rate for each subset of task trials (see Fig. 16) was collapsed across target shapes and locations (see highlighted histograms in Fig. 15).

View larger version (26K): [in this window] [in a new window]   FIG. 2. Sorting of data for analyses. A: for analyses of sample and delay periods, data were sorted by the sample conditions (purple box). Specifically, data were sorted by the matching task instruction (shape task in red, location task in blue), sample shape (3 possible shapes), and sample location (3 possible locations). B: for analyses of test and eye movement periods, data were sorted by the test array and eye-movement conditions (green box). Specifically, data were sorted by the matching task instruction (shape task in red, location task in blue), target shape (for these trials, the square), and target location (for these trials, to the right of fixation).

View larger version (29K): [in this window] [in a new window]   FIG. 4. Single unit with shape-selective delay activity. Conventions are the same as in Fig. 3. The highlighted histograms in the bottom margin nicely illustrate a main effect of shape. That is, this neuron was much more active during the sample presentation and during the delay period that followed for the H-like shape (shape effect during delay period, F test, P < 0.0001; shape SI = 0.62). This neuron also showed a small but significant increase in delay activity when the animal was required to remember the location of the sample (task instruction effect during the delay, P < 0.043; task SI = –0.06). Note that this is the same neuron that was previously shown in Fig. 3B of Sereno and Maunsell (1998). In this figure, we show data from the 3 conditions (i.e., the first 3 red histograms in the top row) that were depicted in Fig. 3B of Sereno and Maunsell (1998), as well as the activity of the neuron for the other 15 trial conditions and the activity of the neuron averaged across these different shape and/or location conditions (14 histograms).

View larger version (33K): [in this window] [in a new window]   FIG. 5. Single unit with shape-selective delay activity. Conventions are the same as in Fig. 3. The highlighted histograms in the bottom margin nicely illustrate a main effect of shape. That is, this neuron was more active during the sample presentation and during the delay period that followed for the O-like shape (shape effect during delay period, F test, P < 0.0001; SI = 0.29). This shape-selective delay activity existed whether or not the animal was required to remember the shape of the sample (task instruction effect during delay period, P > 0.30; task SI = 0.02). This neuron also showed a small but significant increase in sample activity when the animal was required to attend to and remember the shape of the sample (task instruction effect during the sample, P < 0.004; task SI = 0.08).

View larger version (31K): [in this window] [in a new window]   FIG. 3. Single unit with spatial selective delay activity. Each of the 18 histograms within the black outlined rectangle represents the average spike density histogram across 12 trials from 1 neuron in the 18 possible sample trial conditions (2 matching tasks, 3 locations, and 3 shapes). For each combination of sample shape and location trial type, shape-matching task histograms are in red (top), whereas location-matching task histograms are in blue (bottom). The average spike density histograms highlighted in the right margin are histograms averaged across the 3 shapes (i.e., each histogram averaged across 36 trials). The average spike density histograms in the bottom margin are averaged across the 3 locations (i.e., each histogram averaged across 36 trials). The pair of histograms in the lower right corner is averaged across both the 3 locations and the 3 shapes (hence, each histogram is an average of 108 trials). Each histogram shows the activity of the neuron during the presample fixation period (F), the minimum sample period across all trials (S, indicated by the horizontal bar below each histogram) aligned with sample onset, and the minimum delay period across all trials (D; see labels at bottom of figure) aligned with delay onset. The horizontal bar below the histogram represents both the minimum duration of the sample (green) and the duration of the mask (black, if present). The vertical shaded gray region brackets a 600-ms period of the delay period before the test onset. The total duration of the histograms are indicated in the bottom right corner of the figure. For the averaged histograms, mean saccade latency and the number of trials it represents are indicated in the upper left hand corner of the histogram. The highlighted histograms in the right margin nicely illustrate a main effect of location. That is, this neuron was much more active during the sample presentation and during the delay period that followed when the sample appeared in location 1 (location effect during delay period, F test, P < 0.0001; SI = 0.46. This spatially selective delay activity existed whether or not the animal was required to remember the location of the sample [task instruction effect during delay period, P > 0.46; selectivity index (SI) = –0.01].

View larger version (32K): [in this window] [in a new window]   FIG. 6. Single unit with task selective delay activity showing greater activity in the delay period during the location-matching task. Conventions are the same as in Fig. 3. This neuron showed a spatial-selective delay period activity. That is, the neuron was much more active during the delay period that followed when the sample appeared in locations 1 and 2 (location effect during delay period, F test, P < 0.0001; SI = 0.55). In addition, the delay period response of this neuron depended on the previously presented sample shape (shape effect during delay period, F test, P < 0.0004; SI = 0.21). Most importantly, the highlighted histograms in the bottom right corner margin nicely illustrate a main effect of task instruction during the delay period. This neuron was much more active during the delay period when the animal was required to remember the location of the sample (task instruction effect during delay period, F test, P < 0.0001; SI = –0.41).

View larger version (19K): [in this window] [in a new window]   FIG. 16. Model of cognitive processes required in the matching tasks. The trials are grouped according to their different task requirements. In pure trials, the exact same stimuli (sample and test array) are presented and the same motor response is executed in both the location and shape-matching task. On these trials, the sample shape and target shape are identical and the sample location and target location are the same. In the non-pure trials the shape (location task) or the location (shape task) of the target is different from the sample stimulus. The location-matching task pure trials and the shape-matching task pure trials were compared with each other to obtain a measure of the attention to shape process (CHECK); these were identical trials, the only difference being the task requirement either to remember shape or location. The shape-matching task pure trials were compared with the shape-matching task non-pure trials to obtain a measure of the attention shift and change in motor plan processes (CHANGE). The behavioral analyses (saccade reaction time and error rate) of these different trial types across all units recorded is indicated in the figure.

View larger version (33K): [in this window] [in a new window]   FIG. 15. Attention to shape during the EM period in LIP. Single unit with task-selective EM activity showing greater activity in the EM period during the shape-matching task (red). Conventions are the same as in Fig. 13. This neuron was more active during the EM period when the animal was performing the shape-matching task (task instruction effect during EM period, P < 0.001; SI = 0.32; see highlighted histograms).

For statistical analysis with the passive fixation task, the average rate of firing for the matching task, location, and shape that elicited the best response during a period 245–495 ms after the sample offset was compared with the average rate of firing during the passive fixation task following the presentation of the same stimulus presented in the same position (i.e., average rate of firing during the first interstimulus interval, from 245 to 495 ms after the first sample presentation of the trial). For all statistical tests, a criterion level of P < 0.05 was used for significance.

The delay population averaged spike density histograms (see Fig. 8) are the averaged response of the 20 neurons with significant differences in task during the delay period. Separate histograms are shown for units with greater activity in the location task (n = 12) and those units with greater activity in the shape task (n = 8). For each unit, the pair of histograms representing the average response of that unit during the shape- and location-matching tasks for the location and shape that elicited the best response during the delay period was included in the respective population average histogram (shape task or location task). These population average spike density histograms, sampled at 1 kHz, were then convolved with a Gaussian of sigma 15 ms and the average spike train including the delay interval displayed. For the delay period analysis for task effects, described in the preceding text, the SE for each task was calculated for each unit. Error bars in the population histogram represent the SE calculated for the delay period for each unit and task averaged across the units included in the population histogram. The sample population averaged spike density histograms are the average response of the 24 neurons with significant differences in task during the sample period. Separate histograms are shown for units with greater activity in the shape task (n = 15) and those units with greater activity in the location task (n = 9). For each unit, the pair of histograms representing the average response of that unit during the shape- and location-matching tasks for the location and shape that elicited the best response during the sample period was included in the respective population average histogram (shape or location task). These population average spike density histograms, sampled at 1 kHz, were then convolved with a Gaussian of sigma 15 ms and the average spike train including the sample interval displayed. For the sample period analysis for task effects described in the preceding text, the SE for each task was calculated for each unit. Error bars in the population histogram represent the SE calculated for the sample period for each unit and task averaged across the units included in the population histogram.

View larger version (30K): [in this window] [in a new window]   FIG. 8. Task effects during the delay period (A–C) and sample period (D–F) across all units recorded (n = 122). A: histogram showing the magnitude of the delay period task selectivity (during the last 600 ms of the delay period) for the population of units recorded as indexed by each unit's selectivity. Indices >0 (scale highlighted in red) represent units with greater activity in the shape task, indices <0 (scale highlighted in blue) indicate units with greater activity in the location task. Units with significantly different responses between matching tasks during the delay period (significant task instruction effect during delay period) are indicated in dark gray. The number (n) and mean value of the distribution of index values that are either greater or less than 0 for units with average firing rate >1 spike/s are indicated in the upper corners of the figure. B: time course of the averaged response of units with significantly greater responses during the delay in the location-matching task (n = 12). The histogram shows the activity of these units beginning at sample onset (S) until 300 ms after sample onset (indicated by the green bar under the histogram) and then during the last 600 ms of the delay period (indicated by the vertical shaded gray region). The break in the time course is indicated by the diagonal mark in the timeline (abscissa). The total duration of the histogram is indicated in the bottom right corner of the figure and the rate of activity in spike/s at the top right corner of the figure. The number of units, mean sample duration (S bar), and mean minimum delay duration (D bar) of these units are indicated at the top center. The blue curve is the average response from these 12 neurons in the location-matching task for the best sample location and shape. The red curve is the average response from these same units under identical sample conditions in the shape-matching task (error bars represent the SE during the delay period). These LIP units show increased activity during the delay period when the animal is required to remember the sample location (blue) compared with the sample shape (red). C: time course of the averaged response of units with significantly greater responses during the delay in the shape-matching task (n = 8). Conventions are the same as in B. These LIP units show increased activity during the delay period when the animal is required to remember the sample shape (red) compared with the sample location (blue). D: histogram showing the magnitude of the sample period task selectivity for all units recorded, as indexed by each unit's selectivity. Conventions are the same as in A. Units with significantly different responses between matching tasks during the sample period (significant task instruction effect during sample period) are indicated in dark gray. E: time course of the averaged response of units with significantly greater responses during the sample in the location-matching task (n = 9). The histogram shows the activity of these units for 750 ms after the sample onset. The green bar indicates the mean duration of the sample across the averaged units (indicated by the green bar under the histogram) and the average sample period used for analysis (starting 50 ms after onset) indicated by the vertical shaded gray region. The total duration of the histogram is indicated in the bottom right corner of the figure and the rate of activity in spike/s at the top right corner. The number of units and mean sample duration are indicated at the top center. The blue curve is the average response from these 9 neurons in the location-matching task for the best sample location and shape. The red curve is the average response from these same units under identical sample conditions in the shape-matching task (error bars represent the SE during the sample period). These LIP units show somewhat increased activity during the sample period when the animal is required to remember the sample location (blue) compared with the sample shape (red). F: time course of the averaged response of units with significantly greater responses during the sample in the shape-matching task (n = 15). Conventions are the same as in E. These LIP units show increased activity during the sample period when the animal is required to remember the sample shape (red) compared with the sample location (blue).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

Most LIP units (80%, 97/122) responded to the sample stimuli and gave statistically significantly different responses depending on where the sample stimulus appeared (Fig. 3; see also Fig. 13). In addition, most units (71%, 87/122) also had a significant difference in activity during the delay period depending on whether the sample had appeared in the receptive field. As we have previously reported (Sereno and Maunsell 1998) and as illustrated in Fig. 4 (see also Fig. 14), many LIP units also showed a significant difference in activity depending on the sample shape either when it was presented at sample (41%, 50/122) or during the subsequent delay period (34%, 41/122), independent of the following EM.

View larger version (35K): [in this window] [in a new window]   FIG. 13. Single unit (same neuron as Fig. 3) with spatial selective activity during the EM period. Each of the 18 histograms within the black outlined rectangle represents the average spike density histogram across 12 trials from 1 neuron in the 18 possible test conditions (2 matching tasks, 3 target locations or EM responses, and 3 target shapes). For each combination of target shape and target location trial type, shape-matching task histograms are in red (top), whereas location-matching task histograms are in blue (bottom). The average spike density histograms highlighted in the right margin are histograms averaged across the 3 target shapes (i.e., each histogram averaged across 36 trials). The average spike density histograms in the bottom margin are averaged across the 3 target locations (i.e., each histogram averaged across 36 trials). The pair of histograms in the bottom right corner is averaged across both the 3 target locations and the 3 target shapes (hence, each histogram is an average of 108 trials). Each histogram shows the activity of the neuron during the last 800 ms before the onset of the saccade (EM) and 400 ms after the saccade. The activity is aligned with the onset of the EM (dashed line) and the EM time period for analysis is shaded gray (starting 100 ms before the saccade and ending 100 ms after the EM). The green bar under each histogram represents the range of test array onsets across all trials included in the histogram. The vertical bar embedded in this green bar under the histogram indicates the mean test array onset of these trials. The total duration of the histograms are indicated in the bottom right corner of the figure. For the averaged histograms, mean saccade latency and the number of trials it represents are indicated in the upper left corner of the histogram. This neuron was much more active during the EM period when the target location (and EM response) were in location 1 (location effect during EM period, F test, P < 0.0001; SI = 0.54; see highlighted histograms). This neuron also showed a small but significant greater activity during the EM period in the shape-matching task (task instruction effect during EM period, F test, P = 0.01; SI = 0.05).

View larger version (34K): [in this window] [in a new window]   FIG. 14. Single unit (same neuron as Fig. 4) with shape-selective activity during the EM period. Conventions are the same as in Fig. 13. This neuron was much more active during the EM period when the target shape was the H-like shape (shape effect during EM period, F test, P < 0.0001; SI = 0.58; see highlighted histograms). In addition, this neuron showed a spatial selective EM period activity. That is, the neuron was more active during the EM period when the target location (and EM response) were in location 1 or 2 (location effect during EM period, F test, P < 0.0001; SI = 0.25). Finally, this neuron also showed a small but significant greater activity during the EM period in the shape-matching task (task instruction effect during EM period, F test, P = 0.02; SI = 0.04).

As illustrated in Fig. 6 (see also, Fig. 4), some units (16%, 20/122) did show significantly different activity during the delay period depending on whether the animal had to remember the sample shape or location (see also, summary table in Fig. 7). The unit in Fig. 6 was much more active during the delay period when the animal was performing the location task and needed to remember the sample position (in location 1, 20 vs. 8 spikes/s, for location and shape task, respectively; task S.I. = –0.41). We calculated a task SI for each unit using the average rate of firing during the last 600 ms of the delay period during the shape and location tasks (see METHODS). Figure 8A shows the distribution of indices for the units recorded. Among the units with significant differences in delay period activity between the two tasks (units with significant task instruction effect during delay period indicated in dark gray in Fig. 8A), 12 were more active during the location task and 8 were more active during the shape task (see also summary table in Fig. 7). The population histograms of the significant units in Fig. 8, B and C, show the average response of these units during the delay period. The population histogram of units with significantly greater responses during the delay in the location-matching task (Fig. 8B, n = 12) show units that demonstrate enhanced activity when the animal needed to remember the sample location.

View larger version (13K): [in this window] [in a new window]   FIG. 7. Summary table of units with significant task instruction effects across the four different trial intervals (sample, delay, test, and eye movement) for the whole population of recorded units (n = 122 units) and for the subset of units with 9 repetitions (n = 80 units). The purple box indicates trial periods where data were sorted by the sample conditions (as illustrated in Fig. 2). The green box indicates trial periods where data were sorted by the target and response conditions. For each set of units, the percentage and the number of units (in parenthesis) that showed significant task instruction effects are listed for each of the four trial periods. In the top right (in red) and bottom right (in blue) corners are the numbers of units with significantly greater response during the shape and the location tasks, respectively.

View larger version (39K): [in this window] [in a new window]   FIG. 9. Single unit with task selective delay activity showing greater activity in the delay period during the shape-matching task. Conventions are the same as in Fig. 3. After transient increased responsiveness to the onset of the sample and the mask, the response of this neuron was inhibited in the contralateral visual field (locations 1 and 2) during the sample presentation and delay period (location effect during delay period, P < 0.0001; SI = 0.25). This inhibition of response during the delay period was more pronounced when the animal was required to remember the location of the sample in the location-matching task (blue), resulting in a significant but positive SI. As illustrated by the highlighted histograms in the bottom right corner, this neuron showed a significantly lower response in the delay period during the location-matching task (task instruction effect during delay period, P < 0.0008; SI = 0.07).

Because the location and shape match-to-sample trials were randomly intermixed, it is possible that the spatially selective delay period activity existed during the shape task because the monkey elected to actively remember both the shape and position of each sample. We examined this possibility for nine neurons with spatially selective delay period activity by having the animal perform only the shape task for 100 trials (12–22 min). The spatially selective delay period activity remained robust. Each of the neurons tested retained its significant delay selectivity when the shape task was performed alone. Five of these nine units were also recorded while the animal performed only the location task for 100 trials. The mean ratio (best location/worst location) for the delay period activity of these units did not change whether the animal was performing the location task (3.40) or the shape task (3.34), P > .95 (t-test, 2-tailed). Six of these nine units were also recorded from while the animal performed the task with trials from the two matching tasks intermixed. For these units, the mean ratio (best location/worst location) for trials in the shape task also did not differ significantly depending on whether the animal was performing the shape task intermixed with the location task (2.60) or the shape task alone (2.88), P > 0.47 (t-test, 2-tailed). Together these results suggest that changing the task conditions (e.g., shape task only) did not reduce the spatial selectivity of the units.

SPATIAL DELAY PERIOD ACTIVITY IN PASSIVE FIXATION TASK. To examine further whether the spatially-selective delay period activity in area LIP represents a purely passive memory, we also trained the animals on a fixation task that required no behavioral response. While the animal fixated, we presented the shapes from the matching tasks in the isolated neuron's receptive field in the same position used during the matching tasks (see Experimental procedures). We compared the activity for each unit during the fixation period following the presentation of a given shape in this passive fixation condition with that during the delay period in the matching task after the same stimulus had been presented in the same location as a sample (see Data analysis for details). Of the 52 neurons with significant spatially selective delay activity in the matching tasks, 20 (39%) had significantly greater activity in the delay period of the matching task compared with the first interstimulus period of the fixation task (Fig. 10). The failure of these units to show equivalent activation during the interstimulus interval of the fixation task shows that for some of the units with spatially selective delay period activity, this activity is not entirely passive or automatic.

View larger version (16K): [in this window] [in a new window]   FIG. 10. Some LIP units show greater activity following a sample stimulus in a matching task (i.e., during the delay period) than during passive fixation (during the 1st interstimulus interval). Scatter plot of 52 neurons with spatial selective delay activity in the matching task that were also recorded from during a passive fixation task. The ordinate represents mean spike rate during the 1st interstimulus interval in the passive fixation task and the abscissa the mean spike rate during the delay period of the matching task. Units with significant differences in activity between the 2 tasks following the preferred sample stimulus presentation are filled in black. Many of the units (32/52) show no significant difference in their firing rate. However, the majority of cells that did show a significant difference between tasks are more active during the matching task (i.e., below the diagonal line), suggesting that some of the spatial selective delay activity is not entirely passive or automatic.

View larger version (32K): [in this window] [in a new window]   FIG. 11. Task instruction effects during the sample, test, and eye movement periods. A: attention to shape during the sample period in LIP. Single unit with task selective sample activity showing greater activity in the sample period during the shape-matching task (red). Average spike density histograms for the neuron's preferred location for 3 different sample shape presentations during shape (red)- and location (blue)-matching tasks. Each histogram was compiled from 12 trials and shows the activity of the neuron during the presample fixation period (F), the minimum sample period across all trials (indicated by the horizontal bar below each histogram) aligned with sample onset, and the minimum delay period across all trials (D) aligned with delay onset. The total duration of the histograms are indicated in the figure legend. The activity is aligned with the onset of the sample presentation (S, dashed line) and the sample time period for analysis is shaded in gray and bracketed by the solid vertical lines (starting 50 ms after the sample presentation and ending with the offset of the sample for the minimum sample duration). This neuron was more active during the sample presentation when the animal was performing the shape-matching task (task instruction effect during sample period, P < 0.0001; SI = 0.16). B: attention to shape during the test period in LIP. Single unit with task selective test activity showing greater activity in the test period during the shape-matching task (red). Average spike density histograms for the neuron's preferred target location for 3 different test target shape presentations during shape (red)- and location (blue)-matching tasks. Schematic diagram above the histograms indicates the trial conditions and the target shape. Asterisks indicate the 2 other shapes that were presented in the other locations. Average spike density histograms from 1 neuron for its preferred target location (highlighted) with 3 different target shapes presented in the target location during the shape (red)- and location (blue)-matching tasks. Each histogram was compiled from 12 trials and shows the activity of the neuron during the last 500 ms of the delay period (D) before the test array onset (T) and 700 ms after the test array onset. The activity is aligned with the onset of the test array (dashed line) and the test time period for analysis is shaded gray and bracketed by the solid vertical lines (starting 50 ms after the test onset and ending 150 ms after the test onset). The green bar under each histogram represents the range of test array durations across all trials included in the histogram. The blue bar under each histogram represents the range of saccade onsets across all trials included in the histogram. The vertical bar below this blue bar indicates the mean saccade onset of these trials. This neuron was more active during the test presentation when the animal was performing the shape-matching task (task instruction effect during test period, P < 0.0002; SI = 0.18).

DIFFERENCES BETWEEN SAMPLE AND DELAY ACTIVITY. About 16–20% of units had a significant difference in activity between the two tasks during either the sample or delay periods. Only one unit showed significant task effects during both periods. Hence, it seems that these task effects occur in separate populations of units in LIP. Nevertheless, to see if there was a shift in the responses of LIP units between the sample and delay period such that the response during the sample was greater in the shape task, but during the delay, greater in the location task, we compared the task SIs across the sample and delay period, excluding only units with low firing rates (<3 spike/s) during either period (n = 11). We found that for the sample of units we recorded in LIP, there was a significant shift in the SI between the sample and delay periods, P < 0.007, 1-tailed t-test (illustrated in Fig. 12 as a tendency for units to cluster below the diagonal line). Thus there was a significant shift of selectivity indices between the sample and delay periods that is consistent with more activity in the shape task during the sample and more activity in the location task during the delay period.

View larger version (16K): [in this window] [in a new window]   FIG. 12. Task-dependent shift in activity between sample and delay periods. Scatter plot of selectivity indices for all the units recorded in LIP with average activity 3 spike/s during the sample and delay periods. The ordinate represents the selectivity during the delay period and the abscissa represents the selectivity during the sample period. The SI for a given unit is positive when the unit's activity is greater during the shape-matching task shape and negative when the unit's activity is greater during the location-matching task. There is a significant shift of selectivity indices between the sample and delay periods that is consistent with increasing activity in the shape task during the sample (or, equivalently, increasing activity in the location task during the delay period), as evidenced by more units falling below the diagonal line.

MODEL OF COGNITIVE PROCESSES REQUIRED AFTER TEST ONSET. To examine if there were differences in the activity of LIP units during the EM period that reflected these different task requirements, the unit data were sorted by task, target shape, and target location as well as by whether the sample and test target were the same shape and in the same location ("pure" trials in Fig. 16 ). In all the trials, the presentation of the sample should have reflexively or exogenously brought spatial selective attention to the sample location (henceforth, labeled the attended location). During location task trials, after the test onset, the monkey could immediately execute the EM to the attended location (GO trials). These trials can be further divided into trials that presented the identical sample shape in the attended location ("pure" GO trials, one-third of all location task trials), and those trials in which a different shape than the sample shape appeared in the attended location ("non-pure" GO trials, two-thirds of all location task trials). In the shape task, on the other hand, when the test stimulus appeared, the monkey would first need to check the shape in the attended location (to see if it was the remembered shape) before executing the EM (CHECK-GO trials, one-third of all shape task trials). It is important to note that pure shape and pure location trials entailed the same stimulus configurations and the same EMs; however, the animal made the EM for different reasons (matching location vs. matching shape, for location and shape-matching tasks, respectively). In the non-pure trials in the shape task, the correct shape was always in a different location at test than in which it had appeared during the sample. Hence, after the test onset, the monkey would first need to check the shape in the attended location (to see if it was the remembered shape), then shift spatial attention and motor plan, and, finally execute the EM (CHECK-CHANGE-GO trials, two-thirds of all shape task trials).

To examine if there were behavioral (EM latency) and physiological changes in unit activity dependent on these conditions, we made comparisons across these conditions. Because we were looking at a subset of trials for each unit (i.e., 33 or 67% of all trials), we only included units that had more than nine repetitions for each sample shape and location and task (80 of 122 units). The pattern of findings for these 80 units was similar to what we have already reported in the preceding text for the entire population. Namely, 27% (22/80), 19% (15/80), and 59% (47/80), of this subset of units showed differential activity between tasks during the sample, delay, and EM periods, respectively (Fig. 7).

Saccade latencies and error rates.

Despite a large 78-ms difference in EM latency between tasks [P << 0.01; spatial task: 181 ± 3 (SE) ms; shape task: 259 ± 2 ms], there was no significant difference in error rates (1.7%, P > 0.23; spatial task: 21.2 ± 1.0%; shape task: 22.9 ± 0.9%). To examine if behavioral evidence supported these different processes (GO, CHECK, and CHANGE), we examined EM latency and error rates in the different trial conditions (see Fig. 16). We found that EM latencies were fastest in the "pure" GO trials (188 ± 2 ms), 43 ms slower in the CHECK-GO trials (231 ± 2 ms, P << 0.01) and another 44 ms slower in the CHECK-CHANGE-GO trials (275 ± 2 ms, P << 0.01). Thus EMs were delayed by the need to attend to the target shape (CHECK) and further delayed by the need to shift attention and change EM plan (CHANGE). We also found that error rates were consistent with these findings: error rates were lowest in the pure GO trials (1.3 ± 0.2%), 2.4% greater errors in the CHECK-GO trials (3.7 ± 0.4% ms, P << 0.01), and another 25.7% greater errors in the CHECK-CHANGE-GO trials (29.4 ± 1.1% ms, P << 0.01). Reduced error rates were also found in the location task when the identical sample shape appeared in the attended location (pure trials, 1.3 ± 0.2%) as opposed to a different shape (non-pure trials, 27.9 ± 1.2%, P << 0.01). Unexpectedly, however, EM latencies were 9 ms slower in this same condition (P << 0.01). That is, latencies were slower when the identical sample shape appeared in the attended location (pure trials, 188 ± 2 ms) as opposed to when a different shape appeared (non-pure trials, 179 ± 3 ms, P << 0.01). This may be due to the fact that we find that 70% of LIP units (62 of 94 units recorded) show a significantly reduced response to a repeated shape stimulus in their receptive field during a passive fixation task similar to the attenuated responses that have been reported previously in AIT (Miller and Desimone 1994; Miller et al. 1993 1991). Hence, it is possible that these attenuated visual responses are responsible for the slightly longer latencies.

Physiology of attention to feature (CHECK) in LIP.

To investigate whether the differences in activity after the test onset could be due to this additional requirement to check the shape at the attended target location before making an EM, we compared stimulus configurations in the two tasks where the animal made the same EM to the same shape (i.e., pure location GO trials and pure shape CHECK-GO trials, Fig. 16). The only difference in these trials was whether the animal needed to attend to the shape before making the EM. Surprisingly, even with our reduced number of trials, some units during the test (12%, 12/80) and EM (6%, 5/80) periods had significant differences between these task conditions (see Fig. 17A). This unit showed an enhanced activity after the test array onset and at the time of the EM during the shape task when the animal needed to check the shape before making the EM, suggesting that, similar to what we have reported during the sample period, this enhanced activity in LIP is associated with attention to shape.

View larger version (32K): [in this window] [in a new window]   FIG. 17. Task instruction effects during the test and EM periods during shape and location-matching tasks. Conventions are the same as in Figs. 11B (test period) and 13 (EM period), respectively. A: attention to shape (CHECK) during the test period in LIP. Single unit with task selective activity showing greater activity in the test period during the pure trials of the shape-matching task (pink). Average spike density histograms from 1 neuron for its preferred location for 3 different target shape presentations during shape (pink) and location (baby blue) matching tasks. Each histogram was compiled from the 4 pure trials that were identical (i.e., in both stimuli and EM response) in both matching tasks and shows the activity of the neuron during the last 600 ms of the delay period (D) and 700 ms after the test onset (T). The activity is aligned with the onset of the test presentation (dashed line) and the test array time period for analysis is shaded gray and bracketed by the solid vertical lines (starting 100 ms after the test presentation and ending 200 ms after the test array onset). This neuron was more active during the test presentation when the animal was performing the pure trials in the shape-matching task than identical trials in the location-matching task (task instruction effect during test period, P < 0.03; SI = 0.16). This neuron also showed a similar effect during the EM period (task instruction effect during EM period, P < 0.023; SI = 0.33). B: shift in attention and EM motor plan (CHANGE) during the EM period in LIP. Single unit with task selective EM activity showing greater activity in the EM period during the non-pure trials of the shape-matching task (dark pink) than in the pure trials (pink). Average spike density histograms from 1 neuron for its preferred location during the shape-matching task for 3 different target shape presentations: pure or no change, where the target shape happened to be presented in the same location as the sample (pure shape-matching trials, pink, n = 4); non-pure or change, where the target shape happened to be presented in a different location than the sample (non-pure shape-matching trials, dark pink, n = 8). Each histogram shows the activity of the neuron during the last 800 ms before the onset of the EM and 400 ms after the EM onset. The activity is aligned with the onset of the EM (dashed line) and the range of test array onsets for the trials in each histogram are indicated by the green horizontal bar beneath the histogram. The period for analysis is shaded gray (starting 100 ms before the EM and ending 100 ms after the EM). This neuron was more active during the EM period when the animal first needed to shift attention and EM plan before executing the identical eye movement to the identical test array (shape pure vs. shape non-pure task condition effect during EM period, P < 0.005; SI = 0.22).

To investigate whether the differences in activity after the test onset could be due to this additional requirement to shift attention and change the EM motor plan (i.e., CHANGE), we compared the CHECK-GO trials to the CHECK-CHANGE-GO trials from the shape task (Fig. 16). The difference between these two trial types is that in the non-pure shape trials, the target shape would have appeared in a different location at sample so that the animal would have been required to shift his attention and change his EM plan to the new location before making the identical EM to the identical shape. A few units showed significant differences in activity between these two conditions during the test period (11%, 9/80) and around the time of the EM (17%, 13/80). Figure 17B shows a unit that showed a significantly enhanced activity around the EM during the shape task when the animal needed to shift its attention and change its EM plan.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS ACKNOWLEDGMENTS REFERENCES

 
As expected, most neurons in LIP showed spatially selective sensory (80%) and delay period (71%) activity. However, many neurons in parietal cortex that represent remembered locations were active irrespective of whether the animal remembered shape or location. Our findings suggest that many neurons in this region contribute to a more reflexive or automatic process. Nevertheless, in agreement with previous studies, we do find that the delay period activity of some units in LIP reflects a more active process. Specifically, we show for the first time, under highly controlled conditions the following: 1) neurophysiological evidence that the delay period activity of some neurons in LIP represents what the animal is actively attending, remembering, planning, or deciding; and 2) neurophysiological evidence that the responses of some neurons in LIP are modulated when the animal attends to shape providing the first neurophysiological evidence for shape-selective attentional mechanisms in visual cortex.

Much delay period activity is not specific for the to-be-remembered event

Current work suggests that one of the functions of parietal cortex is to contribute to a salience map (Goldberg et al. 2002; Gottlieb et al. 1998). Salience accrues not only from reflexive or intrinsic properties of the stimuli (e.g., abrupt onsets) but also from voluntary or task-related properties (e.g., behavioral relevance: Kusunoki et al. 2000). This study attempts to tease apart the influence of voluntary processes on neuronal responses in LIP. The lack of pronounced differences between the spatially select