Impaired P3 Generation Reflects High-Level and Progressive NeurocognitiveDysfunction in Schizophrenia FREE

One of the most robust biological abnormalities observed in schizophreniais a smaller amplitude of the P3 (or P300) component of the event-relatedpotential (ERP) elicited by using an auditory oddball paradigm in which asubject detects infrequent task-relevant (target) stimuli randomly presentedamong frequent, standard stimuli.^1- 15 BecauseP3 reflects stimulus context and stimulus meaning,^16- 18 theobserved P3 reductions in schizophrenia support the view that dysfunctionof attention and working memory represents a core cognitive deficit in thisdisorder.^19- 23 Severalresearchers have also observed a prolonged P3 latency in schizophrenia,^1,3,5,8- 9 butthe interpretation of this finding has been complicated by potential confoundsof medication.^5,9

An intriguing finding is that the auditory P3 is often selectively ormore severely impaired relative to the visual P3 in schizophrenia.^4- 5,7,13 Also,as opposed to the visual P3, the auditory P3 seems relatively independentof medication status^3,5,7 andclinical symptoms.^3- 4,7 Suchobservations have led to the view that the visual P3 may serve as a statemarker, reflecting the patient's current clinical status, whereas the auditoryP3 may indicate, at least in part, a vulnerability or trait marker, reflectingan enduring, potentially genetically transmitted causative pathophysiologicalfactor or process in schizophrenia.^4,7,13

Another interesting observation is that patients with schizophreniaoften show, in addition to reductions across the midline, a localized left-smaller-than-rightauditory P3 temporal scalp asymmetry, whereas healthy control subjects orpatients with psychotic affective disorder do not show this P3 asymmetry.^6,15,24- 28 Accordingly,the auditory P3 temporal lobe asymmetry may be specific to schizophrenia andhas been linked, by using magnetic resonance imaging, to reduced gray mattervolume of the left posterior superior temporal gyrus and left planum temporale.^6,27

Recent studies have indicated that patients with schizophrenia may alsoexhibit deficits in visual and auditory ERP components that precede P3. Thesedeficits include abnormalities of both relatively early, sensory-evoked components,such as P1,²⁹ N1,^{5,8,25,30- 31} andP2 ^5,25,30 and latecognitive-related components, such as N2^8,25,31 andmismatch negativity (MMN).^31- 36 Thesedeficits, however, do not appear to be as robust as P3 reduction, often varyingas a function of stimulus and task parameters (eg, physical and temporal stimulusproperties)^31,34- 35 andsubject sample characteristics (eg, recent-onset vs chronic schizophrenia).^5,36 Notwithstanding, these results raisethe possibility that schizophrenia is associated with deficits not only ata high (cognitive, semantic) level of information processing, as reflectedin P3 reduction, but also at lower (sensory, perceptual) processing levels,as evidenced by deficits in earlier-occurring components. To address thisissue, we assessed the integrity of several ERP components related to differentlevels of visual and auditory processing in patients with schizophrenia.

The present study was aimed at clarifying whether high-level attention-dependentcognitive deficits, as indexed by P3, in patients with schizophrenia are relatedto, or perhaps even originate from, potential preceding deficits at lowerlevels of information processing, as indexed by earlier-occurring ERP components.On the basis of previous ERP study results, as reviewed earlier, and the theoreticalconceptualization that the pathology of schizophrenia involves multifocaldiffuse abnormalities of brain function and structure,^{22- 23,37- 39} ratherthan a single or specific localized abnormality, we hypothesized that patientswith schizophrenia would probably display widespread independent deficitsat both relatively low and high levels of processing in both the visual andauditory modalities. Additionally, given that the auditory P3 amplitude hasbeen observed to be inversely correlated with illness duration^10,14 and,hence, may potentially track the operation of a putative progressive or neurodegenerativeprocess across the illness course,^40- 41 werecruited patients varying greatly in illness duration to attempt to replicatethis observation in the present study.

Participants consisted of 22 patients, with schizophrenia or schizoaffectivedisorder diagnosed according to DSM-IV criteria,⁴² and 22 age-matched healthy control subjects (Table 1). All subjects provided writteninformed consent. Patients were recruited from psychiatric facilities affiliatedwith the University of North Carolina at Chapel Hill. Diagnoses were establishedby using the Structured Clinical Interview for DSM-IV administratedby a psychiatrist or trained research assistant. Patients were functioningfairly well—most (82%) were outpatients—and they generally volunteeredto participate in more than 1 study, including a functional magnetic resonanceimaging study.⁴³ Clinical symptom ratings,assessed by using the Positive and Negative Syndrome Scale, indicated thatthe patients manifested relatively mild symptoms at the time of testing (Table 1); Positive and Negative SyndromeScale data were missing for 3 patients.

Estimates of illness duration, defined as current age at testing minusestimated age at illness onset, indicated that the mean illness duration wasrelatively short (<6 years), but the variability within the group was large(Table 1). The group consistedof 7 individuals with first-episode schizophrenia, who had only recently experiencedtheir first illness episode and had been ill for less than 1 year (mean ±SD, 0.4 ± 0.3 years), 7 individuals with an illness duration between1 and 5 years (mean ± SD, 3.0 ± 1.9 years), and 6 individualswith chronic (>10 years) schizophrenia (mean ± SD, 15.7 ± 3.8years); estimates of age at onset were not available for 2 patients.

At the time of testing, 18 patients were taking atypical antipsychoticmedication, 2 patients were taking a combination with typical neurolepticmedications, 1 patient was prescribed medication but reported to be not compliant,and 1 patient was medication free. Most (59%) of the patients also had prescriptionsfor antidepressive, anticonvulsive, lithium carbonate, and/or antianxietymedications. Any patient with a history of neurological insult or illness,serious head injury, mental retardation, uncorrected vision or hearing problems,or current (<1 month before study participation) alcohol or drug dependenceor abuse was excluded. Four patients had a history of alcohol and/or drugabuse.

Healthy control subjects were recruited by means of advertisements,and we used the same exclusion criteria, with the addition of no personaland family history of major psychiatric disorder and no current or past alcoholor drug dependence or abuse. The 2 groups did not differ significantly inmean age (P>.98), proportion of female subjects (P>.73), and proportion of right-handed subjects (P>.17), but years of education were significantly fewerin patients (Table 1). However,education was not significantly associated with the performance or ERP measures.

The ERPs were recorded by using a visual and auditory oddball paradigm.In the visual paradigm, subjects were instructed to attend to a series ofvisual stimuli, while ignoring a series of intermixed auditory stimuli. Thevisual stimuli (n = 1428, 25.4° width, 18.8° height, 506-millisecondduration) consisted of standard stimuli (squares, 94.4% of all visual stimuli),target stimuli (circles, 2.9% of all visual stimuli), and novel stimuli (picturesof familiar objects, 2.7% of all visual stimuli) and were presented at a constantinterstimulus interval of 1500 milliseconds. The auditory stimuli (n = 1428,85 dB sound pressure level, 100-millisecond duration, 10-millisecond rise/fall)consisted of standard stimuli (1000-Hz tones, 97.1% of all auditory stimuli)and deviant stimuli (1064-Hz tones, 2.9% of all auditory stimuli) and werepresented binaurally at a variable interstimulus interval of 1300 to 1700milliseconds.

The subject's task was to attend to the visual stimuli, while ignoringthe auditory stimuli, and to make a button-press response with the right indexfinger each time a visual target stimulus was presented. Subsequently, subjectsperformed an auditory oddball task in which they attended to auditory stimuli,while passively viewing visual stimuli. The auditory stimuli (n = 492) werethe same tones as those used in the prior blocks, consisting of standard (91.5%of all auditory stimuli) and deviant (8.5% of all auditory stimuli) targetstimuli presented binaurally at a constant interstimulus interval of 1500milliseconds. The visual stimuli (squares, n = 492, 100% of all visual stimuli)were presented at a variable interstimulus interval of 1300 to 1700 milliseconds.The subject's task was to attend to the auditory stimuli, while ignoring thevisual stimuli, and to make a button-press response with the right index fingereach time an auditory target stimulus was presented.

Electroencephalograms (EEGs) were recorded from 30 electrodes: Fp1,Fp2, F7, F3, Fz, F4, F8, FT7, FC3, FCz, FC4, FT8, T7, C3, Cz, C4, T8, TP7,CP3, CPz, CP4, TP8, P7, P3, Pz, P4, P8, O1, Oz, and O2.⁴⁴ Theright mastoid served as the reference and AFz as the ground. Eye movementsand blinks were measured with bipolar recordings of the vertical and horizontalelectro-oculogram by using electrodes above and below the right eye and onthe outer canthus of each eye, respectively. The EEG and electro-oculogramwere amplified, bandpass filtered between 0.15 and 70 Hz (notch filter at60 Hz), and digitized at 500 Hz.

Performance measures consisted of the proportion of correctly detectedtarget stimuli (hit rate), the proportion of stimuli incorrectly respondedto as target stimuli (false-alarm rate), and the time needed to respond totarget stimuli (reaction time). The EEG recordings associated with incorrectbehavioral responses or containing voltages in excess of ±100 µVwere excluded. Ocular artifacts were controlled for by using regression analysis.⁴⁵ The ERPs were computed for each stimulus type ateach scalp location; the averaging epoch included a 200-millisecond prestimulusbaseline period and a 1000-millisecond poststimulus period. The ERPs werelow-pass filtered at 15 Hz before quantification.

The ERP components were assessed only in those experimental conditionsand only at those scalp locations where they were most clearly present (eg,being most likely uncontaminated by other overlapping potentials) and couldbe most reliably quantified (Table 2).Additional information on the ERP component amplitudes at other scalp locationscan be found at www.nirl.unc.edu. Amplitudes were quantified bycomputing the mean voltage across the latency range during which the componentof interest was maximal. In addition to mean voltage measures, we obtainedbaseline-to-peak voltage measures for estimating the component amplitudes.Although mean amplitude measures were based on multiple data points, whereaspeak measures were based on only a single data point, the mean and peak amplitudemeasures were strongly correlated and yielded essentially the same results.In this study, only the results based on mean voltage measures are presented.The ERP component peak latencies were quantified by detecting the most positiveor negative peak within a specified time window at selected midline locations.For MMN, a statistical onset latency measure was also obtained by determiningthe time point at which the deviant-minus-standard difference wave startedto deviate significantly from zero.⁴⁶

Independent-sample t tests were used to assessbetween-group differences in the performance and ERP latency data. Two typesof analyses were performed on the ERP amplitude data. The first was conventionalanalyses that involved (1) independent-sample t teststo assess between-group amplitude differences only at the midline locationwhere the component of interest is typically largest and (2) 2-way (2 groups[between] × 3 anterior-to-posterior midline [Fz, Cz, Pz] electrode locations[within]) repeated-measures multivariate analyses of variance (MANOVAs) toassess group amplitude and topographic differences across the midline. Thesecond was 3-way (2 groups [between] × 5 anterior-to-posterior midlineelectrode locations [within] × 5 lateral-to-medial coronal electrodelocations [within]) repeated-measures MANOVAs to assess group amplitude andtopographic differences in greater detail. In the 3-way MANOVAs, the electrodelocation factors were arranged such that the coronal electrode chains werenested under the anterior-to-posterior locations (F7-F3-Fz-F4-F8 vs FT7-FC3-FCz-FC4-FT8vs T7-C3-Cz-C4-T8 vs TP7-CP3-CPz-CP4-TP8 vs P7-P3-Pz-P4-P8), which yielded2 orthogonal electrode factors.

Follow-up tests included tests of simple main effects, simple interactioneffects, and simple, simple main effects, and the Bonferroni approach wasused to control for type I errors at the .05 level.⁴⁷ Ifa significant group-by-electrode location interaction emerged, the test wasrepeated on normalized data to determine whether the group-by-electrode interactionreflected real differences in topographic profile or simply differences inoverall amplitude between groups.^6,24,48 Finally,topographic maps were constructed to provide a simple visual means of illustratingthe various topographic patterns. Values in the text represent the mean ±SD. The effect size index, η², is alsoreported. Two-tailed statistics were used.

Within-group correlation and regression analysis was performed to assessthe relationship between P3 and illness duration.^10,14 Also,we explored the relationships between sensory-evoked and cognitive-relatedERP components and the relationships of P3 with symptom severity,^2,5,13 age,¹⁴ andperformance.⁴⁶ For all within-group analyses,significance levels were not corrected for multiple testing to maintain sufficientpower to detect an individual, potentially relevant, true effect, if existing,in the data. Analyses were restricted to the midline electrode where the componentof interest was maximal.

In the visual task, patients showed a longer reaction time to targetstimuli than did control subjects (573 ± 84 milliseconds vs 501 ±87 milliseconds; t₄₂ = 2.8, P = .008); no significant between-group differences were observed inthe hit rate (91.2% ± 15.8% vs 95.5% ± 7.6%; P>.26). In the auditory task, patients displayed a lower hit rate thandid control subjects (86.2% ± 15.9% vs 96.8% ± 7.0%; t₄₂ = 2.9, P = .006), with no groupdifferences seen in reaction time (498 ± 91 milliseconds vs 445 ±110 milliseconds; P>.09). In both tasks, false-alarmrates were low (<0.5%) and did not differ according to group.

Figure 1 illustrates the visual-evokedN1 and P2 and the P3 elicited specifically by visual target stimuli. The N1was accompanied by P1 localized over the occipital scalp (data not shown). Table 3 presents for each group the quantifiedcomponent latencies and amplitudes at a selected midline electrode and theresults of the corresponding statistical analyses. No significant between-groupdifferences were observed in the latencies, topographies, and amplitudes ofP1, N1, and P2. In each group, P1 was maximal occipitally, whereas N1 andP2 were maximal at medial central and frontocentral locations. Additionally,P3 latency and amplitude at Pz did not differ significantly between groups(Table 3). Similarly, MANOVAsyielded no significant main effects of or interactions with group (P>.20 for all), which indicates that no systematic group differencesexisted in P3 topography and amplitude. In each group, the visual target P3was maximal at medial parietal locations and showed left-smaller-than-rightasymmetries at medial frontocentral, central, and centroparietal sites.

Figure 2 illustrates N2 andP3 elicited by visual novel stimuli. No significant group differences wereobserved in the N2 latency, topography, and amplitude. In each group, N2 exhibiteda relatively symmetrical distribution, being maximal across medial frontaland central locations. For P3, no group differences were observed in peaklatency, but patients manifested a significantly smaller amplitude at Pz thandid control subjects (Table 3).The strength of the observed group-P3 relationship, as assessed by η², was moderate to strong, with the group factor accountingfor 10% of the variance. Additionally, 3-way MANOVA yielded a significantgroup-by-coronal electrode interaction effect (Wilks Λ = .68, F_4,39 = 4.40, P = .005). This result reflectedthat group differences were largest at midline and left-medial locations,but results of follow-up tests did not exceed Bonferroni-adjusted significancelevels. No other significant P3 amplitude or topographic differences werenoted. In each group, the novelty P3 was bilaterally symmetrical and largestat medial parietal locations.

Figure 3 presents N1 and P2elicited by auditory standard and deviant stimuli recorded during visual attention.No significant group differences were observed in latencies, topographies,and amplitudes of N1 or P2 elicited by standard stimuli. In each group, N1and P2 were maximal at medial frontocentral, central, and centroparietal locations.Additionally, the auditory deviant stimuli relative to the standard stimulielicited in each group a prominent negativity, the MMN, which overlapped N1and P2 and was most distinct between about 100 and 250 milliseconds afterthe stimulus onset (Figure 3 and Figure 4). An initial 2-way (2 groups ×2 stimulus types) analysis of variance on the unsubtracted data from Fz demonstratedthat the effect of stimulus deviance, as reflected by MMN, was significant(stimulus type, F_1,42 = 87.14, P<.001)and did not vary by group (group-by-stimulus type, P>.60),which provides statistical confirmation that the task-irrelevant auditorydeviant stimuli elicited a significant MMN in both patients and control subjects.Analyses of the MMN time course and size at Fz revealed no significant groupdifferences in peak latency or amplitude (Table 3), but the onset latency was longer in patients than in controlsubjects (110 vs 90 milliseconds). Similarly, MANOVAs produced no significantmain effect of or interactions with group (P>.20for all), which suggests that no systematic group differences existed in MMNtopography and amplitude. In each group, MMN exhibited a relatively symmetricaldistribution, being maximal across medial frontal and central scalp.

Figure 5 displays N2 and P3elicited specifically by auditory target stimuli. No significant group differenceswere detected in N2 latency, topography, and amplitude. In each group, N2was maximal across medial frontal and centrofrontal scalp and exhibited aleft predominance at medial central and centroparietal locations. Additionally,no group differences were established in the auditory P3 latency, but patientsexhibited a significantly smaller amplitude at Pz than did control subjects(Table 3). Also, 2-way MANOVAproduced a significant group-by-midline electrode interaction effect (Wilks Λ= .82, F_2,42 = 4.60, P = .016), whichsignifies that group differences were larger at Pz and Cz than at Fz. Despitethese amplitude differences, MANOVA performed on normalized data demonstratedthat the P3 topographical profile across the midline did not differ accordingto group.

To assess topographic differences in greater detail, 3-way MANOVA wasconducted and yielded a significant 3-way (group-by-coronal-electrode–by–midline-electrode,Wilks Λ = .41, F_16,27 = 2.39, P =.022) interaction effect. Simple interaction effects tests showed that thegroup-by-coronal electrode interaction effect was significant for the centroparietal(Wilks Λ = .64, F_4,39 = 5.60, P =.001) and parietal (Wilks Λ = .65, F_4,39 = 5.28, P = .002) data, while approaching Bonferroni-adjusted significancelevels for the central data (Wilks Λ = .75, F_4,39 = 3.18, P = .024). These interactions reflected that P3 reductionsin patients were most pronounced at midline (Cz, CPz, Pz) and left-medial(C3, CP3, P3) locations (t₄₂ = 2.8-3.7, P = .001-.008, η² =0.16-0.25). Partial correlation coefficients for the Pz data demonstratedthat the group-P3 relationship remained significant after group differencesin auditory hit rate (partial correlation coefficient = 0.32, P = .039) or novelty P3 amplitude (partial correlation coefficient= 0.31, P = .044) had been statistically removed.

To determine whether the significant group-by-coronal electrode interactionsreflected real differences in topographic profile or merely differences inoverall amplitude between groups, similar MANOVAs were performed after theraw data had been normalized. These analyses demonstrated that the 2-way interactionremained significant for the central (P = .005) andcentroparietal (P<.001) coronal electrodes. Theseresults signified that patients displayed a left-smaller-than-right P3 asymmetryat the middle temporal (T7: 2.4 ± 2.3 µV vs T8: 3.6 ±2.3 µV, t₂₁ = 3.7, P = .001) and posterior temporal (TP7: 3.1 ± 2.3 µV vsTP8: 4.2 ± 2.2 µV, t₂₁ =4.1, P = .001) sites, whereas control subjects didnot show such an asymmetry (T7: 4.2 ± 3.1 µV vs T8: 4.3 ±2.6 µV, P>.81; TP7: 4.6 ± 3.0 µVvs TP8: 4.8 ± 2.6 µV, P>.72). No othersignificant amplitude or topographic differences were observed. In each group,the auditory P3 was maximal at medial parietal locations and showed left-smaller-than-rightasymmetries at frontal and frontocentral sites, as well as at medial centraland centroparietal locations.

In patients, the auditory target P3 amplitude showed the anticipatedcorrelation with illness duration (Figure6). The visual target P3 amplitude also correlated with illnessduration (r = −0.49, P =.030). In contrast to the auditory P3 amplitude, the visual target P3 amplitudealso correlated with age in patients (r = −0.54, P = .009, b = −0.24 µV/y)and control subjects (r = −0.45, P = .036, b = −0.27 µV/y). Giventhat age correlated strongly with illness duration (r =0.86, P<.001), these results indicated that thevisual target P3 amplitude was associated more with normal aging than withillness duration.

In control subjects, P3 latency correlated with the reaction time andhit rate in both the visual and auditory modalities, whereas in patients thesecorrelations were significant only in the visual modality (Figure 7). Furthermore, correlations between auditory sensory-evokedand cognitive-related ERP components were observed in control subjects butnot in patients (Figure 7).

The strength of this study is that in it we assessed both sensory-evokedand cognitive-related ERP components and their relationships in both the visualand auditory modalities in patients with schizophrenia and healthy controlsubjects. The major finding is that patients showed smaller P3 amplitudesto visual novel stimuli and auditory target stimuli than did control subjects,whereas small or no significant between-group differences were observed inearlier-occurring ERP components. These data represent compelling evidencethat high-level attention-dependent cognitive deficits central to schizophreniado not originate from potential preceding deficits at lower levels of sensory,perceptual, or cognitive processing. The results support the view that schizophreniais characterized by fundamental deficits in integrative cortical functionsthat preferentially impair the ability to analyze, represent, and use stimuluscontext to guide behavior.^22- 23,49

The observation that the auditory P3 amplitude was associated with illnessduration in patients replicates previous findings^10,14 andindicates that the auditory P3 reflects a basic pathophysiological processin schizophrenia that is present at illness onset and continues across thelong-term course of the illness. These P3 data are in accord with the hypothesisthat the pathophysiology of schizophrenia includes a progressive or neurodegenerativeprocess that operates across the course of the illness in at least a subsetof patients with schizophrenia.^14,40- 41

Another noteworthy finding is that in control subjects, P3 showed meaningfulrelationships to task performance in both the visual and auditory modalities,whereas in patients, significant P3-performance relationships were observedonly in the visual modality. Similarly, significant correlations between auditorysensory-evoked and cognitive-related ERP components were observed in controlsubjects but not in patients. These results suggest that, in the auditorymodality, the functional links between perception, cognition, and action areuncoupled or weakened in schizophrenia. It has been hypothesized, indeed,that such functional disconnections between different levels of informationprocessing in the brain underlie the auditory hallucinations and profounddisintegration of thinking and action that characterize schizophrenia.^50- 51 An important goal for future schizophreniaresearch, possibly with diffusion-tensor magnetic resonance imaging,⁵² is to determine whether such functional disconnectionsare associated in vivo with structural disconnections of long intracorticalwhite matter fiber tracts linking different brain regions, particularly thosetracts connecting the temporal and frontal lobes.^{23,38,41,50- 54}

Patients exhibited a reduced auditory target P3, while manifesting arelatively normal visual target P3. Although a systematic comparison herebetween P3 to auditory and visual target stimuli is not possible because ofmarked differences in the eliciting experimental conditions, this findingadds to the evidence that the auditory target P3 is selectively or more severelyimpaired relative to the visual target P3 in schizophrenia.^4- 5,7,13 Thisobservation substantiates the notion that the auditory target P3 reflectsa basic pathophysiological factor in schizophrenia, whereas the visual targetP3 may index primarily the patient's current clinical state.^4,7,13

The results that small or no between-group differences existed in ERPcomponents preceding P3 are consistent with results of several studies^15,31,55 but are apparentlynot consistent with results of studies in which the authors report markeddeficits in schizophrenia also in these earlier-occurring components, particularlythe auditory P2 ^5,25 and MMN.^31- 36 Theapparent discrepancies among studies are probably related to differences instimuli, task parameters, and patient samples. Specifically, the degree ofstimulus deviance and the stimulus presentation rate are particularly importantvariables that determine whether patients with chronic schizophrenia showan impaired pitch-deviant MMN, with deficits more readily demonstrated whenthe pitch difference between standard and deviant tones is large (eg, >10%)³⁴ and/or when interstimulus intervals are short (<300-400ms).³⁵ Although it is possible that differentattentional strategies (ie, differential attention to the irrelevant auditorydeviant stimuli) may confound MMN measurements in patients with schizophreniaand control subjects,³¹ the effect of thisvariable, if any, seems minimal because the same between-group MMN resultshave generally been obtained regardless of whether a passive-ignore (eg, readinga book) or, as in the present study, an active-ignore (eg, performing a visualtask) experimental procedure was used to direct subjects' attention away fromthe eliciting auditory stimuli.^31- 36

Another variable that does seem to be important is illness durationbecause pitch-deviant MMN reduction has been reported to be present in patientswith chronic schizophrenia but to be absent in patients with first-episodeschizophrenia.³⁶ Thus, our failure to detectMMN reduction in patients with schizophrenia is likely because we used a smallpitch difference and a slow presentation rate to elicit MMN and because theaverage illness duration of the patient sample was relatively short. Indeed,the importance of the latter variable is underlined by the observation that,although we did not detect a significant overall correlation between MMN andillness duration (r = 0.32, P>.16;n = 20), when we classified the patients into 2 distinct, though small, subgroupson the basis of having either first-episode, recent-onset (<1 year; n =7) or chronic (>10 years; n = 6) schizophrenia, the MMN recorded in patientswith chronic schizophrenia was 50% smaller than that observed in patientswith first-episode schizophrenia (Fz: −1.4 ± 2.3 µV vs−2.8 ± 1.5 µV, t₁₁ =1.3, P>.21), who did not differ markedly from controlsubjects. The present data indicate that deficits in ERP components precedingP3 in patients with schizophrenia may not reflect primary pathophysiologicalfeatures of the disease but may be related to illness chronicity or progressionand/or consequential medication effects.

The observation that patients manifested a distinct left-smaller-than-rightauditory P3 temporal scalp asymmetry corroborates and extends results of previousstudies^6,15,24- 28 byindicating that this hemispheric asymmetry in schizophrenia is present onlyin the auditory, and not in the visual, modality. Because the left-lateralizedauditory P3 deficit in schizophrenia has been linked to structural pathologyof the left posterior superior temporal gyrus and the left planum temporale,^6,27 these findings support the conceptthat schizophrenia is characterized by abnormal lateralization of those cerebralfunctions and structures that mediate language and auditory processing.⁵⁶ The ERP data agree with evidence from neuroimagingand postmortem schizophrenia studies,^{23,37- 38,41,43,49- 54} whichimplies that abnormalities in the function and structure of the temporal lobeand of its interaction with other, particularly prefrontal, regions representa core element of the pathophysiology of schizophrenia.

It is important to acknowledge several limitations of the present study.Initially, the patient group was heterogeneous and group sizes were small,and the consequent loss of statistical power increases the probability oftype II errors, which yields an essential ambiguity concerning findings thatwere not significant. Similarly, the within-group analyses involved many testswhile significance levels were not corrected for multiple testing, which increasesthe probability of type I errors. Accordingly, the results of particularlythe exploratory correlational analyses should be considered in need of replication.Moreover, the patients originated predominantly from an outpatient settingand exhibited relatively mild symptoms at the time of testing, so they maynot accurately represent individuals typically encountered in clinical practiceor examined in previous schizophrenia studies, which may limit the generalizabilityand comparability of the present findings. Furthermore, almost all patientswere taking psychotropic medications at the time of testing, so some of thefindings obtained may reflect secondary effects of medication. Finally, thestudy was cross-sectional; therefore, the interpretation that the auditoryP3 may index progressive brain changes in schizophrenia needs to be substantiatedwith longitudinal data.

Corresponding author: Aysenil Belger, PhD, University of North Carolina,Department of Psychiatry, Campus Box 7160, 250 Medical School Wing D, ChapelHill, NC 27599 (e-mail: aysenil_belger@med.unc.edu)

Submitted for publication March 6, 2003; final revision received July11, 2003; accepted July 22, 2003.

This study was supported by grant MH58251 from the National Instituteof Mental Health, Bethesda, Md; grant MH64065 from the University of NorthCarolina Schizophrenia Research Center—National Institute of MentalHealth Silvio O. Conte Center for the Neuroscience of Mental Disorders, ChapelHill, NC; and the Foundation of Hope, Raleigh, NC.

Results of this study were presented in part at Biennial Meeting ofthe International Congress on Schizophrenia Research; April 1, 2003; ColoradoSprings, Colo; and Annual Meeting of the Society of Biological Psychiatry;May 15, 2003; San Francisco, Calif.

We thank Diana O. Perkins, MD, Rajendra Morey, MD, and Margaret Rosemond,BA, for their contribution to patient recruitment and assessment.