Prefrontal Broadband Noise, Working Memory, and Genetic Risk for Schizophrenia

Deficits of frontal lobe-related information processing are common among schizophrenia patients and are thought to be phenotypic traits related to the genetic risk for schizophrenia (1–7). However, it is unclear what pathophysiological mechanisms account for the neuropsychological impairments. In particular, abnormal prefrontal physiological responses in patients with schizophrenia frequently vary from too little activity to too much, depending on factors such as task conditions and clinical state (8–14). This variable expression pattern suggests the existence of a more fundamental underlying deficit.

There are two lines of evidence that the underlying fundamental physiological trait may be related to increased background noise of prefrontal neuronal activity. The first line of evidence is largely circumstantial and comes from studies of the role of dopamine signaling in prefrontal information processing and evidence of abnormal cortical dopamine innervation in schizophrenia. Weinberger et al. (15) reported that a measure of cortical dopamine turnover (CSF concentrations of the dopamine metabolite homovanillic acid) directly predicted physiological activation of the prefrontal cortex during executive cognition in patients with schizophrenia. Akil et al. (16) described decreased dopaminergic prefrontal cortex innervation in the postmortem brain. Abi-Dargham et al. (17) found increased prefrontal cortex dopamine D₁ receptor availability in schizophrenia patients, which strongly predicted working memory performance and was interpreted as a reflection of diminished dopamine innervation. In addition, several functional neuroimaging investigations that were conducted in conjunction with monoaminergic drug administration reported an increase in spatial “focusing” of the prefrontal hemodynamic response during frontal lobe-related tasks (18–23). This hemodynamic cortical response pattern was generally interpreted as an indication of enhanced D₁-receptor-mediated signal-to-noise ratio of prefrontal neuronal activity.

These clinical observations are consistent with basic studies implicating dopamine/D₁ signaling in modulating cortical signal-to-noise ratio. Single unit recordings of action potentials in the cortex of rats and nonhuman primates have shown that the administration of catecholaminergic drugs or stimulation of D₁ receptors abolishes random firing of pyramidal neurons and increases spike rates in selective frequency bands, thereby enhancing the signal-to-noise ratio. These pharmacological effects on single neuron activity are accompanied by improved performance across different task conditions, including choice reaction type and delay type as well as oddball tasks (24–28). The signal-to-noise-ratio increasing effect appears to be mediated via D₁ receptor stimulation, which is thought to enhance selective inputs to both pyramidal cells and inhibitory interneurons (29). D₁ receptors are found on distal dendrites and spines of pyramidal cells adjacent to asymmetric and presumably glutamatergic synapses (30–33). In addition, tyrosine hydroxylase/dopamine immunopositive terminals in the prefrontal cortex form symmetric synapses on dendrites of GABA immunoreactive neurons in layer VI (34, 35). In vitro whole-cell patch-clamp recordings revealed that dopamine D₁ receptor agonists cause large, GABA_A-mediated evoked inhibitory postsynaptic currents in pyramidal neurons and also directly enhance postsynaptic N-methyl-d-aspartic acid currents on pyramidal layer V neurons, while a slight reduction of D₁-mediated presynaptic glutamate release was observed (36, 37). Gao et al. (38) also described that dopamine reduces glutamate release in layer V pyramidal neurons by D₁ activation at presynaptic sites. Recent computational models have incorporated these experimental findings and suggest that dopamine D₁-receptor-mediated synaptic inhibition is critical for shaping the selectivity (i.e., signal-to-noise ratio) of neural activity in prefrontal circuits, in the sense that recurrent excitatory-inhibitory synaptic interactions are balanced toward inhibition (39–43). These various experimental and computational observations on the cortical actions of dopamine have supported the notion that diminished mesocortical dopamine signaling and the resulting reduced signal-to-noise ratio of prefrontal neurons might be involved in cognitive and behavioral deficits that are observed in schizophrenia (15, 29, 44, 45).

A second line of evidence suggesting increased noise of cortical neuronal assemblies during information processing in schizophrenia comes from scalp-recorded electrophysiological investigations of single-trial data in patients. These studies described both increased latency jittering of stimulus-evoked brain events (46–48) and a lack of stimulus-induced synchronization (i.e., phase-resetting) over a wide range of frequencies (49). These findings, which have emerged from two parallel approaches to event-related potential analysis, are compatible with each other. The first approach is based on the traditional notion of event-related potential generation, i.e., a fixed-latency and fixed-polarity neural response within circumscribed cortical areas that requires removal by averaging of background EEG and latency jittering as uninformative noise. Therefore, traditional event-related potential studies paid little attention to the increased latency jittering of cortically evoked responses in schizophrenia. The second, frequency-based approach to event-related potential analysis that was previously taken by us (49), however, took into account recent evidence that event-related potentials can arise—at least in part—from stimulus-locked, dynamic changes of the ongoing neural synchrony generating the scalp electromagnetic fields (50–59). These studies have demonstrated phase resetting of EEG oscillations for different poststimulus time windows using techniques such as frequency filtering and decomposition of the frequency spectrum by principle or independent component analysis. This phase resetting of field potentials is related to the latency of the averaged event-related potential (59), suggesting that event-related potential latency-jittering (“noise”) and lack of stimulus-evoked phase resetting of ongoing oscillations describe the same phenomenon. Of note, measurements taken from visual and auditory event-related potential experiments in normal subjects have demonstrated that event-related potential signal amplitude (power) and signal-to-noise ratio vary across conditions and sensory modalities, whereas noise power is highly stable within subjects across experimental conditions (60), possibly reflecting a task-independent, subject-specific physiological trait factor.

Thus, it is reasonable to hypothesize that prefrontal cortical information processing in schizophrenia is disturbed because of a lack of stimulus-induced synchronous action of prefrontal pyramidal neurons. Even though the physiological and molecular mechanisms are far from clear, it is conceivable that previous observations of reduced phase resetting in schizophrenia patients are not merely a measurement artifact. Nevertheless, validation of these findings is required, which was the main purpose of this study. Specifically, we addressed the following questions: 1) whether stimulus-induced prefrontal field potential variance (i.e., noise estimate) would predict frontal lobe-related cognitive performance, which is commonly deficient in schizophrenia patients; 2) whether prefrontal noise has trait characteristics; and 3) whether increased prefrontal noise would be associated with genetic risk for schizophrenia.

Method

Subjects

Subjects were recruited from national and local sources as part of the “Sibling Study” of the NIMH Clinical Brain Disorders Branch. The details of subject recruitment, evaluation, and potential ascertainment bias are discussed elsewhere (5, 61–63). Briefly, all participants gave written informed consent of an Institutional Review Board-approved protocol. Most families had two eligible full siblings (at least one of whom met DSM-IV criteria for schizophrenia). Participants had to be 18 to 60 years of age, above 70 in premorbid IQ, and able to give informed consent. Applicants with significant medical problems, history of head trauma, and alcohol or drug abuse within the last 6 months were excluded. All subjects were medically screened and interviewed by a research psychiatrist using the Structured Clinical Interview for DSM-IV Axis I Disorders (64).

Task Conditions

Subjects were presented 250 tones of two different pitches by loudspeaker at approximately 80 dB (SPL). Stimulus duration was 50 msec, with 5-msec rise and decay and pseudorandomized interstimulus interval of 1.0–1.5 seconds and order (target: 1500 Hz, probability=20%; nontarget: 1000 Hz, probability=80%). Subjects, who kept eyes closed during the task, were asked to count the number of targets.

As previously outlined (6), a neuropsychological test battery was administered before acquiring the electrophysiological data. The battery included the n-back working memory task (0-back, 1-back, 2-back), which is a test of working memory and executive cognition. The n-back test requires subjects to constantly update their mental set while minimizing interference from incoming stimuli (14). Subjects are told to recall the stimulus (one out of four numbers) seen “n” previously. Stimuli are presented for 1800 msec, with a 2000-msec break at the start of each task epoch. The applied n-back version emphasizes the “executive” aspects of working memory, with continual presentation of incoming stimuli and continual working memory response. For comparison, subjects also had to perform another putative frontal lobe task (Wisconsin Card Sorting Test) as a test of executive function and set shifting. Subjects also received the vigilance and distractibility versions of the Gordon Continuous Performance Test (d-vigilance subscale) as tests for attentional/vigilance processes and the WRAT reading subtest, which is thought to be an estimate of premorbid intellectual levels in schizophrenia patients.

Electrophysiology

Using a Grass model (8-24D) electroencephalograph, event-related EEG was recorded with gold electrodes during the oddball task. Electrode positions were defined according to the international 10/20 system, impedance was kept below 5 kΩ . Eye movements were registered on paper record with an additional channel across electrodes 1 cm lateral and either above or below the corner of the eye. Event-related EEG was measured from 16 electrode positions (Fp1, Fp2, F3, F4, F7, F8, T3, T4, T5, T6, C3, C4, P3, P4, O1, O2). Amplifiers were calibrated by using a 50-μV square pulse. Sampling frequency was 250 Hz with 0.5–70 Hz filters. A relatively high high-pass filter was used, since previous systematic investigations of P300 have suggested that this may improve group contrast between schizophrenia patients and healthy subjects (65, 66). All channels were recorded against separate ears as reference, i.e., left hemispheric electrodes were referenced to the left ear and right hemispheric electrodes were referenced to the right ear. Before data analysis, off-line artifact detection was performed automatically (amplitude criterion >70 μV) and manually repeated afterward by a rater who was blind to diagnosis. Specifically, all EEG segments that contained obvious eye movement, head movement, or muscle artifacts were excluded. Test-retest analyses were performed with a “naturalistic” data set of 15 subjects (three unmedicated, unaffected siblings and 12 schizophrenia inpatients receiving partly changing medication) with a mean duration between sessions of 63.56 days (SD=76.45).

Event-Related Potential/EEG Analysis

Generation of event-related potential grand averages across montage and subsequent topographic EEG magnitude, amplitude, and latency analysis were automatically performed with EEGSYS (67). P300 peak amplitudes against baseline and latencies were determined for the target condition of the event-related potentials by locating the peak within the specified time window: P300 (260–420 msec poststimulus). Baseline was defined as the available 50-msec prestimulus EEG. For quantitative event-related EEG analysis, the recorded signals (–50 msec to 590 msec poststimulus, target condition) were submitted to spectrum analysis using a 256-point fast Fourier transform yielding spectrum values in 1.525-Hz steps. The absolute magnitude in each frequency band was computed expressed in μV (square root of the EEG power). Frequency band partition was: delta=0.5–4.0 Hz, theta=4.5–8.0 Hz, alpha=8.5–12.5 Hz, beta 1=13.0–18.0 Hz, beta 2=18.5–30.0 Hz, gamma=35.0–45.0 Hz. The computation of the noise magnitude, which is subsequently denoted as “noise power,” was calculated following the recommendations of Möcks et al. (68) and Winterer et al. (49, 54). For any given frequency band, the mean magnitude of the single trials is subtracted from the magnitude of the average potential. This way, a quantification of the noise part of the event-related activity is approximated, and “noise” is equivalent with activity that is not time-locked to the stimuli, i.e., spontaneous background activity and jittering of the event-related signal (54).

Statistical Analyses

Statistical analyses were carried out with STATISTICA (69). Comparisons of demographic, clinical, functional, and neuropsychological data were performed with analyses of covariance (ANCOVA), chi-square analyses, regression analyses (Spearman’s r), multiple regression, or Student’s t tests as appropriate.

Exploratory principal component analyses (Varimax rotation) for variable reduction were carried out as described elsewhere (49, 54, 63) across electrode positions for noise power separately for each frequency band as well as P300 amplitudes and were based upon the entire sample of subjects (no major difference of factor solutions between samples). Principal component analysis of the event-related EEG and P300 amplitudes across electrodes revealed a two-factor structure within each frequency band (similar to what has been described before [49, 54]), i.e., one bilateral frontal factor that includes the frontal, central, and anterior temporal electrodes and one bilateral temporoparietal factor that includes mainly the parietal, posterior temporal, and to a lesser extent the occipital electrodes. This factor structure across electrodes was also observed for the P300 amplitudes. However, two notable exceptions were seen. The “frontal” delta factor extended from frontal electrodes, where the maximum was seen, to the posterior, i.e., central, parietal, and occipital electrode positions. The “temporoparietal” delta factor was largely limited to the temporal electrode positions. In the upper frequency bands, we observed a strong trend for a left-right hemispheric factor solution. However, for simplicity, we generally used a two-factor solution calling them “frontal” and “temporoparietal” factors or principal components. The total variance of the temporoparietal noise power factors varied between 8.0% and 19.0%, with eigenvalues in the range between 1.3 and 3.0. The total variance of the frontal noise power factors varied between 45.0% and 75.0%, with eigenvalues in the range between 7.5 and 11.5 (factor loadings available upon request).

Because we were mainly interested in prefrontal processing, ANCOVA group comparisons were only performed using the frontal factor scores of each individual under inclusion of one randomly selected schizophrenia proband and one randomly selected unaffected sibling from families with multiple sibships. Test-retest stability was also calculated for principal components with intraclass correlation (ICC_U) (70). This was achieved by using factor loadings from the principal component analysis as weight factors. These weights were multiplied by the raw magnitude values at each electrode position and subsequently added up for each individual. The resulting individual values for the test and retest set were then subjected to ICC calculations.

For genetic modeling, phenotypic similarity was estimated for sib-pairs (schizophrenia patient and unaffected sibling) using individual factor scores of the electrophysiological variables, treating the electrophysiological measures as quantitative variables, and calculating intraclass correlations (ICC_U) (70). Under the assumption of little influence from shared environment with respect to electrophysiological parameters (71), sib-pair intraclass correlations provide indirect information on the heritability of a phenotype. Sib-pairs with one schizophrenia patient and one clinically unaffected sibling per family were selected and matched as far as possible for age and gender, if sibships with multiple unaffected siblings were present. Relative risk (λ) calculations were performed on the entire sample of patients and siblings for the obtained electrophysiological phenotypes based on “qualitative phenotype” definitions. Accordingly, a cutoff for abnormality for the qualitative phenotype was arbitrarily delimited based on the arithmetic mean of healthy comparison subjects and adding one standard deviation or two standard deviations for the noise power principal components. Relative risk ratio (λ) was calculated for these phenotype definitions for unaffected siblings independently of whether their respective schizophrenia sibling showed the electrophysiological phenotype or not (72, 73). Subsequently, the number of affected siblings of index patients, who were themselves affected with the electrophysiological phenotype, was determined; these were concordant pairs for the electrophysiological phenotype. Unaffected siblings of affected patients were defined as nonconcordant. Based on this concordance definition, relative risk (λ′) was calculated in a modified way as (concordant pairs/[concordant pairs plus nonconcordant pairs])/frequency of affected comparison subjects. A standard chi-square statistic was used to test for significance of relative risk values by comparing distributions of affected status in siblings with the comparison group (5). These calculations were performed under inclusion of all unaffected siblings from families with multiple sibships.

Results

Clinical and Demographic Data

Table 1 depicts demographic and clinical data of the participants. No significant age differences were found between healthy comparison subjects and schizophrenia patients (t=1.2, df=153, p<0.24) or unaffected siblings (t=1.6, df=201, p<0.13). Gender differed significantly between comparison subjects and schizophrenia patients (χ²=15.4, df=1, p<0.001) and between comparison subjects and siblings (χ²=5.45, df=1, p=0.02). Healthy subjects had more years of education compared with schizophrenia patients (t=5.18, df=116, p=0.0001) but not compared with unaffected siblings (t=0.33, df=168, p<0.75). Healthy subjects were less frequently smokers compared with schizophrenia patients (χ²=16.3, df=1, p=0.0001) but not compared with siblings (χ²=0.6, df=1, p<0.44). With respect to handedness as measured by the Edinburgh Inventory (74), no significant differences were found between healthy subjects and schizophrenia patients (χ²=0.1, df=1, p<0.79) or siblings (χ²=2.1, df=1, p<0.16). Comparing the percentage of correctly counted targets, healthy subjects did not differ significantly from schizophrenia patients (t=0.53, df=80, p<0.60) or their unaffected siblings (t=–1.02, df=110, p<0.32).

Topographic Noise Pattern

Figure 1 depicts descriptive maps for each group separately (schizophrenia patients, their unaffected siblings, and comparison subjects) of the adjusted mean noise power of the event-related potential fields across montage and selected frequency bands. In the healthy comparison group, noise power in the delta and theta frequency range is highest over the frontal area, whereas in the other frequency bands, most noise power is observed over the frontocentral region (results from topographic ANCOVA group comparisons of noise power are available upon request).

Group Comparisons

Statistical group comparisons of the frontal noise-power principal components are depicted in Figure 2. This figure shows the results of overall two-way multivariate analyses of covariance with diagnosis and gender as the two factors, age and number of EEG trials as covariates, and the electrophysiological principal components as dependent variables. A significant main effect was seen for diagnosis. A marginally significant gender effect was also observed (Rao’s r_s=2.83, df=6, 201, p<0.02), as was a tendency for a gender-by-diagnosis interaction (Rao’s r_s=1.67, df=12, 402, p<0.08). Women showed lower noise power values than men in the delta-alpha range but higher noise power values in the beta/gamma frequency spectrum. Figure 2 also shows separate ANCOVAs (within the prior multivariate analysis of variance [MANOVA] model) for each frequency band with same factors and covariates as in the MANOVA as well as post hoc comparisons between groups. Schizophrenia patients showed highest noise power values across the entire frequency spectrum; siblings were characterized by intermediate values. Of note, multiple regression analysis across the entire sample revealed that the frontal P300 amplitude principal component was positively correlated with the frontal noise power components (R²=0.41, df=6, 254, p<0.00001).

Prefrontal Information Processing

Neuropsychological performance indices and electrophysiological variables were examined in the delta and alpha-frequency band with rank order correlation analyses (Spearman’s r). Statistically significant negative correlations were observed between frontal noise power in the delta frequency band and 1-back performance (r_s=–0.49, N=185, p<0.0001) (Figure 3), 2-back performance (r_s=–0.34, N=185, p<0.0002), and 3-back performance (r_s=–0.40, N=173, p<0.0001). A significant correlation was also observed with the 0-back condition (r_s=–0.16, N=186, p<0.04) and Wisconsin Card Sorting Test perseverative errors (r_s=0.22, N=223, p=0.001) as well as with the Continuous Performance Test d-vigilance scores (r_s=–0.20, N=224, p=0.003). No significant correlation was observed with WRAT reading scores (r_s=0.05, N=225, p<0.49). Separate correlation analyses between the 1-back condition and noise power in the delta frequency band were performed for the schizophrenia patients, their unaffected siblings, and healthy subjects. In all three groups, frontal noise power in the delta frequency band correlated significantly and negatively: schizophrenia patients (r_s=–0.43, N=39, p=0.006); siblings (r_s=–0.43, N=82, p=0.0006); healthy comparison subjects (r_s=–0.31, N=64, p<0.02). Very similar results were obtained when performing correlation analyses between the same neuropsychological task indices and noise power in the alpha frequency band: 1-back performance (r_s=–0.39, N=185, p<0.0001), 2-back performance (r_s=–0.26, N=185, p<0.0003), 3-back performance (r_s=–0.32, N=173, p=0.0002), 0-back performance (r_s=–0.09, N=186, p<0.25), Wisconsin Card Sorting Test perseverative errors (r_s=0.15, N=223, p<0.03), Continuous Performance Test d-vigilance (r_s=–0.12, N=224, p<0.09), and WRAT reading score (r_s=0.11, N=225, p<0.10).

Genetic Modeling

Test-retest intraclass correlations (ICC_U) of the frontal noise principal components were high in the theta and alpha frequency band (ICC_U=0.85–0.93), intermediate in the delta and beta 1 band (ICC_U=0.62–0.69), and low in the beta 2 and gamma frequency band (ICC_U=–0.05–0.10). Table 2 shows the between sib-pair intraclass correlations ICC_U and relative risk values (λ and λ′) of the frontal noise power principal components for 63 sib-pairs consisting of a schizophrenia patient and an unaffected sibling. On average, intraclass correlations of noise power are high between sib-pairs across frequency bands, with highest values in the alpha and beta frequency range. Relative risk calculations revealed that λ is mostly moderate and that λ′ is high with regard to noise power across the frequency spectrum. It is possible that a somewhat higher risk is conferred by noise power in the medium to upper frequency range (alpha/beta 1).

Discussion

The results of this study suggest that increased, frontally pronounced broadband cortical background noise during cognitive information processing is a characteristic trait pattern of brain activity in schizophrenia patients and to a lesser degree in their clinically unaffected siblings. Moreover, evidence is provided that the amount of frontal background noise predicts poor performance on frontal lobe cognitive tasks as well as genetic risk for schizophrenia.

The negative correlation between frontal background noise and performance on frontal lobe tasks—particularly those involving working memory performance and executive cognition—is intriguing for several reasons. First, it suggests that background noise is not simply an electrical epiphenomenon but a physiological entity with functional implications. Second, background noise is not only inversely related to cognitive performance in schizophrenia patients and siblings but also in healthy comparison subjects. This observation implies that a lack of stimulus-induced synchronization or phase resetting of ongoing field potential oscillations over a wide range of frequencies is generally unfavorable with regard to cortical information processing. Of note, this finding is to some extent in line with a study that has suggested an inverse relationship between parietotemporally generated event-related potential variability and psychometric intelligence (75).

Our findings also suggest that the frontal noise measure qualifies as an interesting quantitative trait for the investigation of susceptibility genes for prefrontal dysfunction in schizophrenia but also in the general population. The prefrontal noise phenotype offers a direct neurophysiological and testable link to theoretical concepts of schizophrenia, e.g., the prefrontal dopamine, glutamate, and GABA hypotheses, which should facilitate candidate gene approaches. From a genetic perspective, it is a particular advantage that the applied prefrontal noise measure manifests excellent test-retest stability as demonstrated in the present study in acutely ill schizophrenia inpatients under changing medication and clinical conditions. In addition, the high intraclass correlation between sib-pairs indirectly suggests high heritability. Other measures of prefrontal function (e.g., behavioral measures of cognitive test performance) have been reported to show considerably lower test-retest stability in patients, especially when symptoms vary (76). Moreover, indices of cognitive ability are generally less heritable (h₂=0.4–0.6) (e.g., reference 77) than electrophysiological parameters (h₂=0.7–0.9), which show little influence of shared environment (e.g., reference 71). Also, a comparison with previously reported relative risk estimates for deficits in executive cognition (Wisconsin Card Sorting Test performance), which were largely obtained from the same study sample (5, 6), indicates substantially higher λ values of our prefrontal noise measures. Also of considerable interest is that a previously reported electrophysiological genetic risk estimate for schizophrenia (63), i.e., the frontal P300 amplitude, which indirectly measures prefrontal noise, apparently is under less genetic control and also is a less strong predictor of risk for schizophrenia than the present noise estimates. This is broadly consistent with earlier findings of Turetsky et al. (60) on the task-independent intraindividual stability of noise measures as compared with the stability of event-related potentials, and it is also in agreement with investigations on the stability of task-related EEG during oddball conditions (78). These observations, together with the simplicity of the oddball task, should therefore render the frontal noise phenotype a promising candidate for developing electrophysiological animal models that could be used in quantitative trait genetic loci identification. In fact, several electrophysiological oddball-type rat and mouse models have been successfully developed that could be used for this purpose (26, 79–82).

An important methodological question is whether our observations reflect a neurophysiological phenomenon or a measurement artifact that could have resulted from eye movement, blinks, or muscle artifacts. Several findings suggest that the increase of background noise in siblings and schizophrenia patients is of cortical origin. First, frontal noise power predicts working memory performance in schizophrenia patients, their unaffected siblings, and healthy comparison subjects, and the high test-retest intraclass correlations within subjects and high intraclass correlations between sib-pairs suggest a major physiological trait component. In addition, the noise power increase in schizophrenia patients and their siblings was observed relatively consistently across the entire frequency spectrum. With regard to muscle or eye movement artifacts, one would expect a more frequency-selective increase of power. Thus, the topographic noise maximum was found in the central region in several frequency bands (i.e., alpha band) where such artifacts are rarely seen. Moreover, principal component analyses did not provide evidence for “artifact” components with major effects. The applied noise measure obtains further validity from the observed positive correlation between noise and frontal P300 amplitude. A novelty task component—such as during oddball paradigms—results in an orienting reaction with activation of the prefrontal cortex as demonstrated by a number of electrophysiological and functional neuroimaging studies (83–89). Electrophysiologically, this orienting response is expressed as an increase of the frontal P300 amplitude around 300 msec after stimulus presentation. Intracortical EEG recordings have shown that the frontal P300 component does not show a polarity inversion or a clear peak between 220 and 360 msec after the oddball stimulus in the scalp-recorded averaged potential (83–85, 90). As opposed to the classical parietotemporal P300, the frontal P300 following the oddball stimulus thus cannot be regarded as a typical cortical signal but rather as a “noisy” component (for a more detailed discussion see reference 63), which is compatible with the observed positive correlation between frontal P300 amplitude and our frontal noise measure. Nevertheless, it is plausible that nonneurophysiological noise (e.g., eye movements) may have some limited effect, as suggested by the somewhat lower intraclass correlations and lower relative risk values in the very low and high frequency range.

Another issue that needs to be addressed is the question about the underlying neurophysiological mechanism that is responsible for the observed increase of prefrontal noise in schizophrenia patients and their siblings. In principal, there are several possibilities as previously explored and discussed by us (49). Perhaps the most important question is whether increased noise can even be observed independently of stimulus processing during resting condition. Thus, an increase of background power across the entire frequency spectrum in resting EEG theoretically could explain our observation of increased noise. In fact, an increase of power, particularly in the slow frequency range, is a frequent observation in schizophrenia patients. However, our prior work (49), which includes a study of resting EEG in a sample that largely overlapped with the present study sample (61) as well as a schizophrenia twin study of resting EEG (91), suggests that background EEG power is 1) not increased in unaffected family members of schizophrenia patients and 2) is not the main reason for the observed increase of noise in schizophrenia. This prior work, which also included the investigation of EEG coherence changes between pre- and poststimulus EEG, also suggested that it is not simply a reduction of event-related signal amplitude in the sense of less activation or cortical recruitment that accounts for the increase of noise but a lack of event-related EEG synchronization (i.e., stimulus-induced phase resetting of EEG activity).

Until very recently, it was difficult to reconcile findings on the single neuron level with observations on the system level made with functional neuroimaging or scalp-recorded electrophysiology in schizophrenia patients. By extension, one could not infer with certainty that a presumed prefrontal dopamine deficit and reduced neuronal signal-to-noise ratio postulated in schizophrenia was in any way related to “noisier” prefrontal hemodynamic responses or field potentials. The reason for this difficulty was that no correspondence could be demonstrated between single neuron activity (mean spike rate) on one hand and electrical field potentials or hemodynamic responses on the other. Rather, oscillating field potentials or event-related potentials—either measured from the scalp or intracortically—are considered to reflect postsynaptic events, produced by synchronous excitation of apical dendrites of pyramidal cells—a process that is in part under thalamic control (92–100). In a similar way, hemodynamic changes have been shown to be related predominantly to these postsynaptic processes rather than single neuron spiking, and both hemodynamic brain responses and electrical field potentials are correlated with each other (101, 102). However, recent work has revealed a relationship between single neuron spike rhythm and postsynaptically generated field potentials, which has been demonstrated for high-frequency gamma activity in the visual cortex (103, 104), beta and alpha activity in the motor cortex (105), and for slower frequency components in theta frequency range in the hippocampus (106). Comparable findings also have been reported by Lauritzen (102), who investigated the relation of cerebral blood flow (CBF) changes and the spike activity pattern of Purkinje cells. Event correlation analysis showed a random spike firing pattern under baseline conditions but rhythmic oscillations after electrical stimulation while the spike rate remained unchanged and CBF increased. These recent findings on the relation of single neuron spike activity and field potentials and hemodynamic responses could explain two seemingly divergent but likely synonymous effects of prefrontal dopamine: 1) increasing neuronal signal-to-noise ratio by shifting the neuron from a random to a rhythmic firing pattern in a selective frequency band; and 2) increasing and “focusing” the hemodynamic cortical response. However, so far, no study has directly addressed this question on the two-level effects of dopamine in cortical neuronal processing.

TABLE 1

Enlarge table

TABLE 2

Enlarge table