PET validation of a novel prefrontal task: Delayed response alteration

Neuropsychology 1996, Vol. 10, No. 1,3-10

In the public domain

PET Validation of a Novel Prefrontal Task: Delayed Response Alternation James M. Gold, Karen Faith Berman, Christopher Randolph, Terry E. Goldberg, and Daniel R. Weinberger National Institute of Mental Health Neuroscience Center at St. Elizabeths Deficits in working memory have been proposed to explain the performance failures of frontally lesioned primates on delayed alternation (DA) and delayed response (DR) tasks. The authors examined a computerized test of delayed response alternation (DRA), which combines elements of DR and DA in a sample of 18 normal volunteers who underwent oxygen-15 PET regional cerebral blood flow scans during the DRA and a sensorimotor control task. Significant activations were observed in a network of frontal, parietal, occipital, and temporal regions during initial task performance. A qualitatively similar but somewhat reduced set of activations was observed in a subset of participants who repeated the task after practice and instruction. These results are consistent with distributed models of working memory derived from studies of nonhuman primates and suggest that the frontal lobes contribute to human working memory function.

There is general consensus within neuropsychology that the frontal lobes mediate a diverse and complex group of cognitive functions commonly subsumed under the term executive control, including planning, judgment, reasoning, and selfmonitoring (Stuss & Benson, 1986; Weinberger, 1993). Clinical testing instruments thought to measure executive functions are similarly complex and varied, such as the Wisconsin Card Sorting Test (WCST), Stroop Color Word Interference, Trail Making, and verbal fluency, and a more componential cognitive view of frontal lobe function have been lacking. However, several recent computational approaches have suggested that working memory may be a critical common denominator of several complex frontal tasks including the Wisconsin Card Sorting Test (Kimberg & Farah, 1993; Pennington, 1994). This perspective is consistent with a long history of research on nonhuman primates, which has suggested that the frontal cortex plays a critical role in the ability to hold perceptually absent information on line to guide action (Goldman-Rakic, 1987). This type of short-term or working memory appears to be an important common feature of the delayed response (DR) and delayed alternation (DA) paradigms that have proven to be sensitive to prefrontal lesions in nonhuman primates (GoldmanRakic, 1987). In these paradigms, the animal must track either the location of a stimulus or of its own prior response choice in order to gain a food reward following the imposition of a delay period. Thus, there are no environmental cues available at the time the animal makes a response, necessitating the use of an James M. Gold, Karen Faith Berman, Christopher Randolph, Terry E. Goldberg, and Daniel R. Weinberger, Clinical Brain Disorders Branch, Intramural Research Program, National Institute of Mental Health Neuroscience Center at St. Elizabeths. Christopher Randolph is now at the Departments of Psychiatry and Neurology, Northwestern University. Correspondence concerning this article should be addressed to James M. Gold, who is now at the Maryland Psychiatric Research Center, P.O. Box 21247, Baltimore, Maryland 21228.

internal representation to guide action. This representation must be updated on every trial. In single cell recordings from the frontal cortex of behaving monkeys, increased firing of prefrontal neurons has been observed during the delay intervals of DR supporting the concept that these neurons are involved in the maintenance of information (Funahshi, Bruce, & Goldman-Rakic, 1989; Fuster, 1973). Goldman-Rakic and Friedman (1991) in summarizing a series of studies with 2-deoxyglucose (2DG) technique noted that the performance of delayed response tasks resulted in increased glucose utilization in the frontal cortex, hippocampus, thalamus, and parietal cortex, suggesting that a widely distributed network of interconnected anatomic regions participate in the performance of these tasks (see also Friedman & Goldman-Rakic, 1994). These paradigms have had limited use in human populations, but the available data are encouraging. Patients with frontal lobe lesions have been found to perform poorly on DA and DR, and the extent of DR and DA impairment correlated with perseverative responses on the WCST (Freedman & Oscar-Berman, 1986). Impairments have also been observed most often on DA in patients with Parkinson's disease, KorsakofFs syndrome, and Alzheimer's disease (see OscarBerman, McNamara, & Freedman, 1991, for review). Of note, Verin et al. (1993) reported that patients with frontal lesions performed poorly on a classical delayed response test, but showed supernormal performance on a delayed alternation test, as these patients demonstrated an unusual alternation "bias." Although this result is somewhat at odds with the other studies in this area, their overall pattern of results on a series of alternation and reversal tasks suggest that the frontal cortex plays an important role in the sequential organization of behavior in response to changing environmental cues. Thus, their results can be seen as broadly compatible with a working memory framework. Initial applications of physiological imaging techniques have also suggested the involvement of the frontal cortex in elementary working memory tasks. Jonides et al. (1993) recently reported that normal subjects studied with PET demonstrate

GOLD, HERMAN, RANDOLPH, GOLDBERG, AND WEINBERGER

enhanced blood flow in a network of right hemisphere prefrontal, parietal, occipital, and premotor areas while performing a spatial delayed response task. Paulesu, Frith, and Frackowiak (1993) reported that phonological storage involves the left supramarginal gyrus, whereas subvocal rehearsal was associated with activation of Broca's area, illustrating the distributed nature of verbal working memory. Petrides, Alivisatos, Evans, and Meyer (1993) reported evidence of frontal lobe activation during the performance of tasks that required the monitoring of self-generated response sequences. Additionally, they reported frontal activation while subjects monitored a scrambled series of numbers ranging from 1 to 10 and were required to detect the one missing number (Petrides, Alivisatos, Meyer, & Evans, 1993). Herman et al. (1995) found that performance of the WCST resulted in widespread activation of frontal, parietal, temporal, and occipital structures similar to that activated in the simpler visual working memory paradigms cited above. Thus, preliminary human clinical and physiological studies suggest that delayed response and working memory paradigms involve prefrontal and posterior brain regions with somewhat differing activation patterns depending on specific task demands. One potential clinical limitation of delayed response paradigms is that reported error rates have consistently been low, suggesting that humans may be able to use other cognitive processes to mediate or support task performance. In an effort to increase the processing demand of these paradigms, we developed a hybrid computerized version of DA and DR that we call delayed response alternation (DRA). In DRA, the subject is first presented for 1 s with a simple display of two boxes, one colored and one uncolored. After a 2-s presentation of an empty screen, the subject next sees two empty boxes and at this point makes a response. Feedback (the words right and wrong) was presented on the screen for 1.5 s as soon as the response was registered. The next trial then began 1.5 s after the feedback was cleared from the screen with a new display of two boxes, one of which is colored, and the sequence of stimulus delay-response was repeated. The correct answer alternates every trial between being on the same side as the colored box to being on the same side as the uncolored box. The position of the colored box varied randomly over trials. Prior to doing the task, subjects were told that there is a simple rule that determines the correct answer on each trial and that the correct answer is sometimes on the same side as the colored box and sometimes on the same side as the empty, uncolored box. A sample stimulus sequence is shown in Figure 1. Thus, initial performance of the task involved the discovery of a simple rule through the use of feedback. After discovering the rule, the performance of the task required the simultaneous updating of stimulus location and updating of the alternating response sequence. Both pieces of information must be held on line over the delay peri6d and integrated to select a response. Both initial rule discovery and rule application stages of performance appear to require the operation of working memory. We examined the regional cerebral blood flow of normal volunteers during DRA task performance with the oxygen-15 PET technique. Both initial task acquisition and practiced performance were studied with this method.

Delayed Response Alternation Task

D DH

D Right

D ED

D

D

Right

D D

D

Right

Figure 1. Sample of six delayed response alternation trials with stimulus at top of each pair of boxes and correct response choice below indicated by circle.

Method Participants Eighteen right-handed (12 men and 6 women) between the age of 20 and 38 (M = 27) took part in the study. They were highly educated, having completed a mean of 19 years of education. Participants, recruited from the normal volunteer office of the National Institute of Health (NIH) and by word of mouth, were paid. All participants provided written informed consent in accordance with NIH Institutional Review Board guidelines. They were screened for past and present medical, neurological, and psychiatric problems. Participants were taking no medications and were asked to refrain from caffeine and nicotine for 4 hr prior to the study.

Task Conditions Tasks were presented on a computer monitor suspended from a gantry after participants were positioned inside the PET scanner. Participants indicated their responses by pressing one of two buttons on a response box held in their right hands. They were trained in the use of the response box prior to the scan. Two task conditions were compared: the DRA and a sensorimotor control task. The order of the two tasks was counterbalanced across participants. In the control task, they were presented with two boxes, one blue and one uncolored, and were instructed to press the button on the same side as the blue box while the box was visible on the screen. Feedback was provided for each response. The DRA used the same stimulus presentation (except

DELAYED RESPONSE that the colored box was yellow) and response apparatus as the control task. Thus, the two conditions involved nearly identical stimuli, identical responses, and similar number of responses. Nine of the participants performed the DRA again later in the PET session. The second trial followed after allowing time for explicit instruction and practice between the two scans. Participants began each task approximately 60 s before the injection of 40 to 46 mCi oxygen-15 water and continued the tasks for approximately 4 min after the bolus injection.

PET Methodology Participants were scanned in a recumbent position with a Scanditronix PC2048-15B brain tomograph, which obtains 15 slices 6.5 mm thick (full width at half maximum) through the brain with in-plane resolution after reconstruction of 6.0 to 6.5 mm. The slices were oriented parallel to a line drawn between the outer canthus of the eye and the external auditory meatus, and the scan data were reconstructed with corrections for attenuation, random coincidences, scatter, and deadtime. A total of 16 dynamic scans (12 x 10 s and 4 x 30 s) were obtained during the 4 min following the arrival of tracer in the brain, allowing for the determination of the time course of regional cerebral radiation concentration. Transmission scans were obtained to correct for tissue and skull attenuation of radioactivity. A thermoplastic head restraint was used to maintain a stable head position during the scans. The arterial input function was obtained by automated arterial blood sampling throughout each scan (Daube-Witherspoon, Chon, Green, Carson, & Herscovitch, 1992). The arterial time-activity curve along with the 16 dynamic scans was analyzed with a pixel-by-pixel least squares method to produce quantitative images of regional cerebral blood flow (rCBF; Koeppe, Holden, & Ip, 1985).

Statistical Analysis We used the Statistical Parametric Mapping (SPM) technique developed by Friston, Frith, Liddle, and Frackowiak (1991; Friston et al., 1990), which evaluates the probability of significant change throughout the entire brain on a pixel-by-pixel basis. In this technique, the original 15-slice data were interpolated to 43 planes, creating nearly cubic voxels. The data were then stereotactically normalized with nonlinear warping such that all scans were in the same threedimensional space allowing for interindividual averaging. The standard space corresponds to the stereotactic atlas of Talairach and Tournoux (1988). The images were smoothed with a filter of 10 mm in the X and Y directions and 6 mm in the Z direction to reduce error variance, resulting in a final resolution of 11.7, 11.7, and 8.5 mm, estimated by summing in quadrature. An analysis of covariance was performed on the stereotactically normalized rCBF images to adjust for individual differences in global metabolic activity. Between-task comparisons were performed on a pixel-by-pixel basis. The / value for each pixel was calculated and transformed to a normal standard distribution. Maps of the z statistic were made, and the stereotaxic coordinates of the maximum significant change for each contiguous area were determined. Only maxima with between-tasks changes having a z value of 3.62 or greater (significant at or above thep < .0001 level, one-tailed) were tabulated for the initial trial of the DRA; given the 50% loss of sample size for the second, practiced trial, a more liberal cutoff was used (z = 2.58 or greater, corresponding to a p < .005, one-tailed). Only activation values (i.e. DRA > control) are reported in the tables. It should be noted that areas of deactivation were found (see Results section), but because the interpretation of deactivations remains ambiguous, as deactivations may sometimes result artifactually from the process of normalizing data (Berman,

Carson, Holt, Herscovitch, & Weinberger, 1993), we have chosen to focus our results on areas of activation.

Results Participants taking the DRA for the first time made an average of 69% correct responses (range; 34.9%-98.2%) during the first PET data acquisition; participants performed at 99% correct during the second trial. Global flow, determined as the average of every intracerebral pixel, did not differ between the initial DRA and the control task (DRA = 54.71 ml/100 g/min; control = 53.64 ml/100 g/min; paired t test - 0.79,p = .44) or between the initial and practiced DRA conditions (t = 0.34, p — .74). The SPM projection images for the initial trial in the total group are shown at the top of Figure 2. As shown in Figure 2, there was a widespread activation of multiple cortical areas during initial task performance, including large areas of the frontal cortex (areas 8, 9, 10, 11, 44, 45, 46, and 47), cingulate (area 32), parietal cortex (area 40), temporal cortex (area 20), and occipital cortex (area 19). Areas of relative deactivation were noted in the superior portion of the temporal lobe, posterior cingulate, and in the amygdala/ hippocampus (see Figure 2). A total of 23 activation maxima were identified with z values greater than 3.62, which are listed in Table 1. Figure 2 (bottom) shows both the initial and practiced trial results from the group of 9 participants with repeat data. There is clear similarity of overall activation pattern between the initial and practiced performance of this smaller cohort seen at the bottom of Figure 2 to that observed at the top of Figure 2 despite the dramatic difference in behavioral performance between initial and practiced scans. For the practiced DRA, a total of 27 activation maxima met or exceeded the z cutoff of 2.58, which are listed in Table 2. Thus, widespread activation spanning bilateral areas of the frontal, temporal, parietal, and occipital lobes was observed even when participants were performing the task basically without error, suggesting that this metabolic signal cannot be attributed solely to problem solving, rule learning, or cognitively nonspecific factors, such as mental effort, adaptation to task novelty, or confusion. To assess the magnitude of frontal activation in the initial and practiced scans, we calculated the total number of pixels that were more active in each condition with a p < .05 criterion. Of the 17,735 frontal lobe pixels, 1,608 pixels (or 9.07% of all frontal pixels) were more active in the initial trial than in the practiced condition. Alternately, 964 pixels (5.44% of all frontal pixels) were more active in the practiced condition than in the initial trial. Thus, the initial trial appears to have resulted in a broader total extent of activation. Of interest, the initial activation was broader in the right frontal lobe (a total of 1,089 activated pixels on the right and 519 left activated pixels), whereas the practiced condition resulted in more activation on the left (left frontal = 725 activated pixels and right frontal = 239 activated pixels). Discussion We observed significant metabolic activation of widespread areas of the cortex during both the initial and practiced

GOLD, BERMAN, RANDOLPH, GOLDBERG, AND WEINBERGER

DELAYED RESPONSE

Table 1 Statistically Significant Areas of Activation During the First Delayed Response Alternation (DRA) Task Coordinate Region Frontal lobe

Y

Z

Z score

Hemisphere

Location

Description

34 42 -42 -32 -26 34 22 20 -8 -54 38 -34 28 -32

8 24 20 52 52 54 58 50 22 14 30 24 16 18 46 42 26 -50 -46 -62 -58 -68 -82

44 28 28 8 -8 4 -4 -12 32 8 20 28 -4 0 -8 -4 -12 40 40 40 24 -24 -24

3.71 4.10 3.87 4.60 4.98 4.84 4.86 5.14 3.71 3.65 4.12 3.78 3.82 3.92 5.43 4.28 3.65 4.40 4.03 3.94 4.40 4.25 4.18

Right Right Left Left Left Right Right Right Left Left Right Left Right Left Right Left Right Left Right Right Left Right Right

6/8 9 9 10 10 10 10 11 32 44 45 46 47 47 47 47 20 40 40 19

Mid. frontal gyrus Inferior and mid. frontal gyrus Mid. frontal gyrus Mid. frontal gyrus Mid. frontal gyrus Mid. frontal gyrus Sup. frontal gyrus Sup. frontal gyrus Cingulate gyrus Precentral gyrus Inferior frontal gyrus Mid. frontal gyrus Insula Insula Mid. frontal gyrus Mid. frontal gyrus Mid. temporal gyrus Inferior parietal lobule Inferior parietal lobule Precuneus Hemisphere Hemisphere Posterior

40

Temporal lobe Parietal lobe

-44 60 -42 40

Occipital lobe Cerebellum

Brodmann area

X

30 -30 38 10

Note. Maxima of activations noted during initial trial of the DRA. Coordinates X, Y, and Z refer to atlas of Talairach and Tournoux (1988). Mid. = middle; sup. = superior.

performance of the DRA, including frontal and parietal areas discussed by Goldman-Rakic (1987) in her work on the anatomic network that subserves working memory in the nonhuman primate. Similar to the work of Jonides et al. (1993) and Petrides, Alivisatos, Meyer, and Evans (1993), these data support the basic idea that the frontal cortex participates in working memory tasks. Indeed the most surprising aspect of these data is the extent of cortical activation that we observed and the fact that this activation was fairly stable in the face of changing behavioral performance. That is, although it is certainly true that the initial performance of many participants was relatively uncontrolled as they were performing poorly on the task, the fact that the practiced state was relatively similar in terms of the pattern of metabolic activation is highly suggestive that this basic working memory network is involved in both active problem solving and more elementary working memory. The fact that the practiced state frontal lobe activation was somewhat smaller than the initial trial activation and was somewhat more left sided in focus may be consistent with the proposal of Goldberg and Costa (1981) that the dimension of cognitive novelty versus routinization may be a helpful

heuristic for interpreting laterality effects. In essence, the initial learning trial presents the participant with the challenge of mapping the spatial stimuli into a conceptual representational framework, a process that is heavily weighted toward novelty, and according to their model involves right hemisphere bias. In the process of developing a more efficient schema (a "descriptive system" or code) for meeting the demands of the task, the left hemisphere is hypothesized to become preferentially involved. In this instance, the code utilized may be inner speech, as suggested by the activations near Broca's area. The frontal cortical activation we observed in both trials included lateral, orbital, and medial regions. This result is not altogether surprising given the hybrid nature of the DRA that involves an alternating response set tied to spatial cues. The lateral and medial activations should be expected in light of the task demand to use spatial cues following a delay (Passingham, 1993). The activations extending to orbital regions may be consistent with both animal and human data suggesting that these areas may be particularly sensitive to the demand to alternate responses (Oscar-Berman, McNamara, & Freed-

Figure 2 (opposite). Top: Statistical parametric maps (SPMs) showing pixel-based analyses of 18 participants. The / values for each pixel have been converted to z scores linked to a color scale. Sagittal, coronal, and transverse views are shown at upper left. Reds and yellows indicate delayed response activations (DRA) greater than control with z values significant at a value of/? < .01 for lateral and medial quadrants of the left and right hemispheres projected onto an MRI scan. Below the SPMs are shown planes parallel to the intercommissural anterior-posterior line. Reds and yellows indicate DRA greater than control, and blues indicate control greater than DRA. Numbers indicate the number of millimeters above or below the anterior-posterior commissural line. Bottom: SPMs showing pixels in which activations are significant (DRA greater than control only) atp < .05 for naive (left) and practiced trials (right).

8

GOLD, HERMAN, RANDOLPH, GOLDBERG, AND WEINBERGER

Table 2 Statistically Significant Areas of Activation During the Second Delayed Response Alternation (DRA) Task Coordinate Region Frontal lobe

Temporal lobe Parietal lobe Occipital lobe

X

Y

-14 36 34 24 -8 -28 16 -6 -42 -44 -50 -44 -50 -38 -42 40 16 16 0 28

-32 -4 4 14 46 52 42 0 24 36 16 -48 10 -60 -46 -44 -72 -78 -68 -88 -58 -22 12 -88 -80 -58 -58

-6

Basal ganglia Cerebellum

-18 8 -22 -30 -16 -32

Z 48 28 44 48 24 -4 12 28 24 -4 0 8 -8 8 44 24 44 12 12 16 0 -4 4

-20 -24 -24 -24

Z score 2.58 2.62 2.72 2.73 3.18 3.37 2.72 2.60 3.79 2.60 3.07 3.50 2.94 2.65 3.31 2.68 2.97 2.59 2.69 3.34 3.62 2.91 2.83 3.41 3.10 2.85 2.75

Brodmann area Hemisphere

Location

Description

Left Right Right Right Left Left Right Left Left Left Left Left Left Left Left Right Right Right

5 6 6 8 9 10 10 24 46 47 47 21 38/12 37 40 40 7 17 17/31 18 18

Paracental lobule Precentral gyrus Mid. frontal gyrus Sup. frontal gyrus Sup. frontal gyrus, medial Sup. frontal gyrus Sup. frontal gyrus, medial Cingulate gyrus Mid. frontal gyrus Inf. and mid. frontal gyrus Inf. frontal gyrus Mid. temporal gyrus Sup. temporal gyrus Mid. temporal gyrus Inf. parietal lobule Inf. parietal lobule Precuneus Cuneus

Right Left Left Right Left Left Left Left

Fusiform gyrus Lingual gyrus Subtstantia nigra Caudate Posterior Posterior Hemisphere Hemisphere

Note. Maxima of activations noted during practiced trial of the DRA. Mid. = middle; sup. = superior; inf. = inferior.

man, 1991). The cognitive process involved in alternation has remained unspecified in the animal literature, and it remains unclear the extent to which the deficit observed in frontally lesioned animals is dependent on delay or reflects a more primary deficit in learning conditional responses (see Passingham, 1993, for review). Activations in the occipital and parietal cortex were also observed, presumably reflecting the visual and spatial demands of the task, demands that were identical in the control task. The enhanced activation of these areas, which was also present in the practiced state, may well represent the prolongation and attentional enhancement of processing in these areas as these signals are integrated with the conceptual alternating response set. The temporal activation that was observed (with a peak in area 20 on the first trial and in area 21 in the practiced state) is somewhat ambiguous. These locations appear to be outside of any language areas that could be explained as representing some type of phonological storage or subvocal rehearsal (Paulesu, Frith, & Frakowiak, 1993). Instead these temporal areas likely represent visual processing, presumably as part of the "what," or so-called object processing stream (Haxby et al., 1991). Given the fact that the stimuli involved in this study would not appear to demand extensive processing of this type as only rather simple features are involved and are identical between the control and task conditions, the nature of this bilateral activation remains to be explained. In examining our results in comparison to other functional imaging studies of working memory tasks, we are struck by the

fact that these tasks typically activate similar networks of frontal and parietal cortex with some variation in the specific activations likely reflecting differing stimulus characteristics (Jonides et al., 1993; Petrides, Alivisatos, Evans, & Meyer, 1993; Petrides, Alivisatos, Meyer, & Evans, 1993). Compelling direct evidence for this view was recently reported by Wilson, Scalaidhe, and Goldman-Rakic (1993) who found that working memory for spatial information contrasted with working memory for visual patterns, were mediated by different areas of the frontal cortex, areas that have different connectivity to posterior regions thought to mediate either coding of spatial location or object identity. Thus it appears likely that the frontal cortex may be a critical component of the functional anatomy of working memory, but that the exact connectivity involved in different types of working memory may be dependent on the type of stimulus material involved. It is important to point out that there are nearly as many different concepts of working memory as there are writers on the subject, and the implications of these models, both cognitive and anatomic, may be quite different. The primate models, such as those of Fuster (1991) and Goldman-Rakic (1987), emphasize the short-term memory aspects of the construct, perhaps analogous to briefly remembering a phone number that is forgotten as soon as the number is dialed. Cognitive psychological models such as those of Baddeley (1986) are concerned with a general purpose, multicomponent system that allows for the simultaneous storage and processing of information. Baddeley proposed that the frontal cortex serves as the central executive of this system, coordinating the

DELAYED RESPONSE

activity and manipulating the output of modality specific storage systems that likely involve more posterior brain areas. This is a broad construct, sometimes difficult to distinguish from attention, that implies a role for working memory in a wide variety of cognitive functions that require multiple operations. Thus, the most fully developed concepts of working memory in the human and primate literatures are quite different despite use of the same term. The data from this study may provide a slightly different perspective on the construct of working memory. In DA and DR, the information to be remembered is only relevant to select an action for a single trial. Indeed, remembering other trials is only likely to be a source of interference. Furthermore, the information to be remembered has no distinctive features or associative connections with already established knowledge. This may serve to distinguish working memory from the type of short-term memory demanded by a delayed nonmatch-tosample type task where the old sample and novel objects have distinctly different identities and where novelty is a powerful cue to guide response selection. It is this largely content-free, context, or trial specific memory in the service of action selection that appears to be critical to DA, DR, and DRA. It is difficult to specify precisely the concept of context or trial specific memory. However, it is intriguing that some have suggested that the frontal cortex plays an important role in the encoding of temporal order and sequential information as these features appear to have little necessary connection to the actual specific identity of the item to be remembered or the information to be manipulated (Schacter, 1987). That is, temporal order and sequential cues are inherently features of the context in which a stimulus or object is encountered rather than necessary features of the object itself. The critical role of the frontal cortex in the use of such context cues has recently been demonstrated with particular clarity by Petrides (1991) who found that monkeys able to perform delayed nonmatch to sample with novel distractors were unable to perform the same task when the choice at test was between two previously exposed objects with the potentially rewarded choice object having been exposed first. In the latter condition, the task requires a continual updating of item information in terms of the recency of presentation. Knowledge of the identity of the item alone cannot guide performance accurately; the identity must be placed in a temporal context. We suggest that it is the maintenance and manipulation of context-specific representations that require maximal frontal involvement and that underlies the persistent frontal metabolic signal of the DRA. Our conclusions need to be considered in light of the inherent spatial-temporal limitations of PET and the SPM approach to data analysis. Through the averaging of participants and transformation into a standard reference space, the resulting group activation image may not accurately reflect the specific functional anatomy of a single individual participant. The image represents perhaps the lowest common cognitiveanatomic denominator, and some have argued that such maps are potentially misleading (Steinmetz & Seitz, 1991). The development of techniques such as functional magnetic resonance imaging that may allow for the reliable imaging of single participants should provide fresh data on the issue of indi-

vidual functional-anatomic variability, an individual neuropsychology. It is hoped that this future work will build on the results of the accumulated PET literature that has used group data and clarify the relationship of group average and single subject activations. Working memory would appear to be a fruitful area for such study given the convergence between the nonhuman primate and human PET literatures that have consistently demonstrated a network of frontal and posterior activation sites, results consistent with those we observed with the DRA.

References Baddeley, A. D. (1986). Working memory. Oxford, England: Oxford University Press. Berman, K. F., Carson, R. E., Holt, D., Herscovitch, P., & Weinberger, D. R. (1993). A comparison of methods for quantitation of absolute and relative blood flow for cognitive activation studies with PET. Journal of Cerebral Blood Flow and Metabolism, 13 (Suppl. 1), 326. Berman, K. F., Ostrem, J. L., Randolph, C, Gold, J., Goldberg, T. E., Coppola, R., Carson, R. E., Herscovitch, P., & Weinberger, D. R. (1995). Physiological activation of a cortical network during performance of the Wisconsin Card Sorting Test: A positron emission tomography study. Neuropsychologia, 33, 1027-1046. Daube-Witherspoon, M. E., Chon, K. S., Green, S. L., Carson, R. E., & Herscovitch, P. (1992). Factors affecting dispersion correction for continuous blood withdrawal and counting systems. Journal of Nuclear Medicine, 33, 1010. Freedman, M., & Oscar-Berman, M. (1986). Bilateral frontal lobe disease and selective delayed-response deficits in humans. Behavioral Neuroscience, 100, 337-342. Friedman, H. R., & Goldman-Rakic, P. S. (1994). Coactivation of prefrontal cortex and inferior parietal cortex in working memory tasks revealed by 2DG functional mapping in the Rhesus monkey. Journal of Neuroscience, 14, 2775-2788. Friston, K. J., Frith, C. D., Liddle, P. F., & Frackowiak, R. S. J. (1991). Comparing functional (PET) images: The assessment of significant change. Journal of Cerebral Blood Flow and Metabolism, 11, 690-699. Friston, K. J., Frith, C. D., Liddle, P. F., Lammertsman, A. A., Dolan, R. D., & Frackowiak, R. S. J. (1990). The relationship between local and global changes in PET scans. Journal of Cerebral Blood Flow and Metabolism, 10, 458-466. Funahashi, S., Bruce, C. J., & Goldman-Rakic, P. S. (1989). Mnemonic coding of visual space in the monkey's dorsolateral prefrontal cortex. Journal of Neurophysiology, 61, 331-349. Fuster, J. M. (1973). Unit activity in prefrontal cortex during delayedresponse performance: Neuronal correlates of transient memory. Journal of Neurophysiology, 36, 67-78. Fuster, J. M. (1991). Role of prefrontal cortex in delay tasks: Evidence from reversible lesion and unit recording in the monkey. In H. S. Levin, H. M. Eisenberg, & A. L. Benton (Eds.), Frontal lobe function and dysfunction (pp. 59-71). New York: Oxford University Press. Goldberg, E., & Costa, L. D. (1981). Hemispheric differences in the acquisition and use of descriptive systems. Brain and Language 14, 144-173. Goldman-Rakic, P. S. (1987). Circuitry of the primate prefrontal cortex and regulation of behavior by representational memory. In Handbook of physiology: The nervous system (Vol. 5, pp. 373-417). Bethesda, MD: American Physiological Society. Goldman-Rakic, P. S., & Friedman, H. R. (1991). The circuitry of working memory revealed by anatomy and metabolic imaging. In H. S. Levin, H. M. Eisenberg, & A. L. Benton (Eds.), Frontal lobe function and dysfunction (pp. 72-91). New York: Oxford University Press.

10

GOLD, HERMAN, RANDOLPH, GOLDBERG, AND WEINBERGER

Haxby, J. V., Grady, C. L., Horwitz, B., Ungerleider, L. G., Mishkin, M., Carson, R. E., Herscovitch, P., Schapir, M. B., & Rapaport, S. I. (1991). Dissociation of object and spatial visual processing pathways in human extrastriate cortex. Proceedings of the National Academy of Sciences, USA, 88, 1621-1625. Jonides, J., Smith, E. E., Koeppe, R. A., Awh, E., Minoshima, S., & Mintun, M. A. (1993). Spatial working memory in humans as revealed by PET. Nature, 363, 623-625. Kimberg, D. Y., & Farah, M. J. (1993). A unified account of cognitive impairments following frontal lobe damage: The role of working memory in complex, organized behavior. Journal of Experimental Psychology: General, 122, 411-428. Koeppe, R. A., Holden, J. E., & Ip, W. R. (1985). Performance comparison of parameter estimation techniques for the quantitation of local cerebral blood flow by dynamic positron computed tomography. Journal of Cerebral Blood Flow and Metabolism, 5, 224-234. Oscar-Berman, M., McNamara, P., & Freedman, M. (1991). Delayedresponse tasks: Parallels between experimental ablation studies and findings in patients with frontal lesions. In H. S. Levin, H. M. Eisenberg, & A. L. Benton (Eds.), Frontal lobe function and dysfunction (pp. 230-255). New York: Oxford University Press. Passingham, R. E. (1993). The frontal lobes and voluntary action. Oxford, England: Oxford University Press. Paulesu, E., Frith, C. D., & Frackowiak, R. S. J. (1993). The neural correlates of the verbal component of working memory. Nature, 362, 342-345. Pennington, B. F. (1994). The working memory function of the prefrontal cortices: Implications for developmental and individual differences in cognition. In M. M. Haith, J. Benson, R. Roberts, & B. F. Pennington (Eds.), The development of future oriented processes (pp. 243-289). Chicago: Chicago University Press. Petrides, M. (1991). Monitoring of selections of visual stimuli and the

primate frontal cortex. Proceedings of the Royal Society of London B, 2436, 293-298. Petrides, M., Alivisatos, B., Evans, A. C., & Meyer, E. (1993). Dissociation of human mid-dorsolateral from posterior dorsolateral frontal cortex in memory processing. Proceedings of the National Academy of Sciences, USA, 90, 873-877. Petrides, M., Alvisatos, B., Meyer, E., & Evans, A. C. (1993). Functional activation of the human frontal cortex during the performance of verbal working memory tasks. Proceedings of the National Academy of Sciences, USA, 90, 878-882. Schacter, D. L. (1987). Memory, amnesia, and frontal lobe dysfunction. Psychobiology, 15, 21-36. Steinmetz, H., & Seitz, R. (1991). Functional anatomy of language processing: Neuroimaging and the problem of individual variability. Neuropsychologia, 29, 1149-1161. Stuss, D. T., & Benson, D. F. (1986). The frontal lobes. New York: Raven Press. Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain. New York: Thieme. Verin, M., Partiot, A., Pillon, B., Malapani, C., Agid, Y., & Dubois, B. (1993). Delayed response tasks and prefrontal lesions in man: Evidence for self-generated patterns of behavior with poor environmental modulation. Neuropsychologia, 31, 1379-1396. Weinberger, D. R. (1993). A connectionist approach to the prefrontal cortex. Journal of Neuropsychiatry and Clinical Neurosciences, 5, 241-253. Wilson, F. A., Scalaidhe, S. P., & Goldman-Rakic, P. S. (1993). Dissociation of object and spatial processing domains in primate prefrontal cortex. Science, 260, 1955-1958.

Received October 4,1994 Revision received June 1,1995 Accepted June 8,1995 •

New Editor Appointed The Publications and Communications Board of the American Psychological Association announces the appointment of Kevin R. Murphy, PhD, as editor of the Journal of Applied Psychology for a sixyear term beginning in 1997. As of March 1, 1996, submit manuscripts to Kevin R. Murphy, PhD, Department of Psychology, Colorado State University, Fort Collins, CO 80523-1876.