| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Physiology, The University of Tokyo School of Medicine, Tokyo 113-0033, Japan
Submitted 28 March 2003; accepted in final form 25 October 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
; Goldman-Rakic 1987; Miller and Asaad 2002; Miyashita and Hayashi 2000; Passingham 1993; Tanji and Hoshi 2001; Wise and Murray 2000). In humans, the lateral prefrontal cortex is activated when subjects are given information and asked to hold it for brief periods, and the dorsolateral prefrontal cortex is further activated when the subjects are required to process the information (Fletcher and Henson 2001; Owen 1997; Postle et al. 1999). In monkeys, although electrophysiological studies have shown that prefrontal neurons exhibit sustained activity bridging the temporal gap between a cue and a response (Funahashi et al. 1989; Fuster and Alexander 1971; Kubota and Niki 1971; Rainer et al. 1999), the results of lesion studies suggest that the dorsolateral prefrontal cortex also plays a critical role in manipulation and monitoring of information cued in seconds past (Petrides 1994, 1995). Moreover, it was recently pointed out that the macaque prefrontal cortex engages in decision-making processes based on the physical properties of visual stimuli (Hoshi et al. 2000; Iba and Sawaguchi 2003; Kim and Shadlen 1999; Schall 2001) and in cognitive control of associative memory (Hasegawa I et al. 1998; Tomita et al. 1999).
How the brain processes sensory information to prepare for proper responses has been addressed in monkeys using various paradigms (Fuster 1997; Goldman-Rakic 1987; Miller and Asaad 2002; Miyashita and Hayashi 2000; Ohbayashi et al. 2003; Passingham 1993; Tanji and Hoshi 2001; Wise and Murray 2000). However, few studies focused on the update of internal representations, which would often intervene between sensory processing and response preparation. In the present study, a novel sequential target-shift (STS) task was devised (Fig. 1). In the STS task, after a positional cue indicated the initial saccade target among 8 peripheral squares, monkeys were required to successively shift the saccade target by one position on every flash of a nonspatial target-shift cue (Fig. 1, inset). The target shift was required of the monkeys 0 to 3 times within a single trial, and the number of shifts was pseudorandomly determined from trial to trial so that the monkeys could not predict the final saccade target. When the fixation point disappeared at the end of any delay, the monkeys had to make a saccade to the target position at that moment. To perform the STS task, counting or numerical information is not explicitly required.
|
; Schlag-Rey et al. 1997; Zhang and Barash 2000) or rotatory oculomotor delayed response (rODR) task (Takeda and Funahashi 2002). In the context of sensorimotor transformation, these variations are greatly advantageous in the separation of visual processing and response preparation. However, their task-switching cues (the shape or color of a fixation spot), which instructed the subjects to make prosaccades or antisaccades, were presented from the beginning of trials before the visual peripheral cue appeared. Therefore in the antisaccade trials, the construction of internal representation for the saccade target could occur immediately after spatial sensory processing, or possibly at any time between sensory processing and response preparation. To overcome this point, as mentioned above, in the STS task, we sequentially presented an identical nonspatial target-shift cue after the presentation of the initial positional cue. This sequential structure is advantageous in controlling the timing of the target shift and investigating the neuronal activity related to the updated internal target.
While the monkeys performed the STS task, we recorded the neuronal activity from the dorsolateral prefrontal cortex. We found that macaque prefrontal neurons were able to actively follow the task demand for the sequential shift of the internal target. A preliminary report of the results of the present study was previously presented in abstract form (Fukushima et al. 2000).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Two male monkeys (Macaca fuscata) served as subjects. Their care and use were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and with the regulations of the University of Tokyo School of Medicine. The monkeys were first implanted with a scleral eye coil (Judge et al. 1980) and head-holding devices; then after training on the task, they were implanted with a recording chamber (Crist Instrument). All surgery was performed under aseptic conditions and general anesthesia (pentobarbital sodium 30 mg/kg iv).
Glass-coated Elgiloy electrodes (approximately 1 M
at 1 kHz) were used in the present study, and they were advanced into the cortex through the dura mater with a hydraulic microdrive manipulator (MO-95C, Narishige). We recorded from neurons in the posterior principalis region (mainly areas 46 and 8Ar; Fig. 2). When we preliminarily recorded from area 9 in the most anterior part of the chamber, we rarely found responsive neurons. While the monkeys carried out their task (see following text), we searched for neurons that showed any change in activity during any trial epoch by monitoring the neuronal activity with on-line rastergrams and a sound monitor. We studied neurons for which there were data from 
50 correct trials. The recording sites were localized using magnetic resonance imaging (Fig. 2, B and C, right).
|
Figure 1 shows the sequence of the STS task (Fukushima et al. 2000). Trials started with presentation of 8 dark squares (1.1° in size) evenly distributed at an eccentricity of 10° (peripheral squares). A white fixation spot (0.1° in size) appeared in the center (Fig. 1, Fix), and the monkeys maintained central fixation until the spot disappeared. One second after the onset of the spot, a white positional cue (0.5° in size) flashed for 300 ms in one of the peripheral squares, which was pseudorandomly chosen from trial to trial (Fig. 1, Cue). The positional cue indicated the initial saccade target. In 0-shift trials, after a 1-s cue delay the fixation spot disappeared (Fig. 1, i), and the monkeys had to make a saccade to the target position. In the other trials (Fig. 1, ii–iv), the fixation spot remained and a target-shift cue (1.0° in size) flashed in the center for 200 ms (Fig. 1, Shifts1, 2, and 3). This signaled the monkeys to internally shift the saccade target clockwise (or counterclockwise) by one position. The target-shift cue flashed a maximum of 3 times in a trial, and each was followed by a 1-s shift delay. When the fixation spot disappeared at the end of any shift delay, the monkeys had to make a saccade to the target position at that moment (Fig. 1, Saccade). The 0- to 3-shift trials were pseudorandomly interleaved, and the final target position could be 0, 45, 90, or 135° away from where the positional cue was presented (Fig. 1, inset). The monkeys were rewarded for correct saccades that terminated within 2.5 or 3° of the center of the target. When the eye position deviated by more than a predetermined distance (1.0–1.5°) before the offset of the fixation spot, the trial was immediately aborted. When a monkey made an incorrect saccade, correction trials were inserted, if necessary, until the monkey responded correctly. The correction trials were not included in any analysis in this study. Monkey G performed a clockwise version of the task with red target-shift cues; monkey H did a counterclockwise version with green target-shift cues. We did not use the bidirectional (clockwise and counterclockwise) version because of difficulty in training. The shift direction (clockwise or counterclockwise) had no effect on the neuronal properties of selectivity (early phase of the shift delays: Fisher's exact test, P = 0.127; late phase: Fisher's exact test, P = 0.754; see Table 3). The monkeys' performance was expressed as the percentage of correct saccades over correct plus incorrect saccades, excluding the aborted trials. The target position for a possible impending saccade is referred to as "current" target in the present study. The 8 peripheral positions were referenced in polar angles as shown in the inset of Fig. 1.
|
We classified a neuron as "task-related" as follows. We evaluated neuronal activity during the presentation of the positional cue (Fig. 1, Cue), the delay periods after the positional and target-shift cues (Fig. 1, Cue delay and Shift delays 1, 2, and 3), and 300 ms after the offset of the fixation spot (Fig. 1, Saccade); then we compared the average activity during each epoch with the baseline activity, which was assessed using the average activity during the 500 ms before the onset of the positional cue. When a significant difference was observed in at least one comparison (Mann–Whitney test, P < 0.01), the neuron was judged to be task-related.
MULTIPLE REGRESSION ANALYSIS. The activity during the shift delays (shift-delay activity) could be influenced by 3 factors: the location of the positional cue (cue factor), the location of the "current" target (target factor), and the "sequential position" of the shift delay (sequential factor). The sequential position indicates the condition of a shift delay being the 1st, 2nd, or 3rd within a given trial. We evaluated the influence of these factors on the shift-delay activity using multiple linear regression (Fu et al. 1995; Grunewald et al. 2002; Hayashi 1952). Each shift delay was divided into 2 phases, early and late, and the regressions were separately performed. This was because we attempted to examine the neuronal activity that culminated just after the presentation of the target-shift cue (see Figs. 3, 4, 5 and Figs. 8 and 9) as well as the activity that built up toward the end of the delay (see Figs. 6 and 7). We judged a neuron to be influenced by a factor when that factor made a significant contribution to the shift-delay activity in the multiple regression (F-test, P < 0.01). We then used the partial R2 values of the factors to estimate the strength of the factors' influence on the shift-delay activity. The regressions were performed using the GLM procedure of the SAS/STAT software (SAS Institute). Using the general linear model, our data were analyzed essentially the same way as ANOVA [cue and target factors, degree of freedom (df) = 7; sequential factor, df = 2]. With this software and MATLAB/Statistics Toolbox (MathWorks), all other statistical analyses were carried out. With the exception of performance calculations and error-trial analyses, only data from correct trials were used. Error correction trials were excluded from all analyses.
|
|
|
|
|
|
|
We also used a "selectivity vector" to estimate spatial selectivity during any given period. The vector was obtained by means of the first-degree sinusoidal regression (Georgopoulos et al. 1982; Kakei et al. 2001; Mardia 1972)—that is, by fitting a cosine curve to the neuronal activity that was plotted against the positional cue location for every trial. The direction of the vector was defined as the one at which the cosine curve peaked, and the amplitude was defined as the height from the bottom of the curve to the peak. The spatial selectivity of a given vector was judged significant when the original regression was significant (F-test, P < 0.01).
Temporal properties of the spatial selectivity of a neuron were examined by calculating instantaneous selectivity vectors for a series of 100-ms time bins having 50 ms of overlap (Fu et al. 1995; Johnson et al. 1999), beginning at the onset of the fixation spot and continuing until 300 ms after its offset. The vectors represent the transition of the neuron's momentary preference with reference to the location of the positional cue. If neuronal activity was dependent on the location of the positional cue irrespective of target shifts, the instantaneous selectivity vectors of the neuron would remain pointing to its preferred cue location. On the other hand, if the activity was dependent on the location of the "current" target, the vectors would change their direction on every flash of the target-shift cue.
POPULATION ANALYSIS OF TARGET-RELATED NEURONS. We judged a neuron to be "target-related" when the target factor influenced the shift-delay activity of the neuron in the multiple regression analyses. Target-related neurons were divided into 3 groups: early-dominant, late-dominant, and intermediate. Neurons whose target-related activity occurred only in the early phases of the shift delays were classified as the early-dominant type, and those whose target-related activity occurred only in the late phases were classified as the late-dominant type. Those neurons whose target-related activity occurred during both phases were further divided into the 3 groups: neurons that showed significantly greater activity in the early phase than in the late phase (Wilcoxon test, P < 0.01; compared were mean discharge rates during the first 300 ms and last 300 ms of each shift delay) were classified as early-dominant; neurons that showed significantly higher activity in the late phase (Wilcoxon test, P < 0.01; compared were mean discharge rates during the period from 150 to 450 ms and the last 300 ms of each shift delay) were classified as late-dominant; the rest of the neurons were classified as the intermediate type. When a neuron simultaneously met both of the above Wilcoxon tests, we operationally classified it as late-dominant. This was because neuronal activity that increased toward the end of the delay sometimes spilled over into the next delay, which could cause misjudgment of the neuronal classification based on the first test. Aligning the neurons of each group with respect to the individual preferred target location (Johnson et al. 1999; Kakei et al. 2001), we calculated collective instantaneous selectivity vectors, just as in the instantaneous selectivity vectors of an individual neuron. The individual preferred target location was calculated by averaging the selectivity vectors during the shift delays collocated by the location of the "current" target.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
50 correct trials (7 neurons were not tested in 3-shift trials). Of the 233 neurons, 173 were task-related (Table 2). There was no significant difference between the monkeys in distribution of the task-related neurons (Fisher's exact test, 2 monkeys x 23 categories, P = 0.227). Of these task-related neurons, 130 were delay-responsive, which we examined in detail in the present study. Their delay-period activity after the target-shift cues (shift-delay activity) was found influenced by the positional cue location (cue factor), the "current" target location (target factor) and/or the "sequential position" of the shift delay (sequential factor). The last indicates whether the activity was influenced by the condition of a shift delay being the 1st, 2nd, or 3rd within a given trial. We evaluated the influence of these factors on the shift-delay activity using the multiple regression analysis: each shift delay was divided into 2 phases, early and late, and regressions were separately performed (see METHODS). Table 3 shows the numbers of delay-responsive neurons whose activity was influenced by the combinations of the factors. Of the 130 delay responsive neurons, 125 neurons underwent 3-shift trials. The "other" group in the table contains those neurons whose activity was biased during delay periods (e.g., a general load of task performance) or failed to reach the significance level of P < 0.01 in the regression analyses. Table 3 also shows that there was no significant difference between the 2 monkeys in distribution of the neurons categorized against the influential factors in both the early (Fisher's exact test, P = 0.127) and late (Fisher's exact test, P = 0.754) phases of the shift delays. As expected from the task demand, during the shift delays, a substantial number of neurons exhibited selective activity related to the location of the "current" target.
|
Figure 3 shows the activity of a target-selective neuron recorded while the subject performed the clockwise version of the task. As shown in the eye tracing data (Fig. 3A, H and V for each panel), the monkey maintained good central fixation during trials. When the fixation spot disappeared at the end of any delay, the monkey made ballistic saccades to the "current" target 0, 45, 90, or 135° away from the initial position, complying with the task requirement (Fig. 3A, gray and black deflections in the eye traces). The neuron exhibited transient activity on presentation of the positional cue around the upper-right (45°) location, and weak activity persisted during the subsequent delay (Fig. 3A, Cue and D0 in the 1st, 2nd, and 3rd panels). After the first target shift, the neuron exhibited delay period activity that culminated early, in the trials in which the positional cue had flashed around the top (90°) location (Fig. 3A, D1 in the 2nd, 3rd, and 4th panels). After the second target shift, similar activity was observed in the trials in which the positional cue was located around the upper-left (135°) location (Fig. 3A, D2 in the 3rd, 4th, and 5th panels). After the third target shift, activity was observed in the trials in which the positional cue was around the left (180°) location (Fig. 3A, D3 in the 4th, 5th, and 6th panels). Thus the neuron's preference with respect to the location of the positional cue appeared to rotate counterclockwise on every target shift, which suggests that the neuronal activity was dependent on the location of the "current" target: for instance, when the monkey was required to shift the saccade target to the 90° location, the neuron started to exhibit the specific activity during the following delay periods (Fig. 3A, D1 in the 4th panel, D2 in the 5th panel, and D3 in the 6th panel). The target preference of the neuron is clearly illustrated in Fig. 4 where the spike density functions (SDFs) for each target shift were drawn together for each location of the positional cue or "current" target. The shift-delay activity fitted almost perfectly if collocated by the location of the "current" target (target collocation, Fig. 4B) but not if collocated by the location of the positional cue (cue collocation, Fig. 4A). During the cue delay, the neuron showed low activity compared with its shift-delay activity (Fig. 3A). This could arise from the fact that the cue delay followed the initial cue spatially indicating the target, whereas the shift delays followed the nonspatial target-shift cues. The neuron showed less activity around saccade execution (Fig. 3B).
Figure 5A shows the spatial tuning of the neuron during the early phases of the shift delays (see METHODS). As in Fig. 4, the polar charts were inconsistent in cue collocation (Fig. 5A, left), whereas they corresponded well in target collocation (Fig. 5A, middle) and their comparable size showed little sequential influence. The spatial tuning of the neuron during error trials was distorted (Fig. 5A, right), although a small number of error trials precluded detailed analysis. The neuronal activity during error trials appeared smaller compared with correct trials when the "current" target was shifted to the 45 and 90° locations, to which the neuron was roughly tuned in the correct trials (Fig. 5A, right). However, there was no significant difference in activity between the correct and error trials (P = 0.635).
The instantaneous selectivity vectors, which represent the neuron's momentary preference with respect to the positional cue location (see METHODS), showed that the spatial selectivity of the neuron was not intense during presentation of the target-shift cues (Fig. 5B, pink shadings). Thereafter, however, the amplitude of the vectors peaked and their directions stepped by about 45° between neighboring peaks, which suggests the neuron's preference for the location of the "current" target. During the rest of the delay, the spatial selectivity declined again.
Figure 6 shows another target-selective neuron. During the first shift delay, the activity of this neuron gradually increased in the trials in which the positional cue was located around the 180° location, whereas the activity was suppressed in the other trials (Fig. 6A, D1). During the 2nd shift delay, the neuronal activity showed a similar gradual increase in the trials in which the positional cue was around the 225° location, being suppressed in the other trials (Fig. 6A, D2). The increasing activity sharply declined at the end of the delay periods. During the 3rd shift delay, and the cue delay as well, the neuron exhibited less or suppressed activity (Fig. 6A, D3 and D0). From the activity during the 1st and 2nd shift delays, this neuron is considered target-selective; the suppressed activity during the 3rd shift delay suggests modulation by the sequential factor. Because the monkey showed over 80% correct responses in all trial types (0-shift trials, 93.9%; 1-shift, 90.4%; 2-shift, 89.7%; 3-shift, 84.2%) during recording of this neuron, the suppression in the 3rd shift delay could not be ascribed to the monkey's performance. After execution of saccades, the neuronal activity gradually recovered from the suppression, although the recovery was delayed after the saccades to the 90 and 135° locations (Fig. 6B).
In the polar charts in target collocation (Fig. 7A, right), the spatial tuning for the 1st and 2nd shift delays was consistent and comparable, whereas the spatial tuning during the 3rd shift delay was much less intense, being influenced by the sequential factor. The instantaneous selectivity vectors for this neuron gradually developed during the 1st and 2nd shift delays (Fig. 7B), and the directions of the vectors stepped by about 45° between the delays, suggesting the target preference of the neuron. On the other hand, the neuron exhibited weak selectivity during the cue delay and the 3rd shift delay.
The neuron shown in Fig. 8 was nearly silent before target shifts (Fig. 8A, Cue and D0), and then responded transiently after the target-shift cues (Fig. 8A, D1, D2, and D3). During the rest of each shift delay and around saccade execution, the neuron was again nearly silent (Fig. 8, A and B). Prominent transient activity emerged only after the 3rd target shift in trials with the positional cue around the 270° location (Fig. 8A, D3). The polar charts of the neuron were aligned in order in target collocation compared with cue collocation (Fig. 9A), even though the sizes of the polar charts were substantially modulated by the sequential factor. The amplitude of the instantaneous selectivity vectors for this neuron peaked just after presentation of the target-shift cues, and the 3rd peak was the highest (Fig. 9B). The directions of the vectors stepped by about 45° between the 2nd and 3rd peaks, which suggests the target preference of the neuron.
Population analysis of target-related neurons
Of 130 delay-responsive neurons tested, the target factor significantly influenced the shift-delay activity (multiple regression, P < 0.01) of 55 neurons during the early phases of the shift delays and 48 during the late phases (target-related neurons; Table 3). Of these, 35 were target-related in both phases, and in total 68 were target-related. We classified the 68 target-related neurons into 2 groups and the remainder (see METHODS): one group with early-dominant activity ("early-dominant," 30 of 68 neurons) and the other with late-dominant activity ("late-dominant," 22 of 68); the remainder was referred to as "intermediate" (16 of 68). For example, the neurons in Figs. 3, 4, 5 and in Figs. 8 and 9 were classified as early-dominant, whereas one in Figs. 6 and 7 was late-dominant. Of the whole delay-responsive neurons (130 neurons), 41 were early-dominant, 37 were late-dominant, and 52 were intermediate.
Figure 10 shows the collective instantaneous selectivity vectors for the target-related neurons. The direction of the vectors for the early- and late-dominant neurons stepped by about 45° as the "current" target was shifted. The figure confirms the existence of the prefrontal neurons most activated just after the target-shift cue and the neurons gradually activated toward the end of the delay. The figure suggests that the early- and late-dominant neurons in the prefrontal neuronal population could cooperate together in performance of the task in an alternate or complementary way. The vectors for the intermediate neurons changed direction smoothly rather than stepwise.
|
As shown in Table 3, the shift-delay activity was also influenced by the cue factor as well as the target factor in some neurons. Figure 11 shows an example of "cue, target, and sequential"–dependent neurons. Although each of the 3 factors had significant influence on neuronal activity in the regression analysis (cue: df = 7, F = 4.28, P < 0.0002; target: df = 7, F = 22.46, P < 0.0001; sequential: df = 2, F = 17.92, P < 0.0001), the spatial tuning of the neuron during the shift delays appeared to correspond better in target collocation (Fig. 11, right). In fact, the partial R2 for the target factor was greater than those for the other 2 factors (cue, 0.12; target, 0.42; sequential, 0.14), suggesting major influence of the target factor on the neuron's shift-delay activity.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
In the present study, to investigate the update of internal representations, we devised a novel STS task, introducing the target-shift cue into the memory-guided saccade task (Hikosaka and Wurtz 1983). After the positional cue that directly informed the monkeys about the location of the initial saccade target, the target-shift cue appeared 0 to 3 times within a single trial, prompting the monkeys to internally shift the target. The target-shift cue was always the same in shape, size, and color and was presented in the center so that the cue itself carried no spatial information. The identical target-shift cue could not discriminatingly indicate any particular position in the present study, whereas nonspatial cues in spatial conditional tasks (Boussaoud and Wise 1993; Kurata and Hoffman 1994; Quintana and Fuster 1999) signaled the response direction (e.g., a red cue signaling a leftward movement and a green one signaling a rightward movement). The STS task has a structure similar to that of other variations of the memory-guided saccade task, such as the antisaccade task (Funahashi et al. 1993; Schlag-Rey et al. 1997; Zhang and Barash 2000) or rODR task (Takeda and Funahashi 2002), in which subjects are required to make delayed saccades in the direction opposite to or 90° away from the location of peripheral visual cues; the task switch between normal saccades and antisaccades (or rotatory saccades in the rODR task) is indicated by nonspatial cues (the shape or color of a fixation spot). However, the task-switching nonspatial cues in these tasks appear from the beginning of trials before the presentation of the peripheral cues and, furthermore, these tasks do not have a sequential structure. Therefore the neuronal activity related to the construction of the internal target for anti- or rotatory saccades could be covered with the visual response to the cue stimuli, or could be blurred over the delay period between the cue and response. On the other hand, in the STS task, after the spatial sensory processing of the positional cue had been completed, the nonspatial target-shift cue signaled the monkeys to internally shift the saccade target. Therefore target-selective activity after the target-shift cues, if it emerged, is likely related to the updated internal representation for the saccade target.
In the STS task, the number of target shifts within a single trial was pseudorandomly determined from trial to trial so that the monkeys were not able to predict the final saccade target. Consequently, the simple and economical way for the monkeys to perform the task is probably to keep track of the "current" target preparing for a possibly impending saccade during each shift delay and to shift the target on every flash of the target-shift cue. In this "keep-and-shift" strategy, the initial cue location was no longer relevant during the shift delays. That the monkeys actually used the keep-and-shift strategy in the STS task is supported by the neuronal activity observed. A substantial number of prefrontal neurons showed target-selective activity during the performance of the STS task. Moreover, the shift-delay activity of the neurons tested was mainly influenced by the target factor under the sequential shift conditions rather than the cue factor, and only a few neurons showed the shift-delay activity that depended mostly on the location of the positional cue. The results described here show that the macaque prefrontal neurons actively followed the task demand for the sequential shift of the internal target, changing their activity on every flash of the target-shift cue.
Target-selective shift-delay activity
While the monkeys performed the STS task, we found the selective neuronal activity related to the location of the "current" target in the dorsolateral prefrontal cortex. When the target-shift cue implied the target shift to particular peripheral positions, a group of prefrontal neurons exhibited the early-dominant activity, which culminated just after the presentation of the target-shift cue, whereas another group exhibited the late-dominant activity, which built up toward the end of the delay. In the prefrontal cortex, similar discharge profiles, phasic and tonic, have been known during the performance of delayed response tasks (Fuster 1997; Goldman-Rakic 1987). Therefore in the aspect of the basic discharge profiles of neurons, our results were consistent with the properties of prefrontal neurons previously reported.
With respect to what the early-dominant activity reflected in the present study, there would be several possibilities. First, in general, the phasic neuronal activity that was time-locked to the presentation of visual stimuli has been considered as a visual response, and when the activity was modulated by the sensory properties of the stimuli, the activity was further considered as a sensory coding of the stimuli (Constantinidis et al. 2001; Quintana and Fuster 1999; Rainer et al. 1999). In the present study, however, the early-dominant activity could not be a mere visual response to nor sensory coding of the target-shift cue because the early-dominant activity was spatially tuned; the target-shift cue to which the activity was locked, on the other hand, carried no spatial information as previously explained. Although the 8 peripheral squares were on the screen throughout the trials, the squares were all identical and there was no external clue indicating the response target among the 8 possible positions. Second, the early-dominant activity in the present study could be regarded as the activity that emerged at the end of the previous delay period. It is known that some prefrontal neurons exhibit target-selective "perisaccadic" activity at the end of the delay period in delayed saccade tasks (Boch and Goldberg 1989; Funahashi et al. 1993). In the STS task, however, the monkeys did not make a saccade during the time period when the early-dominant activity emerged. Furthermore, as shown in Figs. 3 and 8, some early-dominant activity in the present study did not occur around the execution of the actual saccades. Third, considering the spatial and temporal properties of the early-dominant activity together with the demand of the STS task, the early-dominant activity could reflect the transitory representation for the saccade target that was triggered by the target-shift cue. This view may be supported by the fact that, in the error trials, the early-dominant activity tended to be smaller at the preferred target shift than in the correct trials, suggesting the failure of the construction of an internal representation for the next "current" target. In the STS task, it should be noted that the visible peripheral squares would facilitate the update of the internal representation.
The late-dominant target-selective activity gradually built up during the delay period, and then sharply declined once the monkeys were allowed to make a saccade or were signaled to shift the "current" target. This observation is consistent with other studies in which, after a variety of sensory inputs, the prefrontal neurons gradually developed the selective activity related to the location of the response target or the direction of the response (Funahashi et al. 1993; Hasegawa R et al. 1998; Kim and Shadlen 1999; Niki and Watanabe 1976; Quintana and Fuster 1999; Takeda and Funahashi 2002). The increasing trend of the neuronal activity suggests that the activity reflects the formation of the target representation in the working memory or the preparation for the forthcoming response (Constantinidis et al. 2001; Quintana and Fuster 1999; Rainer et al. 1999). The fact that the sequential factor modulated the late-dominant activity in some neurons suggests the cognitive nature of the activity rather than pure motor preparation.
Recursive processing in the STS task
The STS task aimed to model recursive processing of information, in which an internal representation formerly constructed can be recycled as materials and be updated repeatedly. While the monkeys performed the STS task, we found the early- and late-dominant neurons: the latter could reflect the sustained target representation in the spatial working memory as previous studies suggested (Constantinidis et al. 2001; Quintana and Fuster 1999; Rainer et al. 1999) and we now suggest that the former reflects a transitory target representation made from the previous representation having been maintained in the working memory, and that the transitory representation could be copied in the working memory to update its content. In the STS task, recognition of the target-shift signal triggered the preparation of an internal representation for the next "current" target. The neurons in charge of this initiation process would be activated whenever a target shift is required, irrespective of the target location. However, we did not find typical neurons of such type at least in the dorsolateral prefrontal cortex that we investigated. Such neurons could be found in other cortical areas, particularly in the ventrolateral prefrontal cortex (Barbas 1988; Pandya and Yeterian 1996) or in the anterior prefrontal cortex, area 10 (Koechlin et al. 1999). Alternatively, the same internal representation could be used in all situations irrespective of the location of the "current" target and, to shift it spatially, activities in the neural network can be updated each time the shift signal is presented. Moreover, the STS task may demand attentional resources in addition to the mnemonic ones. While the monkeys maintained the target representation in their short-term storage, they could locate their spatial attention on the very location (Boussaoud and Wise 1993; di Pellegrino and Wise 1993; Everling et al. 2002) as well as on the central fixation spot, of which offset signaled the monkeys to "GO." In psychological studies, it has been argued that the spatial attention and spatial working memory are closely correlated (Awh and Jonides 2001; Schneider 1999; Smyth 1996). Neuronal activities having attentional and mnemonic aspects could closely cooperate while the monkeys performed the STS task.
In the present study, we found the prefrontal neuronal activity related to the recursive update of internal representations in the sequential target-shift paradigm. Such dynamic update of representations would be one of the essential abilities of the primate prefrontal cortex, and the present results may provide a clue to the neuronal mechanisms underlying the dynamic processing of internal representations.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
GRANTS
This work was supported by a Grant-in Aid for Specially Promoted Research (14002005) to Y. Miyashita from the Ministry of Education, Culture, Sports, Science and Technology of Japan.
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: Y. Miyashita, Department of Physiology, The University of Tokyo School of Medicine, Tokyo 113-0033, Japan (E-mail: yasushi_miyashita{at}m.u-tokyo.ac.jp).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Barbas H. Anatomic organization of basoventral and mediodorsal visual recipient prefrontal regions in the rhesus monkey. J Comp Neurol 276: 313-342, 1988.[CrossRef][ISI][Medline]
Boch RA and Goldberg ME. Participation of prefrontal neurons in the preparation of visually guided eye movements in the rhesus monkey. J Neurophysiol 61: 1064-1084, 1989.
Boussaoud D and Wise SP. Primate frontal cortex: neuronal activity following attentional versus intentional cues. Exp Brain Res 95: 15-27, 1993.[ISI][Medline]
Constantinidis C, Franowicz MN, and Goldman-Rakic PS. The sensory nature of mnemonic representation in the primate prefrontal cortex. Nat Neurosci 4: 311-316, 2001.[CrossRef][ISI][Medline]
di Pellegrino G and Wise SP. Visuospatial versus visuomotor activity in the premotor and prefrontal cortex of a primate. J Neurosci 13: 1227-1243, 1993.[Abstract]
Everling S, Tinsley CJ, Gaffan D, and Duncan J. Filtering of neural signals by focused attention in the monkey prefrontal cortex. Nat Neurosci 5: 671-676, 2002.[CrossRef][ISI][Medline]
Fletcher PC and Henson RN. Frontal lobes and human memory: insights from functional neuroimaging. Brain 124: 849-881, 2001.
Fu QG, Flament D, Coltz JD, and Ebner TJ. Temporal encoding of movement kinematics in the discharge of primate primary motor and premotor neurons. J Neurophysiol 73: 836-854, 1995.
Fukushima T, Hasegawa I, and Miyashita Y. Modulation of neuronal activities in macaque dorsal periarcuate area by target-shift cues. Soc Neurosci Abstr 26: 404.2, 2000.
Funahashi S, Bruce CJ, and Goldman-Rakic PS. Mnemonic coding of visual space in the monkey's dorsolateral prefrontal cortex. J Neurophysiol 61: 331-349, 1989.
Funahashi S, Chafee MV, and Goldman-Rakic PS. Prefrontal neuronal activity in rhesus monkeys performing a delayed anti-saccade task. Nature 365: 753-756, 1993.[CrossRef][Medline]
Fuster JM. The Prefrontal Cortex: Anatomy, Physiology, and Neuropsychology of the Frontal Lobe (3rd ed.). Philadelphia, PA: Lippincott-Raven, 1997.
Fuster JM and Alexander GE. Neuron activity related to short-term memory. Science 173: 652-654, 1971.
Georgopoulos AP, Kalaska JF, Caminiti R, and Massey JT. On the relations between the direction of two-dimensional arm movements and cell discharge in primate motor cortex. J Neurosci 2: 1527-1537, 1982.[Abstract]
Goldman-Rakic PS. Circuitry of the prefrontal cortex and the regulation of behavior by representational memory. In: Handbook of Physiology. The Nervous System. Higher Functions of the Brain. Bethesda, MD: Am. Physiol. Soc., 1987, sect. 1, vol. V, pt. 1, chapt. 9, p. 373-417.
Grunewald A, Bradley DC, and Andersen RA. Neural correlates of structure-from-motion perception in macaque V1 and MT. J Neurosci 22: 6195-6207, 2002.
Hasegawa I, Fukushima T, Ihara T, and Miyashita Y. Callosal window between prefrontal cortices: cognitive interaction to retrieve long-term memory. Science 281: 814-818, 1998.
Hasegawa R, Sawaguchi T, and Kubota K. Monkey prefrontal neuronal activity coding the forthcoming saccade in an oculomotor delayed matching-to-sample task. J Neurophysiol 79: 322-333, 1998.
Hayashi C. On the prediction of phenomena from qualitative data and the quantification of qualitative data from the mathematico-statistical point of view. Ann Inst Stat Math 3: 69-98, 1952.
Hikosaka O and Wurtz RH. Visual and oculomotor functions of monkey substantia nigra pars reticulata. III. Memory-contingent visual and saccade responses. J Neurophysiol 49: 1268-1284, 1983.
Hoshi E, Shima K, and Tanji J. Neuronal activity in the primate prefrontal cortex in the process of motor selection based on two behavioral rules. J Neurophysiol 83: 2355-2373, 2000.
Iba M and Sawaguchi T. Involvement of the dorsolateral prefrontal cortex of monkeys in visuospatial target selection. J Neurophysiol 89: 587-599, 2003.
Johnson MT, Coltz JD, Hagen MC, and Ebner TJ. Visuomotor processing as reflected in the directional discharge of premotor and primary motor cortex neurons. J Neurophysiol 81: 875-894, 1999.
Judge SJ, Richmond BJ, and Chu FC. Implantation of magnetic search coils for measurement of eye position: an improved method. Vision Res 20: 535-538, 1980.[CrossRef][ISI][Medline]
Kakei S, Hoffman DS, and Strick PL. Direction of action is represented in the ventral premotor cortex. Nat Neurosci 4: 1020-1025, 2001.[CrossRef][ISI][Medline]
Kim JN and Shadlen MN. Neural correlates of a decision in the dorsolateral prefrontal cortex of the macaque. Nat Neurosci 2: 176-185, 1999.[CrossRef][ISI][Medline]
Koechlin E, Basso G, Pietrini P, Panzer S, and Grafman J. The role of the anterior prefrontal cortex in human cognition. Nature 399: 148-151, 1999.[CrossRef][Medline]
Kubota K and Niki H. Prefrontal cortical unit activity and delayed alternation performance in monkeys. J Neurophysiol 34: 337-347, 1971.
Kurata K and Hoffman DS. Differential effects of muscimol microinjection into dorsal and ventral aspects of the premotor cortex of monkeys. J Neurophysiol 71: 1151-1164, 1994.
Mardia KV. Statistics of Directional Data. London: Academic Press, 1972.
Miller EK and Asaad WF. The prefrontal cortex: conjunction and cognition. In: Handbook of Neuropsychology (2nd ed.), edited by Boller F and Grafman J. Amsterdam: Elsevier, 2002, vol. 7, p. 29-54.
Miyashita Y and Hayashi T. Neural representation of visual objects: encoding and top-down activation. Curr Opin Neurobiol 10: 187-194, 2000.[CrossRef][ISI][Medline]
Niki H and Watanabe M. Prefrontal unit activity and delayed response: relation to cue location versus direction of response. Brain Res 105: 79-88, 1976.[CrossRef][ISI][Medline]
Ohbayashi M, Ohki K, and Miyashita Y. Conversion of working memory to motor sequence in the monkey premotor cortex. Science 301: 233-236, 2003.
Owen AM. The functional organization of working memory processes within human lateral frontal cortex: the contribution of functional neuroimaging. Eur J Neurosci 9: 1329-1339, 1997.[CrossRef][ISI][Medline]
Pandya DN and Yeterian EH. Comparison of prefrontal architecture and connections. Philos Trans R Soc Lond B 351: 1423-1432, 1996.[ISI][Medline]
Passingham R. The Frontal Lobes and Voluntary Action. Oxford, UK: Oxford Univ. Press, 1993.
Petrides M. Frontal lobes and working memory: evidence from investigations of the effects of cortical excisions in nonhuman primates. In: Handbook of Neuropsychology, edited by Boller F and Grafman J. Amsterdam: Elsevier, 1994, vol. 9, p. 59-82.
Petrides M. Impairments on nonspatial self-ordered and externally ordered working memory tasks after lesions of the mid-dorsal part of the lateral frontal cortex in the monkey. J Neurosci 15: 359-375, 1995.[Abstract]
Postle BR, Berger JS, and D'Esposito M. Functional neuroanatomical double dissociation of mnemonic and executive control processes contributing to working memory performance. Proc Natl Acad Sci USA 96: 12959-12964, 1999.
= 25 ms). SDFs were collocated by the location of the positional cue (A) or "current" target (B). S1–S3 (pink shadings): presentation of the target-shift cue.
