Prefrontal–posterior parietal networks in schizophrenia: primary dysfunctions and secondary compensations

Introduction 

In primates, the prefrontal (PFC) and posterior parietal (PPC) cortices are main components of a distributed neural network for working memory (WM), perception–action integration, and goal-oriented behavior Belger et al 1998, Braver et al 1997, Carlson et al 1998, Collette et al 1999, Courtney et al 1996, Fuster 1997, Fuster 1998, Goldman-Rakic 1988, Jonides et al 1998, Quintana and Fuster 1992, Quintana and Fuster 1993, Sarnthein et al 1998, Selemon and Goldman-Rakic 1988, Smith and Jonides 1999, Smith et al 1998. Working memory involves the protracted activation, in both those cortices, of neuronal ensembles supporting perceptual representations and goal-oriented action planning Fuster 1997, Fuster 1998, whose coordination supports correct task performance by closing the perception–action cycle. Neuronal activity related to visuo-motor WM occurs in both the PFC and the PPC Carlson et al 1998, Collette et al 1999, Courtney et al 1996, Courtney et al 1998, Funahashi et al 1993, Goldman-Rakic 1996, Jonides et al 1998, Owen 1997, Owen et al 1996a, Owen et al 1996b, Owen et al 1999, Petrides et al 1993, Quintana and Fuster 1992, Quintana and Fuster 1993, Quintana and Fuster 1999, Quintana et al 1988, Quintana et al 1989, Sarnthein et al 1998, Selemon and Goldman-Rakic 1988, Smith and Jonides 1999, Wilson et al 1993, Yajeya et al 1988, yet—particularly in the case of visuo-motor, nonsemantic WM—the distribution of its representational (i.e., mnemonic retention of a percept) and operational (i.e., anticipation and planning of a response contingent upon that percept) components is different between these cortices. In monkeys, the PFC houses mechanisms that support the mnemonic and response anticipatory components of visuo-motor WM and their integration across time, whereas the PPC appears more specialized in supporting anticipatory response planning by providing information processing mechanisms necessary for the spatial organization of behavioral responses Quintana and Fuster 1992, Quintana and Fuster 1993, Quintana and Fuster 1999. It has been postulated that, in the presence of PFC deficits, correct performance of visuo-motor delay tasks requiring WM may be efficiently supported by PPC mechanisms, as long as visual information need not be retained (i.e., correct responses can be fully anticipated) Quintana and Fuster 1992, Quintana and Fuster 1993, Quintana and Fuster 1999. The balance of WM activity between PFC and PPC seems to be in part dependent on mnemonic or anticipatory demands of the task being performed.

Schizophrenic patients exhibit deficits in WM and several cognitive processes supported by it, such as decision making, behavioral planning, and response execution Morice and Delahunty 1996, Pantelis et al 1997, Stone et al 1998, Sullivan et al 1994. It has been proposed that dysfunctions of the PFC may underlie some of these deficits Carter et al 1996, Fuster 1999, Goldman-Rakic 1994, Weinberger and Berman 1996. Early functional neuroimaging studies, whose results were later confirmed in several reports, identified abnormally diminished PFC activation (termed hypofrontality) during WM performance in schizophrenic patients when compared to normal control subjects Berman et al 1986, Callicott et al 1998, Carter et al 1998, Franzen and Ingvar 1975, Ingvar and Franzen 1974, Weinberger et al 1986; however, several other recent studies have failed to elicit such hypofrontality in schizophrenia, and some even have found abnormally increased PFC activation (hyperfrontality) during cognitive performance Callicott et al 2000, Manoach et al 1999. It has been proposed that the PFC is functionally inefficient in schizophrenia and that the degree or the direction of PFC activation abnormalities in the condition might depend on task demands, such as WM load, and on performance levels Callicott et al 2000, Curtis et al 1999, Manoach et al 1999. Abnormal PFC activation in schizophrenia has been correlated with local neurochemical abnormalities (Callicott et al 2000) and with symptoms of cognitive disorganization (Perlstein et al 2001).

We wanted to examine whether differences in mnemonic or anticipatory WM demands could be contributing to changes in PFC metabolic state in schizophrenic patients during WM task performance. Our hypothesis was that, within a given group of schizophrenic patients, the WM-related activity seen in their primarily dysfunctional PFC might be increased, decreased, or normal (i.e., comparable to control subjects) as a function of whether specific task demands could be met by alternate—and functionally preserved—available resources (i.e., areas) within a distributed network for WM. We thus used functional magnetic resonance imaging (fMRI) to study activity changes in PFC and PPC during WM performance in a group of schizophrenic patients and compared them to those seen in a group of control subjects. Functional MRI data were acquired while the subjects in both groups performed a simple visuo-motor delay task with different degrees of contingency between cues (colors, face diagrams) and the responses they prompted. The task either required the retention of cue information in WM (when correct response could not be predicted) or allowed the anticipation of responses (when they were fully predictable) for its correct performance.

Methods and materials 

We recruited eight healthy volunteers (six male, two female, mean age: 29.25 ± 5.13 years) and eight schizophrenic patients (six male, two female, mean age: 35.22 ± 10.69), all right handed according to self-report. Diagnoses were confirmed by two board-certified psychiatrists, according to DSM-IV criteria. The control subjects (average education level: 16.75 ± 2.96 years) had no personal or familiar history of neurologic or psychiatric disorders. The schizophrenic subjects were all chronic, stable outpatients (i.e., no hospital admissions of treatment changes for at least 1 year before their participation in the study) maintained on standard doses of modern antipsychotic medications (olanzapine 10–20 mg/day or risperidone 2–6 mg/day). None of the patients received concurrent anticholinergic, sedative, or any other medications. They were free of neurologic conditions or symptoms, as revealed by a review of their past and current medical records and by personal and family interviews. Their average length of illness was 8.5 years, and their mean level of education 12.88 ± 1.46 years. All subjects received a detailed explanation of the nature and possible consequences of the study and gave written informed consent to participate in it. The study was approved by the University of California, Los Angeles School of Medicine and the West Los Angeles Department of Veterans Affair (VA) Health Care Center Institutional Review Boards (IRB).

During the fMRI sessions, the healthy control subjects and schizophrenic patients performed a series of block-design paradigms counterbalanced for order across subjects of each group. Each paradigm consisted of three resting blocks of 24 sec (during which the subjects were presented with a black visual field) interleaved with two series of six trials, each series lasting approximately 70 sec (Figure 1). Each trial consisted of a cue (presented in the center of the screen on a black background for 0.5 sec), a 7-sec delay (during which the subjects saw a black screen), and a choice (between two response items presented side by side on the middle level of a black screen for up to 2.5 sec) sequence. The subject’s response, via button pressing (left or right) on a mouse device, required minimal movement of two fingers of the dominant (right) hand and terminated the display of the choice prompts. If a response was completed in less than 2.5 sec, a black screen was shown for the remaining choice time up to 2.5 sec. During any standard block of six trials, 3.5 sec (5% of total) corresponded to cue presentations, 42 sec (60% of total) to delay periods, and, on average, 7–10 sec (10%–15% of total) to choice execution and 14–17 sec (20%–25% of total) to intertrial periods. Thus, our design maximized the weight of delay activity in the task-related signal changes measured by fMRI. Overall, for the purpose of this study, each subject underwent four runs, each containing 12 trials.

Figure 1. (A) Example of time course of average changes in signal intensity over an entire scanning run in the posterior (inferior) parietal cortex (PPC)—Brodmann’s area 40 (Talairach coordinates: x = −50, y = −34, z = 44)—of control subjects during anticipatory task performance with facial diagram stimuli. Observe the prominent, protracted signal intensity increases during the activation periods of the run. The middle section of the figure shows a diagram of the temporal events of an experimental run, indicating the duration of resting (24 sec) and activation periods (69 sec), as well as the temporal structure and epochs of each of the six trials composing an activation period. A1, A2 indicate activation periods numbers 1 and 2 (see Methods for details). The bottom graph shows the time course of averaged changes in signal intensity during periods of activation (average of two activation periods of a run for a specific task and area). Observe that the relative increases in signal intensity coincide with the delay periods of the six trials (light gray bands). (B) Diagrams of the two tasks—remembering and anticipating—with both color and facial diagram stimuli (see Methods). A double arrow between the two choices in the case of the “remembering” task indicates the random presentation (i.e., right or left) of the correct response in that task. A small “c” under one of the choices for each diagram indicates the correct response according to task contingencies.

We used two sets of cues for the trials: 1) colored circles or 2) white line drawings of faces, either happy or sad. At the end of each trial, the subjects had to choose between 1) two colored circles or 2) two face diagrams presented side by side on a black background respectively (Figure 1). The cues indicated the location— right or left— of the correct choice either at chance levels (uncertain, 50% probability) or with certainty (100% probability), according to predetermined contingencies. In one of the task variations, a red or a green circle, or a smiling or a sad face diagram, preceded a choice (7 sec later) between a red and a green circle or between a smiling and a sad face diagram, respectively. The response choices were presented side by side in random position. The subjects were instructed to choose the response stimulus—right or left—that matched the cue in each trial. In this case, the predictability of the correct response side was 50% (chance level, correct choice randomly on right or left side). Hence, this task demanded the retention of cue information in WM for correct—i.e., above chance—performance.

In the other task variation, a blue or a yellow circle preceded a choice (7 sec later) between two side-by-side white circles; and a smiling or a sad face diagram whose nose pointed at the right or left of the subject’s visual field preceded a choice between a smiling and a sad face diagram (also presented side by side at the end of the 7 sec delay, the one on the left with its nose pointing to the left of the subject, the one on the right with its nose pointing to the right). The subjects had been instructed before the scanning session that after a blue color cue the correct response was always the left white circle, and after a yellow color cue, the right white circle. They had also been instructed that the direction where the nose of the face diagram cue was pointing (from the subject’s perspective) always determined the correct response between the two face diagram choices, left or right. Thus, the predictability of correct choice—right or left—was 100%, although the subjects could not execute a response until the two choices appeared following the cue and the delay (Figure 1). Hence, this second task allowed for the anticipation of the correct response choice on the basis of cue information. Each six-trial series randomly used two cues of the same category (color or face) and same level of predictability (certain or uncertain correct choice side) (e.g., blue and yellow, red and green, smiling and sad faces, faces with right- and left-pointing noses).

Response times and correctness were recorded by microcomputer and analyzed using SPSS software (SPSS Inc., Chicago, IL) and multiple and logistic regression models, respectively. Neurofunctional images were acquired using a General Electric 3T scanner (General Electric, Waukesha, WI) with an echo-planar imaging upgrade. First, we obtained co-planar structural images consisting of 26 slices (4 mm thick, 1 mm gap) that covered the entire brain volume (repetition time/echo time [TR/TE] = 4000/54 msec). Then we acquired gradient echo functional images covering 14 slices (4 mm thick, 1 mm gap), beginning near the middle of the temporal lobes and moving upward (TR/TE = 2000/45 msec). For each subject, images were first realigned to correct for head motion and then normalized into a standard stereotaxic space using automated image registration (AIR; Woods et al 1998a). We smoothed the imaging data using a Gaussian filter set at 6 mm full-width at half-maximum to minimize noise and residual differences in gyral anatomy. Finally, a group analysis of the AIR-processed images was completed using a general linear model with a delayed box-car reference function (SPM96 [Statistical Parametric Mapping, Wellcome Foundation, London, UK]; Friston et al 1994). For each group and task, we first applied a fixed-effects model comparing activation and resting periods and using explicit definitions of baseline periods and appropriate masking contrasts. We then pooled together data related to mnemonic or anticipatory WM performance respectively (i.e., independently of stimulus type, colors or face diagrams) within each group of subjects, and conducted a fixed-effects model analysis followed by a conjunction analysis contrasting activation versus resting periods (SPM99) to further characterize the activation patterns at the group level and test our hypothesis. Differences in activation patterns between patients and control subjects were statistically analyzed using a random effects model followed by a two-tailed t test contrasting the two groups (SPM99). We paid special attention to the level of subjects’ movement during the imaging procedures. Despite the described effects of motion on individual and group data Bullmore et al 1999, Friston et al 1996, Hajnal et al 1994, we found only a progressive and insignificant drift of less than 0.4 mm in our data samples, which was unrelated to task periods or behavioral events and well within the range of motion correction provided by AIR algorithms Woods et al 1998a, Woods et al 1998b.

Results 

No statistical differences in performance accuracy were found between groups (F = 3.38, p = .066). Importantly, performance differences between tasks were similar in both groups. Within each group, the number of errors (less than 6% and 10% for control subjects and patients, respectively) was equally divided between the two tasks. In both groups, response times were shorter by a similar nonsignificant magnitude (141.34 and 134.32 msec for control subjects and patients, respectively) in the anticipatory task than in the mnemonic task. Overall, there was a statistically nonsignificant tendency toward longer reaction times in the patients group, particularly among female subjects, and more response omissions (5% vs. 2.3% in control subjects). We identified well-defined frontal and parietal activations in both control subjects and schizophrenic patients (Figure 2, Table 1) that were related— as further confirmed by time-series analysis— to the performance of either one of the two modalities of the task (i.e., 100% vs. 50% of response predictability); however, several differences in distribution pattern of the activations between both subject groups were observed.

Figure 2. Sections of brain templates with overlaid group analyses results of significant prefrontal cortex activations (see text for data preprocessing and other details) by task and stimulus type in schizophrenic subjects and normal control subjects. Talairach coordinates, voxel statistical significance, and cluster extensions are described in Table 1. A red ring circles significant activations. Yellow lettering: R, right; L, left; BL, bilateral. Small graphs near the brain templates represent the time course of average signal changes during activation periods for the specific activated area. These changes occur within the general signal increase observed throughout the entire six-trial activation period (see Figure 1). Blue bars represent the periods of delay of each of the six trial of one activation period. Note the congruency between increased signal and delay period, which indicates the relationship of such signal and working memory activity.

Table 1. Frontal Cortex Activations in Schizophrenic Patients and Control Subjects During Working Memory Performance

	Test	Anatomical area	Brodmann area	Patients			Control subjects
				Talairach coordinates (x y z)	Voxel-level corrected p score {Z score in brackets}	Cluster extent (number of voxels	Talairach coordinates (x y z)	Voxel-level corrected p score {Z score in brackets}	Cluster extent (number of voxels)
	Remembering
	Logical	Left dorsolateral prefrontal cortex	9	−48 −4 40	.022 {4.88}	26	—	—	—
		Right dorsolateral prefrontal cortex	9	34 32 30	.020 {4.96}	34	—	—	—
		Left inferior frontal cortex	47	—	—	—	—	—	—
		Left inferior frontal cortex	45/46	−54 28 18	.026 {4.70}	12	−40 20 4	.000 {6.07}	183
		Left inferior frontal cortex	44	−58 4 22	.000 {6.24}	191	—	—	—
	Emotional	Left dorsolateral prefrontal cortex	9	−40 22 34	.000 {6.38}	773	—	—	—
		Right dorsolateral prefrontal cortex	9	40 12 28	.000 {6.74}	1158	44 6 38	.001 {5.70}	298
		Left inferior frontal cortex	47	—	—	—	−42 24 −4	.000 {5.81}	195
		Left inferior frontal cortex	45/46	−48 26 18	.000 {5.85}	773	—	—	—
		Right inferior frontal cortex	44	—	—	—	48 20 24	.002 {5.53}	298
	Anticipating
	Logical	Left inferior frontal cortex	44	—	—	—	−58 10 32	.000 {6.47}	521
		Right inferior frontal cortex	44	44 6 16	.014 {5.07}	168	54 6 14	.002 {5.45}	414
	Emotional	Left dorsolateral prefrontal cortex	9	−40 24 36	.001 {5.66}	53	−56 14 32	.000 {7.24}	750
		Right inferior frontal cortex	47	—	—	—	40 18 −6	.000 {6.47}	185
		Left inferior frontal cortex	47	—	—	—	−52 16 −2	.000 {6.58}	155
		Left inferior frontal cortex	44	−56 4 22	.002 {5.52}	121	−50 6 26	.000 {7.10}	750

Voxel size: 2 × 2 × 2 (mm).

Retention task 

Prefrontal cortex 

We found an overall increase in dorsolateral PFC (DPFC) activity in our patients group when compared to control subjects during mnemonic task performance. The schizophrenic subjects activated Brodmann’s area 9 bilaterally, extending to areas 45 and 46 on the left hemisphere, independently of stimulus type. Control subjects, on the other side, showed only right area 9 activation, of less magnitude than the patients’ activation, and restricted to performance under facial diagram stimuli (Table 1, Figure 2). Also in control subjects, we observed activation of left area 47 during performance based on facial diagrams as cues, activation that was absent in the patient group. Direct between-group comparisons confirmed a statistically significant increased activation of left PFC in patients when compared to control subjects (t = 3.26, p < .001) corresponding to inferior aspects of the overall area of activation.

Posterior parietal cortex 

We observed bilateral PPC activation in our patients during mnemonic WM performance with either type of stimuli. This activation was much less extensive, restricted to the left PPC and only during performance with facial diagram stimuli in the control group (Table 2, Figure 3), although direct intergroup statistical comparisons did not reach statistical significance.

Table 2. Posterior Parietal and Temporal Cortex Activations in Schizophrenic Patients and Control Subjects During Working Memory Performance

	Test	Anatomical area	Brodmann area	Patients			Control subjects
				Talairach coordinates (x y z)	Voxel-level corrected p score {Z score in brackets}	Cluster extent (number of voxels	Talairach coordinates (x y z)	Voxel-level corrected p score {Z score in brackets}	Cluster extent (number of voxels)
	Remembering
	Logical	Left superior parietal lobule	7	−34 −60 56	.000 {6.66}	225	—	—	—
		Right inferior parietal lobule	40	40 −26 52	.001 {5.50}	43	—	—	—
		Left inferior parietal lobule	40	−40 −38 40	.000 {5.53}	73	—	—	—
		Left superior temporal gyrus	22	−50 4 0	.000 {6.44}	138	−50 14 0	.000 {7.14}	183
		Right superior temporal gyrus	22	—	—	—	48 6 2	.003 {5.36}	203
	Emotional	Left superior parietal lobule	7	−2 −58 −32	.000 {5.58}	155	−28 −64 50	.000 {6.19}	137
		Right superior parietal lobule	7	20 −66 48	.000 {6.73}	1624	—	—	—
		Left inferior parietal lobule	40	−46 −32 48	.000 {7.15}	1722	−46 −30 48	.000 {6.51)	170
	Anticipating
	Logical	Left superior parietal lobule	7	−32 −44 60	.000 {6.48}	50	—	—	—
		Right superior parietal lobule	7	28 −60 46	.000 {6.86}	805	18 −60 58	.000 {6.13}	51
		Left inferior parietal lobule	40	−38 −32 52	.000 {7.45}	1958	−46 −32 46	.000 {6.90}	682
		Right inferior parietal lobule	40	—	—	—	40 −30 52	.000 {7.73}	441
		Left superior temporal gyrus	22	—	—	—	−50 12 0	.000 {6.30}	521
		Right superior temporal gyrus	22	—	—	—	46 10 −4	.000 {7.59}	414
	Emotional	Left superior parietal lobule	77	−44 −52 52−32 −72 50	.000 {7.40} .000 {5.79}	1185 80	−32 −62 48−32 −62 48	.000 {7.84}.000 {7.84}	771771
		Right superior parietal lobule	7	38 −52 48	.000 {7.26}	392	—	—	—
		Left inferior parietal lobule	4040	−32 −52 40−32 −52 40	.000 {7.55}.000 {7.55}	11851185	−50 −34 44−42 −56 46	.000 {7.55} .000 {7.31}	771 2229
		Right inferior parietal lobule	40	34 −56 42	.000 {6.88}	392	30 −52 42	.000 {7.34}	2229

Voxel size: 2 × 2 × 2 (mm).

Figure 3. Sections of brain templates with overlaid group analyses results of significant posterior parietal cortex activations (see text for data preprocessing and other details) by task and stimulus type in schizophrenic subjects and normal control subjects. Talairach coordinates, voxel statistical significance, and cluster extensions are described in Table 1. The figure labels and graphs follow the same conventions as in Figure 2.

Other cortical association areas 

Significant activation of the anterior portion of the superior temporal gyrus (STG, Brodmann’s area 22) was found only during task performance based on color cues. The control group showed bilateral area 22 activation, whereas the patients exhibited area 22 activation only on the left side (Table 2).

Discussion 

Our results provide evidence of different patterns of activation of a PFC–PPC network between schizophrenic patients and normal control subjects during performance of simple visuo-motor WM tasks that emphasize remembering percepts or anticipating responses. The differences observed support the proposition that schizophrenic patients may show hyper- or hypo-activity of DPFC during WM performance as a result of variations in certain task demands. Our results further indicate that hypofrontality occurs in schizophrenia when these task demands allow for other cortical areas within a distributed network for visuo-motor WM performance to “take over” the necessary processing effort, in response to DPFC functional deficits. On the other hand, if those task demands require the functional integrity of the DPFC, deficits in this area result in hyperfrontality, likely compensatory.

A role of the PFC–PPC network in distributed or parallel processing during WM performance is supported by a long series of studies in monkeys and humans Belger et al 1998, Braver et al 1997, Carlson et al 1998, Collette et al 1999, Courtney et al 1996, Ferreira et al 1998, Fuster 1997, Fuster 1998, Goldman-Rakic 1988, Jonides et al 1998, Quintana and Fuster 1992, Quintana and Fuster 1993, Quintana and Fuster 1999, Quintana et al 1988, Quintana et al 1989, Sarnthein et al 1998, Selemon and Goldman-Rakic 1988, Smith and Jonides 1998, Smith and Jonides 1999. Single-unit recording studies in monkeys performing visuo-motor WM tasks indicate that WM activity related to response anticipation is present in both PFC and PPC, whereas activity related to mnemonic retention of percepts in WM can only be detected in DPFC Quintana and Fuster 1992, Quintana and Fuster 1999. Additional monkey experiments indicate that, under conditions of low load or difficulty, WM tasks that allow anticipation of responses can be efficiently performed in the presence of transient lesions (by cooling) of the DPFC (Quintana and Fuster 1993). The same is not true, however, when the tasks demand retention of information in WM. Hence, it seems to follow that lesions of the DPFC in primates render the PFC–PPC circuit dysfunctional, which results in visuo-motor WM processing that must rely solely on PPC function. If this increased role of the PPC—manifested by its hyperactivity—can successfully handle WM task demands, the PFC is relieved from its contribution and thus will exhibit hypoactivity. If, on the other hand, the PPC compensatory attempt is unsuccessful, the result will then be PFC and PPC hyperactivity. Multiple studies have confirmed DPFC dysfunctions in schizophrenia Berman et al 1986, Callicott et al 1998, Callicott et al 2000, Carter et al 1996, Carter et al 1998, Curtis et al 1999, Franzen and Ingvar 1975, Goldman-Rakic 1994, Ingvar and Franzen 1974, Manoach et al 1999, Morice and Delahunty 1996, Pantelis et al 1997, Stone et al 1998, Sullivan et al 1994, Weinberger and Berman 1996, Weinberger et al 1986; if our model is correct, these dysfunctions should result in hypofrontality during response anticipation and hyperfrontality during information retention. Our data fully confirm these outcomes and thus provide solid support for our hypothesis.

Both decreases and increases in DPFC function during WM performance have been reported before in schizophrenia Callicott et al 1998, Callicott et al 2000, Carter et al 1996, Carter et al 1998, Curtis et al 1999, Franzen and Ingvar 1975, Goldman-Rakic 1994, Ingvar 1979, Ingvar and Franzen 1974, Weinberger and Berman 1996, and even alternate recruitment of PPC has been observed in the presence of frontal deficits (Crespo-Facorro et al 2001). It is difficult to ascertain, given the different nature of these studies and the experimental tasks they employed, whether our model could explain their results. In particular, our intentional use of simple visuo-motor WM tasks (designed to reproduce previous monkey studies and to maintain low task difficulty, thus minimizing performance differences and education background effects) likely precludes the application of our hypothesis to the more complex semantic WM network Jonides et al 1998, Smith et al 1998 and its abnormalities in schizophrenia. Yet in cases where mnemonic or anticipatory demands could be clearly identified, the results of those studies do not contradict or disagree with our proposition, and some are largely in agreement with it Callicott et al 2000, Manoach et al 1999. At the same time, our results in control subjects do not disagree with other functional divisions of WM representation reported elsewhere Collette et al 1999, Courtney et al 1996, Courtney et al 1998, Goldman-Rakic 1996, Owen et al 1996a, Owen et al 1999, Petrides et al 1993, Sarnthein et al 1998, Smith and Jonides 1999, Wilson et al 1993.

Structural Barta et al 1997, Gur et al 2000, Hirayasu et al 2000, Marsh et al 1997, McCarley et al 1999, Menon et al 1995, Highley et al 1999, Rajarethinam et al 2000, Shenton et al 1992, Shenton et al 2001, Sigmundsson et al 2001, neurochemical Le Corre et al 2000, Radewicz et al 2000, Shirakawa et al 2001, Sokolov et al 2000, Tune et al 1996, and functional Crespo-Facorro et al 2001, Kiehl and Liddle 2001, Kircher et al 2001, Menon et al 2001b, O’Donnell et al 1999, Wible et al 2001 abnormalities of the STG have been consistently described in schizophrenia Pearlson 1997, Pearlson et al 1997. Most of the these dysfunctions correspond to the most posterior part of the STG and have been linked to auditory or semantic processing deficits observed in schizophrenic patients Barta et al 1990, McCarley et al 1993, Nestor et al 1993, O’Donnell et al 1993, O’Donnell et al 1999. We found activation deficits in the most anterior part of STG (Brodmann’s area 22) during processing of nonemotional visual cues in our schizophrenic subjects when compared to control subjects. In monkeys, area 22 receives its main afferent cortico-cortical projections from the DPFC (Shiwa 1987), and it has been shown to be involved in complex visual processing or in associative processing of visual cues as needed for goal-oriented behavior Bancaud et al 1994, Brunet et al 2000, Faillenot et al 1997, Hermann et al 1999, Pardo et al 1990, Partiot et al 1996, Paulesu et al 1995. Deficits in different steps of visual processing have also been described in schizophrenia, and some of them have even been proposed as markers of the condition, because deficits at early stages of that processing appear to also be present in healthy first-degree relatives of schizophrenic patients (Nuechterlein and Dawson 1984). A functional disconnection between PFC and temporal lobe structures has been suggested as a pathophysiological mechanism in schizophrenia Fletcher et al 1996, Sigmundsson et al 2001.

Which of the changes identified can be seen as primary abnormalities and which ones as secondary compensations in schizophrenia? The PFC–PPC model used here involves only a simplified part of the circuits for WM in the brain. It entertains the idea that, within such restricted view of brain circuitry, schizophrenia involves primary DPFC and IFC functional abnormalities, but not PPC ones. Our results provide evidence for these propositions as well as for deficits, also likely primary, in association areas of the temporal cortex linked to processing of visual cues devoid of emotional content. There must be, of course, additional brain dysfunctions that form part of the primary abnormalities in schizophrenia. The anterior cingulate cortex (ACGC), closely interconnected with the DPFC (Paus 2001) and complementing its role during goal-oriented behavior, exhibits structural Albanese et al 1995, Benes 1998, Benes et al 1997, Benes 1998, Deicken et al 1997, Dolan et al 1995, Goldstein et al 1999, Sigmundsson et al 2001, Wyatt et al 1995 and functional Carter et al 1997, Haznedar et al 1997, Liddle et al 1992, Tamminga et al 1992, Tamminga et al 2000 abnormalities in schizophrenia, and might also be considered as primarily dysfunctional in the condition. Recent reports have characterized motor system functional abnormalities at the cortical Mattay et al 1997, Schroeder et al 1999 and subcortical (Menon et al 2001a) level in schizophrenia, although other studies have shown no differences between control subjects and patients (Braus et al 2000) or interpreted differences as compensatory (Quintana et al 2001a). The changes we observed in PPC function, however, better fit the concept of secondary (perhaps compensatory) abnormalities. No reports of structural or neurochemical abnormalities of the PPC have been published so far in schizophrenia. The argument can be made that a uniformly present increase in PPC activity during WM cannot by itself explain any of the numerous cognitive deficits seen in the condition, yet it should be expected if our PFC–PPC activity balance model is correct. If schizophrenic subjects perform WM tasks at similar levels to control subjects, any pervasive hyperactivity observed in those subjects on brain areas whose integrity is not critical for that performance could be considered compensatory (Quintana et al 2001a). Nonetheless, further specific studies addressing this point must be conducted to reach definitive conclusions.

We did not study structures within the limbic system and the cerebellum that have been considered as components of pathophysiological models of cognitive dysfunctions in schizophrenia by others (Andreasen et al 1998). We focused on the PFC–PPC network, because previous studies in monkeys had provided us with a good, parallel experimental model of that network, and also because of limitations in the contiguous brain volume we were able to cover with our fMRI protocol and scanning facilities. Notwithstanding, new studies are underway that address whether those other areas may influence the balance of activity between cortical areas involved in WM processing. For example, our current results point out to a pervasive activation deficit—i.e., hypoactivity throughout all types of situations tested—of area 47 in the IFC of schizophrenic patients during WM processing of emotional information from face diagrams. This area is heavily interconnected with limbic structures, yet not directly linked to PPC. Hence, in the presence of functional abnormalities of area 47, limbic structures rather than PPC may be the ones that show parallel deficits or compensatory activity during processing of emotional information in WM. It is known that schizophrenic subjects exhibit difficulties processing facial affect information. Our results provide evidence of neural dysfunctions related to those deficits and complement the results from recent preliminary reports (Quintana et al 2001b).

Deficits of DPFC function in schizophrenia may also result in activation abnormalities of other cortical and subcortical structures—including the limbic system—as a result of impaired “top-down” modulation. It has been reported that the PFC may modulate limbic structures in normal subjects during semantic processing of emotional information (Hariri et al 2000). In addition, a fundamental role of the PFC is motor inhibition, which might be at the basis of certain motor and premotor functional abnormalities in schizophrenia Mattay et al 1997, Menon et al 2001a, Schroeder et al 1999. Thus, it seems reasonable to assume that the picture of functional abnormalities in the brain of schizophrenic patients performing WM tasks may be far more complex than what is presented in this study. Nonetheless, because of the central roles of PFC and PPC in WM and the adherence of our results to the model predicted by monkey studies, we feel confident that future results involving additional cortical and subcortical areas will also be congruent with that model. In any case, the current results serve to further characterize fundamental unbalances in processing resources for WM in schizophrenia. (Benes 1999)