Cortical Abnormalities in Schizophrenia Identified by Structural Magnetic Resonance Imaging

Background  Relatively few magnetic resonance imaging studies of schizophrenia have investigated the entire cerebral cortex. Most focus on only a few areas within a lobe or an entire lobe. To assess expected regional alterations in cortical volumes, we used a new method to segment the entire neocortex into 48 topographically defined brain regions. We hypothesized, based on previous empirical and theoretical work, that dorsolateral prefrontal and paralimbic cortices would be significantly volumetrically reduced in patients with schizophrenia compared with normal controls.

Methods  Twenty-nine patients with DSM-III-R schizophrenia were systematically sampled from 3 public outpatient service networks in the Boston, Mass, area. Healthy subjects, recruited from catchment areas from which the patients were drawn, were screened for psychopathologic disorders and proportionately matched to patients by age, sex, ethnicity, parental socioeconomic status, reading ability, and handedness. Analyses of covariance of the volumes of brain regions, adjusted for age- and sex-corrected head size, were used to compare patients and controls.

Results  The greatest volumetric reductions and largest effect sizes were in the middle frontal gyrus and paralimbic brain regions, such as the frontomedial and fronto-orbital cortices, anterior cingulate and paracingulate gyri, and the insula. In addition, the supramarginal gyrus, which is densely connected to prefrontal and cingulate cortices, was also significantly reduced in patients. Patients also had subtle volumetric increases in other cortical areas with strong reciprocal connections to the paralimbic areas that were volumetrically reduced.

Conclusion  Findings using our methods have implications for understanding brain abnormalities in schizophrenia and suggest the importance of the paralimbic areas and their connections with prefrontal brain regions.

SEVERAL MODELS have been proposed to explain the widespread brain abnormalities in patients with schizophrenia. Early anatomical models were based largely on hypothesized focal abnormalities in particular brain regions,¹ derived mainly from adult lesion models of neuropsychiatric disorders. Schizophrenia has more recently been understood as, in part, a neurodevelopmental disorder^2,3 in which altered connectivity or multifocal abnormalities are more likely than focal disorders.^4-6 The most frequently replicated findings have been in subcortical structures, such as the hippocampal region^7-10 and thalamus.^10,11 An increasing number of studies, however, have demonstrated abnormalities in cortical brain regions,^12-21 indicative of developmental origins.^{12,13,20,22-24} Neuropathologic and neural network findings have suggested that schizophrenia may involve a defect in neuronal migration,^22-24 myelination,⁵ and/or corticocortical pruning.^25-27

Despite the current emphasis on the importance of cortical abnormalities in schizophrenia, structural imaging studies of the entire cortex are relatively few. In fact, of 67 studies recently reviewed,⁸ only 16 examined more than one cortical brain region in more than one lobe. Studies that have examined the entire cortex have sampled primarily men and acquired images with relatively large slices—5-mm slices with 2.5-mm gaps.^17,20,21 The relatively small number of cortical structural imaging studies may be due, in part, to the difficult and time-consuming nature of segmenting these relatively small areas of the brain. Furthermore, until recently, methods to assess in vivo the subtle volumetric reductions in small cortical regions, which require fine distinctions between brain regions, were not available for the analysis of brain images. Thus far, most of the subtle abnormalities in the cortex have been identified at the cellular level using postmortem techniques.

Neuropathologic studies of schizophrenia have identified abnormalities of cell size, orientation, and receptor density in the anterior cingulate gyrus²² and prefrontal areas.^12,13,28 Structural magnetic resonance imaging (MRI) studies have shown volumetric abnormalities in schizophrenia in the frontal lobe and prefrontal subregions (eg, orbital¹⁴ and dorsolateral^15,29); the prefrontal areas in general²¹; and the temporal lobe, including the parahippocampal gyrus^7,16 and auditory cortex (eg, the superior temporal gyrus,^7,18,29-31 planum temporale,^19,32 and related sylvian fissure region³³). A few studies—eg, Schlaepfer et al²⁹—have implicated the parietal cortex and occipital lobe (reviewed in Shenton et al⁸), but findings in these areas have been equivocal. Cortical volume reductions have been estimated^34,35 to range from only 4% to 6%. In general, previous cortical studies suggested that prefrontal, paralimbic, and left frontotemporal lobe areas are subtly but significantly reduced in patients with schizophrenia.

We introduce the application of a new brain segmentation technique to study patients with schizophrenia, with the goal of the better identification of subtly altered cortical tissue. The method, based on conceptual models of the cortex, was developed to divide the entire neocortex into 48 topographically defined brain areas^36,37 and has been applied successfully to healthy subjects.^37-39 The unique advantage of this technique is that it allows an estimation of the relative differences in volume in specific areas of the entire cortex between patients with schizophrenia and normal controls.

Neuropsychological studies,^40-44 including our own, some^45-47 using the same patients as in this study, have demonstrated impairments in working memory and other executive functions, verbal and visual short-term memory, attention, olfaction, and motivation. These functions rely heavily on circuitry that primarily includes the frontal lobes (ie, working memory, executive functions, and olfaction) and frontolimbic (verbal and visual short-term memory) or paralimbic (attention and motivation) regions. Based on previous reports, we hypothesize that the primary cortical abnormalities in schizophrenia will be in the prefrontal, especially dorsolateral prefrontal cortex, and paralimbic⁴⁸ regions (eg, cingulate and parahippocampal gyri and the frontal orbital cortex)—ie, areas involved in communication between the prefrontal and limbic brain regions. Although these hypothesized areas are not unique, this study is the first to examine all cortical areas in one study in a substantial number of subjects with schizophrenia.

Patients were systematically sampled from the universe of outpatients at 3 public psychiatric hospitals in the Boston, Mass, area that serve primarily patients with psychotic disorders.^45-47 Inclusion criteria consisted of ages between 25 and 66 years, at least an eighth-grade education, English as the first language, and an estimated IQ of at least 70. Exclusion criteria for subjects were substance abuse within the past 6 months, history of a head injury with documented cognitive sequelae or loss of consciousness longer than 5 minutes, neurologic disease or damage, mental retardation, medical illnesses that substantially impair neurocognitive function, and a history of electroconvulsive treatment. Written informed consent was obtained after a complete description of the study was given to the subjects.

Healthy control subjects were recruited through advertisements in the catchment areas and notices posted on bulletin boards at the hospitals from which the patients were recruited. They were proportionately matched to patients by age, sex, ethnicity, parental socioeconomic status,⁴⁹ reading ability, and handedness. Control subjects were screened for current psychopathological disorders using a short form of the Minnesota Multiphasic Personality Inventory⁵⁰ and a family history of psychoses or psychiatric hospital admissions. We excluded potential controls if they had a current psychopathological disorder; a lifetime history of any psychosis; a family history of psychosis or psychiatric hospitalization; or a score on any clinical or validity scale on the Minnesota Multiphasic Personality Inventory, except the Masculinity-Femininity scale, above 70.

Patients were included if they had a DSM-III-R⁵¹ clinical diagnosis of schizophrenia. (Patients were rediagnosed by research criteria, as described in the subsection "Diagnostic Procedures.") The sample consisted of 29 patients, 17 (59%) of them male. Table 1 presents a summary of the sociodemographic and clinical characteristics of the patients and controls.

The patients were a middle-aged sample, primarily non-Hispanic white (25 [86%]), with an average education of partial college, who came from a middle to lower-middle socioeconomic status. Measures of premorbid and current IQ were in the average range. They had a mean±SD age at illness onset of 23.6±5.8 years (range, 16-45 years), with 4.2±3.1 hospital admissions, reflecting 22.0±9.9 months of hospitalization and 20.9±10.2 years of illness. The daily chlorpromazine-equivalent dose was 689.9±591.6 mg of typical neuroleptic medications. In general, the patients were clinically stable, being treated long term as outpatients, although they were rated as having mild to moderate negative and positive symptoms.^52,53

The healthy controls were a proportionately matched comparison group. There were no significant differences in sex distribution, age, ethnicity, parental socioeconomic status, education, Wide Range Achievement Reading⁵⁴ score, Wechsler Adult Intelligence Scale–Revised^55,56 vocabulary score, and handedness. There was a significant (P=.01) difference in IQ, which is typical for schizophrenia.

Research DSM-III-R diagnoses were based on the Schedule for Affective Disorders and Schizophrenia⁵⁷ and a systematic review of the medical record. Patients primarily had undifferentiated or paranoid subtypes (Table 1). Interviews were obtained by master's level interviewers with extensive diagnostic interviewing experience. Senior investigators (Drs Goldstein and Seidman) reviewed the transcripts from the interview and the medical records to determine the consensus, best-estimate, lifetime diagnosis. Blindness of assessments among psychiatric and MRI data was maintained.

Magnetic resonance imaging scans were acquired at the Nuclear Magnetic Resonance Center of the Massachusetts General Hospital, Boston, with a 1.5-T MRI scanner (General Electric Signa scanner; General Electric Corporation, Waukesha, Wis). Contiguous 3.1-mm coronal spoiled-gradient echo images of the entire brain were obtained using the following parameters: repetition time, 40 milliseconds; echo time, 8 milliseconds; flip angle, 50°; field of view, 30 cm; matrix, 256×256; and averages, 1. The MRI scans were processed and analyzed at the Massachusetts General Hospital Center for Morphometric Analysis for further processing and analysis.

Data were analyzed using image analysis workstations (Sun Microsystems Inc, Mountain View, Calif). Images were positionally normalized by imposing a standard 3-dimensional coordinate system on each 3-dimensional MRI scan, using the midpoints of the decussations of the anterior and posterior commissures and the midsagittal plane at the level of the posterior commissure as points of reference for rotation and (nondeformation) transformation.^38,58 Scans were then resliced into normalized 3.1-mm coronal, 1.0-mm axial, and 1.0-mm sagittal scans and were analyzed. Positional normalization overcomes potential problems caused by variation in head position of subjects during scanning.

Each slice of the T₁-weighted, positionally normalized, 3-dimensional coronal scans was segmented into gray and white matter and ventricular structures using a semiautomated intensity contour–mapping algorithm³⁸ and signal-intensity histogram distributions. This technique, described in detail elsewhere^38,58,59 and illustrated in Figure 1, yields separate compartments of neocortex, subcortical gray nuclei, white matter, and ventricular system subdivisions that generally correspond to the natural tissue boundaries distinguished by signal intensities in the T₁-weighted images.

The focus of the present study was the subdivision of the neocortex into parcellation units (PUs).

The neocortex, defined by the gray-white matter segmentation procedure, was subdivided or "parcellated" into 48 bilateral PUs based on the system originally described by Rademacher,³⁶ modified in Caviness et al,³⁷ and shown in Figure 1. This is a comprehensive system of neocortical subdivision designed to approximate architectonic and functional subdivisions and based on specific anatomical landmarks present in all brains.³⁷

Two types of landmarks specify the boundaries of the PUs: major fissures of the hemisphere (Figure 1) and anatomically specified single nodal points along the longitudinal axis of the brain. The fissures and nodal points are easily identifiable and normally present in all brains. Nodal points are specified by diverse anatomical structures, most of which lie in the cortex itself (eg, the intersection of 2 sulci or a sulcus within the hemispheric margin). Four nodal points are specified by subcortical landmarks—the splenium and genu of the corpus callosum, the decussation of the anterior commissure, and the lateral geniculate bodies. The PUs are mainly bounded by the major fissures of the brain (Figure 1). Where the anterior or posterior border of a PU is not completely specified by major fissures, this boundary is closed by a coronal plane through a nodal point. Following parcellation, volumes were calculated for each PU by multiplying the area measurement of the PU on each slice by the slice thickness and then summing all slices on which the PU appeared.

Our collaborators at the Massachusetts General Hospital Center for Morphometric Analysis (Drs Kennedy, Makris, and Caviness), who developed these procedures, trained and maintained quality control of the segmentation of the data. In previous studies³⁷ using these procedures, reliability was good. For our study, the brains of 10 subjects were completely parcellated into 48 PUs in the right and left hemispheres by 2 well-trained image analysts who had a background in neuroanatomy. Table 2 presents the intraclass correlation coefficients (ICCs) for the 48 PUs, which were generally good.

Forty percent of the PUs had excellent reliabilities (ICC≥0.80), and about 69% were very good (ICC≥0.70). There were several PUs with ICCs of 0.55 or less. The brains of 10 additional subjects were analyzed, after a discussion of areas in which raters disagreed, and ICCs of these areas were presented in the third column of Table 2. The ICCs remained low for only 3 areas: frontal operculum, basal forebrain, and occipital pole. Finally, intrarater reliability was conducted using images from 6 subjects (Table 2, last column) and was generally excellent. Only the fusiform gyrus, lingual gyrus, and superior parietal lobule had fair reliabilities (ICC=0.50, ICC=0.57, and ICC=0.62, respectively).

Three approaches were used to test for volumetric differences between patients and controls, given disagreement among the investigators in the field as to the ideal statistical model. Adjusted PU volumes were expressed as a percentage of the total cerebral volume—PU volume divided by total cerebral volume—to control for individual variations in brain size. The total cerebral volume was also used as a covariate in an analysis of covariance (ANCOVA). Finally, the z-score method⁶⁰ was used to adjust for a normal variation in age- and sex-corrected head size, as measured by the total cerebral volume. Thus, z scores reflected volumetric estimates for the patients relative to volumes expected from normal subjects of a particular head size, age, and sex.^17,21,34,60

Adjusted brain volumes and the z scores of the PUs were analyzed using ANCOVA to assess the effect of the group (ie, subjects with schizophrenia vs controls), controlled for age, sex, and sex-by-group interaction. An ANCOVA was appropriate because tests of normality showed that the PU volumes were, in general, normally distributed. A multivariate ANCOVA using all PUs would not be statistically powerful with our sample size and, thus, would not provide accurate covariance structure estimates. However, we had some specific hypotheses about certain PUs. Thus, as suggested by Rothman,⁶¹ a multivariate ANCOVA was used that included PUs hypothesized to be different between groups (the middle frontal gyrus, frontomedial cortex, fronto-orbital cortex, divisions of the cingulate and parahippocampal gyri, and insula), to control for a type I error. We were also interested in whether other areas, which were reflected in exploratory analyses, distinguished patients from healthy controls. In addition, relative volume differences and effect sizes⁶² between patients and controls were estimated (Table 3). Effect sizes are unaffected by the sample size and, thus, can be compared across studies.

The unadjusted (mean±SD) total cerebral volume was significantly different between groups (patients, 1140.9±116.9 cm³; and controls, 1077.3±97.4 cm³) (t₅₃=2.2, unpaired, 2-tailed test of significance; P=.03) but was not significant when adjusted for the total brain volume (patients, 87.1±0.9 cm³; and controls, 88.7±0.9 cm³). The total neocortex was 591.4±60.2 cm³ for patients and 575.2±65.8 cm³ for controls. Adjusted for the total cerebral volume, the cortex was 51.9%±2.3% for patients and 53.4%±2.7% for controls (t₅₃=−2.2; P=.03).

Table 3 presents the unadjusted mean volumes in cubic centimeters for the PUs and total volumes by lobe. Tests of differences between the groups were based on the adjusted values and z scores, and these did not differ across statistical methods, with significant differences illustrated in Figure 2. First, the overall F test (Wilks λ) from the multivariate ANCOVA for our hypothesized regions was significant (F_9,45=3.17; P≤.005), suggesting that patients and controls significantly differed in volumes of these areas. Table 3 shows that, within the frontal lobe, the middle frontal gyrus (F_1,52=4.87; P=.03) and the frontomedial cortex (F_1,52=8.41; P=.005) were significantly reduced in patients with schizophrenia. The relative volumetric reductions for patients compared with controls were 8.80% and 14.89%, respectively. When we examined the right and left hemispheres separately for hypothesized prefrontal areas, only the right fronto-orbital cortex was significantly reduced in patients and not the left (F_1,52=3.94; P=.05) (relative volumetric reduction of the right fronto-orbital cortex among the patients was 8.5% compared with controls; effect size, 0.56). The middle frontal gyrus and frontomedial cortex were bilaterally reduced, although the middle frontal gyrus showed greater reduction on the left.

Patients also exhibited significant volumetric reductions in PUs approximating the medial paralimbic cortex, in particular, the anterior cingulate gyrus (F_1,52=3.79; P=.05) and paracingulate gyrus (F_1,52=9.46; P=.003), with reductions of 11.14% and 7.99%, respectively, compared with controls. In addition, there was a large and significant volumetric reduction of the insula in patients (F_1,52=11.90; P=.001) by almost 1 SD below that of controls (effect size, 0.88). Exploratory analyses of other cortical temporal areas showed no significant volumetric reductions in patients.

The parietal and occipital cortices also showed few significant differences in individual PUs. The posterior supramarginal gyrus, however, was significantly and bilaterally reduced in patients compared with controls (F_1,52=3.95; P=.05). When combined to encompass the inferior parietal cortex as a whole—ie, anterior and posterior supramarginal gyri, angular gyrus, and parietal operculum—the volume was not significantly reduced among the patients. All analyses were rerun after omitting 8 subjects aged 60 years or older, equally distributed across the groups, and results were unchanged. In addition, in patients, cortical volumes of the significant PUs were uncorrelated with the neuroleptic medication dose.

This study provides new evidence that there are cortical abnormalities in schizophrenia, detectable using structural MRI. The brain areas that showed a significant reduction in patients with schizophrenia compared with well-matched healthy controls were primarily in PUs that approximated regions of the prefrontal and medial paralimbic cortices. Significant frontal lobe reductions were in the middle, medial, and right-sided fronto-orbital cortices, the last 2 of which are considered part of the paralimbic cortices.⁴⁸ The anterior cingulate and paracingulate gyri and insula were also significantly reduced in patients compared with controls, the last 2 by almost 1 SD. The relative differences in volumes between patients and controls were primarily in the range of 7% to 15%. The largest effect sizes were in the middle frontal gyrus and right fronto-orbital cortex, the insula, and the anterior cingulate and paracingulate gyri.

In general, our findings are consistent with those of several previous MRI and neuropathological studies. Previous MRI studies that segmented specific prefrontal areas reported significant reductions in dorsolateral prefrontal (ie, middle frontal gyrus)^15,29,63 and orbital¹⁴ cortices. A recent study⁶⁴ using sophisticated segmentation procedures did not find significant prefrontal volumetric reductions. The investigators' sample size of 15, however, would have only 26% to 38% statistical power to detect a medium effect size of 0.50, and effect sizes of the frontal gyri in our study ranged from 0.10 to 0.62. Their findings also suggested that connectivity with prefrontal areas was abnormal in schizophrenia, which may be consistent with our findings.

The largest effect sizes in our study were demonstrated in paralimbic cortices, eg, anterior cingulate and paracingulate gyri and the insula. This is consistent with previous postmortem studies that reported cytoarchitectonic and structural abnormalities in the cingulate^22,65 and imaging studies that showed anterior cingulate gyral volumetric reductions^66,67 and hypofunction⁶⁸ in patients. Lesion and functional neuroimaging studies of the anterior cingulate gyrus have shown a variety of neurobehavioral deficits^69,70 and associations with internally initiated thought and behavior,^71,72 attention to action,⁷³ divided attention,^74,75 and learning.⁷⁶ Thus, it has been suggested that abnormalities of the cingulate gyrus may be central in understanding schizophrenia.²²

We would argue that our findings are not due to sample bias because patients were representative of a large outpatient network serving 3 major hospitals in the Boston area. Furthermore, the healthy controls were similar in sociodemographic background, including education, a variable typically affected by the illness. Finally, as in Pearlson et al,¹⁶ we used 3 different statistically analytic approaches to test our hypotheses, and these demonstrated consistent findings across methods, suggesting their validity.

What is most striking about our results is that, although we included 48 cortical PUs, the significant reductions were in paralimbic cortices (ie, cingulate gyrus, insula, and frontomedial and fronto-orbital cortices) and the prefrontal area—the middle frontal gyrus—which has strong reciprocal connections to paralimbic and limbic brain regions.^48,77 Furthermore, in a previous presentation of these patients,¹⁰ it was shown that among 10 subcortical regions tested, the thalamus and amygdala-hippocampal complex were the only subcortical areas significantly volumetrically reduced—areas with reciprocal connections^48,77,78 to the cortical areas that we found reduced in patients. (The pallidum and ventricles were significantly increased.¹⁰)

The functional consequences of the reduced size of the insula in patients is as yet unclear. The insular cortex, however, is highly interconnected with somatosensory and other cortical areas and limbic structures (perirhinal and entorhinal cortex and amygdala). It plays a role in sensory, motor, and language functions and has been described⁷⁹ as a limbic-integration area. In addition, the insula has connections to several brain areas that were found to be volumetrically reduced in this sample of patients, including the orbital and medial frontal cortices, the cingulate gyrus, and the amygdala. Insular abnormalities in schizophrenia may be related to an impairment in integrating sensory stimuli with internal motivational states, a hypothesis that warrants further investigation in patients with schizophrenia.

The importance of the paralimbic areas is also suggested by 2 of our previous analyses^44,46 of neuropsychological deficits of patients, including those in the present study. Olfactory and executive function deficits appeared to separate patients into 2 groups with distinct cognitive profiles, suggesting some heterogeneity among patients with cognitive deficits that may be associated with the fronto-orbital and hippocampal divisions^48,77 of the paralimbic cortices. Furthermore, these patients have exhibited other significant cognitive deficits that are dependent on limbic and paralimbic cortices, ie, executive function, attention, olfaction, and memory functions.^45,46

We also found a significant volumetric reduction in the posterior supramarginal gyrus, which is densely connected to prefrontal and cingulate cortices. This is a potentially important finding, given the area's role in attention and memory identified in lesion and functional imaging studies.^75,80,81 Although most studies^8,82 have shown that the parietal cortex as a whole is not significantly reduced in patients with schizophrenia, previous work²⁹ reported significant reductions in specific areas, such as the inferior parietal cortex—ie, the supramarginal and angular gyri. We have further suggested that volumetric reductions in this cortex may be in area 40, rather than the angular gyrus (area 39). Structural deficits in area 40 are consistent with functional studies^71,72,75,81 of attention in patients with schizophrenia and healthy subjects. These studies showed abnormalities in perfusion or activation in this area during sustained attention tasks.

We also found that several cortical PUs—the basal forebrain, subcallosal cortex, operculum, and the frontal pole—were larger, although not significantly, in patients than in controls. These subtle volumetric increases were in cortical areas that have strong reciprocal connections to the areas we found significantly reduced. This may reflect an orchestrated developmental mechanism representing brain plasticity. There is some precedent for posing this hypothesis because it is consistent with animal models and functional studies^5,25,42,83 of schizophrenia that showed hypofunction in particular cortical areas (eg, dorsolateral prefrontal cortex) and hyperfunction in subcortical (eg, hippocampus) or other cortical areas. It is also consistent with a recent study of schizophrenia⁸⁴ that suggested disruptions in the dopaminergic regulation of intracortical and corticostriatal connectivity associated with the orbitofrontal cortex. Thus, there may not be a focally acting pathological process producing volumetric reductions in specific brain regions in patients with schizophrenia. Rather, there may be cortical systems, ie, limbic-paralimbic, that exhibit subtle abnormalities in their connections that result in reductions in some cortical areas and increases in others. Although this study did not test this hypothesis, it warrants further investigation.

There were no significant volumetric reductions in other cortical regions, and the effect sizes of volumetric reductions in other cortical areas were negligible, except for visual primary and association cortices, ie, 0.34 and 0.40, respectively, in the cuneus and superior lateral occipital gyrus. This is consistent with a recent study¹³ that reported cytoarchitectonic abnormalities in neuronal density in patients with schizophrenia in the primary visual cortex, area 17.

Schizophrenia, however, is a heterogeneous disorder. Other brain areas—in particular, the superior temporal, Heschl and parahippocampal gyri, and planum temporale—have been found^{7,16,18,29-32,85,86} to be volumetrically reduced in patients. In all studies of the superior temporal gyrus, however, the samples were men only or a 2:1,¹⁶ 3:1,³¹ or approximately 5:1³⁰ ratio of men to women. Furthermore, some of these samples were patients with primarily positive symptoms,⁷ and positive symptoms, such as hallucinations and thought disorder, have been associated with superior temporal gyral abnormalities.^7,18

Our sample was a mix of male and female patients with mixed symptomatology. In fact, DeLisi et al¹⁹ had a similar sex distribution to ours and found no significant reduction in the superior temporal gyrus. Reite et al⁸⁷ found that only men had reduced superior temporal and Heschl gyri,⁸⁵ whereas Vogeley et al,⁸⁸ in a postmortem study, found that men had increased volumes in anterior and posterior superior temporal gyri, but women had a reduced superior temporal gyrus in the middle compartment. A reduction in women only was also found by Schlaepfer et al,²⁹ whereas men showed atypical laterality in the planum temporale.³³ These studies raise the possibility that only subgroups of patients exhibit particular volumetric abnormalities (eg, one sex, patients with primarily positive or negative symptoms, or those with asymmetry abnormalities rather than volumetric reductions^18,19,85,87). Given that our method allows for the assessment of topographically identified cortical and subcortical areas of the entire brain, we are in an advantageous position to ask questions about abnormalities in anatomical and functional neural systems as they relate to the heterogeneity of the expression of schizophrenia.

In this study, we demonstrated cortical volumetric abnormalities in the paralimbic system that have important implications for understanding the development of and functional consequences for schizophrenia. In future studies, we will investigate the effect of these anatomical abnormalities on symptoms and cognitive function in subgroups of patients with schizophrenia.

Accepted for publication February 10, 1999.

This work was supported by grants K21 MH00976 (1992-1994) and RO1 MH56956 (1997-2000) (Dr Goldstein) and Merit Award MH 43518 (Dr Tsuang) from the National Institute of Mental Health, Bethesda, Md; a grant from the Fairway Trust (Dr Kennedy); and a Dissertation Award from the Scottish Rite Foundation, Lexington, Mass (Dr Goodman).