Abstract

The anatomical and functional organization of the lateral prefrontal cortex (LPFC) is one of the most debated issues in cognitive and integrative neurosciences. The aim of this study is to determine whether the human LPFC is organized according to the domain of information, to the level of the processing or to both of these dimensions. In order to clarify this issue, we have designed an experimental protocol that combines a functional magnetic resonance imaging study in healthy subjects (n = 12) and a voxel-by-voxel lesion mapping study in patients with focal prefrontal lesions (n = 37) compared with normal controls (n = 48). Each method used the same original cognitive paradigm (“the domain n-back tasks”) that tests by a cross-dimensional method the domain of information (verbal, spatial, faces) and the level of processing (from 1- to 3-back). Converging data from the 2 methods demonstrate that the left posterior LPFC is critical for the higher levels of cognitive control and is organized into functionally different subregions (Brodman's area 9/46, 6/8/9, and 44/45). These findings argue in favor of a hybrid model of organization of the left posterior LPFC in which domain-oriented (nonspatial and spatially oriented) and cross-domain executive-dependent regions coexist, reconciling previously divergent data.

Introduction

In retrorolandic regions, integration of perceptual information is partially segregated into a dorsal stream, oriented toward spatial cognition, and a ventral stream, involved in cognition based on object features (Ungerleider and Haxby 1994). An essential issue in understanding the general architecture of the cerebral cortex is to determine whether this functional orientation is maintained in more anterior cortical areas (i.e., the lateral prefrontal cortex [LPFC]), involved in the elaboration of nonautomatic goal-directed actions and cognitive control. Yet, there is a debate on the anatomical and functional organization of the LPFC. One model stipulates an anatomical and functional segregation based on the domain of the information being processed (Goldman-Rakic 1987; Awh et al. 1995; Courtney et al. 1996, 1998; Smith and Jonides 1999; Levy and Goldman-Rakic 2000). Alternatively, several other models postulate a segregation based on the nature of processing or underline the functional importance of supra- or cross-modal integration in the LPFC (Owen et al. 1996; Duncan and Owen 2000; Miller and Cohen 2001; Koechlin et al. 2003; Petrides 2005). More recently, some studies or review articles suggest that these schemes of prefrontal organization are not mutually exclusive, and that both domain-specific and domain-general regions may coexist within the LPFC (Curtis and D'Esposito 2003; Sakai and Passingham 2003, 2006; Mohr et al. 2006; Sala and Courtney 2007). It is worth of note that the debated models mostly rely on data obtained in the monkey and functional imaging studies in humans. However, 1) the precise localization of functions cannot be directly transposed from monkeys to humans because of significant interspecies macroscopic anatomical differences (Fuster 1997; Petrides and Pandya 1999), and 2) in humans, functional imaging cannot formally demonstrate whether one given region of a functional network is critical (i.e., the cognitive process cannot be implemented without this region) or accessory (i.e., the cognitive process can be partially or fully functional without it).

Lesions studies in humans are useful to complement the above methodological approaches. Indeed, they rely on a principle which holds that a sustained functional deficit occurring after a focal brain lesion implies that the damaged area is critical for the tested function (Rorden and Karnath 2004). New techniques of lesion studies, such as voxel-by-voxel lesion mapping allow precise clinical–radiological correlations by testing all damaged voxels and do not require a pathological control group in order to establish anatomical–functional dissociations. The use of such methods has produced consistent results for assessment of the contribution of given brain regions to cognitive processes (Rorden and Karnath 2004).

Here, we address the issue of the functional and anatomical organization of the LPFC using a combined approach that includes: 1) a voxel-by-voxel lesion mapping study applied to patients with focal prefrontal lesion; 2) an functional magnetic resonance imaging (fMRI) study in healthy subjects; and 3) the use of the same cognitive paradigm (“the domain n-back” paradigm; Fig. 1) in the fMRI and lesion studies, that allows to test 2 factors of interest: the domain of information and the level of processing.

Figure 1.

Experimental tasks. (a) Schematic representation of the verbal, figurative (faces) and spatial n-back tasks performed by patients and controls in the lesion study. (b) Temporal organization of a run of the n-back tasks performed by subjects in the fMRI study.

Figure 1.

Experimental tasks. (a) Schematic representation of the verbal, figurative (faces) and spatial n-back tasks performed by patients and controls in the lesion study. (b) Temporal organization of a run of the n-back tasks performed by subjects in the fMRI study.

With the convergence of these methodological approaches, we can expect to answer the following question: Is the LPFC anatomically and functionally organized according to the domain of information, to the level of the processing or to both of these dimensions as recently brought up?

Material and Methods

These studies were approved by an institutional ethics committee for biomedical research (CCPPRB of the Pitié Salpêtrière Hospital) and all subjects provided informed consent.

The Experimental Cognitive Paradigm: the Domain n-Back Tasks

The Domain n-Back Tasks Adapted for the Lesion Study

Tasks were based on the n-back procedure (Braver et al. 1997; Cohen et al. 1997; Owen et al. 2005), in which the subject has to indicate whether a visual stimulus presented on the screen (the “target” stimulus) is similar to or different from a previously presented stimulus (the “cue” stimulus) (Fig. 1a). This procedure requires the relevant information to be maintained and updated in working memory (WM). Two dimensions were explored: 1) the level of processing (“complexity”) within WM, with 3 different levels: 1-back (maintenance of 1 item of information in WM in the interval between the cue and target stimuli), 2- and 3-back (interposition of 1 or 2 “distractors,” respectively, between the cue and target stimuli, each “distractor” becoming the cue for the next trial); 2) the nature of the stimuli being processed (“domain”), with 3 different materials: different locations of squares on a matrix of several squares (the “spatial n-back” task), different preselected men's faces (the “face n-back” task) and different preselected letters (the “letter n-back” task). Therefore, the domain n-back procedure followed a factorial design, crossing the 2 dimensions (complexity and domain processing), yielding a total of 9 task conditions.

All tasks were computerized and participants were seated in front of a computer screen. The examiner stood behind the subject throughout the testing procedure. Each n-back task started with the first cue stimulus (a square, a man's face or a letter) being presented on the screen for 3 sec. The subject had 3 s to answer aloud “same” or “different.” After a 1-s interstimulus interval, a new stimulus appeared on the screen. In the spatial n-back task, the visual stimulus was a blue square presented randomly in 1 of 6 possible locations on a dark screen. In the face n-back task, the stimulus was a man's face among 8 possible faces, and was presented centrally in the visual field. These faces were selected from Warrington's Recognition Memory Test for Faces. In the letter n-back task the stimulus was a capital letter among 7 possible letters selected for their frequency of occurrence, and was presented centrally in the visual field. Each task consisted of 3 blocks of 15 recorded responses to cue/target stimuli (16, 17, and 18 stimuli were presented in the 1-, 2-, and 3-back tasks, respectively). The order of spatial, face, and letter n-back tasks were counterbalanced when given to the patients and to the controls. All subjects were given a training block of the tasks for each of 3 levels of n-back using a material (different colored shapes) not subsequently used in the testing procedure. The total duration of 1 session was about 70 min.

Tasks were designed and controlled in a pilot study involving 23 healthy controls (detailed in du Boisgueheneuc et al. 2006), in order to match performances between different domain stimuli (i.e., there was no statistically significant difference between spatial, face, and letter tasks at each step of the n-back).

The Domain n-Back Tasks Adapted for the fMRI Study

Subjects performed a variant of the domain n-back tasks using letters and spatial items as memoranda (Fig. 1b). The domain n-back tasks were in most aspects similar to the tasks performed by the patients in the lesion study. Only a few changes were made in order to fit with the fMRI session: 1) a control task was added in order to perform comparisons with a condition that was similar in sensorimotor parameters but that did not include WM processing. The control task, called “0-back” consisted of detecting a prespecified item (i.e., letter “X” for the letter n-back task) when it appeared on the screen; 2) because of time constraints related to an fMRI session, it was necessary to shorten the paradigm. The face n-back tasks were not performed, given the fact that spatial and verbal domains were the most contrasted domains and should be kept for the fMRI study; therefore, it was possible to build a factorial design crossing 4 levels of complexity (0-, 1-, 2-, and 3-back) and 2 domains (letter and spatial), yielding a total of 8 conditions (90 stimuli were presented per condition); 3) At difference with the lesion study, where trials were blocked by levels of complexity, in the fMRI study, each of the 8 conditions was presented pseudorandomly during a block. Spatial and verbal n-back tasks were pooled in separate runs, and alternated. Half of the subjects began with a verbal run and the other with a spatial one. Each condition was performed 6 times across the fMRI session. The total duration of a session was approximately 45 min and included 6 trials of each of the 8 conditions. Each trial lasted 50 s and consisted of a series of 15 stimuli preceded by a 3.5-s presentation of task instructions, and a 1.5-s pause. Stimuli were presented for 2 s followed by a 1-s cross-fixation interstimulus interval. For each stimulus, the subject had 3 s to provide a response by clicking on a right (“same”) or a left (“different”) button. Trials were separated by a 7.5-s intertrial interval, providing a rest period.

Subjects

Lesion Study

Patients were recruited from the Neurosurgery and the Neurovascular Departments of La Pitié-Salpêtrière Hospital (Paris, France) (see Table 1, giving patient details, and Fig. 2 for structural imaging). Patients included in the study were selected on the basis of the following inclusion criteria: 1) the presence of a single frontal focal lesion, excluding lesions extending to other lobes and confirmed by an anatomical T1-weighted 3D MRI; 2) the frontal lesion was acquired in adulthood and was not due to an evolving disease (i.e., only patients with a sequellae from an hemorrhage or an ischemic stoke and those with a removal of a low-grade glioma leaving clear margins on postoperative MRI scans were included); 3) participants were tested at distance (> 1 month) from the onset of the episode responsible for the observable frontal lesion; 4) participants were able to understand and perform the cognitive tasks; 5) the absence of a prior history of neurological or psychiatric disease. It is important to note that every patient who matched the above criteria was included, regardless of the location of the lesion within the frontal lobes and the pattern of the cognitive deficit. At the time of the inclusion, none of the patients demonstrated sensorimotor or instrumental impairments that would interfere with the n-back tasks.

Table 1

Characteristics of the different populations included in the studies

 N Mean ± SD Min–max 
    
Lesion study    
    Patients 37   
        Age (years)  45.6 ± 11.3 23–72 
        Educational level (years)  13.5 ± 3.7 6–20 
        Lesion volume (cm3 26.43 ± 23.54 1.62–97.12 
        Time interval (months)a  18.3 ± 16.8 1–240 
        Gender    
            Male 24   
            Female 13   
        Lesion side    
            Right 12   
            Left 25   
        Etiology    
            Stroke/hemorrhage 20   
            Surgery 17   
    Controls 48   
        Age (years)  49.2 ± 10.6 23–65 
        Educational level (years)  14.1 ± 2.9 4–18 
        Gender    
            Male 18   
            Female 30   
fMRI study    
    fMRI subjects 12   
        Age (years)  45.5 ± 6.9 37–58 
        Educational level (years)  14.4 ± 2.8 7–17 
        Gender    
            Male   
            Female   
 N Mean ± SD Min–max 
    
Lesion study    
    Patients 37   
        Age (years)  45.6 ± 11.3 23–72 
        Educational level (years)  13.5 ± 3.7 6–20 
        Lesion volume (cm3 26.43 ± 23.54 1.62–97.12 
        Time interval (months)a  18.3 ± 16.8 1–240 
        Gender    
            Male 24   
            Female 13   
        Lesion side    
            Right 12   
            Left 25   
        Etiology    
            Stroke/hemorrhage 20   
            Surgery 17   
    Controls 48   
        Age (years)  49.2 ± 10.6 23–65 
        Educational level (years)  14.1 ± 2.9 4–18 
        Gender    
            Male 18   
            Female 30   
fMRI study    
    fMRI subjects 12   
        Age (years)  45.5 ± 6.9 37–58 
        Educational level (years)  14.4 ± 2.8 7–17 
        Gender    
            Male   
            Female   
a

Time interval: period of time separating the onset of the pathological event (stroke or surgical resection) from the inclusion in the study.

Figure 2.

Lesion diagrams of each of the 37 patients with a focal prefrontal lesion. The lesions represented in white were reconstructed from native MR scans (according to radiological convention, i.e., left is right).

Figure 2.

Lesion diagrams of each of the 37 patients with a focal prefrontal lesion. The lesions represented in white were reconstructed from native MR scans (according to radiological convention, i.e., left is right).

Statistical analyses for the voxel-by-voxel lesion mapping study required to obtain normative data for the domain n-back tasks from a group of healthy normal subjects matched for age and level of education (Table 1). As a consequence, a group of 48 control subjects was included. Subjects with a history of neurological or psychiatric disease were not included.

fMRI Study

Twelve right-handed healthy volunteers with no history of neurological or psychiatric disease were included in the study. In order to make comparisons between the regions showing activation in the fMRI study and the results of the lesion study, subjects included in the fMRI study were matched for age and level of education with the patients and normal controls included in the lesion study (Table 1). Therefore, it is important to note that the subjects included in the fMRI study were older than the ones usually included in fMRI studies with normal control subjects. This adaptation of the protocol seemed essential to avoid biases in comparing the fMRI study to the lesion study (where patients and controls were older than in most of the previously published fMRI studies regarding the n-back tasks).

Data Acquisition and Analysis of Results

Lesion Study

The study was performed using a voxel-by-voxel lesion mapping method developed and validated in our laboratory (Anatomo-Clinical Overlapping Maps: AnaCOM; see Kinkingnéhun et al. 2007 for a full description of the method). The AnaCOM method allows for statistical analysis to indicate which voxels contribute the most to a given cognitive or behavioral deficit. Inversion-recovery 3-dimensional fast SPGR contiguous T1-weighted axial slices (thickness: 1.5 mm) were acquired using a 1.5-Tesla scanner. All MRI scans were obtained at the time of evaluation. The T1 images were preprocessed in SPM99 (http://www.fil.ion.ucl.ac.uk/; Wellcome Institute of Cognitive Neurology, London) by spatially normalizing them to the Montreal Neurological Institute (MNI) template. As spatial normalization can be affected by the presence of a brain lesion, each lesion (all signal abnormalities due to the lesion) was manually segmented (using MRIcro, http://www.sph.sc.edu/comd/rorden/mricro.html) and was used as a mask during the normalization procedure to optimize the brain normalization. This masking procedure was used to weight the normalization to brain rather than nonbrain tissue or lesions (Brett et al. 2001). The spatial normalized images were resliced with a final voxel size of 1.5 × 1.5 × 1.5 mm3. The normalized images were then compared with the MNI template to evaluate normalization accuracy. Brain lesions were manually segmented again, this time on the normalized anatomical MRI. For both vascular lesions and surgical removal, the manual segmentation only included damaged areas in which MRI signal abnormalities was close to the cerebrospinal fluid signal. This second segmentation, which was used for further statistical analyses, did not include other signal abnormalities at the margin of the infarct or the resection cavity, assuming that these abnormal tissues may remain at least partly functional. Five patients had a lesion occurring less than 6 months (but more than 1 month) before inclusion in the study. One should consider that, in these cases, the lesion may extend farther than the boundaries of the final lesion. However, it is important to note that our method of segmentation only took into account the necrotic part of the lesion, representing the portion of the lesion that would not diminish with time.

Following these steps, each voxel of each lesion was weighted by the score of the patient in the task of interest (for instance, if a patient scored 28/45 in a given task, all the voxels included in his brain lesion were set at 28), whereas the rest of the image was set to zero, assuming that the brain lesion was responsible for the patient's deficit. Volumes representing each patient's lesion (n = 37) were then superimposed on a normalized brain, to built the “Maximum Overlap Map.” This map gave, for each voxel the number of lesions that covers this voxel. In these maps, the overlaps of the segmented lesions defined clusters (group of contiguous voxels covered by the same lesions, and thus with the same value). Statistical analyses were performed in the clusters that were composed of at least 3 lesions (Kinkingnéhun et al. 2007). For these clusters, a Student t-test corrected for multiple comparisons (Bonferroni correction) was performed between the mean performances obtained for each voxel of the cluster and those of the control subjects. Only regions where significance was still present after Bonferroni corrections were considered. It was then possible to obtain statistical maps for each n-back task. In other words, each statistical map represented brain regions where the patients’ performance statistically differed from that of the control subjects for a given task. These statistical maps thus indicated the clusters of voxels within the areas covered by at least 3 overlaps that contributed the most to a given impairment in the n-back tasks.

Regions within the prefrontal cortex (PFC) where there were at least 3 overlaps covered the anterior half of the precentral gyrus (Brodman's area [BA] 6) including the rolandic operculum, most of the left lateral PFC (BA 8, 9, 44, 45, 46, and part of 47 and lateral 10), the inferior, middle, and superior orbital gyri, the anterior insula and subcortical white matter adjacent to the above cortices (Figs 2 and 3f). By contrast, the most ventral and medial PFC and a portion of the lateral frontal pole, as well as the right dorsolateral PFC, were poorly covered. Consequently, conclusions are drawn from the analysis of the left lateral frontal cortex. For these areas, statistical maps were generated for each n-back task (n = 9) and for each level of complexity (n = 3) by pooling tasks from the same level of complexity (1-, 2-, and 3-back tasks). These maps identified regions where performances differed significantly from the controls.

Figure 3.

AnaCOM maps. Statistical maps of the 3-back tasks for all 3-back tasks pooled together (a), verbal (b), spatial (c), and faces (d) domains are superimposed on serial axial sections of a normalized brain. Significant areas at a Bonferroni threshold are represented in red-yellow. Maps in the right column (e) show the location of the frontal gyri (from Marina toolbox, http://www.bion.de/) superimposed on serial axial sections in the same z axis than n-back maps. Left side of the brain is on the right of each section. Surface rendering of a normalized brain showing regions where at least 3 overlaps of lesions occurred (i.e., where statistical analyses were performed) (f).

Figure 3.

AnaCOM maps. Statistical maps of the 3-back tasks for all 3-back tasks pooled together (a), verbal (b), spatial (c), and faces (d) domains are superimposed on serial axial sections of a normalized brain. Significant areas at a Bonferroni threshold are represented in red-yellow. Maps in the right column (e) show the location of the frontal gyri (from Marina toolbox, http://www.bion.de/) superimposed on serial axial sections in the same z axis than n-back maps. Left side of the brain is on the right of each section. Surface rendering of a normalized brain showing regions where at least 3 overlaps of lesions occurred (i.e., where statistical analyses were performed) (f).

Functional MRI

Visual stimuli were generated by a PC 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 2 buttons designed for fMRI experiments, connected to the computer and placed in their right and left hand. Imaging was carried out on a 1.5-T scanner (GE Medical Systems, Milwaukee, WI) using gradient echo-planar imaging, sensitive to blood oxygen level–dependent contrast (repetition time: 2500 ms, echo time: 60 ms, flip angle: 90°, matrix: 64 × 64, field of view: 240 × 240 mm). Functional images consisted of 16 contiguous axial slices with an in-plane resolution of 3.75 × 3.75 mm and a 5-mm slice thickness that fully covered the frontal lobes. The lower part of the temporal and occipital lobes and the cerebellum were not imaged. For each subject, anatomical high-resolution T1-weighted images were acquired in the same session (inversion-recovery sequence, inversion time 400 ms, echo time 2 ms, matrix 256 × 256, field of view 240 × 240 mm2, slice thickness 1.5 mm).

Subjects were required to perform 6 separate runs of 8 trials (Fig. 1b). Each trial consisted of an instruction appearing for 5 s followed by 15 consecutive stimuli of the n-back tasks (45 s). This latter part (the n-back per se) was modeled as an epoch of interest. Other parts were modeled but not analyzed. For each run, 186 volumes of 16 slices were continuously acquired over a total duration of 7 min and 45 s. The first 3 images were discarded from further analysis to await a steady state of tissue magnetization.

All analyses were carried out 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 9 linear rigid transformations to the MNI system. Functional imaging data from each run were corrected for motion (6-parameters, rigid-body realignment), normalized to the MNI coordinates and then smoothed with a 5-mm full-width half-maximum (FWHM) Gaussian filter.

Individual (first level) and group (second level) analyses were performed. For both analyses, each voxel was processed using the general linear model. For all individuals, epoch of interest was modeled with a delayed hemodynamic response function (convolution of a standard hemodynamic response with a box-car function). Overall signal differences between runs were also modeled. A temporal cut-off of 230 s (i.e., twice the duration of the interval between 2 similar trials) was applied to filter subject-specific low-frequency drift, mostly related to biological rhythms and to magnetic field drift. Motion effects during scanning were modelled by inclusion of the realignment parameters into the model. To test the hypothesis about regionally specific condition effects, the model estimates associated with the n-back phases were compared using linear contrasts. The resulting set of voxel values for each contrast was used to build SPM {t} maps. The threshold for individual activation was set to P < 0.001 uncorrected.

Individual contrast images of interest were smoothed with a 8-mm FWHM Gaussian kernel and used as dependent variables for the random effect group analysis (1-sample t-test). The threshold for significance was set to P < 0.05, corrected for multiple comparisons, taking into account the spatial extent of activation based on a cluster-size criterion (i.e., 50 contiguous voxels). Statistical maps were also analyzed at an uncorrected threshold of P < 0.001.

The main effect of the n-back was obtained by subtracting the 0-back (i.e., the control task) from the 1-, 2-, and 3-back conditions. A linear and parametric contrast of the estimates was built to isolate brain areas that increasingly vary with complexity (complexity effect). Effects of spatial and verbal domains were assessed by contrasting the letter to the spatial n-back tasks and vice versa.

To analyze the congruency between the fMRI and the lesion studies, we compared the location of specific regions obtained by each method and evidenced by statistical maps. All statistical maps generated by AnaCOM and those generated by fMRI analysis were coregistered together allowing comparison within the same normalized MNI space. Maps were superimposed on a normalized brain using Anatomist software (http:/brainvisa.info/index.html) to compare the location of the involved regions. More specifically: 1) the AnaCOM maps showing regions where performance significantly differed from the controls as a function of the complexity were compared with the fMRI maps showing regions activated as a function of the complexity. This comparison was aimed at determining the PFC subregions that were complexity-dependent, regardless of the domain of the information being processed; 2) AnaCOM maps showing regions where performance significantly differed from the controls for the verbal n-back tasks (1-, 2-, and 3-back maps) were compared with the fMRI maps contrasting the verbal to the spatial n-back tasks for each level of the n-back, at both the group and the individual levels. These comparisons were aimed at determining whether one or several PFC subregions were oriented toward verbal WM at one or several levels of complexity; 3) AnaCOM maps showing regions where performance significantly differed from the controls for the spatial n-back tasks (1-, 2-, and 3-back maps) were compared with the fMRI maps contrasting the spatial to the verbal n-back tasks for each level of the n-back (for instance, verbal 3-back vs. spatial 3-back), at both the group and the individual levels. These comparisons were aimed at determining whether one or several PFC subregions were oriented toward spatial WM at one or several levels of complexity.

Results

Lesion Study: Voxel-by-Voxel Lesion Mapping (AnaCOM)

The mean performance in each task of all the patients taken as a single group is provided in Table 2. It should be noted that comparisons between controls and patients (analyses of variance [ANOVAs]) showed a “group effect” (F1,83 = 19.31; P = 0.000033; mean performance for n-back tasks of the patient group being inferior to that of the control group), an interaction “group × complexity” (F2,166 = 7.81; P = 0.000573; performance being lower in the group of patients as a function of the increase of the n in the n-back tasks). No interaction was observed for the “group × domain” (F2,166 < 1; not significant [NS]) and for the “group × modality × complexity” (F4,332 < 1; NS). More importantly, no statistical correlation was found between performance in each of the n-back tasks and 1) the delay between the onset of the stroke or the surgical procedure and the experimental study, and 2) the volume of the lesions.

Table 2

Performances in the n-back tasks for the 3 groups of subjects

Tasks All n-back tasks Verbal tasks Spatial tasks Face tasks 
 1-back 2-back 3-back 1-back 2-back 3-back 1-back 2-back 3-back 1-back 2-back 3-back 
fMRI 
    Subjects (n = 12) 98.1 ± 2.4 89.8 ± 6.6 81.9 ± 6.8 98.0 ± 2.2 88.2 ± 7.5 80.0 ± 6.6 98.2 ± 2.7 91.5 ± 5.4 83.9 ± 6.8 — — — 
AnaCOM 
    Patients (n = 37) 97.1 ± 3.6 85.4 ± 14.0 65.4 ± 15.5 97.5 ± 3.1 83.8 ± 15.7 66.8 ± 15.9 97.7 ± 2.8 87.7 ± 13.0 63.8 ± 17.1 96.0 ± 4.5 84.6 ± 12.4 65.6 ± 12.7 
    Controls (n = 48) 98.9 ± 2.0 92.5 ± 7.6 75.2 ± 12.0 98.8 ± 2.3 91.8 ± 7.8 76.0 ± 10.7 99.4 ± 1.3 95.0 ± 6.4 74.4 ± 13.9 99.4 ± 2.2 90.7 ± 8.0 75.0 ± 11.3 
Tasks All n-back tasks Verbal tasks Spatial tasks Face tasks 
 1-back 2-back 3-back 1-back 2-back 3-back 1-back 2-back 3-back 1-back 2-back 3-back 
fMRI 
    Subjects (n = 12) 98.1 ± 2.4 89.8 ± 6.6 81.9 ± 6.8 98.0 ± 2.2 88.2 ± 7.5 80.0 ± 6.6 98.2 ± 2.7 91.5 ± 5.4 83.9 ± 6.8 — — — 
AnaCOM 
    Patients (n = 37) 97.1 ± 3.6 85.4 ± 14.0 65.4 ± 15.5 97.5 ± 3.1 83.8 ± 15.7 66.8 ± 15.9 97.7 ± 2.8 87.7 ± 13.0 63.8 ± 17.1 96.0 ± 4.5 84.6 ± 12.4 65.6 ± 12.7 
    Controls (n = 48) 98.9 ± 2.0 92.5 ± 7.6 75.2 ± 12.0 98.8 ± 2.3 91.8 ± 7.8 76.0 ± 10.7 99.4 ± 1.3 95.0 ± 6.4 74.4 ± 13.9 99.4 ± 2.2 90.7 ± 8.0 75.0 ± 11.3 

Note: For each task, performances are expressed in percentage of correct responses (mean ± standard deviation).

After corrections for multiple statistical comparisons, AnaCOM maps of the 1- and 2-back tasks (for each domain tested) revealed no voxel within the covered PFC regions that showed a statistical difference when compared with the performance of the control subjects. By contrast, AnaCOM maps showed several clusters of voxels significantly associated with a deficit in the 3-back tasks (Fig. 3; Table 3). In particular, we found within the left LPFC 1) regions commonly involved for all domains, and 2) topographical dissociations according to domain.

Table 3

Anatomical regions identified by AnaCOM maps to be significantly associated with a deficit in the different 3-back tasks

n-back tasks Anatomical regions Brodmann areas MNI coordinates of the epicenter P values 
All 3-back MFG (posterior part) 46 −30, 16, 39 2.3 × 10−5 
 MFG (posterior and superior part) 8/9 −32, 14, 51 2.16 × 10−5 
 PreC 6/44 −54, 4, 18 <10−7 
Verbal 3-back IFG (posterior part) 45 −52, 28, 11 7.3 × 10−6 
 PreC 6/44 −54, 4, 18 <10−7 
 MFG (inferior part) 46 −42, 27, 38 1.8 × 10−6 
Spatial 3-back SFS and SFG (posterior part) −22, 12, 42 1.9 × 10−5 
 MFG (posterior part) 46/9 −36, 22, 45 1.8 × 10−7 
  −38, 18, 52 4.5 × 10−6 
 Cingulate gyrus 32 −10, 29, 36 1.5 × 10−5 
 Medial SFG and adjacent SMA 8/6 −10, 15, 51 1.0 × 10−6 
 WhM — −20, 4, 31 1.3 × 10−5 
 PreC 6/44 −54, 4, 18 <10−7 
Face 3-back PreC 6/44 −54, 4, 18 2.051 × 10−6 
n-back tasks Anatomical regions Brodmann areas MNI coordinates of the epicenter P values 
All 3-back MFG (posterior part) 46 −30, 16, 39 2.3 × 10−5 
 MFG (posterior and superior part) 8/9 −32, 14, 51 2.16 × 10−5 
 PreC 6/44 −54, 4, 18 <10−7 
Verbal 3-back IFG (posterior part) 45 −52, 28, 11 7.3 × 10−6 
 PreC 6/44 −54, 4, 18 <10−7 
 MFG (inferior part) 46 −42, 27, 38 1.8 × 10−6 
Spatial 3-back SFS and SFG (posterior part) −22, 12, 42 1.9 × 10−5 
 MFG (posterior part) 46/9 −36, 22, 45 1.8 × 10−7 
  −38, 18, 52 4.5 × 10−6 
 Cingulate gyrus 32 −10, 29, 36 1.5 × 10−5 
 Medial SFG and adjacent SMA 8/6 −10, 15, 51 1.0 × 10−6 
 WhM — −20, 4, 31 1.3 × 10−5 
 PreC 6/44 −54, 4, 18 <10−7 
Face 3-back PreC 6/44 −54, 4, 18 2.051 × 10−6 

Note: All areas were left-lateralized (PreC: inferior third of the precentral sulcus, that is, from the operculum [z = 10] to the junction between the precentral sulcus and the inferior frontal sulcus [z = 38]; MFG: middle frontal gyrus; SFS: superior frontal sulcus; SFG: superior frontal gyrus; SMA: supplementary motor area; WhM: white matter between callosal fibers and corona radiata).

Indeed when pooling the spatial, verbal, and face 3-back tasks together (Fig. 3a; Table 3), 2 clusters in the posterior part of the middle frontal gyrus were statistically significant (P < 6.12 × 10−5): 1 cluster was located in BA 46 and the other in BA 9 and 8. A third cluster was located along the inferior third of the precentral sulcus (from the operculum to the junction between the precentral sulcus and the inferior frontal sulcus).

Areas specifically associated with a deficit in the verbal 3-back task (P < 6.56 × 10−5; Fig. 3b; Table 3) were a subregion of the posterior part of the left inferior frontal gyrus (pars triangularis; area 45) and a portion of the middle frontal gyrus (BA 46).

For the spatial 3-back task, areas specifically associated with a deficit (P < 6.47 × 10−5; Fig. 3c; Table 3) were observed within the posterior part of the left superior frontal sulcus and adjacent superior frontal gyrus (BA 8), extending to the surrounding white matter, the posterior part of the middle frontal gyrus (BA 46 and 9), the left cingulate gyrus (BA 32), the medial superior frontal gyrus and adjacent supplementary motor area (BA 8 and 6) and an area in the left white matter (located between the corona radiata and callosal fibers, above the body and head of the left caudate nucleus).

For the face 3-back task, only a small cluster in the inferior third of the precentral sulcus reached statistical significance (P < 8.04 × 10−5), in an area that overlapped with the other 3-back maps (Fig. 3d; Table 3).

In sum, the voxel-by-voxel lesion study showed that 1) a portion of the middle frontal gyrus (BA 46 and BA 9/8) and an area located in the inferior third of the precentral sulcus contributed to the deficit observed in the 3-back tasks, irrespective of the domain tested; 2) a left posterior portion of the inferior frontal gyrus (BA 44/45) contributed to the verbal 3-back deficit; 3) a posterior portion of the superior frontal gyrus and sulcus (BA 6/8/9) contributed to the deficit in the spatial 3-back task.

fMRI Study

Behavioral performances are provided in Table 2. ANOVAs performed in healthy fMRI subjects showed no effect of the domain (F1,11 = 4.25; NS), but a significant complexity effect (F3,33 = 55.21; P < 0.001), (i.e., performance decreased as the complexity increased).

The comparison between the performance of healthy subjects included in the fMRI study and that of the control subjects included in AnaCOM study showed no group effect (F1,58 = 0.24; NS), no interaction group × domain (F1,58 = 0.75; NS) but an interaction group × complexity (F2,116 = 8.52; P < 0.001; performances in 3-back tasks were lower in the AnaCOM control group than in the fMRI healthy group).

When all the n-back tasks were pooled and compared with the control task (the “0-back”), activation was observed bilaterally in parietal, medial, and lateral premotor, and lateral prefrontal regions (Fig. 4a).

Figure 4.

fMRI activation showing main, complexity and domain effects of the n-back tasks. Significant activation are superimposed on the surface rendering of a normalized brain. (a) Main effect of all n-back tasks compared with the control 0-back task, at a 0.05 threshold corrected for multiple comparisons. (b) Complexity effect showing regions in which the signal increases parametrically with the complexity for both verbal and spatial n-back tasks (P < 0.05 corrected for multiple comparisons). (c) Domain effect (verbal) contrasting the verbal 3-back tasks and the spatial 3-back tasks (P < 0.05 corrected for multiple comparisons). (d) Domain effect (spatial) contrasting the spatial 3-back tasks and the verbal 3-back tasks (P < 0.001 uncorrected).

Figure 4.

fMRI activation showing main, complexity and domain effects of the n-back tasks. Significant activation are superimposed on the surface rendering of a normalized brain. (a) Main effect of all n-back tasks compared with the control 0-back task, at a 0.05 threshold corrected for multiple comparisons. (b) Complexity effect showing regions in which the signal increases parametrically with the complexity for both verbal and spatial n-back tasks (P < 0.05 corrected for multiple comparisons). (c) Domain effect (verbal) contrasting the verbal 3-back tasks and the spatial 3-back tasks (P < 0.05 corrected for multiple comparisons). (d) Domain effect (spatial) contrasting the spatial 3-back tasks and the verbal 3-back tasks (P < 0.001 uncorrected).

Regions where the hemodynamic signal varied parametrically with the increase in complexity of the n-back tasks (“complexity effect”), for both domains of information, included the left and right inferior and superior parietal lobules and the intraparietal sulcus (BA 7 and 40), the left and right lateral premotor cortex (BA 6), supplementary motor area (BA 6/8), the left and right posterior portions of the middle frontal gyrus (BA 46 and 9/46), and the right frontal pole (BA 10) (Fig. 4b; Table 4).

Table 4

Coordinates of significant cluster maxima in the fMRI group analysis for the different n-back conditions of interest

n-back tasks Anatomical regions BA Left coordinates Z scores Right coordinates Z scores 
All n-back Parametric effect OrbG 10 −42 54 0 3.22   
 MFG anterior 46/10 −39 54 12 3.98 36 54 9 4.14 
 SFS posterior 6/9 −33 6 54 4.70 30 6 48 5.17 
 MFG posterior 46 −45 27 30 4.57 42 36 24 5.09 
 IFG 44 −39 12 27 3.95   
 Cingulum 32 −3 18 48 4.67   
 SPL −33 −66 51 5.32   
 Precuneus −12 −72 54 4.83   
 IPL 40 −48 −48 48 4.97   
3-back Verbal > spatial IFG 45 −51 36 6 4.77   
 IFG 44/45 −51 18 21 4.29   
 Precuneus 23 −9 −51 24 4.57   
 STS 22   48 −27 6 3.92 
 Posterior cingulum 23/24   9 −3 39 4.53 
 Posterior insula  −35 −6 18 3.30 45 −12 0 3.52 
 Putamen  −24 6 15 4.14   
3-back Spatial > verbal SFS posterior 6/9 −33 0 54 3.58* 33 6 48 3.38* 
 SPL −27 −53 57 5.32   
 Precuneus −9 −72 54 4.65 3 −57 51 4.33 
n-back tasks Anatomical regions BA Left coordinates Z scores Right coordinates Z scores 
All n-back Parametric effect OrbG 10 −42 54 0 3.22   
 MFG anterior 46/10 −39 54 12 3.98 36 54 9 4.14 
 SFS posterior 6/9 −33 6 54 4.70 30 6 48 5.17 
 MFG posterior 46 −45 27 30 4.57 42 36 24 5.09 
 IFG 44 −39 12 27 3.95   
 Cingulum 32 −3 18 48 4.67   
 SPL −33 −66 51 5.32   
 Precuneus −12 −72 54 4.83   
 IPL 40 −48 −48 48 4.97   
3-back Verbal > spatial IFG 45 −51 36 6 4.77   
 IFG 44/45 −51 18 21 4.29   
 Precuneus 23 −9 −51 24 4.57   
 STS 22   48 −27 6 3.92 
 Posterior cingulum 23/24   9 −3 39 4.53 
 Posterior insula  −35 −6 18 3.30 45 −12 0 3.52 
 Putamen  −24 6 15 4.14   
3-back Spatial > verbal SFS posterior 6/9 −33 0 54 3.58* 33 6 48 3.38* 
 SPL −27 −53 57 5.32   
 Precuneus −9 −72 54 4.65 3 −57 51 4.33 

Note: Coordinates are in millimeters, relative to the anterior commissure, in the MNI space. These maxima were thresholded for significance at 0.05 corrected for multiple comparisons by using a cluster-size criterion of 50 voxels, except for *, thresholded at P < 0.001 uncorrected (MFG: middle frontal gyrus; SFS: superior frontal sulcus; IFG: inferior frontal gyrus; SPL: superior parietal lobule; IPL: inferior parietal lobule; STS: superior temporal sulcus).

When all the verbal n-back tasks (1-, 2-, and 3-back) were compared with all the spatial n-back tasks, activation was found in a large area within the left inferior frontal gyrus that extended to the adjacent inferior precentral gyrus (BA 6, 44, and 45). This area was also found activated when the verbal 3-back task was compared with the spatial 3-back task (Fig. 4c; Table 4). Activation was also observed in the right posterior insula, the right posterior superior temporal gyrus and the left posterior cingulate region. The reverse contrast (all spatial vs. all verbal) evidenced bilateral activation in the superior parietal lobules, the intraparietal sulcus and the precuneus. In addition, the comparison of the spatial to the verbal 3-back tasks evidenced a bilateral activation of the posterior half of the superior frontal sulcus (BA 8; Fig. 4d; Table 4).

In sum, activation in the LPFC increased in intensity and size along with the increase in complexity in areas usually ascribed to WM. Within the LPFC, an area located in the middle frontal gyrus was found sensitive to the complexity of the task, irrespective of the domain tested, whereas a relative dissociation was observed in other subregions according to the domain of the information being processed: the verbal 3-back task specifically activated the left inferior frontal gyrus and the spatial 3-back task specifically activated the superior frontal sulcus.

Comparison of the Voxel-by-Voxel Lesion and fMRI Maps for the 3-Back Tasks

Both methods showed regions in the LPFC involved in the n-back tasks in a complexity-dependent manner (Figs 5 and 6). Among these regions, some were common to verbal and spatial domains and others were domain-specific.

Figure 5.

Comparative MNI coordinates of AnaCOM and individual fMRI results for verbal and spatial domains. Individual maxima of activation was projected on the MNI space. Squares represent the contrast 3-back spatial versus 3-back verbal and triangles the reverse contrast (Individual level: P < 0.001 uncorrected). Circles represent the MNI coordinates of the epicenter of the regions critical for the spatial 3-back (dark circle) and the verbal 3-back (light circle) tasks in the AnaCOM study.

Figure 5.

Comparative MNI coordinates of AnaCOM and individual fMRI results for verbal and spatial domains. Individual maxima of activation was projected on the MNI space. Squares represent the contrast 3-back spatial versus 3-back verbal and triangles the reverse contrast (Individual level: P < 0.001 uncorrected). Circles represent the MNI coordinates of the epicenter of the regions critical for the spatial 3-back (dark circle) and the verbal 3-back (light circle) tasks in the AnaCOM study.

Figure 6.

Schematic representation of the convergence of fMRI and AnaCOM results. Regions of interest identified by both methods are visualized on the surface rendering of a normalized, inflated left hemisphere. The purple area corresponds to the region associated with the complexity effect, irrespective to the domain of the tasks (the left middle frontal gyrus). The blue area corresponds to the region specialized for the verbal 3-back task (the posterior part of the left inferior frontal gyrus). The red area corresponds to the region specialized for spatial 3-back (the posterior part of the left superior frontal gyrus and sulcus).

Figure 6.

Schematic representation of the convergence of fMRI and AnaCOM results. Regions of interest identified by both methods are visualized on the surface rendering of a normalized, inflated left hemisphere. The purple area corresponds to the region associated with the complexity effect, irrespective to the domain of the tasks (the left middle frontal gyrus). The blue area corresponds to the region specialized for the verbal 3-back task (the posterior part of the left inferior frontal gyrus). The red area corresponds to the region specialized for spatial 3-back (the posterior part of the left superior frontal gyrus and sulcus).

AnaCOM study showed that a lesion in the left posterior middle frontal gyrus (BA 8/9/46) was associated with an impairment of both verbal and spatial 3-back tasks. When compared with the fMRI study, this AnaCOM region overlapped with the activated area (BA 46) associated with the complexity effect, irrespective of the domain tested. Data from both AnaCOM and fMRI thus indicate that the left and posterior middle frontal gyrus is a complexity-dependent area, regardless of the domain being tested (Fig. 6).

AnaCOM study showed that a lesion in the left posterior inferior frontal gyrus (BA 44/45) was associated with an impairment of the verbal 3-back task. When compared with the fMRI study, this AnaCOM region overlapped with the activated area (BA 44/45) observed when the verbal 3-back task was contrasted to the spatial 3-back tasks (Figs 4c and 5). Moreover, Figure 5 shows that for 9 of the 12 subjects, the individual maxima of activation for this contrast overlapped or are located at a very close distance to the AnaCOM region. These converging data from both AnaCOM and fMRI thus indicate that the left and posterior inferior frontal gyrus is preferentially oriented toward verbal WM, in a complexity-dependent manner (Fig. 6).

AnaCOM study showed that a lesion in the left posterior superior frontal gyrus (BA 6/8/9) was associated with an impairment of the spatial 3-back task. The maxima of activation in the fMRI study, when the spatial 3-back task was contrasted to the verbal 3-back task, was also located in the posterior superior frontal gyrus, caudal to the AnaCOM region (MNI coordinates: −33, 0, 54, for the maxima of activation in the fMRI study; −22, 12, 42 for the AnaCOM region) (Figs 4d and 5). It is important to note that, in the fMRI study, 11 out of the 12 subjects showed a maxima of activation that overlapped or are located at a very close distance to the AnaCOM region (Fig. 5). Altogether, the combination of the AnaCOM and fMRI data suggests that the left posterior superior frontal gyrus (BA 6/8/9) is preferentially oriented toward spatial WM, in a complexity-dependent manner (Fig. 6).

Discussion

The present study argue for a hybrid model of organization in which the posterior half of the left LPFC is only critical for the higher levels of processing in WM, and where domain-oriented and cross- (or supra-) domain subregions coexist. Our data fit well with recent data and theories of LPFC organization, considering that domain-oriented and domain-general regions coexist (Curtis and D'Esposito 2003; Sakai and Passingham 2003, 2006; Courtney 2004; Mohr et al. 2006; Postle 2006b; Sala and Courtney 2007). Moreover, our study reinforces and clarifies this view by the use of a voxelwise lesion-behavior correlation, and its convergence with fMRI activation.

The Posterior LPFC and Executive Processing

The fMRI study demonstrates a parametric effect of complexity (i.e., the higher the complexity, the more activated is the WM network) in the intraparietal sulci, the lateral premotor cortex and the LPFC, as previously reported in younger adult subjects (Braver et al. 1997; Owen et al. 2005). Accordingly, the comparison of frontal patients and control subjects showed a “group × complexity” interaction. Moreover, the data obtained in AnaCOM showed that LPFC lesions only induce a significant decrease in performance in the 3-back tasks, not observed in the 1- and 2-back tasks. These results suggest that the posterior LPFC, although contributing to all aspects of WM processing (as shown by fMRI), is a critical region only for WM operations requiring a high demand in executive processing. This interpretation is supported by lesion (D'Esposito and Postle 1999; Muller et al. 2002; du Boisgueheneuc et al. 2006) and functional imaging (Braver et al. 1997; Manoach et al. 1997; Pochon et al. 2001; Volle et al. 2005) studies showing that the LPFC is necessary for tasks involving a high demand on executive processing but not for tasks that only require the simple maintenance of items in short-term memory. The 3-back tasks involve several executive processes such as building the temporal ordering of mental representations and updating them, linking and compressing items in memory, inhibiting distractors, selecting and organizing the forthcoming responses… All these processes have been linked to the LPFC and particularly to the mid-dorsolateral PFC-BA (9/46) (Owen et al. 1998; Rowe et al. 2000; Pochon et al. 2001; Sakai et al. 2002; Volle et al. 2005; Postle 2006a). Although our study was not designed to specifically determine which of these mechanisms was particularly related to the functions of the LPFC, the fact that a deficit was only observed for the 3-back tasks suggests that the left posterior LPFC is critical for one or several of the executive processes that are necessary to permanently rehearse and update 3 items or more in WM. In sum, the left posterior LPFC is recruited specifically when processing demand increases, a condition represented in the present study by the 3-back tasks.

The Functional Heterogeneity of the Left Posterior LPFC

Converging data from fMRI and AnaCOM show that the subregion contributing the most to the spatial 3-back deficit is located in the left posterior portion of the superior frontal sulcus and gyrus, encompassing most of BA 8 and the caudal portion of BA 9. From an anatomical and functional point of view, this spatial subregion is rostral to the frontal eye field (FEF) (Paus 1996; Lobel et al. 2001). The epicenter of activation in the fMRI study for the spatial 3-back task is located a few millimeters caudally to the cluster found in the lesion study and partially overlaps with the FEF. It is, however, unlikely that the foci (lesion or activation) found in our study involve an area whose main function is to control saccadic eye-movements. Indeed, if this was the case, one would also expect the deficit to occur in the spatial 1-back and 2-back tasks that require to control or trigger eye-movements at the same level as in the 3-back task. On the contrary, the deficit suggests that the incriminated area is involved in cognitive rather than in motor control. In support of this interpretation, several studies have shown that an area in the posterior superior frontal gyrus, at coordinates that overlap with the loci found in our study, is specialized in spatial cognitive processing including the shift of spatial attention and spatial WM (Courtney et al. 1998; Haxby et al. 2000; Husain and Rorden 2003; Awh et al. 2006), particularly when the load and complexity increase, as in 3-back tasks (Carlson et al. 1998; Cohen et al. 1997; Nystrom et al. 2000). Furthermore, patients whose PFC lesion spares the posterior and lateral superior frontal gyrus performed normally in spatial WM tasks (D'Esposito and Postle 1999). Interestingly, this area is more activated in the spatial than in the shape n-back tasks (Nystrom et al. 2000). In addition, evidence from fiber tracking techniques indicates that the discussed area is part of a parietal–frontal network involved in spatial WM (Olesen et al. 2003; Klingberg 2006). The AnaCOM results also showed an extension to the white matter of the foci associated with impairments in the 3-back tasks, suggesting that the subcortical extensions contribute to the impairment by disconnecting several functional subregions involved in spatial WM. The anatomical connectivity in humans between the evidenced subcortical and cortical regions needs further explorations (Schmahmann et al. 2007). Altogether, our study and others argue for the existence of a cognitive spatial area located in the posterior portion of the superior frontal gyrus (mostly in BA 8) involved in spatial WM when the executive demand surpasses a certain threshold, such as in the spatial 3-back task.

Converging results from the fMRI and AnaCOM also evidence that the posterior segment of the left inferior frontal gyrus is oriented toward verbal WM. This area is involved in verbal WM, and particularly in the rehearsal component of the phonological loop (Paulesu et al. 1993; Awh et al. 1995). As rehearsal processing is also present in the 1- and 2-back tasks, an alternative and likely explanation for the deficit only occurring in the 3-back task is that, again, this region is sensitive to the increase in executive demand, as previously suggested (Braver et al. 1997; Cohen et al. 1997). As a whole, these data suggest that the posterior portion of the left inferior frontal gyrus is oriented toward verbal WM when executive demand surpasses a certain threshold of executive control. This involvement in verbal WM shares similarities with that of the posterior superior frontal gyrus for spatial WM, suggesting that both regions may perform similar WM processing but for different domains.

In AnaCOM, no cluster was found to be specifically face-oriented. This may be due to a lesser degree of specialization of the left inferior PFC or to a right frontal dominance for face processing in humans not evidenced because of the few overlaps in the right hemisphere in the present study.

Another LPFC region, located in the middle frontal gyrus (in the posterior “mid-dorsolateral” PFC—rostral BA 8, posterior BA 9/46) is found by the 2 methods to contribute to the 3-back tasks. In the fMRI study, this region is activated by the 2 domains tested (verbal and spatial), and in AnaCOM, this region contributes to the deficit observed in the spatial, letter and face 3-back tasks. This set of data argues for a supra- or cross-domain involvement of this area for cognitive processing. This finding is in line with functional studies in humans failing to demonstrate a domain-specific segregation within the mid-dorsolateral PFC, but rather suggesting an organization according to processing distinctions (Owen et al. 1998; Petrides 2005) or a holistic organization in which the dorsolateral PFC plays a global and flexible role in executive control, regardless of any distinctions in terms of processing or domain (Duncan and Owen 2000; Miller and Cohen 2001; Muller and Knight 2006). In addition, a subregion located along the inferior portion of the precentral sulcus was also found to be involved in the 3-back tasks, irrespective of the domain being processed. This subregion encompassed an area (the so-called “inferior frontal junction”) that has recently been found activated in a series of tasks requiring executive control but not specific of 1 given process or domain (Derrfuss et al. 2005).

In sum, these data show that the involvement of the left posterior LPFC in tasks involving a high level of executive processing is topographically heterogeneous: some subregions are preferentially oriented toward 1 domain of information, whereas others are sensitive to all domains tested, suggesting for the latter subregions a cross- or supradomain involvement in executive control. These convergent lesion and fMRI data, in line with other recent studies (Curtis and D'Esposito 2003; Sakai and Passingham 2003, 2006; Courtney 2004; Mohr et al. 2006; Postle 2006b; Sala and Courtney 2007), lead to describe a hybrid model of posterior LPFC organization, combining domain-dependent, domain-independent and complexity-dependent subregions. Finally, these data are of importance to neurologists and neurosurgeons as they provide more precise information on the expected cognitive sequellae following left posterior LPFC lesions.

Funding

Societé Française de Neurologie and Académie Nationale de Médecine.

The authors thank Dr Laurent Renié and Dr Foucaud du Boisgueheneuc for their contribution, Dr Mathias Pessiglione and Pr Stephane Lehéricy for their helpful comments and valuable discussions and thank the subjects who have participated in this research. Conflict of Interest: None declared.

References

Awh
E
Smith
E
Jonides
J
Human rehearsal processes and the frontal lobes: PET evidence
Ann N Y Acad Sci.
 , 
1995
, vol. 
769
 (pg. 
97
-
119
)
Awh
E
Vogel
EK
Oh
SH
Interactions between attention and working memory
Neuroscience.
 , 
2006
, vol. 
139
 (pg. 
201
-
208
)
Braver
TS
Cohen
JD
Nystrom
LE
Jonides
J
Smith
EE
Noll
DC
A parametric study of prefrontal cortex involvement in human working memory
Neuroimage.
 , 
1997
, vol. 
5
 (pg. 
49
-
62
)
Brett
M
Leff
AP
Rorden
C
Ashburner
J
Spatial normalization of brain images with focal lesions using cost function masking
Neuroimage.
 , 
2001
, vol. 
14
 (pg. 
486
-
500
)
Carlson
S
Martinkauppi
S
Rama
P
Salli
E
Korvenoja
A
Aronen
HJ
Distribution of cortical activation during visuospatial n-back tasks as revealed by functional magnetic resonance imaging
Cereb Cortex.
 , 
1998
, vol. 
8
 (pg. 
743
-
752
)
Cohen
JD
Perlstein
WM
Braver
TS
Nystrom
LE
Noll
DC
Jonides
J
Smith
EE
Temporal dynamics of brain activation during a working memory task
Nature.
 , 
1997
, vol. 
386
 (pg. 
604
-
608
)
Courtney
SM
Attention and cognitive control as emergent properties of information representation in working memory
Cogn Affect Behav Neurosci.
 , 
2004
, vol. 
4
 (pg. 
501
-
516
)
Courtney
SM
Petit
L
Maisog
JM
Ungerleider
LG
Haxby
JV
An area specialized for spatial working memory in human frontal cortex
Science.
 , 
1998
, vol. 
279
 (pg. 
1347
-
1351
)
Courtney
SM
Ungerleider
LG
Keil
K
Haxby
JV
Object and spatial visual working memory activate separate neural systems in human cortex
Cereb Cortex.
 , 
1996
, vol. 
6
 (pg. 
39
-
49
)
Curtis
CE
D'Esposito
M
Persistent activity in the prefrontal cortex during working memory
Trends Cogn Sci.
 , 
2003
, vol. 
7
 (pg. 
415
-
423
)
D'Esposito
M
Postle
BR
The dependence of span and delayed-response performance on prefrontal cortex
Neuropsychologia.
 , 
1999
, vol. 
37
 (pg. 
1303
-
1315
)
Derrfuss
J
Brass
M
Neumann
J
von Cramon
DY
Involvement of the inferior frontal junction in cognitive control: meta-analyses of Switching and Stroop studies
Hum Brain Mapp
 , 
2005
, vol. 
25
 (pg. 
22
-
34
)
du Boisgueheneuc
F
Levy
R
Volle
E
Seassau
M
Duffau
H
Kinkingnehun
S
Samson
Y
Zhang
S
Dubois
B
Functions of the left superior frontal gyrus in humans: a lesion study
Brain.
 , 
2006
, vol. 
129
 (pg. 
3315
-
3328
)
Duncan
J
Owen
AM
Common regions of the human frontal lobe recruited by diverse cognitive demands
Trends Neurosci.
 , 
2000
, vol. 
23
 (pg. 
475
-
483
)
Fuster
J
The prefrontal cortex
 , 
1997
New York
Raven Press
Goldman-Rakic
P
Plum
F
Circuitry of primate prefrontal cortex and regulation of behavior by representational memory
Handbook of physiology
 , 
1987
Washington (DC)
The American Physiological Society
(pg. 
373
-
417
)
Haxby
JV
Petit
L
Ungerleider
LG
Courtney
SM
Distinguishing the functional roles of multiple regions in distributed neural systems for visual working memory
Neuroimage.
 , 
2000
, vol. 
11
 (pg. 
380
-
391
)
Husain
M
Rorden
C
Non-spatially lateralized mechanisms in hemispatial neglect
Nat Rev Neurosci.
 , 
2003
, vol. 
4
 (pg. 
26
-
36
)
Kinkingnéhun
S
Volle
E
Golmard
JL
Lehéricy
S
du Boisguéheneuc
F
Zhang-Nunes
S
Sosson
D
Duffau
H
Samson
Y
Levy
R
, et al.  . 
A novel approach to clinical-radiological correlations: Anatomo-Clinical Overlapping Maps (AnaCOM) Method and Validation
Neuroimage
 , 
2007
, vol. 
37
 (pg. 
1237
-
1249
)
Klingberg
T
Development of a superior frontal–intraparietal network for visuo-spatial working memory
Neuropsychologia.
 , 
2006
, vol. 
44
 (pg. 
2171
-
2177
)
Koechlin
E
Ody
C
Kouneiher
F
The architecture of cognitive control in the human prefrontal cortex
Science.
 , 
2003
, vol. 
302
 (pg. 
1181
-
1185
)
Levy
R
Goldman-Rakic
PS
Segregation of working memory functions within the dorsolateral prefrontal cortex
Exp Brain Res.
 , 
2000
, vol. 
133
 (pg. 
23
-
32
)
Lobel
E
Kahane
P
Leonards
U
Grosbras
M
Lehericy
S
Le Bihan
D
Berthoz
A
Localization of human frontal eye fields: anatomical and functional findings of functional magnetic resonance imaging and intracerebral electrical stimulation
J Neurosurg.
 , 
2001
, vol. 
95
 (pg. 
804
-
815
)
Manoach
DS
Schlaug
G
Siewert
B
Darby
DG
Bly
BM
Benfield
A
Edelman
RR
Warach
S
Prefrontal cortex fMRI signal changes are correlated with working memory load
Neuroreport.
 , 
1997
, vol. 
8
 (pg. 
545
-
549
)
Miller
EK
Cohen
JD
An integrative theory of prefrontal cortex function
Annu Rev Neurosci.
 , 
2001
, vol. 
24
 (pg. 
167
-
202
)
Mohr
HM
Goebel
R
Linden
DE
Content- and task-specific dissociations of frontal activity during maintenance and manipulation in visual working memory
J Neurosci.
 , 
2006
, vol. 
26
 (pg. 
4465
-
4471
)
Muller
NG
Knight
RT
The functional neuroanatomy of working memory: contributions of human brain lesion studies
Neuroscience.
 , 
2006
, vol. 
139
 (pg. 
51
-
58
)
Muller
NG
Machado
L
Knight
RT
Contributions of subregions of the prefrontal cortex to working memory: evidence from brain lesions in humans
J Cogn Neurosci.
 , 
2002
, vol. 
14
 (pg. 
673
-
686
)
Nystrom
LE
Braver
TS
Sabb
FW
Delgado
MR
Noll
DC
Cohen
JD
Working memory for letters, shapes, and locations: fMRI evidence against stimulus-based regional organization in human prefrontal cortex
Neuroimage.
 , 
2000
, vol. 
11
 (pg. 
424
-
446
)
Olesen
PJ
Nagy
Z
Westerberg
H
Klingberg
T
Combined analysis of DTI and fMRI data reveals a joint maturation of white and grey matter in a fronto-parietal network
Brain Res Cogn Brain Res.
 , 
2003
, vol. 
18
 (pg. 
48
-
57
)
Owen
AM
Doyon
J
Petrides
M
Evans
AC
Planning and spatial working memory: a positron emission tomography study in humans
Eur J Neurosci.
 , 
1996
, vol. 
8
 (pg. 
353
-
364
)
Owen
AM
McMillan
KM
Laird
AR
Bullmore
E
N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies
Hum Brain Mapp.
 , 
2005
, vol. 
25
 (pg. 
46
-
59
)
Owen
AM
Stern
CE
Look
RB
Tracey
I
Rosen
BR
Petrides
M
Functional organization of spatial and nonspatial working memory processing within the human lateral frontal cortex
Proc Natl Acad Sci. USA.
 , 
1998
, vol. 
95
 (pg. 
7721
-
7726
)
Paulesu
E
Frith
CD
Frackowiak
RS
The neural correlates of the verbal component of working memory
Nature.
 , 
1993
, vol. 
362
 (pg. 
342
-
345
)
Paus
T
Location and function of the human frontal eye-field: a selective review
Neuropsychologia.
 , 
1996
, vol. 
34
 (pg. 
475
-
483
)
Petrides
M
Lateral prefrontal cortex: architectonic and functional organization
Philos Trans R Soc Lond B Biol Sci.
 , 
2005
, vol. 
360
 (pg. 
781
-
795
)
Petrides
M
Pandya
DN
Dorsolateral prefrontal cortex: comparative cytoarchitectonic analysis in the human and the macaque brain and corticocortical connection patterns
Eur J Neurosci.
 , 
1999
, vol. 
11
 (pg. 
1011
-
1036
)
Pochon
JB
Levy
R
Poline
JB
Crozier
S
Lehericy
S
Pillon
B
Deweer
B
Le Bihan
D
Dubois
B
The role of dorsolateral prefrontal cortex in the preparation of forthcoming actions: an fMRI study
Cereb Cortex.
 , 
2001
, vol. 
11
 (pg. 
260
-
266
)
Postle
BR
Distraction-spanning sustained activity during delayed recognition of locations
Neuroimage.
 , 
2006
, vol. 
30
 (pg. 
950
-
962
)
Postle
BR
Working memory as an emergent property of the mind and brain
Neuroscience.
 , 
2006
, vol. 
139
 (pg. 
23
-
38
)
Rorden
C
Karnath
HO
Using human brain lesions to infer function: a relic from a past era in the fMRI age?
Nat Rev Neurosci.
 , 
2004
, vol. 
5
 (pg. 
812
-
819
)
Rowe
JB
Toni
I
Josephs
O
Frackowiak
RS
Passingham
RE
The prefrontal cortex: response selection or maintenance within working memory?
Science.
 , 
2000
, vol. 
288
 (pg. 
1656
-
1660
)
Sakai
K
Passingham
RE
Prefrontal interactions reflect future task operations
Nat Neurosci.
 , 
2003
, vol. 
6
 (pg. 
75
-
81
)
Sakai
K
Passingham
RE
Prefrontal set activity predicts rule-specific neural processing during subsequent cognitive performance
J Neurosci.
 , 
2006
, vol. 
26
 (pg. 
1211
-
1218
)
Sakai
K
Rowe
JB
Passingham
RE
Active maintenance in prefrontal area 46 creates distractor-resistant memory
Nat Neurosci.
 , 
2002
, vol. 
5
 (pg. 
479
-
484
)
Sala
JB
Courtney
SM
Binding of what and where during working memory maintenance
Cortex.
 , 
2007
, vol. 
43
 (pg. 
5
-
21
)
Schmahmann
JD
Pandya
DN
Wang
R
Dai
G
D'Arceuil
HE
de Crespigny
AJ
Wedeen
VJ
Association fibre pathways of the brain: parallel observations from diffusion spectrum imaging and autoradiography
Brain.
 , 
2007
, vol. 
130
 (pg. 
630
-
653
)
Smith
EE
Jonides
J
Storage and executive processes in the frontal lobes
Science.
 , 
1999
, vol. 
283
 (pg. 
1657
-
1661
)
Ungerleider
LG
Haxby
JV
‘What’ and ‘where’ in the human brain
Curr Opin Neurobiol.
 , 
1994
, vol. 
4
 (pg. 
157
-
165
)
Volle
E
Pochon
JB
Lehericy
S
Pillon
B
Dubois
B
Levy
R
Specific cerebral networks for maintenance and response organization within working memory as evidenced by the ‘double delay/double response’ paradigm
Cereb Cortex.
 , 
2005
, vol. 
15
 (pg. 
1064
-
1074
)