Shima, Keisetsu and Jun Tanji. Both supplementary and presupplementary motor areas are crucial for the temporal organization of multiple movements. J. Neurophysiol. 80: 3247–3260, 1998. To study the involvement of the supplementary (SMA) and presupplementary (pre-SMA) motor areas in performing sequential multiple movements that are individually separated in time, we injected muscimol, a γ-aminobutyric acid agonist, bilaterally into the part of each area that represents the forelimb. Two monkeys were trained to perform three different movements, separated by a waiting time, in four or six different orders. First, each series of movements was learned during five trials guided by visual signals that indicated the correct movements. The monkeys subsequently executed the three movements in the memorized order, without the visual signals. After the injection of muscimol (3 μl, 5 μg/μl in 10 min) into either the SMA or pre-SMA bilaterally, the animals started making errors in performing the sequence of movements correctly from memory. However, when guided with a visual signal, they could select and perform the three movements correctly. The impaired memory-based sequencing of movements worsened progressively with time until the animals could not perform the task. Yet they still could associate the visual signals with the different movements at that stage. In control experiments on two separate monkeys, we found that injections of the same amount of muscimol into either the SMA or pre-SMA did not cause problems with nonsequential reaching movement regardless of whether it was visually triggered or self-initiated. These results support the view that both the SMA and pre-SMA are crucially involved in sequencing multiple movements over time.
The involvement of cerebral motor areas in the sequential limb movement of primates recently has been the subject of numerous studies (Barone and Joseph 1989; Boecker et al. 1998; Gerloff et al. 1997; Harrington and Haaland 1992; Jenkins et al. 1994; Kermadi and Joseph 1995; Mushiake and Strick 1995; Mushiake et al. 1990, 1991; Sadato et al. 1996). In these reports, the central issue has been the mechanisms controlling sequential and continuous movements with multiple spatial components. A related, but separate, issue is how to arrange multiple, separately performed movements in the correct temporal order. In daily life, it is critically important to perform multiple, discrete movements in a proper sequence to achieve a certain behavioral goal. Previous reports from our laboratory have shown that neurons in the supplementary motor area (SMA) and the presupplementary motor area (pre-SMA) exhibit a variety of activities that are useful for processing information for the purpose of temporally sequencing multiple movements (Shima and Tanji 1994; Tanji and Shima 1994). The properties of neuronal activity in these areas suggest that they are particularly prominent in retrieving the memorized information necessary for organizing the temporal sequence of future movements. Subsequently it also was reported that neuronal activity in the pre-SMA suggests that it plays a role in updating the information required to perform the upcoming motor sequence (Shima et al. 1996). If the participation of these neurons is crucial for sequencing multiple movements over time, then inactivation of these areas should impair behavior requiring the correct planning and performance of multiple movements based on memorized information. The primary aim of this study is to test this hypothesis by temporarily inactivating a substantial number of neurons in either the SMA or pre-SMA with the local application of muscimol, a γ-aminobutyric acid agonist. We show that bilateral injection of muscimol into either the SMA or pre-SMA profoundly affects the temporal sequencing of three different movements based on memory, whereas the ability to associate visual signals with the same three movements is not affected. In control experiments, we also show that inactivation of either area by the same technique does not alter the reaction or motor time in simple visually triggered or self-initiated motor tasks. A preliminary account of this study appeared in abstract form (Shima and Tanji 1997).
We trained four monkeys (Macaca fuscata) to perform the two different motor tasks described in the following text. The animals were cared for in accordance with the National Institute of Health's Guide for the Care and Use of Laboratory Animals, and the Guidelines for Institutional Animal Care and Use published by our Institute. In the main part of the present series of experiments, we trained two monkeys to perform three movements (push, pull, or turn a manipulandum) in four (for the 1st monkey) or six (for the 2nd monkey) different orders with their right arms. The monkeys sat in a primate chair and were required to place a manipulandum in a neutral position and wait 2.5–4.5 s for the first movement-triggering signal (high-pitched tone). When the animal performed the first movement, a mechanical device returned the manipulandum to the neutral position. While keeping the manipulandum in this position, the animal had to wait 1–1.4 s for each of the second and third movement-triggering signals. A series of three correct movements was rewarded with applesauce, 500 ms later. The average time interval between motor sequences was 7–9 s. To teach the monkey the sequence, the correct movement was initially indicated with green (for turn), red (for push), or yellow (for pull) lights. The lights came on individually at the time for each movement, together with the movement-triggering tone signal. The animal had to learn the correct sequence after only five visually guided trials, after which the sequential motor task was performed from memory. For the memory-guided trials, only the tone signal was given as the movement trigger, without any lights. After completing the memorized sequential task six times, randomly flashing lights for 2 s signaled the end of the current sequence and the beginning of the next. After the flashing lights, there was a waiting period of 2.5–4.5 s before the next trial began. Thus a particular sequence of movements was performed in blocks. Each block consisted of 11 trials for a given sequence of the three movements, 5 trials with visual guidance, and 6 with no visual cues, followed by the next block with a different sequence of movements. The order in which different sequences were used was varied unpredictably. At least two blocks of trials for each sequence of movements were included in the data file while recording from each individual cell.
Electromyographic (EMG) recordings were made from muscles in the digits, wrist, elbow, shoulder, neck, chest, abdomen, thigh, and paravertebral muscles. EMG analysis showed that the activity of the forelimb muscles changed briefly during the execution of individual movements but not during the period when the animals were waiting for the next movement-triggering signal. Although the area examined did not include the supplementary eye field (Mushiake et al. 1996; Schlag and Schlag-Rey 1987), vertical and horizontal eye movements also were monitored electrooculographically with a resolution of 2° at 20° from the primary position of the eye.
In the second series of experiments, we trained two monkeys to perform a reaching task under two different conditions. The task began when the monkey pressed a key in front of the monkey chair. In 75% of the trials, an LED that was installed inside a push button on the panel in front of the animal was illuminated after a waiting period of 2.5–4.5 s. This was the go signal to release the key and press the push button with the right hand. The monkey was rewarded if it pressed the push button within 800 ms of the visual trigger (visually triggered task). In the remaining 25% of the trials, no trigger signals were given. In this case, the animal had to keep pressing the hold key in front of the chair for ≥4.5 s, after which it was rewarded if it reached and pressed the push button (self-initiated task).
Surgery, neuronal recording, and muscimol microinjection
After completing the behavioral training, an acrylic recording chamber (40 × 30 mm) and head fixation bolts were implanted on the skull under aseptic conditions. Markers indicating reference points using Horseley-Clarke's stereotactic coordinates also were placed on the skull. Anesthesia in the monkeys was induced with ketamine hydrochloride (8 mg/kg im) with atropine sulfate, followed by pentobarbital sodium (30 mg/kg im). Antibiotics and analgesics were used to prevent postsurgical infection and pain. Standard electrophysiological techniques for single-cell recording were used (Matsuzaka and Tanji 1996; Shima et al. 1991) to record from the left pre-SMA, SMA, and the MI (primary motor cortex). After complete recovery from the surgery, neuronal activity was recorded in the medial part of the frontal cortex, using glass-insulated Elgiloy microelectrodes, which were inserted through the dura. The same microelectrodes were used for intracortical microstimulation (ICMS). Each ICMS consisted of a train of either 11 or 22 cathodal pulses of 0.2-ms duration at 333 Hz, in the range of 10–50 μA. First, we mapped the medial frontal cortex to identify the SMA and rostrally adjacent pre-SMA physiologically, by examining neuronal responses to somatosensory and visual stimuli and by observing movements evoked with the ICMS. Neuronal responses were observed when the monkey was sitting quietly in the monkey chair. We examined cutaneous responses by brushing parts of the body with a camel hair brush and by manipulating limb joints. We also used flashing lights and moving objects (pieces of food, tools, and small items in the laboratory) to examine visual responses. We used previously established criteria to differentiate the SMA and pre-SMA (Matsuzaka et al. 1992) on the basis of neuronal responses and the effects of ICMS. After mapping these regions, we recorded neuronal activity while the monkey performed the trained motor tasks. We will report the properties of neuronal activity in the SMA and pre-SMA in relation to the tasks in a separate paper. After recording the neuronal activity from the SMA and pre-SMA bilaterally, 3.0 μl of muscimol was injected into either the SMA bilaterally or the pre-SMA bilaterally. For the injection, we prepared muscimol (Sigma, concentration: 5 μg/μl) dissolved in 0.1 M phosphate buffer at pH 7.4. Using an electronic microinjection pump equipped with dual ejection tubes (Nihon-Kohden, CFV-3100), muscimol was injected at a rate of 0.1 μl/min through a pair of thin (ID and OD: 150 and 300 μm, respectively) stainless steel tubes that were inserted simultaneously into the right and left hemispheres. The tips of the steel-tubes were sharpened so that they would penetrate the dura. A millimeter scale was visible on the surface of the injection tube at the time of insertion and indicated the distance by which the injection tube penetrated the dura. A thin (50–80 μm shaft diameter) Elgiloy microelectrode was attached to the side of the injection tube so that its tip was within 0.5 mm of the tip of the injection tube. The microelectrode told us the approximate depth to which the injection tubes penetrated the cortex and when the muscimol silenced neuronal activity. The steel tubes were positioned in the parts of the SMA or pre-SMA where we recorded a large amount of motor-task related neuronal activity. We used the same hydraulic microdrive used for the recording microelectrodes to position the injection tubes, so that we could inject muscimol at known stereotactic coordinates.
Collection and analysis of behavioral data
In each daily session, a single, simultaneous injection of muscimol was made in either the SMA or pre-SMA bilaterally. First, the injection tubes were inserted into each hemisphere. Then, the monkey's behavior was recorded for three blocks of trials without a muscimol injection. Subsequently, 3 μl of muscimol was injected through each tube into the cerebral cortex. Behavioral events were recorded during and after the muscimol injection. Throughout the pre- and postinjection periods, we measured the success and error rates, the reaction time (RT), and the movement time (MT). The RT was defined as the interval between the beginning of the auditory go signal and the onset of movement (the time when the handle moved out of the hold range in the correct direction for the required movement). The MT was defined as the interval from the onset of movement until the required motor target was attained. In the first series of experiments, target attainment occurred when the handle reached the goal set for each of the three movements. In the control task, target attainment was defined as the time when the button was pressed, and RT was the interval between the illumination of the visual go signal and the onset of movement (the time when the hold key was released). Statistically, these variables were compared using the Mann-Whitney U test or Student's t-test to examine the effects of muscimol injection.
After collecting the neuronal data and studying the effect of muscimol injections, the monkeys were anesthetized deeply with pentobarbital (50 mg/kg im) and perfused through the heart with saline. This was followed by a fixative, containing 3.7% formaldehyde in 0.1 M phosphate buffer at pH 7.4, and then 10 and 20% sucrose solutions in the same buffer. After marking the location of the recording chamber at known electrode coordinates, the brain was removed from the skull and photographed. Then it was sectioned at 50 μm intervals in the frontal plane on a freezing microtome for histological reconstruction of the neuronal recording and muscimol injection sites, with reference to cortical markings made by passing a current through the recording electrodes enough to produce iron deposit (attaining a charge of 300 μC).
Performance before muscimol injection
After being trained, both monkeys made only a small number of errors. The success rates for the trials under visual guidance in the two monkeys were between 98 and 99%. When the three movements were performed sequentially by memory, the success rates for individual trials were 95 ± 3 and 96 ± 4% (means ± SD) in the first and second monkey, respectively. If the animals did make a mistake, they usually succeeded in performing the next trial. The occurrence of two consecutive error trials was extremely rare. A representative example of the time course for performing tasks in successive trials by the second monkey is shown in Fig. 1, in which the movement-triggering signals and the onsets of the three movements are shown for individual trials. EMG analysis revealed that activity in the forelimb muscles increased phasically with the execution of the three movements. The phasic movement-related activity of all the muscles examined did not vary with the temporal sequence of the three movements. Representative muscle activity in four wrist muscles is shown in Fig. 2. We confirmed that the occurrence of saccades was not related to any of the events in the behavioral task.
During the visually triggered reaching task, the monkeys rarely made mistakes (>99% success rate). When required to wait a minimum of 4.5 s and initiate the reaching movement without an external trigger signal, the occurrence of premature reaching was 5.2 ± 2.0 and 4.8 ± 2.3% in the two monkeys.
Location of muscimol microinjection
In the first monkey, muscimol was injected at 10 sites in the SMA and the lateral adjacent area and at 6 sites in the pre-SMA. In the second monkey, microinjections were made at 12 sites in the SMA and the lateral adjacent area and at 14 sites in the pre-SMA and the rostral adjacent area. Figure 3 shows the location of the injection sites along with the location of forelimb areas in the SMA and pre-SMA. These were determined using previously established criteria (Matzusaka et al. 1992): neuronal responses to somatosensory and visual stimuli, effects of ICMS, and relationship to the trained motor task. The microinjections were always bilateral, using a pair of injection tubes inserted into symmetrical points in either the SMA or pre-SMA.
Effects of muscimol microinjection
An effect appeared 21–41 min after the injection of 3 μl of muscimol into the SMA. The first sign of an effect was an increase in the number of errors (incorrect movements) while performing the memory-guided sequential task. The number of errors increased progressively so that the animals began to repeat errors in successive trials. A typical example of task performance after a 3-μl muscimol injection into the SMA in the first monkey is shown in Fig. 4, where the performance 45 min after the injection is displayed serially. In contrast to the frequent errors in performing the memory-guided sequential movements, the task was performed correctly under visual guidance. When individual movements were signaled with an LED, the task was performed with few errors. The memory-guided sequential performance progressively deteriorated until errors were committed continuously. When that occurred, the experimenter had to assist the monkey in performing the task by giving visual signals to tell it which movement to select. When so guided, the monkey could perform the task correctly (blue tick marks in Fig. 4). However, when the visual signals were subsequently not shown, the animal again committed errors frequently. Eventually, the monkey stopped, unless it was assisted with the visual signals. Essentially the same observations were made in the second monkey. The ability of the second monkey to perform the task before and after a 3-μl muscimol injection into the SMA is shown in Fig. 5.
When 3 μl of muscimol was injected into the pre-SMA, its effects appeared 11–33 min after the injection when the monkeys started to commit errors repeatedly. The performance of the first and second monkeys is shown in Figs. 6 and 7. As with after the SMA injection, the monkeys made errors when required to perform the memory-based task, although they could perform the task without difficulty when the visual signals were given.
COMPARING THE EFFECTS OF SMA AND PRE-SMA INJECTION.
Five measures were used to compare the effects of muscimol injection into the SMA and pre-SMA. First, the delay until the onset of effects following the injection was compared. This was called the latency and defined as the time between the completion of the 3 μl muscimol injection and the time when the monkey made two consecutive mistakes or three mistakes in a block of six memory-guided trials. By analyzing the data obtained for all the effective points in the two monkeys, the latencies after the SMA and pre-SMA injection were 28 ± 13 and 14 ± 7 min, respectively (significantly different at P < 0.01 by the Mann-Whitney U test). Second, the rate of errors was compared. Examples of error rates (the percentage of error trials to the total trials) calculated for the two most effective sites in the SMA and pre-SMA of the two monkeys are shown in Fig. 8. As shown, the error rates during the visually guided task were small, even after the muscimol injection. In contrast, during the memory-guided task, the error rates increased progressively after the muscimol injection. The increase was already significant 15–30 min after completing the injection. In the data shown in Fig. 8, the error rates were higher after the pre-SMA injection than after the SMA injection. However, this difference was not significant when the data for all the effective injection sites were statistically analyzed. Third, the monkey's RT and MT after the injection were compared. Examples of the RT and MT after the most effective injections in the SMA and pre-SMA in the first monkey are shown in Figs. 9 and 10. During the memory-guided task (in this particular example, the sequence was push-pull-turn), the RT was slightly longer after the SMA injection but markedly longer after the pre-SMA injection. The RT was not longer in the visually guided task. The MT was not altered after injections into any sites in the SMA or pre-SMA. The same trend was observed performing the motor task with different sequences. Essentially the same observations were made in the second monkey. The RT during the 31–60 min after the muscimol injection into SMA and pre-SMA was compared statistically using the data obtained for all the effective injection sites in the two monkeys (the RT data for push-pull-turn trials were used). When all the data from the two monkeys were combined, the RT was not significantly longer after the SMA injections (P > 0.05, compared with the preinjection data) but was significantly longer after the pre-SMA injections (P < 0.01 by the Mann-Whitney U test). Fourth, the error rate during the first trial in each block of six trials in the memory-guided task was examined (sequence-update errors). With this, we determined whether SMA or pre-SMA injections preferentially impaired performing a new sequence. When the data for all the effective sites in the two monkeys were compared, the error rate for the new trials during the 31–60 min was 69 ± 23% after the pre-SMA injection (compared with an error rate of 34 ± 11% in the remaining trials) and 41 ± 12% after the SMA injection (36 ± 15% in the remaining trials). The probability of errors occurring in the new trial is significantly different in the two areas (P < 0.05, χ2 test). Thus the sequence-update errors were higher after the pre-SMA injections. Fifth, monkeys at times committed a particular type of error that was characterized as repeating the same error. Before the muscimol injection, when the monkeys made an incorrect movement, this resulted in an error signal, and the monkey selected a different movement in the next trial. After the injection, the monkeys kept selecting a particular wrong movement in successive trials despite the error signal. This was called perseverance error. The perseverance error constituted 27% of total errors after SMA injection and 30% after pre-SMA injection. We compared the difference in the relative occurrence rate of perseverance errors to the total errors using the data after effective injections in the pre-SMA and SMA. The difference was not significant (P > 0.05, χ2 test).
To examine whether muscimol injection into the SMA and pre-SMA affects the performance of simple motor tasks that do not require a sequence of multiple movements, the injection effects were examined while performing reaching movements under two conditions, visually triggered and self-initiated. After physiological identification of the forelimb areas of the SMA and pre-SMA in the two monkeys, 3 μl of muscimol was injected bilaterally in the same manner as in the sequencing-task performing monkeys. Eight injection sites were selected in the right and left SMA and pre-SMA of each monkey (Fig. 11). As exemplified by the data shown in Fig. 12 (A and B, left), the RT and MT analyzed during the visually triggered reaching were not altered after the muscimol injection into either the SMA or pre-SMA. This was confirmed by repeating the same procedures in the two monkeys. When the monkeys performed the self-initiated reaching movements, the MT and the time interval during which the animals waited before initiating the reaching movement (waiting time, with a required minimum of 4.5 s) were analyzed. As shown in Fig. 12 (A and B, right), neither the movement time nor the waiting time was altered by the muscimol injection. This also was confirmed by repeating bilateral injections in either the SMA or pre-SMA of the two monkeys. The occurrence of premature reaching did not increase after the muscimol injection.
ADDITIONAL INJECTION EXPERIMENTS.
To obtain information on the spatial extent of effective concentration of muscimol by diffusion, we performed an additional series of control experiments. For this purpose, we constructed an apparatus with which we could insert both the muscimol injection tube and a microelectrode at the same time, separated by a horizontal distance of 1, 2, or 3 mm. We found that the discharge activity of neurons recorded with the electrodes at distances of 1 or 2 mm was silenced, whereas neurons separated by 3 mm continued to be active. Therefore we estimate that neurons located within a 2-mm radius of the injection tube were inactivated but that neurons ≥3 mm away were not much affected. It is possible that the caudal-most injection in the pre-SMA may have inactivated part of the SMA, and, conversely, the rostral-most injection in the SMA may have affected part of the pre-SMA. However, the injections at the most effective sites in the pre-SMA (labeled with arrowheads in Fig. 3) are not likely to have affected a substantial part of the forelimb area of the SMA and vice versa.
In this study we found that transient inactivation of either the SMA or pre-SMA by microinjection of muscimol profoundly impairs the ability to perform a memorized sequence of multiple movements correctly. In contrast, the injection of the same amount of muscimol into either area did not cause any problems when the animals performed simple motor tasks regardless of whether the task was visually triggered or self-initiated. Therefore the impairment was not due to difficulties in performing the limb movement itself or in beginning a self-initiated movement. Furthermore after the muscimol injection, even when the monkeys kept committing errors in memory-guided sequence of movements, they still could associate visual signals with specific movements. These findings suggest that both the SMA and pre-SMA are involved crucially in arranging multiple movements in correct temporal order according to memorized information. The present findings also show that the effects of inactivating the SMA and pre-SMA are dissimilar in some respects, suggesting that the impairments in performing this task with SMA and pre-SMA dysfunction have, at least in part, different causes. On the other hand, in the present study, we injected 3 μl of muscimol into the pre-SMA or SMA at single injection sites in each hemisphere. This inactivated a part of each area. There was the possibility that more widespread inactivation of either area caused more deficits, which were not observed in this study.
Implications for the role of the SMA in different aspects of motor behavior
Previous studies in subhuman primates have suggested that the SMA has a role in planning or controlling sequential movements that have multiple spatial components (Brinkman 1984; Mushiake et al. 1990, 1991), in agreement with reports on human subjects (Boecker et al. 1998; Jenkins et al. 1994; Laplane et al. 1977; Roland et al. 1980; Sadato et al. 1996; Vermersch et al. 1994). In addition, recent reports from our laboratory have revealed neuronal activity in the SMA that seems to be useful in sequencing multiple movements that are performed at different times (Shima and Tanji 1994; Tanji and Shima 1994). The present study provides evidence that the SMA is involved crucially in motor tasks that require the arranging of multiple movements in the correct temporal order. The effects of bilateral dysfunction are so devastating that the monkeys eventually had to stop performing the task. We also showed that the dysfunction had little effect on the nonsequential performance of limb movements. In fact, neither the reaction time nor the movement time in a visually triggered reaching movement was influenced by the SMA dysfunction, indicating that the SMA is not as crucial for performing such simple motor tasks as it is in the sequencing of tasks. Furthermore the monkeys still could wait for the minimum of 4.5 s, and initiate a reaching movement, without an external trigger. The similarity in the waiting times before and after the muscimol injection suggests that the SMA is not crucial for telling the animal when to start a movement. Passingham and his colleagues (Chen et al. 1995; Thaler et al. 1995) systematically studied the effects of ablation of the medial premotor cortex using a battery of behavioral tasks. They found that monkeys with lesions in the medial premotor cortex are able to perform visually triggered movements and quick to relearn a visual conditional motor task to select a correct movement. However, they are poor at selecting between two movements in a simple motor-sequence task and poor at changing between two movements. From these findings they concluded that the fundamental specialization of the medial premotor cortex is for the retrieval of actions that are performed without any external cue. Although these authors did not differentiate between the SMA and pre-SMA, their findings using permanent lesions of the medial premotor cortex and our findings with transient dysfunction of the SMA have a number of behavioral consequences in common. In the present study, by identifying the deactivated part of the cortex as the SMA, our findings support their view that the SMA plays an important role in retrieving correct actions based on memory. In addition, we extended the previous findings by showing that SMA dysfunction leads to severe impairment in retrieving multiple actions in a correct temporal order. It should be mentioned, however, that these findings do not rule out a possibility that the SMA also may be of crucial importance in other aspects of motor behavior, including spatial target selection (Alexander and Crutcher 1990), hand selection (Tanji et al. 1987), motor decision (Romo et al. 1997; Tanji 1994), or motor learning (Deiber et al., 1997; Grafton et al. 1995).
Implications for the role of the pre-SMA in different aspects of motor behavior
The pre-SMA now is viewed as an area distinct from the caudally adjacent SMA (Picard and Strick 1996; Tanji 1996), on the basis of anatomic (Dum and Strick 1991; Luppino et al. 1993; Matelli et al. 1991) and physiological (Matsuzaka et al. 1992) criteria. The anatomic connectivity of the two areas differs in that only the pre-SMA receives massive input from the dorsolateral prefrontal cortex, and only the SMA projects directly to the primary motor cortex. The two areas receive projections from largely nonoverlapping regions of the thalamus (Matelli and Luppino 1996). These anatomic findings led to the view that the two areas are involved, at least in part, in different aspects of motor behavior (Picard and Strick 1996; Tanji 1996). Indeed, analysis of neuronal activity in motor-task performing monkeys revealed different properties of SMA and pre-SMA neurons. In the first report to compare the neuronal activity (Matsuzaka et al. 1992), pre-SMA neurons were found to be more active during preparation for upcoming movements, whereas during execution of movements, many fewer pre-SMA neurons were active than SMA neurons. Subsequently it was found that a group of pre-SMA neurons is particularly active when subjects decide to change the direction of a planned reaching movement well before its execution (Matsuzaka and Tanji 1996). More recently, pre-SMA neurons were found to be active during a number of different phases of performing a motor task that required temporal sequencing of multiple movements (Shima and Tanji 1994, 1997). Studies on human subjects using activity-dependent brain imaging techniques have suggested that the pre-SMA is involved more in planning or controlling complex motor behavior or in mediating cognitive aspects of motor behavior (see Picard and Strick 1996 for review) than the SMA. Hikosaka and his colleagues (1996) recently found that the pre-SMA is activated when subjects are acquiring a new sequential motor task that requires visuospatial and visuomotor learning.
Our present results indicate that the pre-SMA is necessary to perform multiple movements in a correct order. The role the pre-SMA plays in performing the motor task may be explained in a number of possible ways. It is possible that the pre-SMA is involved in planning or preparing the temporal order of movements or in selecting individual movements, one after another, by retrieving a correct movement from memory. Another possibility is that the pre-SMA is involved in changing the temporal order of movements from a current order to the next order, i.e., in updating the temporal order. This is in line with our previous report (Shima et al. 1996) that pre-SMA neurons are particularly active when monkeys are required to acquire information for the next sequence, to be stored in memory. In a recent preliminary account, inactivation of the pre-SMA was reported to cause difficulties in acquiring a new sequential motor task (Miyashita et al. 1996). It is possible that in their motor task, as in ours, the monkeys failed to acquire the information necessary to update the sequence. Relevant to this issue is our additional observation on the nature of errors in the present study. After inactivating the pre-SMA, errors frequently occurred in the first of a block of memory-dependent trials (sequence-updated trials), suggesting the importance of the pre-SMA in renewing the sequence of motor behavior.
Implications for functional differences between the SMA and pre-SMA
In the present study, we found three differences in behavioral effects of muscimol injections into the SMA and pre-SMA. First, errors in selecting the correct movement occurred sooner after the pre-SMA injection. This may mean that the pre-SMA is involved more crucially in sequencing multiple movements, a highly demanding task requiring cognitive control of movements. By summarizing functional imaging studies on human subjects, Picard and Strick (1996) put forth a hypothesis that the pre-SMA is more involved than the SMA in motor tasks requiring higher order aspects of motor control, as in planning piano playing (Zatorre et al. 1994) or in selecting movement direction sequentially (Deiber et al. 1991). Our findings are in line with this view. Second, the reaction time during the memory-guided sequencing task was more lengthened after the pre-SMA injection. This may be due to the presence of more abundant preparation-related activity in the pre-SMA (Matsuzaka and Tanji 1996; Matsuzaka et al. 1992). Alternatively, the sequential selection of correct movement under the memorized condition requires short-term memory information in the pre-SMA that comes from the prefrontal cortex via corticocortical connections (Luppino et al. 1993). Third, when the monkeys were given a new sequence of three movements, they committed more errors on the first trial of a newly memorized sequence, after the pre-SMA injection. As discussed in the previous section, this may reflect the involvement of the pre-SMA in updating the motor sequence (Shima et al. 1996). It is conjectured that the pre-SMA, more generally, may be of importance when subjects are confronted with new requirements or demands for subsequent motor tasks and are finding solutions using information coming from other (possibly prefrontal) sources.
We thank Y. Wang for training monkeys and M. Kurama and Y. Takahashi for technical assistance.
This work was supported by grants from Ministry of Education, Science and Culture of Japan (08279101 and 09308032).
Address for reprint requests: J. Tanji, Dept. of Physiology, Tohoku University School of Medicine, Sendai, 980, Japan.
- Copyright © 1998 the American Physiological Society