Hoshi, Eiji, Keisetsu Shima, and Jun Tanji. Task-dependent selectivity of movement-related neuronal activity in the primate prefrontal cortex. J. Neurophysiol. 80: 3392–3397, 1998. We studied movement-related neuronal activity in the dorsolateral prefrontal cortex from the perspective of a general role for the prefrontal cortex in controlling motor behavior to achieve a specific goal according to the requirements of a given task. Monkeys were trained to perform two delayed motor tasks. The first task involved reaching for a target that matched the shape of a cue. The second task involved reaching for a target that matched the location of the cue. A majority (54%) of 175 movement-related prefrontal neurons exhibited preference for either the target shape or the type of task requirements. Sixty-four neurons (36%) were selectively active while reaching for a circle or a triangle. On the other hand, the activity of 59 neurons (34%) depended on whether the task required matching the shape or the location. These properties, characterizing the movement-related neuronal activity in the prefrontal cortex, rarely were found in the arm area of the primary motor cortex. Only 1 of 130 movement-related neurons (0.8%) showed task selectivity, and none showed target-shape selectivity.
Many lines of evidence support the view that the dorsolateral prefrontal cortex (PF) plays a prominent role in controlling motor behavior by making use of currently available information in the context of learned behavioral requirements that are stored in memory (Fuster 1997; Goldman-Rakic 1987; Passingham 1993). In accordance with this point of view, the activity of neurons in the PF has been studied extensively to look at both neuronal responses to sensory cues prompting a variety of motor behaviors (Watanabe 1986) and the activity of neurons during the delay required for the short-term storage of sensory information (Fuster and Alexander 1971; Miller et al. 1996). However, neuronal activity closely tied to motor performance (movement-related activity) has not been studied sufficiently, although it is known that movement-related activity in the PF reflects the direction or spatial location of movements (Kubota and Funahashi 1982; Quintana et al. 1988). Because neuronal activities related to these aspects of motor behavior have been well described for the motor areas of the frontal cortex (Georgopoulos et al. 1982; Wise 1985), it remains to be determined how behavioral factors other than motor parameters characterize movement-related activity in the PF. As for oculomotor activity, PF neurons were found to exhibit selective activity depending on whether saccades were performed in a delayed-response task or in a visually guided task (Funahashi et al. 1991). However, it is not known whether limb-movement related activity in the PF reflect the nature of motor target or the task requirements to use specific aspects of sensory information. In a series of experiments in our laboratory, we attempted to study movement-related neuronal activity from the perspective of the general role of the PF in controlling skeletomotor behavior conforming to task requirements. We show that the PF neurons exhibit target-shape selectivity and also exhibit motor-task selectivity depending on whether the cue location or cue shape is the crucial information. These properties rarely are found in the primary motor cortex.
We trained two male Japanese monkeys to select a target to reach for under two different sets of conditions and recorded the neuronal activity in the dorsolateral PF and primary motor cortex. The animals were cared for according to the Guidelines for the Care and Use of Laboratory Animals of the National Institutes of Health. During recording sessions, each subject sat in a monkey chair with its head and left arm restrained. We installed a 14-in color monitor screen equipped with a touch-sensitive front panel in front of the subject, which the monkey could reach with its right arm. Each task started with a blank screen for a 3-s intertrial interval (ITI). Subsequently, three cue frames measuring 8 × 8 cm appeared at the top, bottom left, and bottom right of the screen (Fig. 1 A). When the subject pressed a hold button in front of the chair with its right hand for 0.5 s, a red sample cue, either a circle or a triangle, appeared in one of the three frames for 1 s. The diameter of the circle was 4.5 cm, and the length of the sides of the triangle was 5 cm. The sample cue disappeared and only the cue frames remained visible for a 3-s delay period. After this delay, a red choice cue appeared. There were two different sets of cues, each one required that a different task be performed. If the choice cue was a combination of a triangle and a circle, the subject had to select an object with the same shape as the sample cue (shape-matching task). On the other hand, if the choice cue was either three triangles (after a triangle sample cue) or three circles (after a circle sample cue), the subject had to select the triangle or circle that was in the same location as the sample cue (location-matching task). If the subject continued to press the hold button for another 1.5 s, the color of the choice cue changed from red to green. This was the go signal and the subjects then had to release the hold key and press the correct triangle or circle. Shape- and location-matching tasks were interlaced randomly. In addition to these standard tasks, we sometimes introduced an additional task in which the reaching movements were guided simply by a single-object cue. This analysis was performed in about two-thirds of cases when we found a task-related neuron. These behavioral tasks were controlled by a laboratory computer (PC-9821AP, NEC).
We recorded single neuron activity with conventional electrophysiological techniques (Shima et al. 1991). Before starting to record from the dorsolateral PF, we mapped the frontal eye field (FEF) in the anterior bank of the arcuate sulcus with intracortical microstimulation (ICMS) (cf. Bruce et al. 1985). We penetrated the PF around the principal sulcus and inferior convexity anterior to the FEF where ICMS did not evoke eye movements with currents <50 μA. We also made recordings from the arm area of the primary motor cortex (M1) that we identified by ICMS (Sato and Tanji 1989).
Electromyographic (EMG) monitoring of activity from the arm, shoulder, neck, and paravertebral muscles revealed that muscle activity differed depending on the location of the target for which the subject reached. However, neither the type of task nor the target shape produced any detectable differences in muscle activity. We recorded the position and movement of the eye with an infrared monitoring system with a 0.5° and 10-ms resolution in 10 recording sessions. Eye movements were calibrated at each session before the animal started the behavioral task. The animals made saccades to the reaching target a few hundred milliseconds before the onset of the go signal. During the arm-reach to targets, the eyes were fixated on the reaching target for a period of ∼1 s. The eye movements did not vary with the type of task or the target shape.
To analyze the neuronal activity, we first divided the data file according to trials with six different sample cues and further sorted them depending on whether the task was to match the location or the shape. Thus we obtained 6 × 2 data files for each neuron and constructed raster displays and histograms for individual files. To define the movement-related activity, we used a standard time window that started when the go signal appeared and ended when the target was touched. Neuronal activity was defined as movement related when the number of discharges in at least four successive 20-ms bins of the histogram during the time window deviated from the mean value during a control period by >2 SD. The control period was the last 500 ms of the ITI period. For 12 M1 neurons, the control period was taken during the first 500 ms of the delay period because of elevated activity during the ITI. In one of the monkeys, recording sites were verified by histological analysis.
We recorded the activity of 384 task-related neurons from the dorsolateral PF around the principal sulcus (Fig. 3, B and C). Of these, 175 met the criteria for movement-related neurons given in methods. In this report, we do not deal with 204 neurons that responded to the appearance of the sample cue or 132 neurons exhibiting long-lasting activity during the delay period. We focus here on two salient properties of movement-related neuronal activity. The first was preferential activity that depended on whether the target was the circle or the triangle. A representative neuron with this property is shown in Fig. 1 B. This neuron was significantly more active (ANOVA, P < 0.01) when the subject was reaching for a circle (Fig. 1 Ba) than when reaching for a triangle (Fig. 1 Bb), although in both cases the target was in the bottom left of the screen. The same neuron showed similar activity when the animal reached for the circle in the shape-matching task (Fig. 1 Bc). We call this property the “target-shape selectivity” because the shape of the target is the primary determinant of the neuronal activity. The second finding was differences in movement-related activity that depended on the type of task. Typical examples of two prefrontal neurons with this property are shown in Fig. 2. The neuron shown in Fig. 2 A was clearly active when reaching for the top target in the shape-matching task (a) but less active when reaching for the same target in the location-matching task (b). In contrast, the prefrontal neuron shown in Fig. 2 B was preferentially more active when the subject was reaching for the bottom-left target in the location-matching task (a) than in the shape-matching task (b). Furthermore the same neuron was not very active when the animal reached for the bottom-left target if the reach was guided by a single object (Fig. 2 Bc: a simple reaction to a visual cue). We call this property the “type of task selectivity” because the type of task was the prime determinant of changes in neuronal activity.
To systematically determine which of these two properties each of the 175 movement-related prefrontal neurons had, we first determined the direction of target-reaching that gave rise to the most prominent activity for each neuron. Subsequently, we generated four data files for each neuronal data set in a 2 × 2 matrix according to the type of task and the target shape, as shown in Fig. 2 C. We quantified the movement-related neuronal activity by counting the numbers of discharges that appeared after the go-signal onset and before the animal touched the target in individual trials, thereby obtaining trial-by-trial data. Using the matrix for each neuron, we performed a two-way analysis of variance (ANOVA) looking at the type of task and the target shape. The neuron shown in Fig. 2 C was most active when the subject was reaching for a triangle in the bottom right of the screen during the shape-matching task. The activity showed significant effects for both target shape (P < 0.01) and type of task (P < 0.01). We performed ANOVA for 175 neurons, and a majority of prefrontal neurons exhibited a type of task effect (significant at P < 0.05 in 59 neurons), a target shape effect (64), or both (29) while reaching for a target in a particular location (Table 1). The movement-related activity was unaffected by either of these factors in a minority of 81 neurons (41%). Among those neurons having no selectivity for the target-shape or the task-type, some neurons showed selectivity for the location of a target. Therefore we analyzed the target-location selectivity of the 81 neurons by performing ANOVA. We found that 38 neurons had the target-location effect (P < 0.05). Furthermore some of the PF neurons having either the target shape effect or type of task effect also exhibited the target-location effect. We will describe the detailed account of the relationship between the task-dependent selectivity and the target location effect for these neurons in a separate paper. Of 175 movement-related neurons, 35 (20%) exhibited activity during the delay period, and 81 (46%) responded briefly to the sample cue. The relationship of these activities and the movement-related activity will be a subject of a separate paper.
All movement-related neurons were recorded from the principal sulcus (PS) and the inferior prefrontal convexity (IC) regions (Fig. 3 B). Most of movement related neurons were found in the ventral bank of the principal sulcus and IC. Recording sites of three neurons documented in Figs. 1 B and 2, A and B, are shown in Fig. 3 C (coronal sections). These neurons were found in the ventral bank of the principal sulcus.
We performed the same two-way ANOVA analysis for the 130 movement-related neurons recorded in the arm area of the primary motor cortex. Only one neuron exhibited the type of task effect. A typical example of an M1 neuron exhibiting nonselective movement-related activity is shown in Fig. 3 A.
In this study, we identified two novel aspects of movement-related neuronal activity in the dorsolateral PF. First, we found that the activity differed significantly depending on whether the subject was reaching for a circle or a triangle. Second, we found that the activity also depended on whether the animals were required to select a target on the basis of its location or shape. These properties do not seem to be ascribable to any aspect of the movement itself because no differences in the limb or eye movements resulted from the target shape or the type of task. Instead, these observations seem to suggest that the neuronal activity reflects behavioral factors such as the goal or concept of the motor activity.
Previous studies of movement-related activity in the PF have shown a relationship between the activity and the direction or spatial location of a target to be captured by the limb (Kubota and Funahashi 1982; Quintana et al. 1988). Niki (1974) had reported earlier that the prefrontal neurons also carry information about the relative spatial location of targets. Subsequently, saccade-related PF activity was found to be selective during a delayed-response task as opposed to a visually triggered task (Funahashi et al. 1991). Further, the saccade-related activity appeared stronger in an anti-saccade delayed task (Fig. 2 in Funahashi et al. 1993). These studies revealed that the oculomotor related activity in the PF differ depending on whether the motor task requires memorized information or requires usage of memorized information to suppress as well as prescribe a response. The present study extends these findings and indicates that the neuronal activity expresses a much broader range of information about motor behavior. This information may be useful in constructing a motor output that conforms to a specific behavioral goal or requirement, or conversely, prefrontal neuronal activity based on a variety of information may be useful in monitoring the performance of motor behavior within a behavioral context (Petrides 1991). Although the majority of movement-related neurons found in the present task was not active during the sample cue or delay periods, ∼50% of them did show cue- or delay-related activity. Thus these neurons may be involved in short-term working memory (Fuster 1997; Goldman-Rakic 1987) in the context of sensory-motor integrative function (Quintana and Fuster 1992). We mainly encountered movement-related neurons in the ventral bank of the principal sulcus, where there are extensive connections with the inferior temporal cortex. It seems important to note that, in addition to the target-shape and task selectivity, we also found selectivity to the location of the target in some of the movement-related neurons. The target location selectivity was found ∼20% of the 175 movement-related neurons. This proportion to corresponds to 19% of inferior convexity neurons that showed direction selectivity during an oculomotor delay period (Wilson et al. 1993). These findings are consistent with a view that visuomotor cognitive processing occurs in the PF (Quintana and Fuster 1992).
Our study also showed that the movement-related neuronal activity in the primary motor cortex is related primarily to the direction of the target-reaching movement. As far as the present task conditions are concerned, the wealth of information expressed in the neuronal activity in the PF is scarcely carried over into the primary motor cortex. In the course of its transmission through neural pathways, it seems that this information is transformed largely into signals encoding motor output, although this finding does not deny a possible participation of the M1 in cognitive aspects in other behavioral conditions (Georgopoulos 1994).
We thank K. Kurata for advice in setting up our equipment, H. Mushiake for valuable discussion, and M. Kurama and Y. Takahashi for technical assistance.
This work was supported by Ministry of Education, Science, and Culture of Japan Grants 08279101 and 09308032.
Address for reprint requests: J. Tanji, Dept. of Physiology, Tohoku University School of Medicine, Sendai 980–8575, Japan.
- Copyright © 1998 the American Physiological Society