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: 95/2/1008    most recent 00552.2005v1 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 (4) Citing Articles via Google Scholar Google Scholar Articles by Inoue, M. Articles by Mikami, A. Search for Related Content PubMed PubMed Citation Articles by Inoue, M. Articles by Mikami, A.

Prefrontal Activity During Serial Probe Reproduction Task: Encoding, Mnemonic, and Retrieval Processes

Department of Behavioral and Brain Sciences, Primate Research Institute, Kyoto University, Inuyama, Aichi, Japan

Submitted 26 May 2005; accepted in final form 28 September 2005

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS REFERENCES

 
To study the prefrontal neuronal mechanism for the encoding and mnemonic processing of multiple objects, the order of object presentation, and the retrieval of an object among objects in the working memory, we recorded neuronal activity from the lateral prefrontal cortex while two monkeys performed the serial probe reproduction task. In the task, two objects (C1 and C2) were presented sequentially interleaved with a delay (D1) period, and after the second delay (D2) period, a color cue was presented. Monkeys were trained to select one target object on the basis of the color stimulus. During the C1 and C2 periods, we found responses that depended on the order of presentation (order-selective response). During the D1 and/or D2 periods, two-thirds of the neurons with object-selective delay-period activity showed order-selective activity coding either C1 or C2. Neurons with larger response magnitudes during the C2 period showed order-selective delay-period activity during the D2 period. These order-selective responses during the C2 period could also contribute to order-selective delay-period activity, and order-selective delay-period activity during the D1 and D2 periods could play an essential role in storing information on both the object and the temporal order of presentation. During the color cue period, two-thirds of the neurons with responses showed target object selectivity (CT and T responses), although the target object was not presented during this period. The CT and T responses could play a critical role in the retrieval of an item among various items in the working memory.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS REFERENCES

 
Previous studies provided evidence that the lateral prefrontal cortex is involved in memory for temporal order. Patients with lesions involving the mid-dorsolateral prefrontal cortex perform more poorly than control subjects or patients with other cortical lesions in a task requiring judgment on which of two items was presented more recently (Kesner et al. 1994; McAndrew and Milner 1991; Milner 1971; Milner et al. 1991; Shimamura et al. 1990). Animal lesion studies also strongly implicated the prefrontal cortex in temporal order memory. Petrides (1991) found that monkeys with lesions involving the mid-dorsolateral prefrontal cortex were impaired in judging the order of object stimuli that had been presented in the middle of a list of objects. Similarly, rats with medial prefrontal lesions performed poorly on tests of temporal order memory for spatial locations (Chiba et al. 1994; Kesner and Holdbrook 1987). These studies of humans and animals with focal brain damage support the idea that the prefrontal cortex plays an important role in memory for temporal order.

On the other hand, many electrophysiological studies have provided evidence that the lateral prefrontal cortex participates in the process of temporary storage of spatial or object information (Chafee and Goldman-Rakic 1998; Constantinidis et al. 2001; di Pellegrino and Wise 1991, 1993; Funahashi et al. 1989, 1993b; Fuster 1973; Fuster et al. 1982; Kubota and Niki 1971; Kubota et al. 1974; Miller et al. 1996; Niki 1974; Niki and Watanabe 1976; Rainer et al. 1998a; Rao et al. 1997; Sawaguchi and Yamane 1999; Takeda and Funahashi 2002; Watanabe 1981; Wilson et al. 1993). Tonic delay-period activity with directional selectivity or object selectivity has been considered to be a neuronal correlate of temporal storage mechanism for spatial or object information in the prefrontal cortex.

A few studies showed prefrontal neuronal activities when monkeys retain information on multiple items simultaneously. Barone and Joseph (1989) recorded prefrontal neuronal activities while a monkey performed a delayed-response task in which a monkey was required to memorize three target positions and the temporal order of presentation and to respond by performing sequential saccades and hand-reaching movements toward the targets in the same temporal order. They found that one class of prefrontal neurons exhibit tonic activity with spatial and temporal selectivity, such that tonic activity was observed only when the first visual cue was presented at one particular position out of three. Funahashi et al. (1997) recorded prefrontal neuronal activities while a monkey performed a delayed sequential reaching task, in which the monkey is required to memorize two of three target positions and the temporal order of presentation and to respond by performing hand-reaching movements toward the targets in the same temporal order. They found two types of delay-period activity: position-dependent and pair-dependent activities. Position-dependent delay-period activity is a selective response to one of three positions. Most of the position-dependent activities arise when a cue is presented at a particular position in a particular temporal order. Pair-dependent delay-period activity is a selective response to a particular combination of two of three positions. Most of the pair-dependent activities arise when two cues are presented in a particular temporal order. Ninokura et al. (2003) found that 43% of delay-period activity is selective for the sequence in which visual objects are presented during the cue period. They further found that 31% of this activity was selective for only one of six sequences, and the remaining activity was selective for multiple sequences. These findings suggest that prefrontal neurons can retain information on multiple items (spatial positions or objects) and the temporal order of cue presentation. However, Ninokura et al. (2003) found the delay-period activity that was selective for multiple sequences, but they did not clarify what information was coded in this delay-period activity. In addition, Funahashi et al. (1997) and Ninokura et al. (2003) used tasks in which monkeys had to memorize spatial locations or objects and to respond by performing sequential movements toward memorized locations or objects, and analyzed neuronal activity during only delay period preceding the sequential movements. Thus these prefrontal neuronal activities could reflect the preparation for the first or second movement or sequential movements. Indeed, there were delay-period activities that coded the direction of movements rather than the location of the cue (Funahashi et al. 1993b; Niki and Watanabe 1976; Takeda and Funahashi 2002). To clarify the neuronal mechanism for the retention of information on multiple items simultaneously, it is necessary to dissociate memorized items and preparation for movements.

In this study, we introduced a serial probe reproduction (SPR) task, in which a monkey had to memorize two objects and their order of presentation, and one target object was selected from two memorized objects on the basis of a color stimulus. In this task, because the target object was determined on the basis of the color cue during the color cue period, monkeys could not determine the target object until the color cue period. Thus the memory during the first and second delay periods must be nothing else but the two objects and their order of presentation.

Previous studies indicated that neurons with cue-period and delay-period activities exhibited a similar spatial or object preference between the cue and delay periods, suggesting that the visual input to the prefrontal cortex plays an important role in contrasting delay-period activity (Funahashi and Inoue 2000; Funahashi et al. 1990; O'Scalaidhe et al. 1999). Recently, Ninokura et al. (2004) have found visual responses in the lateral prefrontal cortex that depend on both the object and the order of object presentation. However, because Ninokura et al. (2004) did not analyze the relationship between visual response and delay-period activity, the functional role of these responses in the process of encoding object information and temporal order information has not been clarified. To examine this issue, we analyzed prefrontal neuronal activities during the first and second cue periods of the SPR task and compared the visual response and delay-period activity.

In addition, the lateral prefrontal cortex (LPFC) plays a crucial role in response selection, which is the ultimate goal of purposeful behavior. Many studies showed that the responses of the prefrontal cortex to visual stimuli are related to the selection of an object from an array of objects (Hasegawa et al. 2000; Iba and Sawaguchi 2002) and the selection of forthcoming movements based on external stimuli (Hasegawa et al. 1998; Hoshi and Tanji 2004; Hoshi et al. 2000; Kim and Shadlen 1999; Sakagami and Niki 1994a,b; Sakagami and Tsutsui 1999; Sakagami et al. 2001; Watanabe 1986). These results suggest that the LPFC contributes to the retrieval of one object from the working memory. Although a few human neuroimaging studies showed that the LPFC is also involved in the retrieval of one item from the working memory (Rowe and Passingham 2001; Rowe et al. 2000), the neuronal mechanism for the retrieval of one item from the working memory in the LPFC has not been clarified. To investigate the neuronal mechanism in the LPFC for this process, we analyzed neuronal activity during the color cue period of the SPR task. In this period, a monkey retrieves one object from two memorized objects. These data were reported elsewhere in abstract form (Inoue and Mikami 2001, 2002a,b, 2003).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS REFERENCES

 
Animals and apparatus

We used one male and one female rhesus monkey (Macaca mulatta; monkey G, 9 kg; monkey H, 5 kg). Experiments were conducted according to the Guide for the Care and Use of Laboratory Animals by the National Institute of Health and the Guide for the Care and Use of Laboratory Primates by the Primate Research Institute, Kyoto University.

The monkey was seated on a primate chair in a dark room, and a head-restraining device was fixed its head. It was trained to look at a 17-in CRT monitor (FlexScan T565, Nanao), which was placed 40 cm from its face. A computer (PC-9821Xa200, NEC) presented a fixation spot and a stimulus on the CRT monitor. The monkey's horizontal and vertical eye positions were sampled at 250 Hz with a monitoring system using an infrared camera (R-21C-AC, RMS Hirosaki). Sampled eye positions were fed into a computer (PC-9801BX, NEC) through an A/D converter to determine whether the monkey maintained its fixation and performed a correct saccade.

Behavioral task

The monkeys were trained to perform an SPR task (Fig. 1A). In this task, after a 1-s intertrial interval, a fixation spot (a white circle; 0.1° diam) was presented at the center of the monitor. After the monkey maintained its fixation for 1.5 s, the first object cue (C1), which was one of three objects (a double cone, a cross, and a circle, 3° x 3° in size), was presented at the center of the monitor for 0.5 s. After 1 s of the first delay (D1) period, the second object cue (C2), which was one of the two remaining objects, was presented at the center of the monitor for 0.5 s. After 1 s of the second delay (D2) period, a color cue (a red or a green rectangle, 3° x 3°) was presented at the center of the monitor for 0.5 s. The presentation of a color cue was followed by the third delay (D3) period of 1–1.5 s. Then the three objects were presented in the upper, lower left, and lower right positions at 9° of eccentricity from the center of the monitor. The fixation spot was extinguished simultaneously. When the color cue was red, the monkey had to perform a saccade to the object presented as the first cue (C1), and when the color cue was green, the monkey had to perform a saccade to the object presented as the second cue (C2). We prepared three patterns of arrangement of target objects (Fig. 1D), and one of the patterns was randomly selected during the response period. Therefore the monkeys could not determine the spatial location to which a saccade should be performed until the appearance of objects during the response period.

View larger version (21K): [in this window] [in a new window]   FIG. 1. Sequence of trial events of serial probe reproduction (SPR) task (A), delayed-matching-to-sample (DMS) task (B), and fixation task (C). R and G indicate the red and green stimuli, respectively. D: arrangement of target objects during response period.

The monkeys also performed a fixation task (Fig. 1C), which allows the examination of neuronal activity when selection was not required. While the monkey maintained its fixation, we presented a color stimulus, which was the same as that used in the SPR task.

Surgical procedure

To fix the head during training, a head-restraining device was attached to the skull. Surgery was performed under aseptic conditions. The monkeys were first administered ketamine (10 mg/kg body weight) intramuscularly, and then an intravenous injection of pentobarbital sodium (20 mg/kg body weight). After partially exposing the skull, polycarbonate screws (3 mm in diameter and 5 mm in length) were used to attach firmly the head-restraining device to the skull. These screws and the head-restraining device were fixed with dental acrylic resin. The monkeys were administered systemic antibiotics for 1 wk after surgery and were allowed free access to water and chow for at least 1 wk after surgery.

After the training was completed, surgery for attaching a recording chamber was performed under aseptic conditions. We performed MRI before surgery, and on the basis of this MRI, we determined the stereotaxic position of the principal sulcus. The position of the recording chamber (anterior-posterior = 32 mm and lateral = 18 mm) was determined by this sterotaxic coordination.

Training procedure

The monkeys were first trained to perform the DMS task. When the monkeys showed an 80% correct performance or higher for 2 wk, the monkeys were trained to perform the SPR task. The SPR task training was divided into three stages. In stage 1, the monkeys were trained to perform the red cue trials, in which the monkeys had to select the C1 object. During the early period of training, the duration of C2 presentation was 100 ms. When the monkeys showed a 70% correct performance or higher for 2 wk, the duration of C2 presentation was progressively extended to 500 ms. In stage 2, monkeys were trained to perform the green cue trials, in which the monkeys had to select the C2 object. During the early period of training, the duration of C1 presentation was 100 ms. When the monkeys showed a 70% correct performance or higher for 2 wk, the duration of C1 presentation was progressively extended to 500 ms. In stage 3, we intermingled the red cue trials and green cue trials. The training was considered completed when the monkeys showed a 70% correct performance or higher for 2 wk. The training process was completed in 18 (monkey G) and 24 (monkey H) mo.

Recording procedure and data analysis

Neuronal activity was recorded using glass-coated Elgiloy microelectrodes (1–2 M). Single-neuronal activity was isolated and converted to pulses by a window discriminator (DIS-1, BAK), and stored with task events as a data file on a hard disk. Recording sites were determined by MRI. The dorsolateral prefrontal cortex (DLPFC) was defined as the region dorsal to the principal sulcus, and the ventrolateral prefrontal cortex (VLPFC) was defined as the region ventral to the principal sulcus (Fig. 2). To determine whether the recording site was in the frontal eye field, we applied intracortical microstimulation (ICMS) through the tips of inserted electrodes (22 pulses of 0.25-ms duration at 333 Hz; current intensity, 100 µA). In this study, there were no recording sites where ICMS evoked saccades.

View larger version (38K): [in this window] [in a new window]   FIG. 2. Surface location of recording sites of neurons. Surface images were obtained by MRI performed before training. , recording sites in the dorsolateral prefrontal cortex (DLPFC); , recording sites in the ventrolateral prefrontal cortex (VLPFC). AS, arcuate sulcus; PS, principal sulcus.

When we analyzed neuronal activity during the D1 and D2 periods, we constructed six histograms and rasters for each combination of C1 and C2 objects aligned at the start of the D1 and D2 periods. We calculated mean discharge rate during the last 1-s interval of the fixation period (control period) and the D1 and D2 periods. First, we analyzed neuronal activity during the D1 period when the mean discharge rate during the D1 period was significantly different from that during the control period (Mann-Whitney U test; P < 0.05); we considered that the neuron had a significant delay-period activity. We considered that delay-period activity was object-selective when the difference was significant as determined by ANOVA (P < 0.05).

A significant response during the D2 period was determined similarly (Mann-Whitney U test; P < 0.05). To determine whether delay-period activity during the D2 period depends on the C1 object, C2 object, or one sequence, we compared delay-period activities during the D2 period under six trial conditions by ANOVA. When the difference in delay-period activity under the six trial conditions was significant (P < 0.05), we compared the highest delay-period activity with other delay-period activities by a post hoc test (Fisher's PLSD) and determined whether delay-period activity was selective in only one sequence. Results showed that there were no neurons activated during the D2 period in only one of six sequences, and the delay-period activity depended on either the C1 object or the C2 object. We calculated selectivity index for the C1 object (SI_C1) and that for the C2 object (SI_C2). SI is defined as

where R_{C1pref, C2pref} = firing rate of the highest delay-period activity during the D2 period, R_{C1pref, C2nonpref} = firing rate in response to the same C1 object and a different C2 object during the D2 period, and R_{C1nonpref, C2pref} = firing rate in response to a different C1 object and the same C2 object during the D2 period. If the delay-period activity depended on the C1 object, SI_C1 is 1, and if the delay-period activity depended on the C2 object, SI_C2 is 1. When SI_C1 was greater than SI_C2, we defined that this delay-period activity depended on the C1 object, and when SI_C2 was greater than SI_C1, we defined that this delay-period activity depended on the C2 object.

We classified object-selective delay-period activities into three types: order-nonselective activity, C1-coding activity, and C2-coding activity. In a neuron with order nonselective activity, delay-period activity during the D1 period depended on the C1 object, delay-period activity during the D2 period depended on the C2 object, and the preferred objects during the D1 and D2 periods were identical. In a neuron with C1-coding activity, delay-period activity during the D1 and/or D2 period depended on the C1 object. In a neuron with C2-coding activity, delay-period activity during the D1 period was not detected or object-nonselective, but delay-period activity during the D2 period depended on the C2 object.

When we analyzed neuronal activity during the color cue period, six averaged perievent time histograms (2 colors and 3 objects to which a saccade should be performed) triggered by the onset of the color cue were constructed (bin width, 10 ms). From these histograms, when a neuron exhibited an excitatory response, the histogram with the highest peak value was chosen for determining the response window, and when a neuron exhibited an inhibitory response, the histogram with the lowest peak value was chosen. In this histogram, the starting point of response (the time at which the 1st 3 consecutive bins differed from discharge rates for 1 s before the color cue presentation by >2 SD or <2 SD) and the endpoint of response (the time of the last bin) were determined. The time from the onset of the color cue to the starting point of the response window was taken as onset latency. When the discharge rate of a neuron during the response window of the color cue period differed significantly (Mann-Whitney U test) from that during the fixation period (for 1 s before the 1st cue period), we concluded that the neuron exhibited a response. We performed two-way ANOVA of the response magnitude during the color cue period in terms of the color (red or green) and target (object to which a saccade should be performed; a double cone, a cross, or a circle) factors. When the difference in response magnitude in terms of the color factor was significant and the difference in terms of the target factor was not significant, we considered the neuron as having a C response. When the difference in response magnitude in terms of the color factor was not significant and the difference in response magnitude in terms of the target factor was significant, we considered the neuron as having a T response. When the difference in response magnitude in terms of both the color factor and target factor were significant or the difference of interaction was significant, we considered the neuron as having a CT response.

When we analyzed neuronal activity during the D3 period, six averaged perievent time histograms (2 colors and 3 objects to which a saccade should be performed) triggered by the offset of the color cue were constructed (bin width, 50 ms). When the discharge rate of a neuron during the first 1 s of the D3 period differed significantly (Mann-Whitney U test) from that during the fixation period (for 1 s before the 1st cue period), we concluded that the neuron exhibited delay-period activity during the D3 period. We performed two-way ANOVA of delay-period activity during the D3 period in terms of the color (red or green) and the target (object to which a saccade should be performed; a double cone, a cross, or a circle) factors. When the difference in response magnitude in terms of the color factor was significant and the difference in response magnitude in terms of the target factor was not significant, we considered the neuron as having a C delay-period activity. When the difference in response magnitude in terms of the color factor was not significant and the difference in response magnitude in terms of the target factor was significant, we considered the neuron as having a T delay-period activity. When the difference in response magnitude in terms of both the color factor and target factor were significant or the difference of interaction was significant, we considered the neuron as having a CT delay-period activity.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION GRANTS REFERENCES

 
Behavioral performance

Two monkeys performed the SPR task well but not perfectly. The chance level of this task is 33%, and the monkeys performed much higher than the chance level. We examined performance accuracy during trials for all recording sessions. The mean correct percent during recording trials was 75 ± 5% (SD) for monkey G and 70 ± 3% for monkey H. There was no tendency for direction preference, but there was tendency for object preference. Both monkeys performed better when the target object was a circle (84 ± 10% for monkey G, 88 ± 9% for monkey H) than when the target object was a double cone (70 ± 6% for monkey G, 62 ± 6% for monkey H) or a cross (72 ± 8% for monkey G, 66 ± 7% for monkey G). Monkey G performed better in the trials in which the C2 object was the target (86 ± 7%) than in the trials in which the C1 object was the target (66 ± 6%). In contrast, for monkey H, there was no difference in performance between the trials in which the C1 object was the target (69 ± 5%) and the trials in which the C2 object was the target (71 ± 5%). The mean latencies for saccades from the onset of target object presentation were 284 ± 105 ms for monkey G and 260 ± 89 ms for monkey H.

Neuronal database

While two monkeys performed the SPR task, we recorded the activity of 611 single neurons from the DLPFC (n = 173) and the VLPFC (n = 438). Of these neurons, 483 responded during at least one epoch of the SPR task. Of these, 119 neurons showed a response during the C1 and/or C2 period, 183 neurons exhibited delay-period activity during the D1 and/or D2 period, 139 neurons showed a response during the color cue period, 260 neurons showed delay-period activity during the D3 period, and 361 neurons exhibited a response during the response period. In this paper, we focused on neuronal activity in the C1, D1, C2, D2, color cue, and D3 periods, and we will deal with the response period in a separate report.

Object-selective response during C1 and C2 periods

During the C1 and/or C2 periods, 119 neurons exhibited responses. Of these, 81 neurons showed the object-selective response during the C1 or C2 period. Of these, 22 neurons showed response magnitudes that were not significantly different between the C1 and C2 periods (order-nonselective). On the other hand, the magnitudes of responses of 59 neurons during the C1 and C2 periods were significantly different (order-selective). Of these neurons, 33 showed larger response magnitudes during the C1 period than during the C2 period (C1-dominant), and 26 showed larger response magnitudes during the C2 period than during the C1 period (C2-dominant).

Figure 3A shows the histograms of the C1-dominant response of a neuron. During the C1 period, the neuron exhibited a large response magnitude (36.92 spikes/s) to the circle, a small response magnitude (20.22 spikes/s) to the cross, and no response to the double cone. Although the activity slightly increased in all trial conditions toward the end of the D1 period, discharge rates during the D1 period were not significantly different from those during the control period. During the C2 period, this neuron did not respond to any of the three visual stimuli. The difference in response magnitude was significant in terms of both the object [F(2,228) = 8.62, P < 0.0005] and order factors [F(1,228) = 29.72, P < 0.0001). Therefore this neuron was considered as having the C1-dominant response, and we defined the circle as the preferred object of this neuron.

View larger version (62K): [in this window] [in a new window]   FIG. 3. Histograms of object-selective and order-selective responses. A: histograms of C1-dominant response of a neuron. The neuron exhibited response when the circle was presented during the C1 period but did not show response during C2 period. B: histograms of C2-dominant response of a neuron. The neuron exhibited a small response magnitude when the circle was presented during C1 period but showed a large response magnitude when the circle was presented during C2 period. Bin width is 20 ms.

To compare the temporal profiles of the order-nonselective, C1-dominant, and C2-dominant responses, we constructed population histograms of these responses (Fig. 4A). Neurons with the order-nonselective response showed large responses to the preferred object presented during both C1 and C2 periods (Fig. 4A, black lines). Neurons with the C1-dominant and C2-dominant responses showed large responses to the preferred object presented during the C1 and C2 period, respectively. Although these differential responses depended on the object and/or the order of presentation, the temporal profiles of the order-nonselective, C1-dominant, and C2-dominant responses were similar. Figure 4B shows the cumulative summation curves of the latencies of the order-nonselective, C1-dominant, and C2-dominant responses. During the C1 period, the mean latencies of responses were 137 ± 44 (order-nonselective), 124 ± 34 (C1-dominant), and 127 ± 51 ms (C2-dominant). During the C2 period, the mean latencies of responses were 160 ± 45 (order-nonselective), 137 ± 70 (C1-dominant), and 141 ± 56 ms (C2-dominant). Although the difference in latency among these responses was not statistically significant [F(2,123) = 1.48, P > 0.05], the latencies of the C1-dominant and C2-dominant responses were slightly shorter than that of the order-nonselective response. The durations of these responses were also not different. During the C1 period, the mean durations of responses were 172 ± 25 (order-nonselective) and 156 ± 15 ms (C1-dominant). During the C2 period, the mean durations of responses were 154 ± 21 (order-nonselective) and 157 ± 14 ms (C2-dominant). The difference in durations among these responses was not statistically significant [F(2,123) = 0.06, P > 0.05].

View larger version (20K): [in this window] [in a new window]   FIG. 4. A: population activities in preferred object trials (black) and nonpreferred object trials (gray) of neurons with order-nonselective, C1-dominant, and C2-dominant responses. F, fixation period; C1, 1st cue period; D1, 1st delay period; C2, 2nd cue period; D2, 2nd delay period. B: cumulative summation curve of latencies of order-nonselective, C1-dominant, and C2-dominant responses during C1 and C2 periods.

C1 objects modulated responses during C2 period

We examined whether responses during the C2 period were affected by the C1 object. Among neurons with the C1-dominant response (n = 33), most of these neurons (n = 32) did not show a significant difference in response magnitude during the C2 period depending on the C1 object presented (Mann-Whitney U test, P < 0.05). Among neurons with the C2-dominant response (n = 26), 35% (n = 9) showed a significant difference in response magnitude during the C2 period depending on the C1 object presented (Mann-Whitney U test, P < 0.05).

Figure 5 shows an example of the C2-dominant response of a neuron that varied depending on the C1 object presented. As shown in Fig. 5A, the neuron exhibited the response when the cross was presented, and the response magnitude during the C2 period (34.43 spikes/s) was significantly larger than that during the C1 period (16.49 spikes/s). The magnitude of response to the cross during the C2 period varied significantly with the C1 (Fig. 5B). When the double cone was presented during the C1 period, the magnitude of response to the cross during the C2 period was 22.31 spikes/s. When the circle was presented during the C1 period, the magnitude of response to the cross during the C2 period was 46.15 spikes/s. These response magnitudes were significantly different (z = 4.13; P < 0.0001).

View larger version (26K): [in this window] [in a new window]   FIG. 5. Histograms of C2-dominant responses of a neuron that varied with C1 object. The neuron showed response when the cross was presented, and magnitude of response during C2 period was larger than that during C1 period (A). Response magnitude during C2 period when the double cone was presented during C1 period was significantly different from that when the circle was presented during C1 period (B).

Order-selective response magnitudes were significantly different between the C1 period and C2 period. This difference could be caused by the enhancement or suppression of responses during the C1 or C2 period. To test this hypothesis, we compared responses during the SPR task with those during the DMS task. Figure 6A shows the C1-dominant response of a neuron during the SPR task and that during the DMS task. The neuron responded when the double cone was presented during the C1 period (34.92 spikes/s), but did not respond when the same double cone was presented during the C2 period. During the DMS task, the magnitude of response to the double cone (35.61 spikes/s) was similar to that during the C1 period of the SPR task. The response magnitude during the cue period of the DMS task was not significantly different (Mann-Whitney U test, z = 0.03, P > 0.05) from that during the C1 period of the SPR task, and was significantly larger (z = 4.18; P < 0.0001) than that during the C2 period of the SPR task. Among seven tested neurons with the C1-dominant response, five showed a similar response magnitude during the DMS task to that during the C1 period of the SPR task (Fig. 6B).

View larger version (51K): [in this window] [in a new window]   FIG. 6. A: histograms of C1-dominant response during SPR task and response during DMS task. Response magnitude during the DMS task was similar to that during C1 period of the SPR task. B: relationship between C1 and C2 periods of SPR task and DMS task. A closed circle shows that response magnitude during the DMS task was not significantly different (P > 0.05) from that during C1 period of the SPR task and was significantly larger (P < 0.05) than that during C2 period of the SPR task. An open circle shows that response magnitude during the DMS task was not significantly different (P > 0.05) from that during C1 and C2 periods of the SPR task. C: histograms of C2-dominant response during SPR task and response during DMS task. Response magnitude during the DMS task was similar to that during C1 period of the SPR task. D: another example of C2-domiant response during the SPR task and response during the DMS task. Response magnitude during the DMS task was similar to that during C2 period. E: relationship among responses during C1 and C2 periods of SPR task and during DMS task. A closed circle shows that response magnitude during the DMS task was not significantly different (P > 0.05) from that during C1 period of the SPR task and was significantly smaller (P < 0.05) than that during C2 period of the SPR task. A closed triangle shows that response magnitude during the DMS task was not significantly different (P > 0.05) from that during C2 period of the SPR task and was significantly larger (P < 0.05) than that during C1 period of the SPR task. An open circle shows that response magnitude during the DMS task was not significantly different (P > 0.05) from that during C1 and C2 periods of the SPR task.

Object-selective response during C1 and C2 periods in error trials

To evaluate the behavioral significance of neuronal activity, we examined neuronal activity when the monkeys incorrectly performed. Although we observed two types of error in the SPR task, breaking the fixation during the cue or delay period and making a saccade to an incorrect target object during the response period, we analyzed error trials in which the monkeys had to make a saccade to a preferred object but made a saccade to an incorrect object.

To evaluate the behavioral significance of the object-selective and order-selective responses during the C1 and C2 periods in encoding information regarding the object and order of presentation, we compared the response to the preferred object in the correct trials with that in the error trials. The order-nonselective response magnitudes in the correct trials were not significantly different from those in the error trials during both the C1 (Wilcoxon signed-rank test, z = 0.43, P > 0.05, mean = 31.87 spikes/s in correct trials, mean = 30.76 spikes/s in error trials) and the C2 (z = 1.61, P > 0.05, mean = 35.26 spikes/s in correct trials, mean = 33.73 spikes/s in error trials) periods (Fig. 7A). The C1-dominant response magnitudes in the correct trials were also not significantly different from those in the error trials during both the C1 (z = 0.64, P > 0.05, mean = 47.35 spikes/s in correct trials, mean = 46.68 spikes/s in error trials) and the C2 (z = 0.27, P > 0.05, mean = 30.88 spikes/s in correct trials, mean = 29.63 spikes/s in error trials) periods (Fig. 7B). C2-dominant response magnitude (Fig. 7C) was not significantly different between the correct and error trials during the C1 period (z = 1.01, P > 0.05, mean = 22.51 s/s in correct trials, mean = 21.00 spikes/s in error trials), but was significantly smaller in the error trials than in the correct trials during the C2 period (z = 2.50, P < 0.05, mean = 38.20 spikes/s in correct trials, mean = 31.48 s/s in error trials).

View larger version (29K): [in this window] [in a new window]   FIG. 7. Comparison of responses during correct trials and error trials. A: order-nonselective response during C1 and C2 periods in correct and error trials. Mean discharge rates during the C1 and C2 periods in correct and error trials were not significantly different. B: C1-dominant response in correct and error trials. Mean discharge rates in correct and error trials were not significantly different during C1 and C2 periods. C: C2-dominant response in correct and error trials. Mean discharge rates in correct and error trials were significantly different during C2 period but not significantly different during C1 period. Closed circles show discharge rate of neurons whose responses were significantly different between correct and error trials (Mann-Whitney U test, P < 0.05). Open circles show discharge rate of neurons whose responses were not significantly different between correct and error trials (P > 0.05).

Thirty-eight neurons showed the object-nonselective response. Of these, 17 neurons showed the order-nonselective response, and 21 neurons showed the C1-dominant response. No neurons showed the C2-dominant response.

To compare the temporal profiles between the object-selective and object-nonselective responses, we constructed the histograms of these population activities (Fig. 8A). We could not find differences in latency and duration between the object-selective and object-nonselective responses in the order-nonselective response. However, in the C1-dominant response, the latency of the object-selective response was slightly shorter than that of the object-nonselective response. Figure 8B shows the difference in response latency. Neurons with the object-selective and order-selective responses were activated early, and the latencies of the object-nonselective and order-nonselective responses were slightly longer.

View larger version (29K): [in this window] [in a new window]   FIG. 8. A: population activities of neurons with object-selective (black) and object-nonselective (gray) responses during C1 and C2 periods, and cumulative summation curve of latencies of object-selective and object-nonselective visual response. B: cumulative summation curve of latencies of 4 types of response.

One hundred eighty-three neurons were activated during the D1 and/or D2 period. Of these, 52 neurons exhibited object-nonselective delay-period activity during the D1 and/or D2 period. In the remaining 131 neurons, delay-period activity during the D1 and/or D2 period showed object selectivity. We classified these delay-period activities into three types: order-nonselective activity, C1-coding activity, and C2-coding activity.

Figure 9A shows the histograms of the order-nonselective delay-period activity of a neuron. The neuron exhibited delay-period activity during the D1 period when the C1 object was a circle (Fig. 9, A, trial conditions 5 and 6, and B). This delay-period activity during the D1 period was object-selective [F(2,142) = 73.29, P < 0.0001], and the activity returned to the baseline level after the appearance of the C2 object. During the D2 period, this neuron exhibited delay-period activity after the circle was presented as C2 (Fig. 9, A, trial conditions 2 and 4, and C). When the double cone or the cross was presented as C2, this neuron showed an increase in activity at the end of the D2 period (Fig. 9A, trial conditions 1, 3, 5, and 6), and discharge rates during the D2 period of these trials were not significantly different from those during the control period (Fig. 9C). The delay-period activity during the D2 period had selectivity [F(5,139) = 57.41, P < 0.0001). This neuron showed the highest delay-period activity during the D2 period in trial condition 4, and this delay-period activity was significantly different (P < 0.05) from those in trial conditions 1, 3, 5, and 6, and not significantly different (P > 0.05) from that in trial condition 2. This indicates that this delay-period activity during the D2 period was not selective in only one sequence. SI_C1 (0.41) was lower than SI_C2 (0.98). Therefore delay-period activity during the D2 period depended on the C2 object. Thus we classified neuronal activities depending on the degree of selectivity for both stimuli, when the delay-period activity during the D2 period was selective in terms of the C1 and C2 objects (see METHODS). In this neuron, delay-period activity was detected after the circle cue was presented as C1 or C2, and the delay-period activity during the D1 or D2 period returned to the baseline level when the C2 object or color cue was presented, respectively. Thus this delay-period activity was order-nonselective, and we defined the circle as the preferred object of this neuron. Thirty-eight neurons showed order-nonselective excitatory delay-period activity, and two neurons showed a significant decrease in activity during the delay periods.

View larger version (48K): [in this window] [in a new window]   FIG. 9. A: histograms of order-nonselective delay-period activity of a neuron. The neuron was activated during the D1 period under trial conditions 5 and 6 and during the D2 period under trial conditions 2 and 4; this neuron was activated after presentation of a circle cue regardless of order of presentation. Bin width is 25 ms. C3, color cue period; D3, 3rd delay period; R, response period. B: average discharge rates during D1 period. Error bar indicates SD. Dashed line indicates mean discharge rate during control period. Asterisk indicates significant increase in activity during delay period. C: average discharge rate during D2 period for 6 trial conditions. Dashed line indicates mean discharge rate during control period. Asterisks indicate significant increase in activity during delay period.

View larger version (37K): [in this window] [in a new window]   FIG. 10. A: histograms of C1-coding (D1) delay-period activity of a neuron. Format is the same as that in Fig. 9A. The neuron was activated during the D1 period under the trial conditions 1–4; this neuron was activated during D1 period when C1 object was the double cone or cross. B: average discharge rates during D1 period. C: average discharge rate during D2 period for 6 trial conditions. Format is same as that in Fig. 9, B and C.

View larger version (49K): [in this window] [in a new window]   FIG. 11. A: histograms of C1-coding (D2) delay-period activity of a neuron. Format is the same as that in Fig. 9A. The neuron was activated during D2 period under trial conditions 5 and 6; this neuron was activated during D2 period when C1 object was the circle. B: average discharge rates during D1 period. C: average discharge rate during D2 period for 6 trial conditions. Format is same as that in Fig. 9, B and C.

View larger version (36K): [in this window] [in a new window]   FIG. 12. A: histograms of C2-coding delay-period activity of a neuron. Format is the same as that in Fig. 9A. The neuron was activated during D2 period under trial conditions 1 and 6; this neuron was activated during D2 period when C2 object was the cross. B: average discharge rates during D1 period. C: average discharge rate during D2 period for 6 trial conditions. Format is same as that in Fig. 9, B and C.

Temporal profile of delay-period activities

To compare temporal profiles among order-nonselective, C1-coding, and C2-coding delay-period activities, we constructed the population histograms of these activities (Fig. 13, A–D). In these figures, red lines indicate population activities in trials in which the preferred object was presented during the C1 period, and green lines indicate population activities in trials in which the preferred object was presented during the C2 period. Blue lines indicate differences between the red and green lines. The starting point of difference in delay-period activity (the time at which the 1st 3 consecutive bins differed from the difference of activity for 1 s of the control period by >2 SD or <2 SD) and the endpoint (the time of the last bin) of the difference in delay-period activity were determined. For order-nonselective delay-period activity, when the preferred object was presented as C1, population activity was elicited at 125 ms after C1 presentation and continued during the delay period until 100 ms after C2 presentation (Fig. 13, A and E, red line). When the preferred object was presented as C2, population activity was elicited at 175 ms after C2 presentation and continued during the delay period until 100 ms after color cue presentation (Fig. 13, A, green line, and E, red line). For the C1-coding (D1) delay-period activity, a difference in population activity was first observed 275 ms after C1 presentation and continued until 200 ms after C2 presentation (Fig. 13, B, blue line, and E, green line). Although population activity showed delay-period activity during the D2 period, a difference could not be found during the D2 period. A significant difference in C1-coding (D2) delay-period activity could not also be found during the D1 period, and it was first observed 325 ms after C2 presentation, and during the D2 period, population activity increased toward the end of the delay period until 600 ms after color cue presentation (Fig. 13, C and E, blue lines). A significant difference in C2-coding delay-period activity could not be found during the D1 period, and it was first observed 275 ms after C2 presentation, and population activity increased toward the end of the delay period until 175 ms after color cue presentation (Fig. 13, D, blue line, and E, light blue line).

View larger version (41K): [in this window] [in a new window]   FIG. 13. Population histograms of order-nonselective delay-period activity (A), C1-coding (D1) delay-period activity (B), C1-coding (D2) delay-period activity (C), and C2-coding delay-period activity (D). Red lines indicate population histograms of trials in which preferred object was presented during C1 period, green lines indicate population histograms of trials in which preferred object was presented during C2 period, and blue lines indicate difference between red and green lines. E: time-course of difference in activity between trials in which preferred object was presented as C1 and C2. F: fixation period. C1, 1st cue period; D1, 1st delay period; C2, 2nd cue period; D2, 2nd delay period; C3, color cue period; D3, 3rd delay period; R, response period.

To evaluate the importance of order-nonselective and order-selective (C1-coding and C2-coding) delay-period activities in retaining information regarding the object cue and order of presentation, we examined neuronal activity during the DMS task, in which the monkey had to memorize only one object information during the delay period. Of 131 neurons with object-selective delay-period activity during the SPR task, 42 neurons were also tested for their activity during the DMS task.

Figure 14 shows two examples of neuronal activity during the DMS task. Figure 14A shows the activity during the DMS task of a neuron, whose activity during the SPR task is shown in Fig. 9. This neuron showed order-nonselective delay-period activity during the SPR task, and the preferred object was the circle. During the DMS task, this neuron exhibited excitatory delay-period activity (early 1 s, z = –4.51, P < 0.0001; last 1 s, z = –3.85, P < 0.0001) when the object cue was the circle. This delay-period activity was object-selective during both the first 1 s [F(2,54) = 20.83, P < 0.0001] and the last 1 s [F(2,54) = 4.39, P < 0.05] of the delay period.

View larger version (36K): [in this window] [in a new window]   FIG. 14. Neuronal activity during DMS task. A: activity during DMS task of a neuron, whose activity the SPR task is shown in Fig. 9. This neuron showed order-nonselective delay-period activity during the SPR task. Because duration of delay was randomly selected to be between 2 and 3 s, rasters and histograms are aligned at onset of cue presentation and at start of response period. This neuron exhibited delay-period activity when the object cue was the circle during the DMS task. B: activity during DMS task of a neuron, whose activity during the SPR task is shown in Fig. 11. This neuron showed C1-coding (D2) delay-period activity during the SPR task. This neuron exhibited delay-period activity during the DMS task, but delay-period activity was not object-selective. C: comparison of delay-period activities between preferred object and nonpreferred object trials during DMS task for neurons with order-nonselective delay-period activity during SPR task. Preferred object trial and nonpreferred object trial were determined by result of the SPR task. D: comparison of delay-period activities between preferred object and nonpreferred object trials during DMS task for neurons with order-selective delay-period activity during the SPR task. Preferred object trial and nonpreferred object trial were determined by result of the SPR task. For example, the neuron shown in B did not show object-selective delay-period activity during the DMS task. However, this neuron showed object-selective and C1-coding delay-period activity during the SPR task (see Fig. 11), and this neuron's preferred object was the circle and nonpreferred object was the double cone.

Nineteen neurons with order-nonselective delay-period activity were also tested during the DMS task. Of these, 18 (95%) neurons exhibited object-selective delay-period activity during the DMS task, and 1 (5%) neuron showed object-nonselective delay-period activity (Fig. 14C). Fifteen neurons with C1-coding delay-period activity and five neurons with C2-coding delay-period activity were tested during the DMS task. Of these, 2 (10%) neurons exhibited object-selective delay-period activity during the DMS task, 13 (65%) neurons showed object-nonselective delay-period activity, and 5 (25%) neurons did not exhibit delay-period activity (Fig. 14D).

Delay-period activity in error trials

To evaluate the importance of order-nonselective and order-selective (C1-coding and C2-coding) delay-period activities in retaining information regarding the object cue and order of presentation during the D1 and D2 periods, we compared discharge rates during the delay period between the correct trials and error trials in all neurons with object-selective delay-period activity (Fig. 15, A–D). In the neurons with order-nonselective delay-period activity, the delay-period activity was not significantly different between the correct and error trials during both the D1 (Fig. 15A; Wilcoxon signed-rank test, z = –0.24, P > 0.05, mean = 20.25 spikes/s in correct trials, mean = 20.60 spikes/s in error trials) and D2 (Fig. 15B; z = –0.98, P > 0.05, mean = 23.43 spikes/s in correct trials, mean = 24.23 spikes/s in error trials) periods. In contrast, neurons with C1-coding (Fig. 15C) and C2-coding (Fig. 15D) delay-period activities showed significantly weaker delay-period activities during the error trials than during the correct trials (C1-coding activity, z = –1.96, P < 0.05, mean = 23.19 spikes/s in correct trials, mean = 21.45 spikes/s in error trials; C2-coding activity, z = –2.23, P < 0.05, mean = 24.18 spikes/s in correct trials, mean = 18.04 spikes/s in error trials).

View larger version (33K): [in this window] [in a new window]   FIG. 15. Comparison between delay-period activities during correct and error trials. A: order-nonselective delay-period activity during D1 period in correct and error trials. Mean discharge rates during D1 period on correct and error trials were not significantly different. B: order-nonselective delay-period activities during D2 period in correct and error trials. Mean discharge rates during D2 period in correct and error trials were not significantly different. C: C1-coding delay-period activities during correct and error trials. Mean discharge rates during correct and error trials were significantly different. D: C2-coding delay-period activities during correct and error trials. Mean discharge rates during correct and error trials were significantly different. E: delay-period activities during correct and error trials in the DMS task. Mean discharge rates during correct and error trials were significantly different. Closed circles indicate discharge rate of neurons whose activities were significantly different between correct and error trials (Mann-Whitney U test, P < 0.05). Open circles indicate discharge rate of neurons whose activities were not significantly different between correct and error trials (P > 0.05).

Relationship between neuronal activities during C1/C2 periods and D1/D2 periods

To study the relationship between neuronal activities during the cue and delay periods, we compared the preferred object of individual neurons that had both visual and delay-period activities. During the SPR task, about one-half of the neurons with the object-selective response to the object cue also exhibited object-selective delay-period activity (Table 1). Among neurons with the order-nonselective object-selective response, eight (36%) also exhibited object-selective delay-period activity. Most of these neurons exhibited order-nonselective delay-period activity. Figure 16A shows an example of the activity of a neuron with both the order-nonselective response during the C1 and C2 periods and order-nonselective delay-period activity during the D1 and D2 periods. In this histogram, black lines indicate activity in trials in which C1 was the preferred object, and gray lines indicate activity in trials in which C2 was the preferred object. This neuron showed a response when the preferred object (circle) was presented and showed delay-period activity following the presentation of the preferred object. Among neurons with the C1-dominant object-selective response, eight (24%) also exhibited object-selective delay-period activity. Figure 16C shows an example of the activity of a neuron with both the C1-dominant response and C2-coding delay-period activity. Among neurons with the C2-dominant object-selective response, 13 (50%) also exhibited object-selective delay-period activity. These neurons exhibited order-nonselective, C1-coding (D2), C2-coding delay-period activity. However, for the majority of these neurons, object preferences were different between the object cue period and delay period. Figure 16E shows an example of the activity of a neuron with the C2-dominant response and C1-coding (D2) delay-period activity. In this neuron, the preferred object during the cue period was the double cone and the preferred object during the delay period was the circle.

View larger version (53K): [in this window] [in a new window]   FIG. 16. Histograms of activity of neurons with order-nonselective response (A), C1-dominant response (C), and C2-dominant response (E), and population activity histograms of neurons with order-nonselective response (B), C1-dominant response (D), and C2-dominant response (F). Black lines indicate activity histograms for trials in which the preferred object of the C1 period was presented, and gray lines indicate activity histograms for trials in which the preferred object of the C2 period was presented. F, fixation period; C1, 1st cue period; D1, 1st delay period; C2, 2nd cue period; D2, 2nd delay period; C3, color cue period.

To examine how many neurons showed anticipatory activity preceding the C2 presentation, we compared activity for 500 ms preceding the C2 presentation with activity for 500 ms preceding the C1 presentation by repeated measures ANOVA. In the neurons with the order-nonselective response, 10 (45%) neurons did not show a significant increase or decrease in activity during the pre-C2 period, 10 (45%) neurons showed a significant (P < 0.05) increase in activity during the pre-C2 period, and 2 (10%) neurons showed a significant (P < 0.05) decrease in activity during the pre-C2 period. In the neurons with the C1-dominant response, 17 (52%) neurons did not show a significant increase or decrease in activity during the pre-C2 period, 9 (27%) neurons showed a significant (P < 0.05) increase in activity during the pre-C2 period, and 7 (21%) neurons showed a significant (P < 0.05) decrease in activity during the pre-C2 period. In the neurons with the C2-dominant response, 10 (38%) neurons did not show a significant increase or decrease in activity during the pre-C2 period, 16 (62%) neurons showed a significant (P < 0.05) increase in activity during the pre-C2 period, and no neuron showed a significant (P < 0.05) decrease in activity