Abstract

Most of the working memory (WM) tasks used in functional imaging studies are based on the principle of the delayed response in which both the storage and the response organization are present during the delay period. It is therefore difficult to isolate activation specific to the storage function from that specific to the organization of the response. To determine the specific neural networks associated with these two WM operations, we performed a functional MRI study in healthy subjects using a new paradigm, ‘the double delay/double response’ tasks. This paradigm isolates maintenance from response organization by dividing the delay into two separate parts, the first being dedicated to memory, while the second includes response organization. Activation within the dorsolateral prefrontal cortex (DLPFC) followed a relative hemispheric dissociation: activation related to maintenance was predominant in the right DLPFC but was only detected when the load exceeded three items. Activation related to response organization was predominant in the left DLPFC, regardless of whether this response was based on information held in WM (‘memory guided’) or was independent of WM (‘visually-guided’). These results suggest that activation of the DLPFC, should be interpreted in terms of executive processing for both maintenance and response organization.

Introduction

The dorsolateral prefrontal cortex (DLPFC) is thought to be involved in the ‘executive functions’ that subserve the mental organization of future actions (Dubois et al., 1995; Fuster, 1997; Miller, 1999; Duncan and Owen, 2000; Fuster, 2000). Some of the processes required for executive functions are gathered within the framework of working memory (WM). These processes include the short-term maintenance of relevant information, their manipulation and the organization of the sequence of actions based on the information held in short-term memory (STM) (Goldman-Rakic, 1987, 1995; Baddeley, 1996, 1998, 2003). To study these WM processes, many researches have used the delayed response tasks, introducing a delay period between the presentation of the stimuli and the response. Nonhuman primate studies have shown that sustained neuronal activity in the DLPFC during the delay period supports WM processes (Niki and Watanabe, 1976; Funahashi et al., 1993). Using delayed-response tasks, functional studies in humans have confirmed the involvement of the DLPFC in WM functions (D'Esposito and Grossman, 1996; Owen, 1997; Courtney et al., 1998; Ungerleider et al., 1998; D'Esposito et al., 2000a; Owen, 2000). However the specificity of DLPFC in WM is unclear, since sustained neural activity has also been observed in other cortical areas (i.e. the posterior parietal and inferior temporal cortices) during such WM tasks (Miller et al., 1991, 1993; Chafee and Goldman-Rakic, 1998). In addition, DLPFC activity may be associated with the maintenance of information as well as with the organization of the forthcoming action, as suggested by electrophysiological studies in monkeys (di Pellegrino and Wise, 1991; Boussaoud and Wise, 1993; Hasegawa et al., 1998; Quintana and Fuster, 1999; Sawaguchi and Yamane, 1999).

Does the activation observed in the DLPFC indicate the role of this structure in the last stage of sensory processing (i.e. the maintenance of information) or in the earliest stage of motor processing (i.e. organizing the response to be executed)? The tasks currently used do not permit to answer the question because maintenance of information in STM and the organization of the forthcoming response are both present during the delay period. In functional imaging studies, paradigms using recognition tasks (Pochon et al., 2001), preventing motor preparation (D'Esposito et al., 2000b; Rowe et al., 2000; Rowe and Passingham, 2001; Simon et al., 2002) or increasing the length of the delay (Toni et al., 1999, 2002; Jha and McCarthy, 2000; Leung et al., 2002) have been used in order to disambiguate maintenance from response organization. These studies have produced disparate results, some indicating an essential role of the DLPFC in selecting and preparing the response to be executed while others support an important role of the DLPFC in the maintenance of information. Discrepancies between these observations may reflect differences in experimental design. Indeed, the type (Courtney et al., 1998; D'Esposito et al., 2000b) and the load (Braver et al., 1997; Jonides et al., 1997) of information to be held in STM, as well as the nature of the response have been demonstrated to differentially recruit the DLPFC (Rowe and Passingham, 2001).

The aim of the present study was to answer the following questions. Is the DLPFC involved in: (i) the general process of maintenance of information; (ii) maintenance, only for high loads in STM; (iii) the general process of organizing the forthcoming response; or (iv) only the response organization in the context of WM? To answer these questions, we developed a cognitive paradigm (the ‘double delay/double response’ task) that was applied in a functional MRI (fMRI) study of healthy subjects.

Materials and Methods

Subjects

Eleven right-handed healthy volunteers (seven men, four women, range 22–34 years, mean 28.3 years) with no history of neurological or psychiatric disease were included in the study. All gave informed consent. The experiment was approved by the Local Ethics Committee for Biomedical Research.

Working Memory Tasks

General Principles

To isolate the maintenance process from the organization of the forthcoming response, we designed a ‘double delay/double response’ paradigm (Fig. 1). Unlike the classical delayed-response tasks, instructions for the response were inserted in the midst of the delay period, thus dividing the delay period into two sections (Delay 1 and 2): Delay 1 for the study of the maintenance processing in WM, Delay 2 for the study of the processes involved in the organization of the forthcoming response. At the onset of each trial, subjects were only instructed to maintain the perceptual stimulus (spatial information) during the forthcoming delay period (Delay 1). During Delay 1, subjects could not anticipate the response since no specific instruction had been given yet. To determine whether the DLPFC is involved in the maintenance process as a general function or only for high loads, the effect of load on DLPFC activities were studied during Delay 1. The tasks were performed with three different loads (one, three and five items to be maintained). An instruction for the response was provided just after Delay 1. During Delay 2 that immediately followed the instruction, subjects were to prepare the forthcoming response. To determine whether the DLPFC was activated during Delay 2 by the general process of response organization or in response organization based on the context of WM, two different conditions were compared: (i) a memory-guided condition (MemG), during which subjects were to organize their response according to a memorized information and (ii) a visually-guided condition (VisG), during which subjects were to organize their response on the basis of information displayed on line, i.e. during Delay 2.

Figure 1.

Schematic representation of the different ‘double delay/double response’ tasks. The arrows that course around the squares during presentation and response phases symbolize the sequential structure of the presentation. The pointed finger appears on the screen during the instruction phase then disappears. The pointed finger drawn at the response phase does not appear on the screen: it only indicates that the subject should click on the screen to respond (see method section for details).

Figure 1.

Schematic representation of the different ‘double delay/double response’ tasks. The arrows that course around the squares during presentation and response phases symbolize the sequential structure of the presentation. The pointed finger appears on the screen during the instruction phase then disappears. The pointed finger drawn at the response phase does not appear on the screen: it only indicates that the subject should click on the screen to respond (see method section for details).

Experimental Tasks (Fig. 1a)

In all tasks, each trial consisted of five different phases: (i) Presentation; (ii) Delay 1; (iii) Instruction; (iv) Delay 2; and (v) Response.

Presentation phase.

In the MemG and VisG conditions, trials started with the presentation of a matrix of twelve blue squares at fixed locations on the screen. After 1 s, one or several squares sequentially turned from blue to red for one second each. The stimulus was either one red square or a temporal–spatial sequence of three or five red squares within the matrix (the presentation phase thus lasted 2 s for one square, 4 s for a three-square sequence and 6 s for a five-square sequence).

Delay 1.

Subjects were asked to fixate a central cross-hair on a blank screen and mentally maintain the stimulus. Delay 1 was similar for each trial and lasted 8 s. Although the design of the task limited the possibility to prepare the response during Delay 1 (the instruction for the response being given after Delay 1), in addition, subjects were asked to memorize the stimulus and to wait for the instruction after the fist delay to prepare their response.

Instruction.

Immediately after Delay 1, instruction appeared on the screen for 2 s. The instruction differed in the MemG and the VisG conditions. In the MemG condition, a symbolic cue indicated to the subjects that, after Delay 2, they should reproduce the spatial stimulus presented before Delay 1 by clicking on the screen with a trackball mouse. In the VisG condition, another symbolic cue instructed the subjects on the new spatial sequence that subjects were to perform after Delay 2. The response consisted of ‘pointing and clicking’ on small squares located in each corner of the screen in a clockwise manner. A pointing finger associated with a central cross-hair indicated which corner was to be touched first.

Delay 2.

In the MemG task, the subjects were instructed to prepare to reproduce the memorized sequence, so the organization of the forthcoming response was directly linked to the memorized sequence (memory-guided). In the VisG task, the subjects were instructed to mentally prepare a response guided by new visual information displayed on the screen during Delay 2. In both tasks, the duration of the delay was 8 s. Subjects were instructed not to move until the visually presented go-signal at the end of the Delay 2, and subjects were asked to fixate on a centrally located cross-hair on the blank screen.

Response.

At the end of Delay 2, the central cross-hair turned to red (‘go-signal’) triggering the motor response. Subjects responded either by clicking on the squares in order to reproduce the memorized sequence (MemG condition) or by clicking on the locations indicated by the symbols present on the screen (VisG condition). The duration of the response period depended on the length of the initial sequence (two s per square). At the end of this period, the screen became blank for 5 s, and then the next trial began.

Control Condition (Fig. 1b).

Subjects performed a control condition designed to match the cognitive tasks in terms of sensory-motor components. The matrix of squares appeared on the screen for 3 s, but none of the squares turned from blue to red. This was followed by a period of 8 s, during which the subjects were asked to fixate a central cross-hair. The instruction following Delay 1 consisted of a centrally located white square, which indicated that no particular action was expected. Then, a second fixation period of 8 s was presented to match Delay 2. At the end of this period, the screen turned to blank for five s until the next trial.

All subjects were trained in each task, first outside and then inside the scanner. Performance scores (number of correct responses and reaction times) were automatically recorded. Afterwards, subjects were debriefed, stating whether they had correctly followed all instructions and describing their strategies while performing the tasks.

MRI Scanning

Visual stimuli were generated by a PC computer and projected using an active matrix video projector connected to the computer located in the control room and presented on a screen positioned at the foot end of the MRI scanner bore. Subjects viewed the screen through a mirror mounted on the head coil. Subjects' head motion was restricted by using a foam-rubber holder. Subjects responded using a trackball mouse designed for fMRI experiments connected to the computer and placed in their right hand. Imaging was carried out on a 1.5 T scanner (GE Medical Systems, Milwaukee, Wisconsin) using gradient echo-planar imaging, sensitive to blood oxygen level-dependent (BOLD) contrast (repetition time 2000 ms, echo time 60 ms, flip angle of 90°, matrix 64 × 64, field of view 240 × 240 mm). Functional images consisted of 14 contiguous axial slices with an in-plane resolution of 3 × 3 mm, and 5 mm slice thickness covering the whole of the frontal lobe. The lower part of the temporal and occipital lobes and the cerebellum were not imaged. The first three images were discarded from further analysis to await a steady state of tissue magnetization.

During fMRI acquisition, subjects were required to perform seven separate runs of nine trials. During each run, 144 volumes were continuously acquired over a total duration of 288 s. Subjects were first familiarized with the noisy environment of the scanner and with the handling of the trackball mouse. In each run, three trials of each cognitive task (MemG and VisG) and three trials of the control task were pseudorandomly presented. Trials were separated by a 5000 ms inter-trial interval.

High-resolution T1-weighted anatomical images were acquired in the same session (gradient-echo sequence, repetition time 400 ms, echo time 2 ms, matrix 256 × 256, field of view 240 × 240 mm2, slice thickness 1.5 mm).

Data Analysis

All analyses were done with SPM'99 software (Wellcome Department of Cognitive Neurology; www.fil.ion.ucl.ac.uk/spm). For each subject, anatomical images were transformed stereotactically with nine linear rigid transformations to the Talairach and Tournoux coordinate system (Talairach and Tournoux, 1988). Functional imaging data from each run were post-processed as follows: image reconstruction, motion correction (six-parameter, rigid-body realignment) (Friston et al., 1995), normalization using the same transformations as the anatomical images and smoothing with a 5 mm full-width half-maximum Gaussian filter.

Individual and group (random effect) analyses were performed. For both analyses, data were processed using the general linear model. Each period of the task was modeled with separate delayed hemodynamic response functions (convolution of a standard hemodynamic response with a box-car function). Overall signal differences between runs were also modeled. A temporal cut-off of 560 s (i.e. twice the duration of the interval between two runs) was applied to filter subject-specific low-frequency drift, mostly related to biological rhythms and to magnetic field drift. An SPM {F} map was obtained, reflecting statistically significant activated voxels interacting with the model used (P < 0.001). To test the hypothesis 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. Separate analyses of the signal changes were performed during the different periods of the tasks (Friston, 1997; Buckner, 1998; Friston et al., 1998). Specifically, the analyses focused on signal variances during Delay 1 and Delay 2. The other task periods (stimuli presentation, instruction and response phases) were modeled but not detailed. All trials were included in the analysis, even trials on which subjects make errors.

Comparisons of interest were performed between the different memory conditions, between Delay 2 and Delay 1 in the MemG condition and between MemG and VisG conditions.

Group analyses were performed using one-sample t-tests on the contrast images resulting from individual analysis. Statistical maps were thresholded for significance by using a cluster-size criterion, which takes into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons. A cluster-size of 72 contigous voxels was used. Data were also examined for activation trends at a more liberal threshold of P < 0.001 not corrected for multiple comparison.

Signal-to-time curves were calculated on voxels presenting the highest t-values in the regions activated in the group study (parietal, premotor, DLPFC). The signals of individual data points were 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

Behavioral Performance

We found a significant ‘length of sequence’ effect on subjects' performance (F = 13.006; P = 0.0003), i.e. the longer the sequence, the lower the accuracy of the response (one square = 100%; three squares = 97.6%; five squares = 80.8%). Reaction times did not significantly differ between memory loads (one square = 1234 ms; three squares = 1271 ms; five squares = 1185 ms). No significant difference in the number of correct responses was observed when comparing memG and visG tasks (92.8 versus 95.3%, respectively; signed rank test for paired data). The reaction times were faster in the memG than in the visG task at each load level (one square, P = 0.002; three squares, P = 0.002; five squares, P = 0.03; signed rank test for paired data).

fMRI Results: Regions Activated during Delay 1

During Delay 1, we compared the cognitive (MemG + VisG) conditions to the control task. Because MemG and VisG conditions did not differ at that point, this contrast was constructed to determine the cerebral network associated with the maintenance of visuospatial (one square) or temporal-visuospatial (three squares and five squares) information in STM. When all memory conditions (one- and three- and five-square memory tasks) were compared with the control condition, a cerebral network including the right DLPFC [Brodmann area (BA) 9/46], the bilateral posterior parietal cortex (PPC; BA 7 and 40), the intraparietal sulci and precuneus, the post-central gyri, the medial [supplementary motor area (SMA)] and lateral premotor cortices (PMC, BA 6 and 8) and the anterior cingulate cortex (ACC; BA 32, 24) was observed.

To determine changes associated with the different memory loads during Delay 1, the following comparisons were performed: one-square memory condition versus the control condition, three-square versus one-square and five-square versus three-square memory conditions. When the one-square memory condition was compared with the control condition, activation was observed in the left PMC (BA 6) (Table 1, Fig. 2a). When the three-square memory condition was compared with the one-square memory condition, a significant activation in the right PPC (BA 7) and the PMC (predominantly left) was observed (Table 1, Fig. 2b). When the five-square memory condition was compared with the three-square memory task, the same brain regions were activated with an increase in size and intensity. In addition, activation in the right DLPFC was detected (Table 1, Fig. 2c).

Figure 2.

Memory conditions during Delay 1. Upper row: SPM {t} maps corresponding to the group study results during Delay 1 of the different memory conditions versus the control task. (a) one-square memory condition versus control condition; (b) three-square versus one-square memory condition; (c) five-square versus three-square memory condition. Statistical maps were thresholded for significance by using a cluster-size criterion, taking into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons. Lower row: fMRI average time-course curves for a region of interest (radius: 5 voxels) in: (d) the right superior parietal lobule (SPL); (e) the right DLPFC.__: control condition; filled squares: 5 squares; filled triangles: 3 squares; filled circles: 1 square.

Figure 2.

Memory conditions during Delay 1. Upper row: SPM {t} maps corresponding to the group study results during Delay 1 of the different memory conditions versus the control task. (a) one-square memory condition versus control condition; (b) three-square versus one-square memory condition; (c) five-square versus three-square memory condition. Statistical maps were thresholded for significance by using a cluster-size criterion, taking into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons. Lower row: fMRI average time-course curves for a region of interest (radius: 5 voxels) in: (d) the right superior parietal lobule (SPL); (e) the right DLPFC.__: control condition; filled squares: 5 squares; filled triangles: 3 squares; filled circles: 1 square.

Table 1

Comparison of activation in the different memory conditions

Memory conditions (Delay 1)
 
  One-square versus control
 
   Three-square versus one-square
 
   Five-square versus three-square
 
   
Anatomical regions Side BA Coordinates
 
  Z score Coordinates
 
  Z score Coordinates
 
  Z score 

 

 

 
x
 
y
 
z
 

 
x
 
y
 
z
 

 
x
 
y
 
z
 

 
Prefrontal cortex               
    Dorsolateral 46         48 42 27 3.34 
 10/46         39 57 18 3.82 
    Medial and cingulate 32     12 15 39 3.93 24 36 3.41 
 32         −3 27 33 3.30 
Motor & premotor cortex               
    Central sulcus     −45 −12 54 4.75     
    Lateral premotor cortex     30 −6 63 3.88     
 −33 −3 60 3.69 −18 63 4.54 −42 48 4.47 
    Superior frontal sulcus −21 57 3.29 −27 −3 48 4.29 −21 54 3.98 
    Medial premotor cortex      −9 12 51 3.91 −6 −3 63 3.88 
          −3 12 51 3.73 
Parietal cortex               
    Superior parietal lobule and precuneus     15 −57 60 4.20 −66 54 5.33 
         −6 −66 57 4.98 
    Inferior parietal lobule
 
L
 
40
 

 

 

 

 

 

 

 

 
−33
 
−39
 
45
 
3.69
 
Memory conditions (Delay 1)
 
  One-square versus control
 
   Three-square versus one-square
 
   Five-square versus three-square
 
   
Anatomical regions Side BA Coordinates
 
  Z score Coordinates
 
  Z score Coordinates
 
  Z score 

 

 

 
x
 
y
 
z
 

 
x
 
y
 
z
 

 
x
 
y
 
z
 

 
Prefrontal cortex               
    Dorsolateral 46         48 42 27 3.34 
 10/46         39 57 18 3.82 
    Medial and cingulate 32     12 15 39 3.93 24 36 3.41 
 32         −3 27 33 3.30 
Motor & premotor cortex               
    Central sulcus     −45 −12 54 4.75     
    Lateral premotor cortex     30 −6 63 3.88     
 −33 −3 60 3.69 −18 63 4.54 −42 48 4.47 
    Superior frontal sulcus −21 57 3.29 −27 −3 48 4.29 −21 54 3.98 
    Medial premotor cortex      −9 12 51 3.91 −6 −3 63 3.88 
          −3 12 51 3.73 
Parietal cortex               
    Superior parietal lobule and precuneus     15 −57 60 4.20 −66 54 5.33 
         −6 −66 57 4.98 
    Inferior parietal lobule
 
L
 
40
 

 

 

 

 

 

 

 

 
−33
 
−39
 
45
 
3.69
 

Shown are the coordinates of significant cluster maxima in the group analysis for one-square memory condition versus control condition, three-square versus one-square memory condition and five-square versus three-square memory condition. Coordinates are in millimeters, relative to the anterior commissure, corresponding to the atlas of Talairach and Tournoux. L = left; R = right; BA = Brodman's area. Statistical maps were thresholded for significance by using a cluster-size criterion, which takes into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons.

Analysis of signal-to-time curves revealed differential patterns in these regions: in the PPC (Fig. 2d) and the PMC, the signal progressively increased from the one- to the five-square memory tasks, whereas in the right DLPFC, signal increased much more in the five-square memory condition than in the other two (Fig. 2e).

fMRI Results: Regions Activated during Delay 2

MemG Condition (one- and three- and five-square Conditions) Compared with the Control Condition

This comparison was aimed at describing the activation observed in classical delayed response tasks since both maintenance in STM and organization of the response were present in MemG condition during Delay 2. This comparison revealed activation already observed during Delay 1, namely the left PPC (BA 7 and 40), the medial and lateral PMC (BA 6 and 8), the ACC (BA 32 and 24) in both hemispheres and the right DLPFC (BA 9/46), and an additional activation in the left posterior DLPFC (BA 9).

To determine the cerebral network associated with the putative additional processes specific to Delay 2 for the MemG condition, we compared Delay 2 to Delay 1 of MemG and vice versa. As the maintenance processes are common to the two delay periods in MemG trials, any activation observed in this comparison should be related to the organization of the forthcoming action indicated by the instruction. Activation was found in the left inferior PPC (BA 40), the SMA, the PMC and the sensorimotor cortex with predominance in the left hemisphere, the medial frontal area (BA 8/9), the ACC (BA 32), the striatum (caudate nuclei and putamen) and the left DLPFC (BA 9; Table 2, Fig. 3). No significant activation was observed when Delay 1 was compared with Delay 2.

Figure 3.

Delay 2 versus Delay 1 during the MemG task. SPM {t} maps are superimposed upon sagittal (upper left), coronal (upper right) and axial (lower left) sections of a standard normalized brain. Statistical maps were thresholded for significance by using a cluster-size criterion, taking into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons. Sections are centered on the dorsolateral prefrontal cortex (a), the supplementary motor area (b), the inferior parietal area (c), the left (d) and right (e) motor/premotor areas and the striatum (f).

Figure 3.

Delay 2 versus Delay 1 during the MemG task. SPM {t} maps are superimposed upon sagittal (upper left), coronal (upper right) and axial (lower left) sections of a standard normalized brain. Statistical maps were thresholded for significance by using a cluster-size criterion, taking into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons. Sections are centered on the dorsolateral prefrontal cortex (a), the supplementary motor area (b), the inferior parietal area (c), the left (d) and right (e) motor/premotor areas and the striatum (f).

Table 2

Delay 2 versus Delay 1 of the MemG condition and MemG versus VisG condition

MemG condition (Delay 2)
 
  versus MemG Delay 1
 
   versus VisG Delay 2
 
   
Anatomical regions Side BA Coordinates
 
  Z scores Coordinates
 
  Z scores 

 

 

 
x
 
y
 
z
 

 
x
 
y
 
z
 

 
Dorsolateral prefrontal area (middle frontal gyrus) 8/9 −42 27 42 3.57     
 −51 15 36 3.35     
Ventrolateral prefrontal area (inferior frontal gyrus) 47 −33 30 −3 3.66     
Medial prefrontal area  39 42 3.74     
  8/9     33 36 3.79 
Anterior cingulate cortex 24/32 −9 39 −3 4.05     
 32     39 18 4.24 
Central sulcus −39 −24 63 4.66     
 39 −24 51 3.20     
Precentral gyrus −12 −27 69 3.97     
Medial premotor cortex 6–8 −6 −15 48 4.16     
Inferior parietal cortex 40 57 −30 27 3.71     
 40 −54 −36 42 3.39     
Parietal operculum 40 −42 −18 15 3.93     
Postcentral gyrus 1–3 57 −21 48 3.64     
 1–3 −51 −21 42 3.78     
Posterior insula  48 −12 3.45     
  −48 −6 3.48     
Striatum
 
L
 

 
−12
 
15
 
0
 
3.80
 

 

 

 

 
MemG condition (Delay 2)
 
  versus MemG Delay 1
 
   versus VisG Delay 2
 
   
Anatomical regions Side BA Coordinates
 
  Z scores Coordinates
 
  Z scores 

 

 

 
x
 
y
 
z
 

 
x
 
y
 
z
 

 
Dorsolateral prefrontal area (middle frontal gyrus) 8/9 −42 27 42 3.57     
 −51 15 36 3.35     
Ventrolateral prefrontal area (inferior frontal gyrus) 47 −33 30 −3 3.66     
Medial prefrontal area  39 42 3.74     
  8/9     33 36 3.79 
Anterior cingulate cortex 24/32 −9 39 −3 4.05     
 32     39 18 4.24 
Central sulcus −39 −24 63 4.66     
 39 −24 51 3.20     
Precentral gyrus −12 −27 69 3.97     
Medial premotor cortex 6–8 −6 −15 48 4.16     
Inferior parietal cortex 40 57 −30 27 3.71     
 40 −54 −36 42 3.39     
Parietal operculum 40 −42 −18 15 3.93     
Postcentral gyrus 1–3 57 −21 48 3.64     
 1–3 −51 −21 42 3.78     
Posterior insula  48 −12 3.45     
  −48 −6 3.48     
Striatum
 
L
 

 
−12
 
15
 
0
 
3.80
 

 

 

 

 

Shown are the coordinates of significant cluster maxima in the group analysis for the MemG condition when Delay 2 was compared with Delay 1 and for Delay 2 of the MemG condition versus Delay 2 of the VisG condition. Coordinates are in millimeters, relative to the anterior commissure, corresponding to the atlas of Talairach and Tournoux. L = left; R = right; SMA = supplementary motor area. Statistical maps were thresholded for significance by using a cluster-size criterion, which takes into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons.

To verify that activation in the right DLPFC was more specifically associated with maintenance and activation in the left DLPFC was more specifically associated with the organization of the response, we analyzed activation in these areas using a more liberal threshold preventing for type-II errors (P = 0.001 uncorrected, no spatial thresholding). Using this threshold, we looked for activation in a mask of 10 mm radius centered on the area of activation peak obtained in the right DLPFC during contrast five- versus three-square conditions (load effect in maintenance) and in the left DLPFC during organization of the response (contrast Delay 2 versus Delay 1 in the memG condition). The contrast of the five-versus three-square memory conditions showed no activated voxels in the left DLPFC mask. Moreover, no significant voxels was detected in the right DLPFC mask for Delay 2 versus Delay 1 contrast.

Figure 4a shows patterns of activation for the MemG condition during Delay 1 (red) and Delay 2 (green).

Figure 4.

(a) Comparison of the network involved in maintenance of information (red) and the network involved in the response preparation (green). The brain network for maintenance of information (red) was obtained by subtracting activation for the control task from that of the MemG and VisG conditions during Delay 1. The network for response preparation (green) was obtained by subtracting activation of Delay 2 from that of Delay 1 in the MemG task: it included the left DLPFC, the lateral and medial PMC and the sensorimotor cortex with a left predominance, the left inferior PPC, the ACC and the striatum. It is noteworthy that, compared with the maintenance network, these additional regions recruited during Delay 2 (when the subjects could prepare their response) are more inferior in the parietal areas, left-sided in the DLPFC and more posterior in the premotor areas. (b) Time-course curves within the left DLPFC (Talairach coordinates −42, 27, 42) for the memG task performed at the five-square (green line) and the control (grey line) conditions. Asterisks indicate time points that are statistically significant in the memG versus control tasks (P < 0.05, Student's t-test, one-tailed). (c) Time-course curves in the right DLPFC (Talairach coordinates 48, 42, 27) for the memG task performed at the five-square (red line) and the control (grey line) conditions. Asterisks indicate time points that are statistically significant in the memG versus control tasks (P < 0.05, Student's t-test, one-tailed).

Figure 4.

(a) Comparison of the network involved in maintenance of information (red) and the network involved in the response preparation (green). The brain network for maintenance of information (red) was obtained by subtracting activation for the control task from that of the MemG and VisG conditions during Delay 1. The network for response preparation (green) was obtained by subtracting activation of Delay 2 from that of Delay 1 in the MemG task: it included the left DLPFC, the lateral and medial PMC and the sensorimotor cortex with a left predominance, the left inferior PPC, the ACC and the striatum. It is noteworthy that, compared with the maintenance network, these additional regions recruited during Delay 2 (when the subjects could prepare their response) are more inferior in the parietal areas, left-sided in the DLPFC and more posterior in the premotor areas. (b) Time-course curves within the left DLPFC (Talairach coordinates −42, 27, 42) for the memG task performed at the five-square (green line) and the control (grey line) conditions. Asterisks indicate time points that are statistically significant in the memG versus control tasks (P < 0.05, Student's t-test, one-tailed). (c) Time-course curves in the right DLPFC (Talairach coordinates 48, 42, 27) for the memG task performed at the five-square (red line) and the control (grey line) conditions. Asterisks indicate time points that are statistically significant in the memG versus control tasks (P < 0.05, Student's t-test, one-tailed).

In addition, in the five-square memG condition, signals-to-time curves were different in the left and right DLPFC (Fig. 4b,c). In the left DLPFC (Talairach coordinates −42, 27, 42), activation for the five-square memG condition was significantly higher than that of the control task during Delay 2 (but not during Delay 1) (t-test; Fig. 4b). In the right DLPFC (Talairach coordinates 48, 42, 27), activation for the five-square memG condition was significantly higher than that of the control condition for the entire task course (t-test; Fig. 4c).

Comparison of Memory-guided and Visually-guided Trials during Delay 2

To specify differences in the cerebral networks between memory- and visually-guided organization of action, we compared the estimates obtained during Delay 2 for the MemG condition to those obtained for the VisG condition. When the MemG task was compared with the VisG task, activation was only found in the left and right medial frontal cortices (BA 8, 9 and 32; Table 2, Fig. 5a). Even at a liberal threshold (P = 0.001 uncorrected) and when looking for activation in a mask of 10 mm radius centered on the area of activation peak obtained in the left DLPFC during the Delay 2 versus Delay 1 contrast, no voxels were detected in the DLPFC mask. The reverse comparison (VisG versus MemG task) did not reveal any significant activation. Taken together, these data indicate that during Delay 2, MemG and VisG tasks shared an overlapping brain network, with the exception of the medial frontal regions, which were more activated in the MemG condition.

Figure 5.

(a) Comparison of MemG and VisG tasks during Delay 2. SPM {t} maps are superimposed on sagittal, coronal and axial sections of a standard normalized brain. Significant activation was observed in the medial prefrontal area. Statistical maps were thresholded for significance by using a cluster-size criterion, taking into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons. (b, c, d) Time-course curves in the left DLPFC (Talairach coordinates −42, 27, 42) for the memG task (blue line) and the visG task (orange line) performed in the five-square (b), three-square (c) and one-square (d) conditions.

Figure 5.

(a) Comparison of MemG and VisG tasks during Delay 2. SPM {t} maps are superimposed on sagittal, coronal and axial sections of a standard normalized brain. Significant activation was observed in the medial prefrontal area. Statistical maps were thresholded for significance by using a cluster-size criterion, taking into account the spatial extent of activation to obtain a threshold of 0.05 corrected for multiple comparisons. (b, c, d) Time-course curves in the left DLPFC (Talairach coordinates −42, 27, 42) for the memG task (blue line) and the visG task (orange line) performed in the five-square (b), three-square (c) and one-square (d) conditions.

No time points out of the signal-to-time curves centered on the left DLPFC (coordinates −42, 27, 42) showed statistical differences between memG and visG conditions (t-test; for one-, three- or five-square conditions) during Delay 2, again indicating that the two conditions equally activated the left DLPFC (Fig. 5b,c,d).

Discussion

Three partially distinct activation patterns were evidenced in this study. The first pattern was observed during Delay 1, with a ‘step-like activation’ of the right DLPFC (i.e. this region showed significant activation only when a sequence of five squares had to be maintained). The second distinct pattern included the activation of the left DLPFC when Delay 2 was compared with Delay 1 during the MemG condition. The third pattern was defined by increased activation of the medial PFC (BA 8/9) and the ACC (BA 32) during the memG when compared with the visG task.

Active Short-term Maintenance and the DLPFC (Delay 1)

A modulation of activation during Delay 1 was observed as a function of the amount of information to be held in memory. Activation in the superior PPC, the lateral and medial PMC and the DLPFC increased with memory load in accordance with previous reports (Braver et al., 1997; Jonides et al., 1997; Jansma et al., 2000; Leung et al., 2002; Pochon et al., 2002). In the PPC and PMC, activation increase was progressive (Fig. 2d). It has been suggested that this parietalpremotor network is the mnemonic buffer for visuospatial WM (Cohen et al., 1997; Smith and Jonides, 1998, 1999; Pochon et al., 2001) and that it overlaps with the visuospatial attentional network (Awh et al., 1995; Awh and Jonides, 2001; Corbetta et al., 2002). These observations suggest that the PPC and the PMC handle information during a brief delay, and that this active maintenance is associated with visuospatial attentional processes. By contrast, time series revealed a significant increase in DLPFC activity for maintenance of five items (as compared with maintenance of one and three items) (Fig. 2c). A step-like activation of the DLPFC as a function of memory load has already been reported during n-back WM tasks (Cohen et al., 1997; Jonides et al., 1997). The increase in DLPFC activation may also result from an increase in the executive processes required for updating and monitoring information in WM and for managing proactive interferences. Here, we show this load effect for the maintenance processes per se, but only when the quantity of information to be maintained surpassed a certain level. We found that the maintenance of five items was associated with poorer performance as compared with the maintenance of one or three items. For the five items condition, maintenance in STM may exceed the capacity of the spatial buffer of the parietalpremotor network and may require ‘strategies’ under the control of the right DLPFC. This interpretation is in line with the suggestion of Rypma et al. (2002) that manipulation may be necessary when memory load exceeds maintenance capacity, resulting in a shift to alternative mnemonic strategies. This is also in agreement with the suggestion of Sakai et al. (2002) that DLPFC activity enhances the strength of memory, allowing a resistance to interference, and may thereby expand the storage capacity in STM. Building a temporal ordering of the mental representations or ‘chunking’ (linking and compression of items in memory) have been proposed as the possible expression of these ‘strategic processes’ for supracapacity maintenance (Rypma and D'Esposito, 1999; Phillips and Niki, 2002; Rypma et al., 2002). However, activation during Delay 1 related to STM could be partially contaminated by activation related to encoding, as Delay 1 began just after the encoding period and lasted 8 s. Neuropsychological (Dubois et al., 1995; Ferreira et al., 1998) and functional imaging (Rypma and D'Esposito, 1999) studies have suggested that the DLPFC plays a significant role in the encoding phase in high load WM tasks. Therefore, a conservative interpretation of our results would be that the supracapacity load effect observed in the right DLPFC is related to maintenance as well as encoding, but only for a high memory load. However, it is noteworthy that other studies on DLPFC activity evidenced the specificity of this load effect in the maintenance process. By increasing the length of the delay in order to avoid contamination of one phase of the task by another, a specific activation of the DLPFC was observed in relation to maintenance processing (Jha and McCarthy, 2000; Leung et al., 2002), and more specifically to the load effect in STM. In addition, Rypma et al. (2002) showed that DLPFC activation was related to a load effect in STM (maintenance phase) by analyzing different periods of delayed-response tasks. Finally, although the right-sided activation of the DLPFC may, at first glance, appear to be explained by the spatial nature of the task (Jonides et al., 1993; McCarthy et al., 1996; Smith et al., 1996), studies using verbal (and non-spatial) delayed-response tasks found a load effect in the right DLPFC (Manoach et al., 1997; Rypma et al., 1999). Therefore, the right-sided activation of the DLPFC might be more related to strategic processes required by load increase rather than to the spatial nature of the material.

Organization of the Forthcoming Response and the DLPFC (Delay 2)

Three main differences were found when the activation in Delay 1 was subtracted from that of Delay 2: (i) an activation in the left DLPFC; (ii) a more caudal and predominantly left activation of the PMC, the primary motor areas and the striatum; and (iii) an activation in the inferior PPC (gyrus supramarginalis; BA 40). In the MemG task during Delay 2, two processes occurred: maintenance and organization of the response. Since maintenance of visual information was common to the two delay periods, the activated areas were likely to be related to the organization of the forthcoming response. This suggests that the concept of WM should be extended to a motor WM component. fMRI and neuropsychological studies have already suggested that the DLPFC is essential for the selection or the organization of the response rather than being involved in sensory maintenance (Ferreira et al., 1998; Pochon et al., 2001; Manoach et al., 2003; Curtis et al., 2004) and for the selection of items in WM to perform the appropriate response (Rowe et al., 2000; Rowe and Passingham, 2001). More specifically, the left DLPFC seems to be preferentially involved in response organization and motor preparation (Simon et al., 2002). Several processes could occur during Delay 2 in association with response organization: selection, planning, inhibition, programming and maintenance of motor representations. Our paradigm did not allow us to isolate the specific operations involved in the processing of ‘response organization.’ However, as all items were necessary for the response in the MemG task, no selection among items in WM was required. It is thus likely that the left lateralization of activation in the DLPFC (as well as in the PMC and PPC) may result from the sequential structure and complexity of the motor response to be organized, as already reported (Barone and Joseph, 1989; Jueptner et al., 1997; Rushworth et al., 1998; Histed, 1999; Harrington et al., 2000; Jenkins et al., 2000), or from the higher executive demand (Henson et al., 1999a,b; Fletcher et al., 2000; Fletcher and Henson, 2001; Nathaniel-James and Frith, 2002). The design of our paradigm cannot totally rule out the possibility of a left DLPFC activation in relation to the preparation of a right-hand movement. However, damage to the left DLPFC disrupts the performance of movement of both the contralateral and the ipsilateral limbs (De Renzi et al., 1983; Jason, 1983; Harrington and Haaland, 1991; Haaland and Harrington, 1994) supporting a left hemispheric dominance for movement. It is unlikely that this left lateralization was due to verbalization since none of our subjects reported a verbal strategy for either the maintenance of information or for the organization of their response. Finally, activation occurring during Delay 2 seems unlikely to be related to the processing of the prior instruction (sensory integration and maintenance) because: (i) activation related to Delay 2 was found in motor and posterior premotor regions; and (ii) it has been shown that within the lateral PMC, the motor related activation was located caudally (falling within the areas activated during Delay 2 of our study), as compared with a more rostral activation for STM (Simon et al., 2002). Taken together, our data emphasize differences within the DLPFC according to the nature of the executive process performed: the right DLPFC being preferentially involved in strategies for maintenance of information and the left DLPFC in response organization.

Memory-guided versus Visually-guided Tasks

Subtracting VisG task activation from MemG task activation did not reveal any significant differences in the DLPFC, and time series showed a similar profile of left DLPFC activity for both tasks. This result may reflect a specific involvement of this region in the executive aspects of the tasks (Stuss and Alexander, 1994), regardless of whether or not the task could be classified as a WM task. Indeed, other studies have indicated that activation in the left DLPFC can be elicited outside the context of WM: activation in this area has been observed in free movement selection in tasks with virtually no WM (Frith et al., 1991; Jueptner et al., 1997; Jenkins et al., 2000; Rushworth et al., 2001a,b). In addition, Rushworth et al. (2001a,b) interpreted the preparatory activity during the delay of a WM task as covert attention oriented to a selective limb movement.

Comparison of MemG and VisG tasks during Delay 2 revealed difference only in the medial frontal cortex, which was more activated in the MemG task. The medial frontal areas have been associated with selection, preparation and execution of motor responses (Botvinick et al., 1999; Turken and Swick, 1999; Carter et al., 2000). While both tasks made the same demand on motor planning and sequencing, and showed no differences in performance, the MemG task relied on an internal representation, whereas the VisG task was based on external cues. These data can thus be interpreted in light of internally versus externally generated actions. The involvement of the medial prefrontal area, the rostral SMA and the ACC in self-initiated movements, as compared with externally triggered movements, has often been emphasized (Deiber et al., 1996, 1999; Jueptner et al., 1997; Jenkins et al., 2000; Weeks et al., 2001; Cunnington et al., 2002; Lau et al., 2004). It is possible that the differential activation in the medial frontal cortex represents the neural substrate that bridges internalized information necessary for one given action and the organization of this particular action. Along the same lines, our results are in accordance with those of Petit et al. (1998), who showed that sustained activity in the pre-SMA and the ACC during the delay period of WM tasks reflected a state of readiness for a motor response based on the information held on-line, but was not elicited when subjects waited for a cue to make a predefined manual response.

Alternatively, in the memG condition, the activation in the ACC may also be explained by its involvement in the monitoring of errors, the detection of conflict and the selection between competitive responses (Botvinick et al., 1999; Carter et al., 2000; MacDonald et al., 2000; Kerns et al., 2004). Although these hypotheses remain possible, it is noteworthy that the number of errors was not higher and that reaction times were even shorter (a result not expected if subjects should manage competition between responses) in the memG as compared with the visG condition.

Finally, although Sweeney et al. (1996) showed that activation in the medial frontal region was associated with response inhibition when memory-guided saccades were compared with reflexive saccades or fixation tasks, the areas they found to be activated were more ventral than in our study. Furthermore, inhibition was similar in the two tasks that we contrasted, as none were reflexive, both were volitional and both included a delay between the instruction and the go-signal.

Conclusion

Is the human DLPFC recruited when information has to be maintained in WM and/or when the response has to be organized on the basis of information maintained in WM? The ‘double delay/double response’ paradigm showed the DLPFC to be involved in the most executive operations of visuospatial WM (strategies for maintenance and organization of the forthcoming response). Our study also emphasizes relative hemispheric differences within the DLPFC according to the nature of the executive process performed (strategies for maintenance of information in the right DLPFC and response organization in the left DLPFC). Finally, the organization of the response may differentially activate medial PFC areas depending on the source of information that precedes response elaboration.

References

Awh E, Jonides J (
2001
) Overlapping mechanisms of attention and spatial working memory.
Trends Cogn Sci
 
5
:
119
–126.
Awh E, Smith EE, Jonides J (
1995
) Human rehearsal processes and the frontal lobes: PET evidence.
Ann N Y Acad Sci
 
769
:
97
–117.
Baddeley A (
1996
) The fractionation of working memory.
Proc Natl Acad Sci USA
 
93
:
13468
–13472.
Baddeley A (
1998
) Recent developments in working memory.
Curr Opin Neurobiol
 
8
:
234
–238.
Baddeley A (
2003
) Working memory: looking back and looking forward.
Nat Rev Neurosci
 
4
:
829
–839.
Barone P, Joseph JP (
1989
) Role of the dorsolateral prefrontal cortex in organizing visually-guided behavior.
Brain Behav Evol
 
33
:
132
–135.
Botvinick M, Nystrom LE, Fissell K, Carter CS, Cohen JD (
1999
) Conflict monitoring versus selection-for-action in anterior cingulate cortex.
Nature
 
402
:
179
–181.
Boussaoud D, Wise SP (
1993
) Primate frontal cortex: neuronal activity following attentional versus intentional cues.
Exp Brain Res
 
95
:
15
–27.
Braver TS, Cohen JD, Nystrom LE, Jonides J, Smith EE, Noll DC (
1997
) A parametric study of prefrontal cortex involvement in human working memory.
Neuroimage
 
5
:
49
–62.
Buckner RL (
1998
) Event-related fMRI and the hemodynamic response.
Hum Brain Mapp
 
6
:
373
–377.
Carter CS, Macdonald AM, Botvinick M, Ross LL, Stenger VA, Noll D, Cohen JD (
2000
) Parsing executive processes: strategic vs. evaluative functions of the anterior cingulate cortex.
Proc Natl Acad Sci
 
97
:
1944
–1948.
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.
Cohen JD, Perlstein WM, Braver TS, Nystrom LE, Noll DC, Jonides J, Smith EE (
1997
) Temporal dynamics of brain activation during a working memory task.
Nature
 
386
:
604
–608.
Corbetta M, Kincade JM, Shulman GL (
2002
) Neural systems for visual orienting and their relationships to spatial working memory.
J Cogn Neurosci
 
14
:
508
–523.
Courtney SM, Petit L, Haxby JV, Ungerleider LG (
1998
) The role of prefrontal cortex in working memory: examining the contents of consciousness.
Philos Trans R Soc Lond B Biol Sci
 
353
:
1819
–1828.
Cunnington R, Windischberger C, Deecke L, Moser E (
2002
) The preparation and execution of self-initiated and externally-triggered movement: a study of event-related fMRI.
Neuroimage
 
15
:
373
–385.
Curtis CE, Rao VY, D'Esposito M (
2004
) Maintenance of spatial and motor codes during oculomotor delayed response tasks.
J Neurosci
 
24
:
3944
–3952.
Deiber MP, Honda M, Ibanez V, Sadato N, Hallett M (
1999
) Mesial motor areas in self-initiated versus externally triggered movements examined with fMRI: effect of movement type and rate.
J Neurophysiol
 
81
:
3065
–3077.
Deiber MP, Ibanez V, Sadato N, Hallett M (
1996
) Cerebral structures participating in motor preparation in humans: a positron emission tomography study.
J Neurophysiol
 
75
:
233
–247.
De Renzi E, Faglioni P, Lodesani M, Vecchi A (
1983
) Performance of left brain-damaged patients on imitation of single movements and motor sequences. Frontal and parietal-injured patients compared.
Cortex
 
19
:
333
–343.
D'Esposito M, Grossman M (
1996
) The physiological basis of executive function and working memory.
Neuroscientist
 
2
:
345
–352.
D'Esposito M, Postle BR, Rypma B (
2000
a) Prefrontal cortical contributions to working memory: evidence from event-related fMRI studies.
Exp Brain Res
 
133
:
3
–11.
D'Esposito M, Ballard D, Zarahn E, Aguirre GK (
2000
b) The role of prefrontal cortex in sensory memory and motor preparation: an event-related fMRI study.
Neuroimage
 
11
:
400
–408.
di Pellegrino G, Wise SP (
1991
) A neurophysiological comparison of three distinct regions of the primate frontal lobe.
Brain
 
114
:
951
–978.
Dubois B, Levy R, Verin M, Teixeira C, Agid Y, Pillon B (
1995
) Experimental approach to prefrontal functions in humans.
Ann N Y Acad Sci
 
769
:
41
–60.
Duncan J, Owen AM (
2000
) Common regions of the human frontal lobe recruited by diverse cognitive demands.
Trends Neurosci
 
23
:
475
–483.
Ferreira CT, Verin M, Pillon B, Levy R, Dubois B, Agid Y (
1998
) Spatio-temporal working memory and frontal lesions in man.
Cortex
 
34
:
83
–98.
Fletcher PC, Henson RN (
2001
) Frontal lobes and human memory: Insights from functional neuroimaging.
Brain
 
124
:
849
–881.
Fletcher PC, Shallice T, Dolan RJ (
2000
) ‘Sculpting the response space’ — an account of left prefrontal activation at encoding.
Neuroimage
 
12
:
404
–417.
Friston K (
1997
) Imaging cognitive anatomy.
Trends Cogn Sci
 
1
:
21
–27.
Friston K, Holmes A, Worsley K, Poline J, Frith C, Frackowiak R (
1995
) Statistical parametric maps in functional imaging: a general linear approach.
Hum Brain Mapp
 
2
:
189
–210.
Friston KJ, Fletcher P, Josephs O, Holmes A, Rugg MD, Turner R (
1998
) Event-related fMRI: characterizing differential responses.
Neuroimage
 
7
:
30
–40.
Frith CD, Friston K, Liddle PF, Frackowiak RS (
1991
) Willed action and the prefrontal cortex in man: a study with PET.
Proc R Soc Lond B Biol Sci
 
244
:
241
–246.
Funahashi S, Chafee MV, 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.
Fuster JM (
2000
) Executive frontal functions.
Exp Brain Res
 
133
:
66
–70.
Goldman-Rakic P (
1987
) Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. In: Handbook of physiology (Plum F, Mouncastle U, eds), Vol. 5, pp. 373–417. Washington, DC: The American Physiological Society.
Goldman-Rakic PS (
1995
) Architecture of the prefrontal cortex and the central executive.
Ann N Y Acad Sci
 
769
:
71
–83.
Haaland KY, Harrington DL (
1994
) Limb-sequencing deficits after left but not right hemisphere damage.
Brain Cogn
 
24
:
104
–122.
Harrington DL, Haaland KY (
1991
) Hemispheric specialization for motor sequencing: abnormalities in levels of programming.
Neuropsychologia
 
29
:
147
–163.
Harrington DL, Rao SM, Haaland KY, Bobholz JA, Mayer AR, Binderx JR, Cox RW (
2000
) Specialized neural systems underlying representations of sequential movements.
J Cogn Neurosci
 
12
:
56
–77.
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.
Henson RN, Rugg MD, Shallice T, Josephs O, Dolan RJ (
1999
a) Recollection and familiarity in recognition memory: an event-related functional magnetic resonance imaging study.
J Neurosci
 
19
:
3962
–3972.
Henson RN, Shallice T, Dolan RJ (
1999
b) Right prefrontal cortex and episodic memory retrieval: a functional MRI test of the monitoring hypothesis.
Brain
 
122
:
1367
–1381.
Histed M (
1999
) Sequences in spatial working memory.
bug journal
 
2
:
1
–9.
Jansma JM, Ramsey NF, Coppola R, Kahn RS (
2000
) Specific versus nonspecific brain activity in a parametric N-back task.
Neuroimage
 
12
:
688
–697.
Jason GW (
1983
) Hemispheric asymmetries in motor function. I. Left-hemisphere specialization for memory but not performance.
Neuropsychologia
 
21
:
35
–45.
Jenkins IH, Jahanshahi M, Jueptner M, Passingham RE, Brooks DJ (
2000
) Self-initiated versus externally triggered movements. II. The effect of movement predictability on regional cerebral blood flow.
Brain
 
123
:
1216
–1228.
Jha AP, McCarthy G (
2000
) The influence of memory load upon delay-interval activity in a working-memory task: an event-related functional MRI study.
J Cogn Neurosci
 
12
(Suppl 2):
90
–105.
Jonides J, Smith EE, Koeppe RA, Awh E, Minoshima S, Mintun MA (
1993
) Spatial working memory in humans as revealed by PET.
Nature
 
363
:
623
–625.
Jonides J, Schumacher E, Smith E, Lauber E, Awh E, Minoshima S, Koeppe R (
1997
) Verbal working memory load affects regional brain activation as measured by PET.
J Cogn Neurosci
 
9
:
462
–475.
Jueptner M, Stephan KM, Frith CD, Brooks DJ, Frackowiak RS, Passingham RE (
1997
) Anatomy of motor learning. I. Frontal cortex and attention to action.
J Neurophysiol
 
77
:
1313
–1324.
Kerns JG, Cohen JD, MacDonald AW, Cho RY, Stenger VA, Carter CS (
2004
) Anterior cingulate conflict monitoring and adjustments in control.
Science
 
303
:
1023
–1026.
Lau HC, Rogers RD, Ramnani N, Passingham RE (
2004
) Willed action and attention to the selection of action.
Neuroimage
 
21
:
1407
–1415.
Leung HC, Gore JC, Goldman-Rakic PS (
2002
) Sustained mnemonic response in the human middle frontal gyrus during on-line storage of spatial memoranda.
J Cogn Neurosci
 
14
:
659
–671.
MacDonald AW, Cohen JD, Stenger VA, Carter CS (
2000
) Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control.
Science
 
288
:
1835
–1838.
Manoach DS, Schlaug G, Siewert B, Darby DG, Bly BM, Benfield A, Edelman RR, Warach S (
1997
) Prefrontal cortex fMRI signal changes are correlated with working memory load.
Neuroreport
 
8
:
545
–549.
Manoach DS, Greve DN, Lindgren KA, Dale AM (
2003
) Identifying regional activity associated with temporally separated components of working memory using event-related functional MRI.
Neuroimage
 
20
:
1670
–1684.
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.
Miller EK (
1999
) The prefrontal cortex: complex neural properties for complex behavior.
Neuron
 
22
:
15
–17.
Miller EK, Li L, Desimone R (
1991
) A neural mechanism for working and recognition memory in inferior temporal cortex.
Science
 
254
:
1377
–1379.
Miller EK, Li L, Desimone R (
1993
) Activity of neurons in anterior inferior temporal cortex during a short-term memory task.
J Neurosci
 
13
:
1460
–1478.
Nathaniel-James DA, Frith CD (
2002
) The role of the dorsolateral prefrontal cortex: evidence from the effects of contextual constraint in a sentence completion task.
Neuroimage
 
16
:
1094
–1102.
Niki H, Watanabe M (
1976
) Prefrontal unit activity and delayed response: relation to cue location versus direction of response.
Brain Res
 
105
:
79
–88.
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.
Owen AM (
2000
) The role of the lateral frontal cortex in mnemonic processing: the contribution of functional neuroimaging.
Exp Brain Res
 
133
:
33
–43.
Petit L, Courtney SM, Ungerleider LG, Haxby JV (
1998
) Sustained activity in the medial wall during working memory delays.
J Neurosci
 
18
:
9429
–9437.
Phillips S, Niki K (
2002
) Separating relational from item load effects in paired recognition: temporoparietal and middle frontal gyral activity with increased associates, but not items during encoding and retention.
Neuroimage
 
17
:
1031
–1055.
Pochon JB, Levy R, Poline JB, Crozier S, Lehéricy S, Pillon B, Deweer B, Le Bihan D, Dubois B (
2001
) The role of dorsolateral prefrontal cortex in the preparation of forthcoming actions: an fMRI study.
Cereb Cortex
 
11
:
260
–266.
Pochon JB, Levy R, Fossati P, Lehéricy S, Poline JB, Pillon B, Le Bihan D, Dubois B (
2002
) The neural system that bridges reward and cognition in humans: an fMRI study.
Proc Natl Acad Sci USA
 
99
:
5669
–5674.
Quintana J, Fuster JM (
1999
) From perception to action: temporal integrative functions of prefrontal and parietal neurons.
Cereb Cortex
 
9
:
213
–221.
Rowe JB, Passingham RE (
2001
) Working memory for location and time: activity in prefrontal area 46 relates to selection rather than maintenance in memory.
Neuroimage
 
14
:
77
–86.
Rowe JB, Toni I, Josephs O, Frackowiak RS, Passingham RE (
2000
) The prefrontal cortex: response selection or maintenance within working memory?
Science
 
288
:
1656
–1660.
Rushworth MF, Nixon PD, Wade DT, Renowden S, Passingham RE (
1998
) The left hemisphere and the selection of learned actions.
Neuropsychologia
 
36
:
11
–24.
Rushworth MF, Ellison A, Walsh V (
2001
a) Complementary localization and lateralization of orienting and motor attention.
Nat Neurosci
 
4
:
656
–661.
Rushworth MF, Krams M, Passingham RE (
2001
b) The attentional role of the left parietal cortex: the distinct lateralization and localization of motor attention in the human brain.
J Cogn Neurosci
 
13
:
698
–710.
Rypma B, D'Esposito M (
1999
) The roles of prefrontal brain regions in components of working memory: effects of memory load and individual differences.
Proc Natl Acad Sci USA
 
96
:
6558
–6563.
Rypma B, Prabhakaran V, Desmond JE, Glover GH, Gabrieli JD (
1999
) Load-dependent roles of frontal brain regions in the maintenance of working memory.
Neuroimage
 
9
:
216
–226.
Rypma B, Berger JS, D'Esposito M (
2002
) The influence of working-memory demand and subject performance on prefrontal cortical activity.
J Cogn Neurosci
 
14
:
721
–731.
Sakai K, Rowe JB, Passingham RE (
2002
) Active maintenance in prefrontal area 46 creates distractor-resistant memory.
Nat Neurosci
 
5
:
479
–484.
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.
Simon SR, Meunier M, Piettre L, Berardi AM, Segebarth CM, Boussaoud D (
2002
) Spatial attention and memory versus motor preparation: premotor cortex involvement as revealed by fMRI.
J Neurophysiol
 
88
:
2047
–2057.
Smith EE, Jonides J (
1998
) Neuroimaging analyses of human working memory.
Proc Natl Acad Sci USA
 
95
:
12061
–12068.
Smith EE, Jonides J (
1999
) Storage and executive processes in the frontal lobes.
Science
 
283
:
1657
–1661.
Smith EE, Jonides J, Koeppe RA (
1996
) Dissociating verbal and spatial working memory using PET.
Cereb Cortex
 
6
:
11
–20.
Stuss DT, Alexander MP (
1994
) Functional and anatomical specificity of frontal lobe functions. New York: Plenum Press.
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
) Co-planar stereotaxic atlas of the human brain: an approach to cerebral imaging. New York: Thieme Verlag.
Toni I, Schluter ND, Josephs O, Friston K, Passingham RE (
1999
) Signal-, set- and movement-related activity in the human brain: an event-related fMRI study.
Cereb Cortex
 
9
:
35
–49.
Toni I, Thoenissen D, Zilles K, Niedeggen M (
2002
) Movement preparation and working memory: a behavioural dissociation.
Exp Brain Res
 
142
:
158
–162.
Turken U, Swick D (
1999
) Response selection in the human anterior cingulate cortex.
Nat Neurosci
 
2
:
920
–924.
Ungerleider LG, Courtney SM, Haxby JV (
1998
) A neural system for human visual working memory.
Proc Natl Acad Sci USA
 
95
:
883
–890.
Weeks RA, Honda M, Catalan MJ, Hallett M (
2001
) Comparison of auditory, somatosensory, and visually instructed and internally generated finger movements: a PET study.
Neuroimage
 
14
:
219
–230.

Author notes

1INSERM U.610, Hôpital de la Salpêtrière, 47 boulevard de l'Hôpital, 75013 Paris, France and 2Department of Neuroradiology, Hôpital de la Salpêtrière, 47 boulevard de l'Hôpital, 75013 Paris, France