Abstract

The dorsolateral prefrontal cortex (DLPFC) plays a key role in working memory (WM). Yet its precise contribution (the storage, manipulation and/or utilization of information for the forthcoming response) remains to be determined. To test the hypothesis that the DLPFC is more involved in the preparation of actions than in the maintenance of information in short-term memory (STM), we undertook a functional magnetic resonance imaging investigation in normal subjects performing two delayed response tasks (matching and reproduction tasks) in a visuospatial task sequence (presenta- tion, delay, response). In the two tasks, the presentation and delay phases were similar, but the expected response was different: in the matching task, subjects had to indicate whether a visuospatial sequence matched the sequence presented before the delay period; in the reproduction task, subjects had to reproduce the sequence and, therefore, to mentally organize their response during the delay. Using a fMRI paradigm focusing on the delay period, we observed a significant DLPFC activation when subjects were required to mentally prepare a sequential action based on the information stored in STM. When subjects had only to maintain a visuospatial stimulus in STM, no DLPFC activation was found. These results suggest that a parietal–premotor network is sufficient to store visuospatial information in STM whereas the DLPFC is involved when it is necessary to mentally prepare a forthcoming sequential action based on the information stored in STM.

Introduction

Higher cognitive functions, such as planning or reasoning, depend on a set of more elementary processes including the maintenance of information in short-term memory (STM) and the ability to manipulate this information to organize an ap- propriate goal-directed behavior. These elementary processes may be included within the framework of working memory (WM) (Baddeley, 1986; Goldman-Rakic,1987). Experimental studies in monkeys (Goldman-Rakic,1987; Petrides, 1994; Fuster, 1997) and humans (Owen, 1997; Ungerleider et al., 1998; Smith and Jonides, 1999) have demonstrated the crucial involvement of the dorsolateral prefrontal cortex (DLPFC) in the cerebral network of WM. One important issue is to determine in which aspects of WM the DLPFC is mainly involved, i.e. the storage, manipulation and/or utilization of information for the forth- coming response. In the DLPFC of monkeys, during the delay period of delayed response tasks, electrophysiological studies have shown sustained neuronal activities that may be attributed to the short-term maintenance of internal representations of perceptual information (Funahashi et al., 1993a; Sawaguchi and Yamane, 1999). In addition, during the delay period, a propor- tion of DLPFC neurons is also tuned to the preparatory set (Barone and Joseph, 1989; Funahashi et al., 1993b; Hasegawa et al., 1998; Quintana and Fuster, 1999), suggesting that one of the critical roles of the DLPFC is to link the information maintained in STM to the organization of the forthcoming actions (Fuster, 1997). This concept has been recently reinforced by neuro- psychological data (Ferreira-Texeira et al., 1998) showing that patients with focal lesions of the prefrontal cortex could maintain visuospatial information in STM, but were mostly impaired when this stored information had to be linked to a forthcoming sequence of actions. We may therefore hypothesize that the DLPFC is part of a neuronal network mostly involved in the preparation of actions based on information stored in WM rather than in the storage of sensory information in STM per se. To test these hypothesis, we used the same paradigm elaborated for patients with prefrontal lesions (Ferreira-Texeira et al., 1998) modified for a functional magnetic resonance imaging (fMRI) investigation in normal subjects. Focusing the analysis on one critical period (i.e. the delay) may enable us to clearly separate the cerebral network involved in short-term storage from that involved in the preparation of a sequence of actions based on the information maintained in WM.

Materials and Methods

Subjects

Eight right-handed healthy volunteers (four men, four women, range 20–25 years) with no history of neurological or psychiatric disease were included in the study. They were paid for their participation and all gave informed consent. The experiment was approved by the Ethics Committee for Biomedical Research of the Salpêtrière Hospital.

Cognitive Tasks

Two different WM tasks, visuospatial matching (MAT) and visuospatial reproduction (REP), were designed (Fig. 1a). Both tasks were based on the same sequence of events: the presentation of a visuospatial pattern, a delay period and a response. In both tasks, the presentation and delay periods were similar, differing only by the expected responses. The presentation consisted of ten open white squares (presented at fixed locations) appearing on a screen for 500 ms. Five of these squares (i.e. the ‘stimulus’) sequentially turned black for 1400 ms each. During the delay period, subjects were asked to fixate a central cross on a blank screen for 6000 ms. In the MAT task, after the delay, a new sequence of five black squares was presented and subjects had only to indicate whether this sequence was identical or different from the one presented before the delay. Subjects responded by pressing the right (for ‘identical’) or left (for ‘different’) mouse button. In the REP task, after the delay, the ten open white squares reappeared on the screen. Using the mouse, subjects were required to sequentially ‘point and click’ on each of the five squares of the stimulus in the correct order. In both tasks, responses depended upon the stimulus presented before the delay. However, only in the REP task could subjects mentally prepare a sequential motor response according to the format of the stimulus held ‘on line’.

Two control tasks were designed to match the two WM tasks in term of visuospatial and motor requirements (Fig. 1b). In the control task for the MAT condition (MAT CONT), two different sequences were used. A first sequence of five squares was presented before an interstimulus interval (ISI) of 6000 ms. This sequence was invariable throughout the trials and matched, in terms of sensory information, the presentation phase of the MAT task. After the ISI, a second sequence of five squares appeared on either the left or the right part of the screen. If the sequence was presented on the left side of the screen, subjects were required to press the left button of the mouse, and so with the right. This second sequence was used to match the sensorimotor components of the response phase of the MAT task. In the control task for the REP condition (REP CONT), two sequences of five squares were presented and separ- ated from each other by an ISI of 6000 ms. The first sequence was the same from one trial to another. When the second sequence appeared on the screen, subjects had to ‘point and click’ on each square when it changed to black. Thus, both the REP and the REP CONT tasks were matched for the presentation and response phases. In both control tasks, the response was triggered by the second sequence and did not depend on the sequence presented before the ISI. Hence, no sensory information had to be memorized during the ISI for the preparation of the forth- coming response.

All subjects were trained in each task, first outside and then inside the scanner before image acquisition. In addition, two other visuospatial WM tasks were performed by all subjects. These tasks are not described here since they are outside the scope of the present study.

MRI Scanning

Visual stimuli were projected using an active matrix video projector (EGA mode, 70 Hz refresh rate, Eiki, Osaka, Japan) connected to a computer located in the control room and presented on a screen that the subjects viewed through mirror glasses. A mouse connected to the computer was placed near the subject's right hand and a mouse-pad was fixed on the subject's chest. The subject's head was firmly positioned in a foam-rubber holder which minimized movement.

Functional images were acquired on a 3 T whole-body scanner (Brucker, Germany), using T2* weighed gradient echo, echo-planar imaging sequence, sensitive to blood oxygen level-dependent contrast (repetition time 2000 ms, echo time 40 ms, flip angle of 90°, matrix 64 × 64, field of view 220 × 220 mm2). The images consisted of 18 contiguous axial slices (interleaved acquistion), with 6 mm thickness and 3.4 × 3.4 mm in plane resolution. The lower parts of the temporal and occipital lobes and the cerebellum were not imaged.

During fMRI acquisition subject were required to perform five separate runs of 13 trials. During each run (Fig. 2), 208 volumes of 18 slices were continuously acquired over a total duration of 416 s. A 30 s rest period was interleaved between two consecutive runs. Each run began with a REP CONT trial. This first trial was used to familiarize the subjects to the noisy environment of the scanner and to the handling of the mouse. Thus, the first 16 volumes corresponding to this trial were always discarded. Following this discarded trial and for each run, two trials of each task were presented in a pseudo-random order separated by a 6000 ms inter-trial interval (including two trials of the two tasks not presented in this study). A total of 10 trials per tasks were analyzed. High-resolution T1-weighted anatomical images were acquired in the same session (gradient-echo inversion-recovery sequence, repetition time 1600 ms, echo time 5 ms, matrix 256 × 256 × 128, field of view 220 × 220 mm2, slice thickness 1 mm).

Data Analysis

Data were analyzed on a individual basis and across subject (group analysis) using across subject variance (random effect analysis) (Friston et al., 1999). All analyses were done with SPM'96 software (Wellcome Department of Cognitive Neurology; www.fil.ion.ucl.ac.uk/spm) modi- fied for fMRI (Friston et al., 1995b). For each subject, anatomical images were transformed stereotactically with nine linear rigid transformations to the Talairach coordinate system (Talairach and Tournoux, 1988). The functional scans, corrected for subject motion (Friston et al., 1995a), were then normalized using the same transformations and smoothed with a 5 mm full-width half-maximum Gaussian filter.

For individual analysis, data from each run were processed using the general linear model with separate delayed box-car functions modeling hemodynamic responses of each period of tasks. Overall signal differences between runs were also modeled. A temporal cut-off of 120 s was applied to filter subject-specific low-frequency drift, mostly related to subject biological rhythms and magnetic field drift. An SPM {F} map was obtained, reflecting significant activated voxels according to the model used (P < 0.05). To test hypotheses about regionally specific condition effects, the estimates were compared using linear contrasts. The resulting set of voxel values for each contrast constituted an SPM {t} map which was transformed to the unit normal distribution to give an SPM {Z} map. The resulting set of Z values was then thresholded at P < 0.001 and P < 0.01. Separate analyses were performed during the presentation, the delay (or interstimulus interval) and the response phases. The MAT and the REP tasks were contrasted with their respective control tasks (MAT/MAT CONT, REP/REP CONT; Fig. 1a,b). Contrasting MAT to its control during the delay should highlight the activated regions related to the short-term storage of temporal–spatial information. During the delay, the contrast between REP and its control (REP/REP CONT) and between the REP and the MAT tasks (REP/MAT) should reveal, in addition to the activation related to the short-term storage, activation specific to the binding between the ‘stimulus’ held ‘on line’ and the forthcoming sequence of actions.

For group analysis, parametric maps were constructed using the same contrast as for the subject per subject analysis. Contrasts during the presentation and response phases were also performed in this analysis, in order to determine whether the activation found during the delay was specific to the cognitive processes studied. Results were obtained using a threshold of P < 0.001 corrected for multiple comparisons across the volume. Activated clusters were considered significant if their spatial extent was >18 voxels (or 172 mm3), corresponding to a risk of error (type I error) of P < 0.05.

For each subject, the signal-to-time curve was calculated for voxels presenting the highest Z values in those regions activated in the group study (parietal, motor, premotor, prefrontal cortices and striatum). These curves were obtain by dividing all data point values with the overall mean value of the subject' voxel signal. The signals of individual data points were next averaged across trials of the same type. The curves obtained were then averaged across subjects, allowing us to obtain a mean time- course of the fMRI signal for each task.

Results

Group Analysis

SPM Analysis during the Delay

When the MAT task was compared with its control condition, significant activation was found in the right premotor cortex [Brodmann's area (BA) 6] and the intraparietal sulcus (IPS; BA 7/40, 19) bilaterally (Fig. 3a, Table 1). No significant prefrontal activation was observed on the basis of the chosen threshold (P < 0.001). When the REP task was compared with its control, the same cortical regions were activated and additional activa- tion was detected in the right DLPFC (BA 9/46; Fig. 3b, Table 1). The REP/MAT contrast showed activation in the right DLPFC (BA 9/46), the left motor (BA 4) and premotor (BA 6) cortices, the left supplementary motor area (SMA; BA 6) and bilaterally in the anterior putamen (Fig. 3c, Table 1).

SPM Analysis during the Presentation and Response

In both the presentation and response phases, no activation reached the level of significance in the left or right DLPFC for the MAT/MAT CONT, REP/REP CONT or REP/MAT contrasts (Fig. 4af).

During the presentation phase, in the MAT/MAT CONT (Fig. 4a) and REP/REP CONT (Fig. 4b) contrasts, significant signal changes were found in the right intraparietal sulcus (BA 7/40) and the premotor cortex bilaterally (BA 6), with a greater extent in the REP/REP CONT contrast. In the latter contrast, activation extended to the superior frontal sulcus (BA 9) and the SMA (BA 6). The right inferior frontal gyrus (BA 45), the left anterior putamen and the precuneus were also activated. In the REP/MAT contrast, no voxel reached statistical significance (Fig. 4c). During the response phase, no significant change was observed for the MAT/MAT CONT and REP/REP CONT contrasts (Fig. 4d,e). In the REP/MAT contrast, significant activation was detected chiefly in the left SMA (BA 6), the premotor (BA 6), sensorimotor (BA 1/2/3/4) and superior parietal (BA 7) cortices and the thalamus (Fig. 4f). Small foci of activation were found in the right premotor (BA 6), superior parietal (BA 5/7/40), superior temporal (BA 42) and superior occipital (BA 19) cortices. Activation in the putamen was found bilaterally.

Individual Analysis

Analysis of individual data is presented in Table 2. Individual analysis allowed us to study inter-subject variability of areas activated in the group study. For the MAT versus MAT CONT comparison (P < 0.001), the left parietal area was activated in all subjects, the right parietal cortex in six subjects and the right premotor in four subjects. At P < 0.01, seven subjects activated the right parietal cortex and six the right premotor cortex.

For the REP versus REP CONT (P < 0.001), the right parietal cortex was activated in all subjects and the left parietal cortex in seven subjects. The left and right premotor cortices were activated in four and three subjects, respectively. The right DLPFC was activated in three subjects. Activation was found in the anterior striatum in two subjects. At P < 0.01, eight subjects had activation in both parietal areas, six subjects in the DLPFC and premotor cortex, four subjects in the anterior striatum of the right hemisphere and five subjects in the left hemisphere.

For the REP versus MAT comparison, all regions found in the group study were activated in all subjects (P < 0.001), including the right DLPFC, the left SMA, the left sensorimotor cortex and the anterior striatum bilaterally.

fMRI Time-courses

Time-courses of activation were studied in the DLPFC, premotor cortex, parietal cortex, sensorimotor cortex and anterior stri- atum in both hemispheres. They were identical in both sides for the premotor and parietal cortices and the anterior striatum. As already mentioned, activation was only found in the left sensorimotor cortex and the right DLPFC. Figure 5 illustrates the time-course curves of theses regions.

In the DLPFC, time-courses of activation were different between the MAT and REP tasks (Fig. 5a). In the MAT task, signal slightly decreased during the presentation phase, then increased slightly but not significantly during the delay, reached a maxi- mum at the end of this phase and remained relatively constant during the response phase. In the REP task, there was a robust signal increase during the delay, starting 2 s after the presenta- tion period and reaching a maximum 4 s after the beginning of the response phase. The signal then continuously decreased to baseline. In both control tasks, the signal was similar. Activation decreased during the first 4 s, remained constant until the end of the delay, then increased during the response period, reaching a maximum 4 s after the end of this period. The signal increase was higher in the REP CONT than in the MAT CONT tasks.

In the premotor cortex (Fig. 5b), the MAT and REP tasks' related activation increased continuously during the presenta- tion period, reaching a maximum 2 s after the onset of the delay period. The signal then decreased slowly during both tasks, more in the MAT than the REP task, and returned to baseline at the end of the ISI. For both control tasks, signal-to-time curves were similar, showing a first increase with a maximum at the end of the presentation period and a slight decrease during the delay period, followed by a second increase during the response period reaching a maximum at the end of this period. During the last 10 s, REP CONT signal amplitude was higher than in the MAT CONT task.

In the parietal cortex (Fig. 5c), time-courses of activation were similar to those observed in the premotor cortex.

In the left sensorimotor cortex (Fig. 5d), the signal increased in all tasks during the delay and response periods, reaching a maximum at the end of the later. The amplitude of signal in- crease was similar in the tasks and in their respective controls but higher in the REP than the MAT ones.

In the anterior striatum (Fig. 5e), the signal increased during the delay and the response periods in the REP task. In all three other tasks it increased only during the response period, reach- ing a maximum at the end of this period.

Discussion

Taking all the data together, the three main results of the present study were observed during the delay period: a parietal– premotor activation in the MAT task, when compared with its control; an additional activation in the right DLPFC in the REP/REP CONT; and, in the REP/MAT comparisons, the same activation in the right DLPFC alone with additional foci of activation in motor-related areas, though no movement had yet been required. Additional information can be obtained from the time-course curves analysis — firstly, the signals observed in the parietal and premotor cortices were similar in the MAT and REP tasks during the presentation and delay phases, and significantly increased in comparison with their respective control tasks (Fig. 5b,c), although the activation in the REP condition was clearly greater than that observed in the MAT condition at the end of the delay and during the response phase. Secondly, for the right DLPFC, time-course curves were significantly different between the REP and MAT tasks during the delay and response phases. Indeed, in the REP task the signal continuously increased from the end of the cue period to the early phase of the response, contrasting with a steady state with a lower level of signal in the MAT task (Fig. 5a). Thirdly, in the REP and REP CONT tasks, signals in more motor-related areas (sensori- motor cortex and anterior striatum) started to increase during the second half of the delay to peak at the end of the response phase. In both tasks, signals were significantly higher than those observed in the MAT and MAT CONT tasks (Fig. 5d,e).

The differential patterns of activation between the REP and MAT conditions during the delay period, i.e. at a time when the subjects had experienced the same events, are probably related to what the subjects were required to do after the delay, namely to recognize the sequence (MAT task) or to program a sequential response (REP task). In the MAT task, which required the sub- ject to maintain ‘on line’ during the delay, the temporal–spatial information that could only be used after the presentation of a second sequence, a sustained activation was found mainly in the parietal and premotor areas when compared with the control condition. The parietal areas activated in this task, namely the intraparietal sulci, correspond to regions classically involved in visuospatial processing (Andersen et al., 1985), including visuospatial WM (Friedman and Goldman-Rakic, 1994; Chafee and Goldman-Rakic, 1998). It is noteworthy that the IPS was also activated during the presentation phase. This may suggest that the IPS is involved in attention by selecting the information to be encoded for further processing (Andersen, 1995) and/or in maintaining information in WM as one may consider that memory phase starts after the presentation of the first square of the temporo-spatial sequence. The IPS may also contribute to more basic perceptual aspects because an increase of activation was also observed, although at a more modest level, in the two control tasks. Regarding the premotor areas, when activation maps obtained in this condition were superimposed on anatom- ical images, we found that the right premotor activation observed during the delay phase was located partly in a cortical region where the precentral and the superior frontal sulci intersect, an area that may include the right frontal eye field (FEF) in humans (Paus, 1996). The FEF is indeed activated in WM tasks requiring saccadic or sequential eye movements (Sweeney et al., 1996) and also when subjects have to sequentially shift attention to different targets (Dias and Bruce, 1994; Corbetta, 1998). A similar activation was also found in the premotor areas during the presentation phase (see Figs 4a,b and 5b), indicating that these areas may also participate in the encoding phase of the temporo-visual stimulus which relies in part on saccadic eye movements (Schall and Hanes, 1993). The persistence of sig- nificant activation in the premotor area during the delay may indicate a ‘spatial rehearsal’ (Smith et al., 1995), as subjects have to mentally shift their attention from one target to another within the visual sequence maintained ‘in mind’. In this case, the premotor activation would be related to the storage of mental representation in terms of visuomotor engrams. In line with these data are the studies of Courtney et al. (Courtney et al., 1997, 1998), who reported activation in the same parietal and premotor areas during the delay phase of a WM task in which subjects had to decide whether a probe matched any of three sequentially presented spatial targets. Regarding the role attri- buted to the DLPFC in WM, it should be noted that no activation above threshold was found in this region in the MAT task. As no specific forthcoming action could be prepared during the delay, it seems likely that the parietal–premotor network is sufficient for the short-term storage of this information. However, we cannot rule out that, even though the DLPFC is not activated above threshold, it may intervene to some extent in the mainten- ance of the information in STM. Indeed, the time-course curves showed that there was a slight increase of the signal during the delay period. Taken together, these data suggest the parietal– premotor network represents a first level of visuospatial WM processing when little demand on executive processing, such as selection and preparation of response, is required. Therefore, it is predicted that additional areas, including the DLPFC, will be triggered above threshold when these executive requirements increase.

In the REP task, in addition to the processes involved in the MAT task, there is the mental representation of the forthcoming action, i.e. the selection and preparation of a sequence of moves, based on the information stored in WM. Several findings from our study argue in favor of the idea that the link between the information held ‘on line’ and a specific program of actions is the upgraded function leading to DLPFC recruitment in this paradigm: (i) the main finding supporting this hypothesis is that the DLPFC was activated during the delay in the REP/REP CONT and, above all, in the REP/MAT comparisons; (ii) this DLPFC activation is mainly due to the mental representation of the forthcoming action. Indeed, because of the short length of the delay period (6000 ms) we cannot totally dissociate activation related to the cue, the memory and the preparation of response. However, if the activation observed in the REP task was only related to cue, the increase in the signal should have started from the beginning of the presentation phase and peaked in the delay period, according to hemodynamic response delay. Analyses of time-course curves demonstrated that this was not the case. Furthermore, no such delayed peak of activation was seen in the DLPFC in the two control tasks, which are identical to WM tasks in terms of sensory cues. If the activation observed in the REP task was related only to memory, one should have noticed a similar hemodynamic response in the MAT task. If activation observed during the delay period of the REP task was due to the overlap of signals from the cue and memory, one should expect a time-course curve and/or increase of signal quantitatively similar to the one observed during the delay period of the MAT task. The significant activation found in the left motor cortex in the REP/MAT comparison during the delay period, at a time when no response has started, also suggests that, during the delay period of the REP task, activation can be partly related to the forthcoming action. However, the peak of activation ob- served in the response phase during the REP task was probably not due only to the response preparation because the activation related to the response occurred more lately in the REP CONT than in the REP tasks; and (iii) to conclude, our findings indicate that the DLPFC contributes to the mnemonic processing, and above all, to the preparatory set. However, the relationship between the preparatory set and the information maintained ‘on line’ is of a complex nature. Indeed, this relationship may require more attentional resources than in the MAT task as the forthcoming response necessitates the holding ‘on line’ of accurate features of the information. An alternative explanation might be that the DLPFC is activated in the REP task in relation to a prominent inhibition process in order to temporally repress the forthcoming action until the ‘go signal’, as it has already been shown in memory-guided saccade tasks, both in humans and monkeys with DLPFC lesions (Funahashi et al., 1993a, b). This seems unlikely, however, given the well-known function of the DLPFC in response programming (Shallice, 1982). Interestingly, it is likely that this activation, which was restricted to the right DLPFC, was related to the nature of the sensory information processed (i.e. spatial domain), as it has already been reported in humans (Smith et al., 1995; McCarthy et al.,1996).

In fact, our data strongly agree with the results of a recent neuropsychological study in patients with focal lesions of the prefrontal cortex in which similar visuo-spatial WM tasks were used (Ferreira-Texeira et al., 1998). In this study, when com- pared with normal subjects or patients with post-rolandic lesions, patients with prefrontal lesions were only impaired in the REP task, whereas they had no deficit when (i) encoding and maintaining visuo-spatial information, as shown by their normal performance in the MAT task with or without a delay; or (ii) executing a sequential response immediately after its presenta- tion, as in the REP task without a delay.

In conclusion, we have identified the neural correlates involved during the delay period of two WM tasks requiring different levels of processing. The results of this study, congru- ent with a previous neuropsychological study in patients with frontal lobe lesions, validates the hypothesis that a parietal– premotor network is sufficient to store visual temporo-spatial sequences in STM; and, in situations when the planning and preparing of a predictable sequence of actions is required, then the DLPFC and additional regions (SMA, sensorimotor cortex, anterior putamen) might be recruited. This is in agreement, at least in humans, with the expected role of the prefrontal cortex, involved in the planning of actions rather than in the short-term storage of information. In this view, the prefrontal cortex would be solicited mainly when a specific or new program of actions is required for an adaptive behavior.

Notes

This work was supported by a grant from the Ministère de l’Education Nationale, de la Recherche et de la Technologie.

Address correspondence to Richard Levy, INSERM EPI 007 Pavillon Claude Bernard, Hôpital de la Salpêtrière, 47 boulevard de l’Hôpital, F-75013 Paris, France. Email: richard.levy@psl.ap-hop-paris.fr.

Table 1

Activation during the delay period for each comparison

 Brodmann's area (BA) Talairach coordinates 
  Z score 
All results presented in the table correspond to areas where statistically significant changes were detected (P < 0.001 corrected for multiple comparison, Z > 3.09). R = right; L = left. 
MAT/MAT CONT      
    R premotor cortex BA 6  27 51 3.88 
    R superior parietal lobule BA 7/40  30 –51 54 4.66 
    L superior parietal lobule BA 7 –24 –51 60 4.39 
    L inferior parietal lobule BA 40 –45 –39 45 4.47 
    R inferior parietal lobule BA 19  27 –63 39 4.23 
REP/REP CONT      
    R dorsolateral prefontal cortex BA 9/46  42  45 27 4.13 
    R premotor cortex BA 6  27 51 3.96 
    R superior parietal lobule BA 7/40  30 –51 54 4.96 
    L superior parietal lobule BA 7 –24 –51 57 4.80 
    R inferior parietal lobule BA 19  27 –63 39 4.68 
REP/MAT      
    R dorsolateral prefontal cortex BA 9/46  42  39 33 3.89 
    L premotor cortex BA 6 –36 –18 60 5.39 
    L SMA BA 6  –6 57 4.29 
    R motor cortex BA 4 –42 –27 60 4.04 
    R anterior putamen   27 4.42 
    L anterior putamen  –24  12 4.29 
 Brodmann's area (BA) Talairach coordinates 
  Z score 
All results presented in the table correspond to areas where statistically significant changes were detected (P < 0.001 corrected for multiple comparison, Z > 3.09). R = right; L = left. 
MAT/MAT CONT      
    R premotor cortex BA 6  27 51 3.88 
    R superior parietal lobule BA 7/40  30 –51 54 4.66 
    L superior parietal lobule BA 7 –24 –51 60 4.39 
    L inferior parietal lobule BA 40 –45 –39 45 4.47 
    R inferior parietal lobule BA 19  27 –63 39 4.23 
REP/REP CONT      
    R dorsolateral prefontal cortex BA 9/46  42  45 27 4.13 
    R premotor cortex BA 6  27 51 3.96 
    R superior parietal lobule BA 7/40  30 –51 54 4.96 
    L superior parietal lobule BA 7 –24 –51 57 4.80 
    R inferior parietal lobule BA 19  27 –63 39 4.68 
REP/MAT      
    R dorsolateral prefontal cortex BA 9/46  42  39 33 3.89 
    L premotor cortex BA 6 –36 –18 60 5.39 
    L SMA BA 6  –6 57 4.29 
    R motor cortex BA 4 –42 –27 60 4.04 
    R anterior putamen   27 4.42 
    L anterior putamen  –24  12 4.29 
Table 2

Individual results for regions activated in the group study during the delay

 Brodmann' area (BA) No. of subjects 
  P < 0.01 P < 0.001 
Results presented in the table correspond to areas where significant changes were detected (P < 0.001 and P < 0.01 uncorrected). R = right; L = left. 
MAT/MAT CONT    
R premotor cortex BA 6 
R parietal lobule BA 7/40 
L parietal lobule BA 7/40 
REP/REP CONT    
    R DLPFC BA 8/46 
    R premotor cortex BA 6 
    L premotor cortex BA 6 
    R parietal lobule BA 7/40 
    L parietal lobule BA 7/40 
    R anterior striatum  
    L anterior striatum  
REP/MAT    
    R DLPFC BA 8/46 
    L SMA BA 6 
    L sensorimotor cortex BA 6 
    R anterior striatum  
    R anterior striatum  
 Brodmann' area (BA) No. of subjects 
  P < 0.01 P < 0.001 
Results presented in the table correspond to areas where significant changes were detected (P < 0.001 and P < 0.01 uncorrected). R = right; L = left. 
MAT/MAT CONT    
R premotor cortex BA 6 
R parietal lobule BA 7/40 
L parietal lobule BA 7/40 
REP/REP CONT    
    R DLPFC BA 8/46 
    R premotor cortex BA 6 
    L premotor cortex BA 6 
    R parietal lobule BA 7/40 
    L parietal lobule BA 7/40 
    R anterior striatum  
    L anterior striatum  
REP/MAT    
    R DLPFC BA 8/46 
    L SMA BA 6 
    L sensorimotor cortex BA 6 
    R anterior striatum  
    R anterior striatum  
Figure 1.

Schematic description of the two WM tasks and their respective controls. (a) WM tasks. Only the response differed between the two tasks: in the MAT task, subjects had to indicate whether a new sequence was identical (‘Yes’) by pressing the right button (R) or different (‘No’) by pressing the left button (L) of a digital pointing device (a mouse). In the REP task subjects were required to reproduce the sequence presented before the delay (i.e. the ‘stimulus’) by sequentially ‘pointing and clicking’ with the mouse the five squares of the stimulus in the correct order. (b) Control tasks. In the MAT CONT task, at the time of the response, subjects had to press the right button (R) of the mouse for a right-sided sequence or the left button (L) for a left-sided sequence. In the REP CONT task, during the response phase, subjects had to ‘point and click' sequentially each square as it turned black. In neither task did the response depend on the sequence presented before the interstimulus interval.

Figure 1.

Schematic description of the two WM tasks and their respective controls. (a) WM tasks. Only the response differed between the two tasks: in the MAT task, subjects had to indicate whether a new sequence was identical (‘Yes’) by pressing the right button (R) or different (‘No’) by pressing the left button (L) of a digital pointing device (a mouse). In the REP task subjects were required to reproduce the sequence presented before the delay (i.e. the ‘stimulus’) by sequentially ‘pointing and clicking’ with the mouse the five squares of the stimulus in the correct order. (b) Control tasks. In the MAT CONT task, at the time of the response, subjects had to press the right button (R) of the mouse for a right-sided sequence or the left button (L) for a left-sided sequence. In the REP CONT task, during the response phase, subjects had to ‘point and click' sequentially each square as it turned black. In neither task did the response depend on the sequence presented before the interstimulus interval.

Figure 2.

An example of the time-organization of a run during a scanning session. The first trial was always discarded for analysis. The time-scale unit is ‘second per volume’ (2 s per volume). I = instruction (4 s); P = presentation (8 s); D = delay (6 s); R = response (8 s); ISI = interstimulus interval (6 s).

Figure 2.

An example of the time-organization of a run during a scanning session. The first trial was always discarded for analysis. The time-scale unit is ‘second per volume’ (2 s per volume). I = instruction (4 s); P = presentation (8 s); D = delay (6 s); R = response (8 s); ISI = interstimulus interval (6 s).

Figure 3.

Z maps corresponding to the group study results superimposed upon the superior and lateral views of cortical surface renders during the delay. Threshold = P < 0.001 corrected for multiple comparison, Z > 3.09. R = right; L = left.

Figure 3.

Z maps corresponding to the group study results superimposed upon the superior and lateral views of cortical surface renders during the delay. Threshold = P < 0.001 corrected for multiple comparison, Z > 3.09. R = right; L = left.

Figure 4.

Z maps corresponding to the group study results superimposed upon the superior and lateral views of cortical surface renders during the presentation and the response. (Upper) Z maps during the presentation (ac) for the MAT/MAT CONT (a), the REP/REP CONT (b) and the REP/MAT (c) comparisons. (Lower) Z maps during the response (df) for the MAT/MAT CONT (d), the REP/REP CONT (e) and the REP/MAT (f) comparisons. Threshold = P < 0.001 corrected for multiple comparison, Z > 3.09. R = right; L = left.

Figure 4.

Z maps corresponding to the group study results superimposed upon the superior and lateral views of cortical surface renders during the presentation and the response. (Upper) Z maps during the presentation (ac) for the MAT/MAT CONT (a), the REP/REP CONT (b) and the REP/MAT (c) comparisons. (Lower) Z maps during the response (df) for the MAT/MAT CONT (d), the REP/REP CONT (e) and the REP/MAT (f) comparisons. Threshold = P < 0.001 corrected for multiple comparison, Z > 3.09. R = right; L = left.

Figure 5.

fMRI averaged time-course curves for five selected voxels in: (a) the right DLPFC; (b) the left premotor cortex; (c) the left parietal; (d) the left sensorimotor; and (e) the left anterior striatum . ▪, REP; □, REP CONT; •, MAT; ○, MAT CONT; the time-scale unit is expressed as TR (repetition time; 2 s per TR).

Figure 5.

fMRI averaged time-course curves for five selected voxels in: (a) the right DLPFC; (b) the left premotor cortex; (c) the left parietal; (d) the left sensorimotor; and (e) the left anterior striatum . ▪, REP; □, REP CONT; •, MAT; ○, MAT CONT; the time-scale unit is expressed as TR (repetition time; 2 s per TR).

References

Andersen RA (
1995
) Encoding of intention and spatial location in the posterior parietal cortex.
Cereb Cortex
 
5
:
457
–469.
Andersen RA, Essick GK, Siegel MR (
1985
) Encoding of spatial location by posterior parietal neurons.
Science
 
230
:
456
–458.
Baddeley A (1986) Working memory. Oxford: Clarendon Press.
Barone P, Joseph JP (
1989
) Prefrontal cortex and spatial sequencing in macaque monkey.
Exp Brain Res
 
78
:
447
–464.
Chafee MV, Goldman-Rakic PS (
1998
)
Matching patterns of activity in primate prefrontal area
 
8a and parietal area 7ip neurons during a spatial working memory task. J Neurophysiol 79
:
2919
–2940.
Corbetta M (
1998
) Frontoparietal cortical networks for directing attention and the eye to visual locations: identical, independent, or overlapping neural systems?
Proc Natl Acad Sci USA
 
95
:
831
–838.
Courtney SM, Ungerleider LG, Keil K, Haxby JV (
1997
) Transient and sustained activity in a distributed neural system for human working memory.
Nature
 
386
:
608
–611.
Courtney SM, Petit L, Maisog JM, Ungerleider LG, Haxby JV (
1998
) An area specialized for spatial working memory in human frontal cortex.
Science
 
279
:
1347
–1351.
Dias EC, Bruce CJ (
1994
) Physiological correlate of fixation disengagement in the primate's frontal eye field.
J Neurophysiol
 
72
:
2532
–2537.
Ferreira-Texeira C, Verin M, Pillon B, Levy R, Dubois B, Agid Y (
1998
) Spatio-temporal working memory and frontal lesions in man.
Cortex
 
34
:
83
–98.
Friedman HR, Goldman-Rakic PS (
1994
)
Coactivation of prefrontal cortex and inferior parietal cortex in working memory tasks revealed by
 
2DG functional mapping in the rhesus monkey. J Neurosci 14
:
2775
–2788.
Friston KJ, Ashburner J, Poline JB, Frith CD, Heather JD, Frackowiak RSJ (
1995
) Spatial registration and normalization of images.
Hum Brain Map
 
2
:
165
–189
Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD, Frackowiak RSJ (
1995
) Statistical parametric maps in functional imaging: a general linear approach.
Hum Brain Map
 
2
:
189
–210.
Friston KJ, Holmes AP, Worsley KJ (
1999
) How many subjects constitute a study?
NeuroImage
 
10
:
1
–5.
Funahashi S, Bruce CJ, Goldman-Rakic PS (
1993
) Dorsolateral prefrontal lesions and oculomotor delayed-response performance: evidence for mnemonic ‘scotomas’.
J Neurosci
 
13
:
1479
–1497.
Funahashi S, Bruce CJ, Goldman-Rakic PS (
1993
) Prefrontal neuronal activity in rhesus monkeys performing a delayed anti-saccade task.
Nature
 
365
:
753
–756.
Fuster J (1997) The prefrontal cortex: anatomy, physiology, and neuropsychology of the frontal lobes. New York: Raven.
Goldman-Rakic PS (1987) Handbook of physiology V (Plum F, Mountcastle V, eds), pp. 373–517. Washington, DC: The American Physiological Society.
Hasegawa R, Sawaguchi T, Kubota K (
1998
) Monkey prefrontal neuronal activity coding the forthcoming saccade in an oculomotor delayed matching-to-sample task.
J Neurophysiol
 
79
:
322
–333.
McCarthy G, Puce A, Constable RT, Krystal JH, Gore JC, Goldman-Rakic P (
1996
) Activation of human prefrontal cortex during spatial and nonspatial working memory tasks measured by functional MRI.
Cereb Cortex
 
6
:
600
–611.
Owen AM (
1997
) The functional organization of working memory processes within human lateral frontal cortex: the contribution of functional neuroimaging.
Eur J Neurosci
 
9
:
1329
–1339.
Paus T (
1996
) Location and function of the human frontal eye-field: a selective review.
Neuropsychologia
 
34
:
475
–483.
Petrides M (
1994
) Frontal lobes and behaviour.
Curr Opin Neurobiol
 
4
:
207
–211.
Quintana J, Fuster JM (
1999
) From perception to action: temporal integrative functions of prefrontal and parietal neurons.
Cereb Cortex
 
9
:
213
–221.
Sawaguchi T, Yamane I (
1999
) Properties of delay-period neuronal activity in the monkey dorsolateral prefrontal cortex during a spatial delayed matching-to-sample task.
J Neurophysiol
 
82
:
2070
–2080.
Schall JD, Hanes DP (
1993
) Neural basis of saccade target selection in frontal eye field during visual search.
Nature
 
366
:
467
–469.
Shallice T (
1982
) Specific impairments of planning.
Phil Trans R Soc Lond B Biol Sci
 
298
:
199
–209.
Smith EE, Jonides J (
1999
) Storage and executive processes in the frontal lobes.
Science
 
283
:
1657
–1660.
Smith EE, Jonides J, Koeppe RA, Awh E, Schumacher E, Minoshima S (
1995
) Spatial versus object working memory: PET investigations.
J Cogn Neurosci
 
7
:
337
–358.
Sweeney JA, Mintun MA, Kwee S, Wiseman MB, Brown DL, Rosenberg DR, Carl JR (
1996
) Positron emission tomography study of voluntary saccadic eye movements and spatial working memory.
J Neurophysiol
 
75
:
454
–468.
Talairach J, Tournoux P (1988) Coplanar stereotactic atlas of the human brain. Stuttgart: Thieme.
Ungerleider LG, Courtney SM, Haxby JV (
1998
) A neural system for human visual working memory.
Proc Natl Acad Sci USA
 
95
:
883
–890.