Dissociating a Common Working Memory Network from Different Neural Substrates of Phonological and Spatial Stimulus Processing

NeuroImage 15, 45–57 (2002) doi:10.1006/nimg.2001.0968, available online at http://www.idealibrary.com on

Dissociating a Common Working Memory Network from Different Neural Substrates of Phonological and Spatial Stimulus Processing Bartosz Zurowski,* Julia Gostomzyk,* Georg Gro¨n,* Rolf Weller,† Holger Schirrmeister,† Bernd Neumeier,† Manfred Spitzer,* Sven N. Reske,† and Henrik Walter* ,1 *Department of Psychiatry, University of Ulm, 89075 Ulm, Germany; and †Department of Nuclear Medicine, University of Ulm, 89070 Ulm, Germany Received February 26, 2001

INTRODUCTION Positron emission tomography was used to investigate common versus specific cortical regions for the maintenance of spatial versus phonological information in working memory (WM). Group and single-subject analyses of regional cerebral blood flow during a new 2 ⴛ 2 factorial n-back task were performed. Eight subjects had to memorize either phonological features or the location of serially presented syllables. Brain activation during phonological judgment and spatial judgment (0-back) was compared with that during two corresponding WM conditions (2-back). We observed a common network associated with the requirement of maintaining and sequencing items in WM. Seven or more subjects activated (posterior) superior frontal sulcus (pSFS, BA 6/8, global maximum) and/or adjacent gyri, posterior parietal cortex, and precuneus (BA 7). Less consistently, bilateral middle frontal gyrus (BA 9/46) was involved. Bilateral anterior (BA 39/40) and posterior (BA 7) intraparietal sulcus, as well as right pSFS, exhibited dominance for spatial WM. Although underlying stimulus processing pathways for both types of information were different, no region specific for phonological WM was found. Robust activation within the left inferior frontal gyrus (BA 44 and 45) was present, during both phonological WM and phonological judgment. We conclude that the controversial left prefrontal lateralization for verbal WM reflects more general phonological processing strategies, not necessarily required by tasks using letters. We propose a stimulus-independent role for the bilateral pSFS and its vicinity for maintenance and manipulation of different context-dependent information within working memory. © 2002 Elsevier Science Key Words: working memory; phonological; spatial; prefrontal cortex; positron emission tomography; functional imaging.

Recent investigations compel us to reconsider some important aspects concerning the neural organization of human working memory (WM). This concept refers to the ability to maintain and manipulate information to make it behaviorally disposable, after its reference is no longer present in the environment (Baddeley, 1986). Logie et al. (1990) provided behavioral evidence for a functional dissociation between spatial and verbal– phonological components of working memory, the “phonological loop” and the “visuospatial scratchpad.” Several observations of a relative independence and noninterference of these two “slave systems” to be controlled by a shared “central executive” (Baddeley, 1986) raised the question of a cortical substrate for these postulated components. Based on subsequent neuroimaging investigations, human WM has been proposed as being organized hemispherically within the prefrontal cortex (PFC), based on spatial (right) and nonspatial (left) (D’Esposito et al., 1998) or spatial (right) and verbal (left) (Smith et al., 1996) information with some meta-analytical evidence for both variants. It is helpful to focus on the three studies that directly tested for regional specificity for verbal (letters) versus spatial (locations) information, all of them involving n-back paradigms (Cohen et al., 1994). Using PET, Smith et al. (1996) found a clear-cut hemispheric double dissociation, with the right dorsal PFC (BA 46) subserving spatial and the left ventral PFC (BA 44) as well as left premotor cortex (BA 6) subserving active maintenance within WM. The left inferior parietal cortex (BA 40) was proposed as a neural substrate for the passive phonological short-term store (Baddeley, 1986). However, two subsequent fMRI investigations failed to support the notion of WM-related hemispheric specificity. Either activation of left inferior frontal cortex was not observed, with right PFC activated in spatial as well as in verbal WM (D’Esposito et al., 1998), or both regions were active bilaterally without showing any significant dominance effects (Nystrom et al., 2000).

1 To whom correspondence should be addressed at Department of Psychiatry, University of Ulm, Leimgrubenweg 12, 89075 Ulm, Germany. Fax: 49-(0)731-50021549. E-mail: [email protected].

45 ©

1053-8119/02 $35.00 2002 Elsevier Science All rights reserved.

46

ZUROWSKI ET AL.

Following another line of evidence, the PFC has been proposed as being organized into a dorsal and a ventral portion according to the nature of the cognitive process (largely irrespective of material to be maintained) or according to the stimulus material (largely irrespective of process aspects). Evidence for the concept of material specificity has been drawn from electrophysiological findings in nonhuman primates, suggesting a regionally distinct, stimulus-dependent increase in delay-related neuronal firing for object memoranda and for spatial memoranda (Goldman-Rakic, 1987; Wilson et al., 1993). However, this distinction has been only inconsistently observed in humans (D’Esposito et al., 1998, for a review). The concept of material specificity has been partially challenged by a series of findings in nonhuman primates indicating a rather polymodal integrative role of PFC, representing abstract environmental context within WM (Fuster et al., 1982; Rao et al., 1997; Miller and Cohen, 2001). These findings are consistent with the second, process-oriented concept of a stimulus-independent role of the dorsal PFC (BA 9/46 or mid-DLPFC (Petrides, 1994)), whenever maintained information has to be manipulated. This is supported by recent neuroimaging approaches in humans (Owen et al., 1996, 1998; Postle et al., 1999; D’Esposito et al., 1998, for review). Finally, there is evidence for a specific stimulusdependent involvement of the (posterior) superior frontal sulcus (pSFS) and adjacent gyri in WM for locations, but not in WM for faces (Courtney et al., 1996, 1998a). However, numerous studies using verbal WM paradigms found activation sites broadly corresponding to the region described by Courtney and colleagues (Petrides et al., 1993; Awh et al., 1996; Schumacher et al., 1996; Smith et al., 1996; Cohen et al., 1997; Jonides et al., 1998; LaBar, 1999). Together with Broca’s area and the supplementary motor area, Smith and Jonides (1998, 1999) relate the BA 6, including the pSFS and vicinity, to a left lateralized verbal rehearsal system. Thus, two rather different concepts are used to explain a highly prevalent WM-related activation within the same region. In the present study, we used a two-tailed approach to quantify the anatomical overlap (via conjunction analysis) and the differences (via interaction analysis) between spatial and verbal–phonological WM tasks, presenting physically identical syllable stimuli under all task conditions. Further, our approach should minimize differences due to perceptual and stimulus processing requirements between WM and non-WM conditions possibly contributing to the above-mentioned inconsistencies in the literature (Courtney et al., 1998b; Postle et al., 2000; Nystrom et al., 2000). For example, it is very likely that the detection of a predefined letter, as used for the baseline condition in verbal WM tasks like the 2-back, is based on object features of the target letter (usually ‘X’), whereas let-

ters in the WM condition are processed phonologically. Further, several investigators reported or assumed verbal coding strategies for the objects or intermediate strategies when comparing object and spatial WM (Postle and Esposito, 1999; Nystrom et al., 2000). In humans, regional differences between WM tasks might depend in great part on the variety of possible encoding strategies (Goldman-Rakic et al., 2000) and the nature of representations generated, thus being not clearly determined by presented stimulus material. The present cognitive paradigm was therefore designed to force subjects to process the same syllable stimuli in two different and verifiable ways and to transfer different aspects of the experimental set into WM: either the phonological features of meaningless syllables (phonological 2-back) or their location on the screen (spatial 2-back). We also designed a phonological and a spatial judgment condition with similar requirements regarding stimulus processing and response selection, but without WM being needed. Thus, a potential WM-related regional dominance should not be attributable to general perceptual or processing differences between 2-back and 0-back conditions, whereas a lack of regional dominance should not be attributable to the lack of implicated processing strategies, phonological versus spatial. METHODS Subjects Eight healthy, male, right-handed volunteers (mean age, 27.3; SD, 6.4) underwent PET scanning. All subjects gave informed, written consent for participation. Subjects were paid for participation. The study was approved by the local ethics committee. Subjects were screened for neurological and psychiatric history as well as substance abuse and they were instructed not to drink alcohol on the day of scanning and not to smoke 2 h prior to the experiment. None of the subjects was excluded from analysis. Sixteen different subjects (mean age, 29.3; SD, 5.4) participated in a behavioral control study. PET Paradigm A 2 ⫻ 2 factorial task design was used in the PET study with WM load (2-back vs 0-back judgment) and attended stimulus feature (phonological vs spatial) as factors. The phonological WM condition (PWM) followed the letter 2-back rationale (Cohen et al., 1994). Instead of letters, the stimulus set consisted of 30 uppercase syllables (15 target–probe pairs) with a consonant–vowel– consonant (CVC) structure. Subjects were instructed to press button 1 with the right index finger, if, when articulated subvocally, the presented syllable corresponded phonologically (“sounded like”)

PHONOLOGICAL AND SPATIAL WORKING MEMORY

to the one presented two syllables back in the sequence. Otherwise they were instructed to press button 2 with the right middle finger. Targets and probes were not identical according to orthography (e.g., not TUP– TUP). In accordance with the pronunciation rules of the German language, phonological similarity without orthographic identity was achieved by exchanging both consonants between target and probe syllables but not the vowel. Moreover, voiced consonants were only replaced by voiceless consonants and vice versa (e.g., TUP–DUB or TEK–DEG). Frequency of letters within the set and frequency of syllables presented were controlled for. To make phonological matches unambiguous, possibly distracting syllables and sequences, e.g., related syllable repeats, were removed. Before scanning, subjects were trained on a parallel version of the experimental tasks until they were able to detect at least two of three targets. Participants were discouraged from overtly vocalizing the syllables or focusing on single letters. Instead, they were instructed to articulate the presented syllable subvocally and compare its acoustic features (“sound”) with those of the syllable presented two back. As a nonmnemonic condition we established a phonological judgment task (PJ). Subjects were instructed to indicate whether the syllable presented closely matched the predefined target phoneme [fop], when articulated subvocally. This was only the case for the following syllables: VOB, VOP, FOP, FOB. All other syllables (30) were the same as in the WM task and served as nontargets. In both tasks, PJ and PWM, stimuli were shown at randomly changing screen locations (13) to enable direct comparisons to a spatial WM task. Stimuli were presented for 1000 ms, followed by a 2000-ms interstimulus interval with central presentation of a fixation cross. The long stimulus presentation time should allow subjects to make a decision during the physical presence of the stimulus, thus minimizing WM requirements during the judgment conditions. In the spatial WM condition (SWM) subjects were presented with the same set of 30 stimuli at changing locations. They were instructed to respond with the right index finger if the screen location of the syllable presented was the same as the one two back in the sequence, regardless of the syllable’s identity. As a non-WM condition we used a spatial judgment task (SJ). Subjects were instructed to respond with the right index finger whenever a syllable occurred on the horizontal meridian of the screen. During PET scanning subjects viewed a 90-s sequence of stimuli during three repetitions of the four conditions (PWM, PJ, SWM, SJ). Each 90-s sequence included four to six target stimuli (a total of 15 targets per condition). Four subjects were presented the two phonological conditions in pseudorandomized order during the first six scans; the other four subjects viewed the spatial conditions first. Participants viewed

47

stimuli projected via a high-resolution head LCD monitor (Sony Glasstron PLM-S700E), which produced a virtual eye-to-monitor distance of 120 cm. This should enable participants to keep all presented stimuli foveally without requiring eye movements. Stimulus presentation and recording of behavioral data were controlled with ERTS Software (Berisoft, Frankfurt, Germany). Participants completed a semistructured postscanning questionnaire asking about strategies used for task completion. Behavioral Control Study To evaluate the new functional paradigm with respect to reliability of behavioral results and strategies used for task completion, both WM tasks, PWM and SWM, were administered to a separate sample of 16 subjects prior to PET investigation. Both tasks consisted of 5 blocks of 48-s duration each, including two to four targets per block (a total of 15 targets per condition). Subjects viewed stimuli sitting in front of a 17-in. PC monitor. All remaining parameters were identical to those of the PET procedure. PET Scanning PET scans were obtained using an ECAT EXACT 47 (Siemens/CTI, Knoxville, TN) scanner in 3D mode. Forty-seven contiguous image slices were reconstructed into a 128 ⫻ 128 ⫻ 47 image matrix (voxel size 2.58 ⫻ 2.58 ⫻ 3.375 mm). For attenuation correction a 10-min transmission scan was acquired prior to the emission scans. A neuroinsert attenuated the contribution of random counts from the rest of the body to the recorded brain images. A total of 12 scans with a 12min interval were obtained for each subject, i.e., three scans for each of the four experimental conditions. Stimulus presentation was initialized 5 s prior to the administration of 370 MBq of 15O-labeled water as a 5to 7-ml bolus through an antecubital vein cannula. Tracer administration over a period of ⬃4 s was followed immediately by a 10-ml saline flush. Consequently, subjects already performed the 90-s task for ⬃20 s prior to bolus arrival in the brain. Each scan was initialized when the rate of true counts exceeded that of random counts. Integrated radioactivity counts were accumulated over a 50-s acquisition period. The subjects’ head was immobilized with the use of a vacuum mold. For each subject, a 3D MRI volume (MPRAGE, voxel size 1 ⫻ 1 ⫻ 1 mm) was acquired using a Magnetom Symphony 1.5-T scanner (Siemens, Erlangen, Germany) to be coregistered with the PET images. Data Analysis Overall accuracy (Table 1) was determined for each subject as follows: (P [targets correct] ⫹ P [nontargets correct]) ⫺2. These values were used to compute the mean accuracy for the

48

ZUROWSKI ET AL.

TABLE 1 Mean (SD) Behavioral Results from the PET Experiment Task condition

PJ

PWM

SJ

SWM

Reaction time Accuracy Targets Nontargets Sensitivity index d⬘ Response index ␤

575 (60.1) 94.1 (3.9) 91.2 (7.0) 97.0 (2.9) 3.67 (1.01) 0.94 (0.08)

780 (220.5) 82.8 (8.8) 69.8 (18.1) 95.7 (1.7) 2.28 (0.59) 0.73 (0.19)

428 (74.9) 93.4 (7.8) 87.9 (16.4) 98.8 (1.3) 4.14 (1.51) 0.89 (0.15)

595 (179.6) 90.2 (5.2) 82.7 (1.4) 97.7 (9.7) 3.19 (0.9) 0.85 (0.09)

Note. Average frequency of targets was 1:5 in all conditions. If participants applied this knowledge, random guessing would yield a 16.7% accuracy rate for the targets and 83.3% for nontargets.

group. Sensitivity index d⬘ and response index ␤ (Grier, 1971) were measured to exclude coarse response tendencies, especially to NTs. Reaction times (RTs) and accuracy in the behavioral control study (n ⫽ 16) were analyzed using a pairwise t test. Each subject’s mean RT and accuracy from the PET experiment were entered into a two-tailed MANOVA for repeated measures with WM load (0-back-judgment, 2-back) and processed stimulus feature (phonological, spatial) as factors. For analysis of the PET images SPM99 Software (Wellcome Department of Cognitive Neurology, London, UK) was used. Individual PET scans were realigned, with the first scan as reference to correct for head movements between scans. Images were spatially normalized into a standard space (Talairach and Tournoux, 1988) using the Montreal Neurological Institute (MNI) template. To accommodate intersubject differences, each image was smoothed with an isotropic Gaussian kernel of 14-mm FWHM. A mixed-effects analysis of covariance (ANCOVA) with global flow (normalized to 50 ml/100 ml/min) as a covariate of no interest was used, treating subjects as random factors and “load” and “stimulus” as within-subjects factors. The resulting foci were thresholded using voxel-based t statistics transformed into z statistics in terms of peak height and spatial extent using distributional approximations from the theory of Gaussian fields (Friston et al., 1995). Our 2 ⫻ 2 factorial task design involved five statistical a priori contrasts to test for the main effect of WM “load” (1), the main effects of “stimulus” (2, 3), and the interaction of “load” and “stimulus” (4, 5). Main effects were assessed via conjunction analyses (Price and Friston, 1997) and a threshold of Z ⫽ 4.85 (P ⬍ 0.05, corrected for multiple comparisons) was applied. In each conjunction analysis, the resulting SPMs include only voxels significant for both contrasts selected. The first conjunction was (SWM ⫺ SJ) conjoined with (PWM ⫺ PJ), testing for the main effect of “load” while ignoring “load” ⫻ “stimulus” interactions. In analogy, main effects of “stimulus” were determined by the following contrasts: [(PWM ⫺ SWM) conj (PJ ⫺ SJ)] and

[(SWM ⫺ PWM) conj (SJ ⫺ PJ)]. To test for regions dominant for phonological or spatial WM, the following interaction contrasts were computed: [(PWM ⫺ SWM) ⫺ (PJ ⫺ SJ)] for PWM dominance and [(SWM ⫺ PWM) ⫺ (SJ ⫺ PJ)] for SWM dominance. For these two interaction analyses, a voxel threshold of Z ⫽ 3.09 (P ⬍ 0.001, uncorrected) was applied. Extent threshold was set up to P ⬍ 0.05 which was equivalent to a cluster size of 47 or more contiguous voxels (1269 mm 3) at the chosen voxel level. Single-subject analyses were performed in two ways. First, the individual contribution to the results from the group was quantified. From each peak voxel of the two bilateral prefrontal group maxima (pSFS and MFG), integrated count images were converted into rCBF equivalents for each subject and condition (averaged across three repetitions). Second, to investigate regional individual variability, separate ANCOVAs for each subject were computed in SPM, using the condition replications as factors. Here, the threshold was set to Z ⫽ 2.39 (P ⬍ 0.01, uncorrected). For the comparison of individual activation foci, spatial normalization was also performed. We defined a cylindrical volume of interest (VOI) for the pSFS, which was centered on the mean Talairach coordinates from 20 working memory studies analyzed by D’Esposito et al. (1998). The very similar coordinates from spatial and nonspatial studies were merged together, resulting in x ⫽ ⫺28, y ⫽ ⫺0.5, z ⫽ 50 for the left pSFS and x ⫽ 27, y ⫽ 2.5, z ⫽ 50 for the right pSFS. Activations from our single-subject analyses, peaking within a 10-mm axial radius and differing less than 20 mm in the Z extension, are reported in Table 3. RESULTS Behavioral Data Behavioral control study. The control sample (n ⫽ 16) showed equivalent accuracy (t(15) ⫽ 1.24, P ⫽ 0.23) during SWM (mean correct: 90.9%, SD 6.4) and PWM (mean correct: 88.1%, SD 8.3) Mean RTs differed sig-

PHONOLOGICAL AND SPATIAL WORKING MEMORY

nificantly (t(15) ⫽ 4.9, P ⬍ 0.0002) between SWM (614 ms, SD 243.5) and PWM (712 ms, SD 281.2). PET study. For the PET sample (n ⫽ 8) (Table 1), the MANOVA revealed a significant main effect for the factor “load” (F(1,7) ⫽ 7.77, P ⬍ 0.03). The second main effect “stimulus” was not significant (F(1,7) ⫽ 1.30, P ⬎ 0.3). The interaction “load” ⫻ “stimulus” was not significant (F(1,7) ⫽ 4.47, P ⬎ 0.07). The MANOVA on mean RTs (Table 1) demonstrated significant main effects of both factors “load” (F(1,7) ⫽ 11.23, P ⬍ 0.01) and “stimulus” (F(1,7) ⫽ 28.59, P ⬍ 0.01). Again, the interaction was not significant (F(1,7) ⫽ 1.75, P ⬎ 0.2). Signal detection measures indicated an appropriate performance in all four conditions (Table 1). A MANOVA for repeated measures revealed a significant effect of only “load” on the sensitivity index d⬘ (F(1,7) ⫽ 6.21, P ⬍ 0.05), with the response index ␤ only close to significance (F(1,7) ⫽ 5.57, P ⬎ 0.05). Thus, subjects were biased toward a nontarget response during both WM conditions, when compared with judgment. Main effects of “stimulus” and “stimulus” ⫻ “load” interactions were not significant for d⬘ and ␤. Subjects did not produce any cross-conditional false alarms, i.e., responses to phonological 2-backs (2) in SWM or to spatial 2-backs (3) in PWM, respectively. Mean RTs in these trials did not differ from mean RTs for nontargets in both PWM (Wilcoxon’s T ⫽ 17, P ⫽ 0.89) and SWM (T ⫽ 17, P ⫽ 0.89), although this comparison must be viewed with caution because of the low rate of cross-conditional 2-backs. Subjective evaluation. According to a semistructured postscanning questionnaire, all subjects scanned (n ⫽ 8) reported a visuospatial memorization of items (“marked a point on the screen”) in the spatial tasks. None of them reported attempts to verbalize the locations. As the memorizing strategy for the syllables, all subjects chose the “as sounds” or “more sounds than images” option. No subject chose the remaining options “as images,” “more images than sounds,” and “other.” One subject (S6) reported a “letter-based strategy, orienting on the middle vowel,” trying it during the first runs of pretesting, but not during scanning. One subject included in the behavioral control study (n ⫽ 16) used this strategy throughout the experiment. PET Data Main effect of WM. The conjunction analysis for the main effect of “load” revealed a highly symmetric bifrontal and biparietal activation pattern (Fig. 1a). The highest activation peaks in the group analysis (Fig. 1a, Table 2) were observed for the (posterior) superior frontal sulcus (pSFS, BA 6/8), including adjacent superior and middle frontal gyri (SFG/MFG) for both hemispheres, in the precuneus (PCU), and in the superior parietal lobule (SPL, BA 7). Left pSFS activation extended to the lateral premotor cortex (BA 4/6). At least

49

seven of eight subjects showed an increase in rCBF in both WM tasks within all these group maxima (Fig. 2 for pSFS). The single-subject analysis for the main effect of WM revealed a more robust right than left pSFS activation, with interindividual variability present mainly along the Z axis (Table 3, Fig. 1b). Further, we observed bilateral, highly homologous foci in the middle frontal gyrus (MFG, BA 9). According to a more recent cytoarchitectonic definition by Rajkowska and Goldman-Rakic (1995), both foci were located within the upper BA 46; therefore we refer to them as BA 9/46. Despite the contribution of six (left) and eight (right) subjects to both group maxima within the MFG, corresponding main individual foci substantially varied in location and significance, compared with those within the pSFS (Table 3), the SPL, and the precuneus as revealed by single-subject conjunction analyses. The SPL (BA 7), the precuneus, and the inferior parietal lobule (supramarginal gyrus, BA 40) were activated bilaterally, with significant activation of the supramarginal gyrus at P ⬍ 0.001, uncorrected, and we report it as a region predicted a priori (D’Esposito et al., 1998). Prefrontal lateralization. In post hoc comparisons of rCBF in the pSFS (Table 2), mixed-effects two-way MANOVAs (see Fig. 2 for data) did not show significant main effects of “hemisphere” or “load” ⫻ “hemisphere” interaction during the phonological conditions (PJ, PWM). During the spatial conditions, (SJ, SWM) the main effect of “hemisphere” was significant (F(1,7) ⫽ 59.8, P ⫽ 0.0001), but not the “load” ⫻ “hemisphere” interaction. For the MFG, main effects of hemisphere and interactions of “load” and “hemisphere” were significant neither in the phonological conditions nor in the spatial conditions. Thus, the only one hemispheric preference was not WM specific, as SJ and SWM both were lateralized to the right pSFS. Main effect of stimulus feature. The conjunction analysis for the main effect of phonological processing revealed a clear left lateralization pattern, with two distinct foci in the left inferior frontal gyrus including inferior frontal sulcus (BA 44 and 45/46) (Table 4) and a third activation within the left lingual gyrus (BA 19). This pattern was observed in seven subjects for the anterior LIFG region (Fig. 3) and could be reproduced when PWM ⫺ SWM and PJ ⫺ SJ were tested separately (Fig. 4). The analogous conjunction contrast for the main effect of spatial processing revealed a rightlateralized activation pattern: (medial) superior parietal lobule (BA 7, r ⬎ l), bilateral occipitotemporal junction (BA 37), right supramarginal gyrus (BA 40), and right superior frontal gyrus (medial anteriormost part of BA 8). “Load” ⫻ “stimulus” interaction. We tested for WM activations related to the processed stimulus feature by calculating two-tailed interaction contrasts of “load”

50

ZUROWSKI ET AL.

FIG. 1. Main effect of working memory. (a) Top: Conjunction analysis for the main effect of WM: resulting activations in the MFG (BA 9/46), the pSFS and adjacent gyri (BA 6/8), and superior parietal lobule (BA 7) at P ⬍ 0.05, corrected. SPMs are rendered on a standard magnetic resonance image. Bottom: Left slice: sagittal view through the right SFS maximum (x ⫽ 30); right slice: sagittal view through the left SFS maximum (x ⫽ 24). SFS, superior frontal sulcus; PM, premotor cortex; DLPFC, dorsolateral prefrontal cortex; SPL, superior parietal lobule; PCU, precuneus. (b) Prevalence and individual variability of pSFS activation as revealed by single-subject analysis of the main WM effect (P ⬍ 0.01, uncorrected). Note that these slices only partially reflect the robustness of parietal activations.

⫻ “stimulus” (see Methods). Dominance for SWM (Fig. 5) was observed within the left superior parietal lobule (BA 7), the left anterior and posterior intraparietal sulcus (IPS; stereotaxic coordinates: ⫺21, ⫺60, 51; Z ⫽ 4.17, and ⫺36, ⫺48, 48; Z ⫽ 3.66) as well as the right posterior and anterior IPS (21, ⫺63, 27; Z ⫽ 3.33, and 45, ⫺51, 42; Z ⫽ 3.60). Foci in the left and right

hemisphere were contiguous, with 149 and 150 voxels, respectively. In right pSFS (37 voxels; 21, 3, 51; Z ⫽ 3.70) preference for SWM was detected without a corrected extent threshold only. No regional dominance was found in the analogous interaction for PWM. When the threshold was lowered to P ⬍ 0.01, uncorrected for multiple comparisons, only regions showing relative

PHONOLOGICAL AND SPATIAL WORKING MEMORY

51

(posterior) superior frontal sulcus (BA 6/8) as well as superior and inferior parietal lobule. The bilateral anterior (BA 39/40) and posterior (BA 7) intraparietal sulcus, as well as the right (posterior) superior frontal sulcus, exhibited dominance for SWM. No analogous dominance for PWM was identified, while both PWM and PJ were lateralized to the left inferior frontal gyrus (BA 44 and 45). Behavior and Subjective Evaluation

FIG. 4. Lateralization to the left inferior frontal gyrus (LIFG) in the phonological conditions. Results of two separate SPM contrasts (voxel level at P ⬍ 0.001, extent threshold at P ⬍ 0.05): PJ ⫺ SJ (left three slices) and in PWM ⫺ SWM (right three slices). Two uniquely resulting frontal foci (Table 3) are shown on the same standard MRI slices for each of the contrasts, with overlapping regions as revealed by the respective conjunction (Table 4).

deactivations in SWM or activations in SJ were present, comprising the cuneus as well as the bitemporal and frontopolar cortex. DISCUSSION We observed a common bilateral working memory network for phonological and spatial representations consisting of the middle frontal gyrus (BA 9/46) and the

Behavioral results and the subjective evaluation indicate that subjects were succesively forced to process stimuli either phonologically or spatially. Difficulty levels were comparable in both the behavioral control study and the PET study. The additional requirement of grapheme–phoneme transduction for response selection (Paulesu et al., 2000, Smith and Jonides, 1998) likely contributed to higher RTs in both PJ (compared with SJ) and PWM (compared with SWM). This finding likely reflects instruction-dependent rather than unselective simultaneous processing of spatial and phonological information. While appropriately attending to all four experimental conditions, in both WM tasks subjects showed a “cautious” response criterion (Grier, 1971), with relatively low hit rate and low false alarm rate (Table 1). The lack of false alarms to uninstructed 2-backs further argues against simultaneous coding of spatial and phonological features of the stimuli in both WM conditions. Furthermore, it is unlikely that subjects tried to compare orthographical features (e.g., the middle vowel) in the phonological conditions, a strategy shown to produce a high rate of false alarm responses in our pretesting experiments. Stimulus Processing Pathways and Components of WM Temporoparietal cortex. Regions involved in perception and processing of the syllabic stimuli in the judgment tasks exhibited a gradual dissociation of activations depending on the features (phonology or location) relevant for task performance. This result is consistent with the distinction of a dorsal and a ventral visual pathway for processing of spatial and object

FIG. 5. Regions showing dominance for SWM in the group interaction contrast (P ⫽ 0.001, uncorrected). AIPS, anterior intraparietal sulcus; PIPS, posterior intraparietal sulcus.

52

ZUROWSKI ET AL.

TABLE 2 Main Effect of WM in the Group Conjunction Analysis (P ⬍ 0.05, Corrected) Cortical region

BA

x

y

z

Z score

L posterior superior frontal R posterior superior frontal L middle frontal R middle frontal L posterior parietal R posterior parietal Precuneus R supramarginal

6/8 6/8 9/46 9/46 7 7 7 40

⫺24 30 ⫺42 45 ⫺27 30 0 45

0 3 36 33 ⫺66 ⫺66 ⫺69 ⫺42

51 54 33 33 33 48 48 36

Inf. (T ⫽ 6.41) 7.31 5.21 4.83 6.54 5.79 5.56 4.88

Note. Activation in the left supramarginal gyrus occurred only at P ⬍ 0.001, uncorrected (stereotaxic coordinates: ⫺54, ⫺54, 45, Z ⫽ 3.85). Activations within cerebellar hemispheres observed in single-subject analyses (four of eight subjects) were not significant in the group analysis.

information, respectively (Ungerleider and Haxby, 1994). Regions related to the dorsal pathway were found in both spatial conditions (SWM, SJ), i.e., the right (medial) superior parietal lobule, the right supramarginal gyrus, and further the bilateral superior occipitotemporal junction (BA 37, 19). In contrast, all regions dominant in the phonological conditions (main effect) are associated with the ventral visual pathway, including the left lingual gyrus and LIFG. Furthermore, activation of the left inferior temporal and the fusiform gyrus was observed in several subjects in both

phonological conditions, although this activation was not significant in the group analysis. These regions are involved in word perception and encoding (Fiez and Petersen, 1998), and may mediate the transduction from graphemes to phonemes (Paulesu et al., 1993, 2000). It should be noted that only the phonological but not the spatial conditions required orthographic (“object”) processing of the syllables for task completion. Thus, context (instruction)-dependent signals from prefrontal areas may have biased activation in posterior regions as early as in these “higher” visual areas (Chelazzi et al., 1993; Miller and Cohen, 2001). Inferior frontal cortex. Both phonological conditions (PJ and PWM) produced a strongly left-lateralized activation pattern within two distinct regions (BA 44 and BA 45), generally accepted as subserving phonological processing (Poldrack et al., 1999). Activation of posterior LIFG (BA 44) and activation of anterior LIFG (BA 45) are two of the most frequently reported findings both in phonological judgment tasks (Poldrack et al., 1999, for review; Burton et al., 2000; Zatorre et

TABLE 3 Main Effect of WM: Individual Peaks within the VOI for pSFS Subject S1 S2 S3 S4 S5 S6 S7 S8 FIG. 2. Single subjects’ contribution to the main effect of WM within pSFS. Mean rCBF at the peak voxel (arbitrary units, SE given) is shown for each subject and each condition separately. (a) rCBF at x, y, z ⫽ ⫺24, 0, 51 (left SFS group maximum). (b) rCBF at x, y, z ⫽ 30, 3, 54 (right pSFS group maximum).

Mean SD

Coordinates, left (x, y, z)

Z score

⫺21, ⫺6, 57

6.36

⫺27, ⫺24, ⫺18, ⫺27,

4.20 4.19 5.91 4.8

9, 72 0, 57 ⫺3, 48 ⫺3, 58

⫺23, 0, 58 3.9, 5.8, 8.6

Coordinates, right (x, y, z) 30, 18, 36, 33, 30, 18, 27, 30,

⫺3, 54 0, 69 9, 63 12, 63 0, 48 6, 48 6, 45 9, 54

Z score 6.92 4.01 4.89 4.50 5.05 5.11 4.09 4.36

26, 5, 55 6.1, 5.3, 8.6

Note. In single-subject analyses a threshold of P ⫽ 0.01 (voxel level; with corrected extent threshold at P ⬍ 0.05) was used.

PHONOLOGICAL AND SPATIAL WORKING MEMORY

TABLE 4 Main Effect of Phonological Processing (Conjunction) Cortical region

BA

Coordinates

Z score

L inferior frontal gyrus L inferior frontal gyrus L lingual gyrus

44 45/46 19

⫺45, 3, 21 ⫺36, 36, 12 ⫺15, ⫺72, ⫺12

5.94 6.56 4.59

al., 1992; Paulesu et al., 1993) and in verbal WM tasks (Petrides et al., 1993; Cohen et al., 1994; Awh et al., 1996; Smith et al., 1996; Braver et al., 1997; Paulesu et al., 1993). Specifically, the executive component of phonetic judgment (Zatorre et al., 1992; Burton et al., 2000) was proposed to activate the LIFG. In the present study, activation of the LIFG was regionally and quantitatively equivalent in judgment (PJ) and working memory (PWM) (Fig. 3), indicating their unspecific involvement whenever phonological discrimination is required (Nystrom et al., 2000). Considering comparable WM investigations, we suggest that observed differences or commonalities within this region vary as a function of perceptual, procedural, and strategical difference/similarity between compared WM and non-WM conditions. Postle et al. (2000) suggested that even differences in perceptual difficulty between compared conditions may sufficiently explain inconsistencies between studies exploring material (domain) specificity. For example, the clear-cut hemispheric dissociation observed in Smith and colleagues’ Expt 1 (Smith et al., 1996) became only a gradual one in Expt 2. First, Expt 2 only included the same stimulus material across instructions and, second, a stronger component of verbal processing and decision when comparing the probe to three possible targets. Also, Expt 2 was a within-subject comparison, thus controlling for sample effects. From this viewpoint, our approach designed to further narrow the search for WM-“specific” versus rather general components via factorial design is similar to those of Paulesu et al. (1993). Both studies used a non-WM condition forcing subjects to make more demanding phonological judgments, which activated very similar sites within LIFG (BA 44, 45) and left lingual gyrus. In contrast, parietal (BA 40) and superior frontal (BA 6) areas were activated by WM maintenance and manipulation of letters as well as of abstract objects (Korean letters). The present study extended the latter evidence to spatial memoranda. Two further studies investigated verbal and spatial WM within subjects. The study by D’Esposito et al. (1998) failed to find specificity of right DLPFC for spatial WM or left inferior frontal cortex for verbal WM. Similarly, Nystrom et al. (2000) did not find evidence for any other region specific for either WM for letters or locations,

53

with the exception of the left (posterior) superior frontal sulcus for locations. It might be argued that at least in part subjects’ encoding and processing strategies define the effective nature of the task employed. For example, during the object WM condition in Nystrom et al. (2000) several subjects employed verbal strategies, with consecutive activation of left BA 44 and 40. The present data strongly indicate that subjects followed phonological versus spatial strategies. Thus, the main effect of WM can be regarded as a result corrected for commonly induced perceptual and processing strategies within each domain. In the light of the Smith et al. (1996) study our findings replicated the preferential involvement of the LIFG in verbal–phonological WM, when compared with spatial WM. However, they argue against its WM specificity, as the LIFG did not show any response to WM demands when compared with judgment (Fig. 3) (D’Esposito et al., 1998; Nystrom et al., 2000, for discussion). Intraparietal Cortex and Spatial WM The intraparietal sulcus has been found to mediate spatial attention and orientation in a large number of studies (e.g., Corbetta, 1998; Nobre et al., 1997; Donner et al., 2000). Here, IPS and right pSFS (Smith and Jonides, 1998) were preferentially activated by spatial WM demands, compared with spatial judgment. Coull and Frith (1998) proposed both, higher demands of spatial attention and WM to explain activation of the IPS in spatial WM. The preferential involvement of the corresponding region for spatial mnemonic processing has been demonstrated

FIG. 3. Individual contribution to the LIFG (BA 45) activation from the group (Table 4). rCBF equivalents (arbitrary units) for all subjects and all four conditions are shown. Bars 1– 4 for each subject: (1) phonological judgment (PJ), (2) phonological WM (PWM), (3) spatial judgement (SJ), (4) spatial WM (SWM). A mixed-effects MANOVA using subject- and condition-specific rCBF equivalents revealed a significant main effect of “stimulus” (F(1,7) ⫽ 28.7, P ⬍ 0.001). WM load had no effect on this region (F(1,7) ⫽ 0.008, P ⬎ 0.93), irrespective of stimuli, as revealed by the interaction (F(1,7) ⫽ 0.82, P ⬎ 0.39). A similar pattern was observed for the posterior LIFG focus (not shown).

54

ZUROWSKI ET AL.

in monkeys and its direct projections to the PFC have been identified (Goldman-Rakic et al., 2000, Wise et al., 1996). The presented IPS activation corresponds to the human homologue of the area LIP in monkeys (Corbetta, 1998; Donner et al., 2000; Nobre et al., 1997), which subserves spatial attentional as well as spatial WM functions, though some specialization on the neuronal level has been described. Dorsolateral Prefrontal Cortex and Executive Functions We observed bilateral activation in the MFG (BA 9/46) or mid-DLPFC (Petrides, 1994), regardless of mnemonic representations (phonological and spatial) held in WM. It is notable that the MFG activation was considerably less significant and more variable across subjects than activation of the pSFS and the parietal cortex. This finding is consistent with the materialindependent involvement of BA 9/46 in human working memory, when manipulation in contrast to “maintenance only” is required (Petrides, 1994; Owen et al., 1996, 1998). However, our design would not allow us to falsify this hypothesis, as it did not include a “maintenance-only” condition. It should be mentioned that all four experimental conditions explicitly included an appreciable component of a context (instruction)-dependent decision in the presence of gradually distracting probe stimuli, here syllables roughly rhyming to the target phoneme in PJ, or close-to-target locations in SJ. The capacity to attend to the actually relevant stimulus class and to select the response while inhibiting others has been ascribed to the PFC (Smith and Jonides, 1999), as one important correlate of the “central executive” (Baddeley, 1986). Thus, our conjunction analysis for the main effect of WM was less sensitive to this component, which likely was “subtracted out” in part. If so, variable presence of BA 9/46 activation would likely reflect individual cognitive resources and additional monitoring requirements. Indeed, respective lateralization patterns on the single-subject level were more individual than they were related to any WM condition: no significant lateralization effects for any WM task were present on the group level. Using EEG, Gevins and Smith (2000) recently found prefrontal involvement in a 2-back task to be inversely correlated with subjects’ general cognitive ability, whereas lateralization patterns reflected subjects’ cognitive styles rather than dependence on task condition. Recent evidence from event-related fMRI studies relates the DLPFC more to response selection (“executive”) than to active maintenance within WM (Rowe et al., 2000; Postle et al., 1999). This is in line with the notion of its significance in representing environmental context, including abstract patterns of response contingencies (Cohen et al., 1997; Miller and Cohen, 2001).

The Role of Superior Frontal Cortex in Working Memory The WM study by Rowe and colleagues identified two sites of activity related to active maintenance of spatial representations: the posterior parietal cortex (BA 7) and a region corresponding to SFS/MFG. Very similar regions exhibited the highest Z scores and subject prevalence in the present PET study, extending this evidence to verbal–phonological WM, too. Beyond the widely established role of the posterior parietal cortex in WM (Smith and Jonides, 1998, 1999), involvement of the pSFS and its vicinity (BA 6/8) is one of the most consistent findings across working memory investigations in humans derived from functional imaging. Activation of the pSFS at coordinates very similar to those in the present study was frequently reported in both verbal–phonological WM (e.g., Awh et al., 1996; Schumacher et al., 1996; Jonides et al., 1998; LaBar, 1999) and spatial WM (e.g., Jonides et al., 1993; Smith et al., 1996; Carlson et al., 1998; Courtney et al., 1996, 1998a; Rowe et al., 2000). Investigating the overlap of spatial attention and verbal working memory, LaBar et al. (1999) reported activation sites corresponding to pSFS as being selectively engaged in WM, in contrast to the DLPFC. As in our investigation (see also Awh et al., 1996; Schumacher et al., 1996, auditory 2-back; Smith et al., 1996, Expt 3) these activations showed the highest Z scores of all reported regions. What follows from this for the role of the pSFS in working memory? Recently, Cabeza and Nyberg (2000) did not find reliable evidence for the involvement of BA 6 in WM studies using objects, especially faces, i.e., iconic and directly meaningful memoranda. In contrast, BA 6 showed the highest scores (70 –100%) of frequency in both verbal, and spatial WM. Notably, verbal–phonological and spatial WM tasks share one aspect, which might explain this puzzling dissociation. Both require the generation and active maintenance of abstract noniconic representations, a process possibly engaging regions different from those for maintenance of more directly stimulus-driven meaningful representations, rather corresponding to stimulus-specific sensory areas (Goldman-Rakic, 1987). Not surprisingly, regions found to be dominant in WM tasks for faces more convincingly correspond in human and nonhuman primates than those of other information domains (Goldman-Rakic et al., 2000). Given the strong reciprocal interconnections between the prefrontal, superior frontal, and posterior parietal cortex in primates (Wise et al., 1996), it is possible that attention and working memory-related bias signals from the PFC not only influence neural activity in visual areas (Chelazzi et al., 1993), but, similarly, the preferential aspect of various information represented in the parietal “shortterm store” (Baddeley, 1986; Smith and Jonides, 1999). Additional involvement of superior frontal cortex might be needed when abstract contextual information

PHONOLOGICAL AND SPATIAL WORKING MEMORY

(e.g., “instruction”) beyond the stimulus itself constitutes the internal representations to be generated and held in WM, i.e., the less stimulus driven they are. WM-related activity in these two areas shows a remarkable covariance across WM studies and there is evidence for a regional distinction between active maintenance (BA 6/8) and passive storage (BA 7, 40) (Smith and Jonides, 1999). Another related requirement included in both of our 2-back WM conditions is the temporal sorting of WM representations (Cohen et al., 1997), similarly activating pSFS and DLPFC, for example, in “self-ordered” WM tasks (Petrides et al., 1993; Petrides, 1994). It is important to note that some subspecializations on the neuronal level (Goldman-Rakic, 1987) may be present despite the common WM network revealed by our conjunction analysis. Further, a posteromedial portion of the right SFS showed significant preference for SWM (Fig. 5), supporting findings by Courtney et al. (1998a; see also Smith and Jonides, 1999). Again, this was not the case for the larger and more anterior-laterally located region showing feature independent WM-related activation at a considerably higher significance level. WM-dependent hemispheric specificity remains, however, questionable as Nystrom et al. (2000) reported the same gradual preference, but for the left pSFS. Both articulatory rehearsal and maintenance of locations might further involve activation of shared motor programs, with information coded in “motor coordinates,” a component proposed for active spatial versus passive visual storage (Logie et al., 1990). Notably, it is difficult to explain activity in pSFS only by saccadic eye movements, as one would expect further regions known to control saccades (Fletcher and Henson, 2001). In our study, potential eye movements during encoding or response would be present in judgment as well as in WM. In contrast, imagined saccades as correlates of maintenance (rehearsal) would be preferentially expected in spatial WM. However, pSFS activation was predominant in SWM as well as PWM, when compared with judgment. Although the compelling question of a functional anatomical relationship between the reported pSFS foci and the frontal eye field (FEF) cannot be resolved here, the human FEF has been localized considerably more posterior and lateral to the reported WM-related pSFS activations (Petit et al., 1997; Petit and Haxby, 1999), and is most probably part of the precentral ´ Scalaidhe and Goldman-Rakic, gyrus. In monkeys (O 1993) as well as in humans (Courtney et al., 1998a) the main locus of sustained activity in spatial WM could be separated from the FEF. It is also very unlikely that our comparison of tasks with equivalent sensorimotor requirements revealed the highest of all observed Z scores in the FEF, whereas activation levels in studies devoted to the investigation of the FEF usually are remarkably lower compared with activations found in

55

parieto-occipital regions (Petit et al., 1997; Petit and Haxby, 1999; Donner et al., 2000). CONCLUSIONS Recent findings of a common bilateral frontoparietal network for working memory have been corroborated and further extended as occurring regardless of different visuospatial or phonological memoranda, derived from physically identical stimuli. Unlike preferential involvement of the intraparietal sulcus in spatial WM, activation within the left inferior frontal gyrus in verbal–phonological WM may be sufficiently explained by the general requirements of phonological processing and judgment. In line with a large number of previous findings, we suggest that the bilateral (posterior) superior frontal sulcus and vicinity play an important role for active maintenance and manipulation of information abstracted from the same environmental set. ACKNOWLEDGMENTS The authors gratefully acknowledge helpful comments from Christian Bu¨chel, Tobias Donner, Markus Kiefer, Leigh Nystrom, and Roberto Viviani.

REFERENCES Awh, E., Jonides, J., Smith, E. E., Schumacher, E. H., Koeppe, R. A., and Katz, S. 1996. Dissociation of storage and rehearsal in verbal working memory: Evidence from PET. Psychol. Sci. 7: 25–31. Baddeley, A. 1986. Working Memory. Clarendon Press, Oxford. Braver, T. S., Cohen, J. D., Nystrom, L. E., Jonides, J., Smith, E. E., and Noll, D. C. 1997. A parametric study of prefrontal cortex involvement in human working memory. NeuroImage 5: 49 – 62. Burton, M., Small, S., and Blumstein, S. 2000. The role of segmentation in phonological processing: An fMRI investigation. J. Cogn. Neurosci. 12: 679 – 690. Cabeza, R., and Nyberg, L. 2000. Imaging cognition. II. An empirical review of 275 PET and fMRI studies. J. Cogn. Neurosci. 12: 1– 47. Carlson, S., Martinkauppi, S., Rama, P., Salli, E., Korvenoja, A., and Aronen, H. J. 1998. Distribution of cortical activation during visuospatial n-back tasks as revealed by functional magnetic resonance imaging. Cereb. Cortex 8: 743–752. Chelazzi, L., Miller, E. K., Duncan, J., and Desimone, R. 1993. A neural basis for visual search in the inferior temporal cortex. Nature 363: 345–347. Cohen, J. D., Forman, S. D., Braver, T. S., Casey, B. J., ServanSchreiber, D., and Noll, D. C. 1994. Activation of prefrontal cortex in a nonspatial working memory task with functional MRI. Hum. Brain Mapping 1: 293–304. Cohen, J. D., Perlstein, W. M., Braver, T. S., Nystrom, L. E., Noll, D. C., Jonides, J., and Smith, E. E. 1997. Temporal dynamics of brain activation during a working memory task. Nature 386: 604 – 608. Corbetta, M. 1998. Frontoparietal cortical networks for directing attention and the eye to visual locations: Identical, independent, or overlapping neural systems? Proc. Natl. Acad. Sci. USA 95: 831– 838. Coull, J. T., and Frith, C. D. 1998. Differential activation of right

56

ZUROWSKI ET AL.

superior parietal cortex and intraparietal sulcus by spatial and nonspatial attention. NeuroImage 8: 176 –187. Courtney, S. M., Petit, L., Maisog, J. M., Ungerleider, L. G., and Haxby, J. V. 1998a. An area specialized for spatial working memory in human frontal cortex. Science 279: 1347–1351. Courtney, S. M., Petit, L., Haxby, J. V., and Ungerleider, L. G. 1998b. The role of prefrontal cortex in working memory: Examining the contents of consciousness. Phil. Trans. R. Soc. London Ser. B 353: 1819 –1828. Courtney, S. M., Ungerleider, L. G., Keil, K., and Haxby, J. V. 1996. Object and spatial visual working memory activate separate neural systems in human cortex. Cereb. Cortex 6: 39 – 49. D’Esposito, M., Aguirre, G. K., Zarahn, E., Ballard, D., Shin, R. K., and Lease, J. 1998. Functional MRI studies of spatial and nonspatial working memory. Cogn. Brain Res. 7: 1–13. Donner, T., Kettermann, A., Diesch, E., Ostendorf, F., Villringer, A., and Brandt, S. A. 2000. Involvement of the human frontal eye field and multiple parietal areas in covert visual selection during conjunction search. Eur. J. Neurosci. 12: 3407–3414. Fiez, J. A., and Petersen, S. E. 1998. Neuroimaging studies of word reading. Proc. Natl. Acad. Sci. USA 95: 914 –921. Fletcher, P. C., and Henson, R. N. A. 2001. Frontal lobes and human memory. Insights from neuroimaging. Brain 124: 849 – 881. Friston, K. J., Holmes, A., Worsley, K. J., Poline, J. B., Frith, C. D., and Frackowiak, R. S. J. 1995. Statistical parametric maps in functional imaging: A general linear approach. Hum. Brain Mapping 2: 165–189. Fuster, J., Bauer, R., and Jervey, J. 1982. Cellular discharge in the dorsolateral prefrontal cortex of the monkey in cognitive tasks. Exp. Neurol. 77: 679 – 694. Gevins, A., and Smith, M. 2000. Neurophysiological measures of working memory and individual differences in cognitive ability and cognitive style. Cereb. Cortex 10: 829 – 839. Goldman-Rakic, P. 1987. Circuitry of the prefrontal cortex and the regulation of behavior by representational memory. In Handbook of Neurobiology (V. Mountcastle and F. Plum, Eds.), pp. 373– 417. Am. Physiol. Soc., Bethesda, MD ´ Scalaidhe, S., and Chafee, M. 2000. Domain Goldman-Rakic, P., O specificity in cognitive systems. In The New Cognitive Neurosciences (M. S. Gazzaniga, Ed.), pp. 733–742. MIT Press, Cambridge. Grier, J. B. 1971. Nonparametric indexes for sensitivity and bias: Computing formulas. Psychol. Bull. 75: 424 – 429. Jonides, J., Schumacher, E. H., Smith, E. E., Koeppe, R. A., Awh, E., Reuter-Lorenz, P. A., Marshuetz, C., and Willis, C. R. 1998. The role of parietal cortex in verbal working memory. J. Neurosci. 18: 5026 –5034. Jonides, J., Smith, E. E., Koeppe, R. A., Awh, E., Minoshima, S., and Mintun, M. A. 1993. Spatial working memory in humans as revealed by PET. Nature 363: 623– 625. LaBar, K. S. 1999. Neuroanatomic overlap of working memory and spatial attention networks: A functional MRI comparison within subjects. NeuroImage 10: 695–704. Logie, R. H., Zucco, G. M., and Baddeley, A. D. 1990. Interference with visual short-term memory. Acta Psychol. 75: 55–74. Miller, E. K., and Cohen, J. D. 2001. An integrative theory of prefrontal cortex function. Annu. Rev. Neurosci. 24: 167–202. Nobre, A. C., Sebestyen, G. N., Gitelman, D. R., Mesulam, M. M., Frackowiak, R. S., and Frith, C. D. 1997. Functional localisation of the system for visuospatial attention using positron emission tomography. Brain 120: 515–533. Nystrom, L. E., Braver, T. S., Sabb, F. W., Delgado, M. R., Noll, D. C., and Cohen, J. D. 2000. Working memory for letters, shapes, and

locations: fMRI evidence against stimulus-based regional organization in human prefrontal cortex. NeuroImage 11: 424 – 446. ´ Scalaidhe, S. P., and Goldman-Rakic, P. S. 1993. Memory fields in O the prefrontal cortex of the macaque. Soc. Neurosci. Abstr. 19: 326.6. ´ Scalaidhe, S. P., and Goldman-Rakic, P. S. 1997. Areal segregation O of face processing neurons in prefrontal cortex. Science 278: 1135– 1138. Owen, A. M., Milner, B., Petrides, M., and Evans, A. C. 1996. Memory for object features versus memory for object location: A positron-emission tomography study of encoding and retrieval processes. Proc. Natl. Acad. Sci. USA 93: 9212–9217. Owen, A. M., Stern, C. E., Look, R. B., Tracey, I., Rosen, B. R., and Petrides, M. 1998. Functional organisation of spatial and nonspatial working memory processing within the human lateral frontal cortex. Proc. Natl. Acad. Sci. USA 95: 7721–7726. Paulesu, E., Frith, C. D., and Frackowiak, R. S. J. 1993. The neural correlates of the verbal component of working memory. Nature 362: 342–345. Paulesu, E., et al. 2000. A cultural effect on brain function. Nat. Neurosci. 3: 91–96. Petit, L., and Haxby, J. V. 1999. Functional anatomy of pursuit eye movements in humans as revealed by fMRI. J. Neurophysiol. 82: 463– 471. Petit, L., Clark, V., Ingeholm, J., and Haxby, J. 1997. Dissociation of saccade-related and pursuit-related activation in human frontal eye fields as revealed by fMRI. J. Neurophysiol. 77: 3386 –3390. Petrides, M. 1994. Frontal lobes and behavior. Curr. Opin. Neurobiol. 4: 207–211. Petrides, M., Alivisatos, B., Meyer, E., and Evans, A. C. 1993. Functional activation of the human frontal cortex during the performance of verbal working memory tasks. Proc. Natl. Acad. Sci. USA 90: 878 – 882. Poldrack, R. A., Wagner, A. D., Prull, M. W., Desmond, J. E., Glover, G. H., and Gabrieli, J. D. 1999. Functional specialization for semantic and phonological processing in the left inferior prefrontal cortex. NeuroImage 10: 15–35. Postle, B. R., and Esposito, M. D. 1999. “What”–then–“where” in visual working memory: An event-related fMRI study. J. Cogn. Neurosci. 11: 585–597. Postle, B. N., Berger, J. S., and D’Esposito, M. 1999. Functional neuroanatomical double dissociation of mnemonic and executive control processes contributing to working memory. Proc. Natl. Acad. Sci. USA 96: 12959 –12964. Postle, B. R., Stern, C. E., Rosen, B. R., and Corkin, S. 2000. An fMRI investigation of cortical contributions to spatial and nonspatial visual working memory. NeuroImage 11: 409 – 423. Price, C. J., and Friston, K. J. 1997. Cognitive conjunction: A new approach to brain activation experiments. NeuroImage 5: 261– 270. Rajkowska, G., and and Goldman-Rakic, P. S. 1995. Cytoarchitectonic definition of prefrontal areas in the normal human cortex. II. Variability in locations of areas 9 and 46 and relationship to the Talairach Coordinate System. Cereb. Cortex 5: 323–337. Rao, S. C., Rainer, G., and Miller, E. K. 1997. Integration of what and where in the primate prefrontal cortex. Science 276: 821– 824. Rowe, J. B., Toni, I., Josephs, O., Frackowiak, R. S., and Passingham, R. E. 2000. The prefrontal cortex: Response selection or maintenance within working memory? Science 288: 1656 – 1660.

PHONOLOGICAL AND SPATIAL WORKING MEMORY Schumacher, E. H., Lauber, E., Awh, E., Jonides, J., Smith, E. E., and Koeppe, R. A. 1996. PET evidence for an amodal verbal working memory system. NeuroImage 3: 79 – 88. Smith, E. E., Jonides, J., and Koeppe, R. A. 1996. Dissociating verbal and spatial working memory using PET. Cereb. Cortex 6: 11–20. Smith, E. E., and Jonides, J. 1998. Neuroimaging analyses of human working memory. Proc. Natl. Acad. Sci. USA 95: 12061–12068. Smith, E. E., and Jonides, J. 1999. Storage and executive processes in the frontal lobes. Science 283: 1657–1661. Talairach, J., and Tournoux, P. 1988. A Co-planar Stereotaxic Atlas of a Human Brain. Thieme, Stuttgart.

57

Ungerleider, L., and Haxby, J. 1994. What and where in the human brain. Curr. Opin. Neurobiol. 4: 157–165. Wilson, F. A. W., O Scalaidhe, S. P., and Goldman-Rakic, P. S. 1993. Dissociation of object and spatial processing domains in primate prefrontal cortex. Science 260: 1955–1958. Wise, S. P., Boussaud, D., Johnson, P. B., and Caminiti, R. 1996. Premotor and parietal cortex: Corticocortical connectivity and combinatorial computations. Annu. Rev. Neurosci. 20: 25– 42. Zatorre, R. J., Evans, A. C., Meyer, E., and Gjedde, A. 1992. Lateralization of phonetic and pitch discrimination in speech processing. Science 256: 846 – 849.