Brain Activation Patterns in Schizophrenic and Comparison Volunteers During a Matched-Performance Auditory Recognition Task

The comparison group consisted of 12 healthy, right-handed volunteers recruited from the community through newspaper advertisements and screened by telephone for good medical health and the lack of any major psychiatric illness. Further screening involved the administration of the Structured Clinical Interview for DSM-III-R—Non-Patient Edition and the Structured Clinical Interview for DSM-III-R Personality Disorders to rule out current or past history of axis I or axis II disorders, respectively. Family history of schizophrenia in any first-degree relative was ruled out by interview. Six comparison group subjects were male, and six were female. Their mean age was 28 years (SD=4.6).

Eighteen medically healthy schizophrenic inpatients were recruited from the Residential Research Unit of the Maryland Psychiatric Research Center in Baltimore. Each patient was administered the Structured Clinical Interview for DSM-III-R at hospital admission. Two research psychiatrists reached a consensus diagnosis of DSM-III-R schizophrenia, which was based on the clinical interview plus all other sources of data. All but one of the patients were right-handed. Fifteen were male, and three were female. Their mean age was 31 years (SD=7.7). The socioeconomic status of their family of origin was 2.7 (SD=1.1). The average age at onset was 21.0 years (SD=5.5), and the mean duration of their illness was 9.5 years (SD=8.2). The average duration of antipsychotic treatment was 6.4 years (SD=5.2).

The schizophrenic patient volunteers, all inpatients, were treated with a fixed dose (0.3 mg/kg) of haloperidol and no other neurologically active medications for at least 4 weeks before the start of the study. The haloperidol treatment was then entirely withdrawn (mean=31 days, SD=17, range=6–61) before the PET scan. Symptoms of schizophrenia were assessed repeatedly at baseline, during haloperidol treatment, and during the subsequent drug-free period. Brief Psychiatric Rating Scale (BPRS) (20) symptoms were rated by two trained raters. At the time of the brain scan studies, the schizophrenic patient volunteers had an average total BPRS score of 32 (SD=9.7). Patients tolerated medication withdrawal without evidence of serious clinical relapse.

All subjects provided written informed consent. The research protocols were reviewed and approved by clinical investigation committees at the University of Maryland School of Medicine and the Johns Hopkins Medical Institutes.

The decision condition was a forced-choice auditory recognition task. Subjects were instructed to press a button held in their right hand when they recognized a high-frequency reference tone (always 1500 Hz) and to press the button in their left hand when they recognized that the tone was of lower frequency (800–1492 Hz). Subjects were told to push the button as soon as they heard the tone. An equal number of high and low tones were randomly presented in each stimulus condition; there were 100 trials per training set and 60 trials per imaging set. Tones lasted 100 msec. Following termination of the signal, the subject had 1.9 seconds to respond before the next trial was presented; the intertrial interval of 2 seconds (from onset of the first tone to onset of the subsequent tone) was fixed. Failure to respond was scored as an error. Response time was calculated from the onset of the stimulus to when the button was pressed.

During the active control task (sensory motor control condition), subjects heard the same sets of tones that were presented during the decision task condition. However, instead of identifying the tones as being of high or lower frequency, they were instructed to simply alternate pressing the right and left hand buttons in synchrony with the tone presentation. During the rest condition, subjects were instructed to lie still and remain quiet. Eyes were open and earphones in place during all scans.

Subjects began training on the decision task 1–5 weeks before the PET scan study. Training occurred in a quiet, isolated room. The purposes of the training were to eliminate stimulus/task novelty, determine for each subject the frequency difference that reliably provided an 80% accuracy performance, and to reduce performance variability within and between subjects. The patients with schizophrenia began training during the haloperidol phase and continued throughout the medication withdrawal period. On average, the comparison subjects participated in 22.1 (SD=7.1) blocks of tone-recognition trials (each block contained 100 trials). The patients with schizophrenia, on average, participated in 29.4 (SD=11.9) blocks of tone trials. The difference was not significant.

During the training, subjects became increasingly accurate in recognizing the two types of stimuli, so we systematically modified the tones presented by decreasing the frequency difference between reference and test tones. In the first training session, the differences in frequency between the reference and test tones ranged from 20 to 700 Hz. If a subject was unable to perform at an accuracy rate of at least 80% at the largest frequency difference (700 Hz) after three training attempts, the subject was dropped from further study. No healthy volunteers were dropped from the study because of performance inadequacies, but one was dropped because of an inability to complete the training schedule. Out of 28 that were tested, 10 of the schizophrenic patient volunteers were unable to participate in the study. Four had begun the training procedure but were not able to continue because of their inability to participate in the full set of training sessions. Six patients attempted to perform the task but were unable to sustain the effort sufficiently well to pass beyond the initial training session.

PET scans were obtained by using the General Electric 4096+ system, which produces 15 brain image slices at an intrinsic resolution of 6.1 mm in each dimension. The bolus [¹⁵O]H₂O method (21) was used without arterial blood sampling. Approximately 62 mCi of [¹⁵O]H₂O were administered on each occasion. Ten to 12 minutes elapsed between scans. The rCBF distribution was measured in each of the three conditions (rest, sensory motor control, and decision) with four scans per condition (12 scans per subject). The order of the scans (rest, sensory motor control, decision) was fixed for all subjects and was consecutively repeated four times. During the rest condition, the subject was inactive, eyes were open, ears were unplugged, and tone stimuli and motor activity were absent, but the earphones were in place. The sensory motor control condition consisted of the subject listening to the same tones presented during the decision condition and alternately pressing the right and left buttons in synchrony with the tones. Subjects were asked to respond as soon as they heard the tone. The decision condition consisted of the subject pressing the right button when the high frequency reference tone (1500 Hz) was recognized and the left button when a lower frequency tone was detected. Within a given scan, the frequencies of the two tones were held constant. The frequencies fell within the range in which that individual performed at an 80% accuracy level during the training session. Sixty trials were delivered during each scan; tones began 15 seconds before tracer delivery, and scan acquisitions began 20 seconds after dose delivery. PET data were acquired for 60 seconds.

The quantitative PET blood flow images were analyzed with statistical parametric mapping (22). The scans from each subject were realigned by using the first as a reference. To optimize the spatial correlation between data sets, the alignment procedure was modified to use 10 iterations (the default procedure uses three iterations). After realignment, all images were transformed into a standard anatomical space (23). Before generating the statistical parametric map, the data were smoothed by using a 12-mm gaussian kernel. Only clusters that contained voxels that were connected (face or edge) and above a threshold (z=2.33, p=0.01, one-tailed, uncorrected for multiple comparisons) were tested by means of spatial extent statistical theory for significance, which was set at an alpha level of <0.05 after correcting for multiple comparisons (22). The full width at half maximum of the average noise subtraction image was used to determine the number of resels (24) available for analysis. The patients with schizophrenia and the comparison subjects were compared for changes in rCBF following shifts in auditory recognition task conditions: sensory motor control versus rest and sensory motor control versus decision. Increases and decreases were assessed. The spatial/anatomical distribution and extent of each cluster were mapped onto matching magnetic resonance images aligned and warped to “fit” statistical data.

Correlation maps to illustrate the linear relationships between response time and regional blood flow were calculated by using the response times of correct choices made during decision scans. Subjects scored about 70%–80% accuracy on each decision block of 60 trials; this provided 42–48 trials, on average, for each block for each subject performing each decision PET scan. With these rCBF-versus-response-time correlation maps, we then contrasted groups voxel-by-voxel by transforming correlation maps to z maps. Difference maps were generated showing where the two groups (comparison subjects and the patients with schizophrenia) significantly differed in their rCBF and response time correlations. The statistical significance of clusters was assessed by spatial extent criteria.

Behavioral outcome measures were extracted from all trials generated in sensory motor control and decision PET scan sessions. These measures are mean accuracy and median response time for accurate trials. The median was used to represent the response time in order to minimize the effect of outliers. Accurate trials were used because inaccurate trials included trials in which no response was made.

For the tone recognition task (decision condition), the patients with schizophrenia provided accuracy scores and response times for accurate trials that were similar to those of the comparison subjects (schizophrenic patients: accuracy=75.6% [SD=10.0], response time: 788.0 msec [SD=169.4]; comparison subjects: accuracy=80.6% [SD=4.7], response time: 735.0 msec [SD=102.0]). However, the frequency disparity (i.e., the difference in frequency required to elicit 80% accuracy in recognizing whether a given tone was lower than the 1500 Hz reference tone) was significantly higher for the patients with schizophrenia (mean=195.8 Hz, SD=265.0) than for the comparison subjects (mean=17.2 Hz, SD=8.3) (t=2.32, df=28, p=0.03, two-tailed with unequal variances). There was no significant correlation between clinical ratings and performance on the tone-recognition task, and variations in medication withdrawal times did not contribute to task performance in any systematic way.

The auditory recognition tasks produced increasing demands on the subjects’ attention and memory resources. Subtraction analyses for the rest and sensory motor control conditions reveal the shift from an inactive state to a simple repetitive motor condition requiring minimal mnemonic demands. Subtraction analyses for the decision and sensory motor control conditions provide brain activity maps of the switch to an active state requiring a high degree of attention and memory from one that only required repetitive motor behavior and a fixed sensory state.

The extremely large frequency range (20 to 700 Hz) required to detect differences in tones with 80% accuracy for the 18 patients with schizophrenia prompted post hoc evaluations of potential behavioral and physiological subgroup patterns. One subgroup of patients with schizophrenia (N=12) were able to score 80% accuracy with tones that fell in the same range (20–100 Hz) as the comparison subjects during the initial testing (mean=36.0 Hz, SD=21.2); these subjects were classified as having “normal frequency disparity.” A second patient group (N=6) required differences greater than 100 Hz to score 80% accuracy (mean=515.3 Hz, SD=232.6) and were classified as having “abnormal frequency disparity.” A one-way analysis of variance used to test for group differences was significant (F=56.14, df=2, 27, p=0.0001). Tukey’s post hoc test showed differences between patients with normal and abnormal frequency disparity at an alpha level of 0.05. The accuracy scores and response times for accurate trials were similar for patients with normal and abnormal frequency disparity (normal disparity: accuracy=78.4% [SD=4.7], response time: 778.4 msec [SD=191.1]; abnormal disparity: accuracy=70.0% [SD=15.5], response time: 807.1 msec [SD=128.7]).

Anterior cingulate cortex rCBF activity was equivalent in the comparison subjects and schizophrenic patients with normal frequency disparity during the sensory motor control condition, but it changed differentially during the decision condition (Table 3). rCBF rose in the comparison subjects, fell in the schizophrenic patients with abnormal frequency disparity, and exhibited a small positive change in schizophrenic patients with normal frequency disparity. The decision minus sensory motor control analysis showed that the cingulate response of the schizophrenic patients with abnormal frequency disparity was significantly lower than that of both the comparison subjects (z score of maximum value voxel within cluster=5.47, p<0.0001, by spatial extent) and the schizophrenic patients with normal frequency disparity (z score of maximum value voxel within cluster=4.32, p<0.05, by spatial extent). In contrast, the schizophrenic patients with normal frequency disparity did not differ significantly from the comparison subjects in the cingulate or any other region.

Figure 4 vividly illustrates this pattern by plotting repeat fluctuations in rCBF at an area in the cingulate cortex chosen to represent the optimal cingulate change scores of all three groups. The comparison subjects exhibited an abrupt positive rCBF change when switching from the sensory motor control to the decision task. The rCBF of schizophrenic patients with normal frequency disparity tended to rise from rest to sensory motor control and continue rising during decision. Schizophrenic patients with abnormal frequency disparity exhibited no consistent cingulate change pattern.

Anterior cingulate rCBF differences between comparison subjects and schizophrenic patients with abnormal frequency disparity were expected because of the initial finding between the comparison subjects and the entire group of patients with schizophrenia. Statistical parametric mapping (22) was used to analyze the entire cortex of these two groups (Table 4 and Figure 4). In addition to the anterior cingulate, the right premotor cortex and left motor cortex also exhibited high, significant change scores. No differences were found between the comparison subjects and the schizophrenic patients with normal frequency disparity in the between-condition contrasts.

rCBF activation changes in schizophrenic patients with normal frequency disparity were similar to those of the comparison subjects; accuracy performance and stimuli characteristics were matched across groups. The changes that accompanied the switch from sensory motor control to decision were similar in the two groups. Within the comparison group, clear and physiologically meaningful relationships appeared between the average response time associated with accurate tone recognitions and the blood flow to a particular brain region (25). These response time-rCBF correlation maps were generated for the comparison subjects and the schizophrenic patients with normal frequency disparity by using blood flow and behavioral data acquired during decision condition PET scans. The correlation maps were then contrasted between groups. Local maxima were then extracted and plotted to inspect the distributions around the regression functions. On the basis of previous analyses with healthy comparison subjects, the right frontal cortex and right auditory cortex were expected to generate the largest between-group contrasts. The comparison subjects exhibited strong negative correlations between response time and rCBF in auditory regions (i.e., short response times were associated with maximal blood flow rates). Robust positive correlations between response time and rCBF were found in the frontal cortex (i.e., long response times were associated with maximal flow rates).

Table 5 and Figure 6 illustrate how the correlations between response time and rCBF differed between the schizophrenic patients with normal frequency disparity and the comparison subjects even though their brain activation patterns were similar. The difference maps illustrate where the two groups exhibited opposite correlations between response time and rCBF. Significant differences were associated with regions in which each group exhibited a significant, but opposite, correlation between average decision response time and rCBF measured during a particular set of decisions. Each subject contributed four decision scans to the analysis.

The comparison subjects exhibited progressively greater flow rates in the frontal cortex the longer the subject took to make an accurate response. In marked contrast, the longer the schizophrenic patients with normal frequency disparity waited to make an accurate response, the lower the flow rate to the frontal cortex. The comparison subjects displayed maximum flow rates to the right auditory cortex in association with short response times; the schizophrenic patients with normal frequency disparity exhibited the opposite pattern in this region.

Most patients were able to distinguish and recognize tones that differed from one another by small frequency levels (about 2%). However, one-third of the schizophrenic patients required tone frequency differences of 10%–50% in order to perform the task at accuracy levels above chance (greater than 65%). This finding is consistent with studies reported by Javitt and colleagues (16). It is important to note that the schizophrenic patients with abnormal frequency disparity performed this task at accuracy levels and response times close to those found in the comparison subjects and the schizophrenic patients with normal frequency disparity. Frequency differences seem to be especially important vulnerabilities for patients with schizophrenia (16, 26). The inability to quickly and accurately ascertain one frequency from another may contribute to speech and language perceptual difficulties commonly ascribed to some schizophrenic patients. In particular, the inability to “understand” the pitch fluctuations of words may contribute to a patient’s social mistakes and contribute to his or her social isolation. If a patient cannot easily detect and track changes in the pitch of a word, he or she may be severely disadvantaged in the ability to understand how the word is being used in a particular social setting. Similarly, patients may exhibit a more restricted affective range in their speech if they are unable to “hear” subtle pitch modulations.

When attending to tones, but not performing a difficult recognition task, patients with schizophrenia activated their auditory cortices, right and left, similar to the comparison subjects. These findings suggest that patients are able to make simple auditory perceptions by using a neural substrate that is not significantly different from that of comparison subjects. Javitt et al. (16) and Holcomb et al. (26) have reported behavioral studies emphasizing the vulnerability to auditory memory of patients with schizophrenia. Shenton et al. (27) and Menon et al. (28) have described strong correlations between psychotic behavior and auditory cortex structures in patients with schizophrenia. The findings reported here suggest the deficits and abnormalities of patients with schizophrenia may not reside in a primary sensory processing site; they may instead be more closely linked to dynamic relationships within the association and heteromodal areas.

The contrast between the sensory motor control and rest conditions in the patients with schizophrenia revealed a striking change in anterior cingulate/supplementary motor area activity. Figure 1 and Figure 3 demonstrate a high level of activity change in conjunction with the simple, repetitive task of pressing alternate buttons that was specific to the motor control condition. The subjects’ rapid responses were consistent with their practiced condition. Work by Carter et al. (29) suggested that cingulate activity is greatest when a subject is performing a task associated with significant error likelihood. However, patients with schizophrenia may not follow this “error context” rule, at least not as it is understood in the normal population. Patients with schizophrenia may be abnormally sensitive to “set maintenance.” When any repetitive task involving side-to-side alternation is engaged, this group may need to use maximum frontal, cingulate, and supplementary motor area activity. This may reflect the group’s difficulty in assessing error likelihood or suppressing distractions. The patients with schizophrenia may “expect” or anticipate greater error than is actually likely to occur, and they may not be able to modulate activity with respect to the likelihood of error. Low-error likelihoods may be treated by the frontocingulate system the same as high-error conditions in volunteer patients with schizophrenia. This may partially explain why no additional change was found in the cingulate cortex/supplementary motor area region when volunteer patients with schizophrenia shifted from the sensory motor control to the decision condition. That behavioral switch did, however, result in a marked activation in the right frontal cortex of patients with schizophrenia. The cingulate/supplementary motor area system may be working optimally in schizophrenic patient volunteers when they are confronted with simple tasks associated with alternate choices; this system may, however, contain fundamental flaws in its dynamic relationships with the frontal lobe (3, 30–34).

Increased frontal rCBF found in the schizophrenic patients and the comparison subjects with the decision minus sensory motor control analysis suggest that the additional demands characteristic of the decision condition are partly represented and mediated by frontal lobe activity. Decision trials are slower and associated with higher error rates than sensory motor control trials. It is not possible to listen to a tone trial in isolation and accurately identify it as high or low frequency. Thus, the decision condition is sufficiently difficult to require a participant to keep in mind what the previous stimuli sounded like. The schizophrenic patient volunteers were probably using frontal regions similar to the comparison subjects for this task. By supplying information about recent events and making rapid comparisons, the frontal regions facilitate subtle perceptual judgments. Over brief time intervals, the frontal lobes of patients with schizophrenia and comparison subjects may function in similar ways. This interval, however, may last no more than about 600 msec; beyond that time the two groups may use substantially different strategies for problem solving.

The schizophrenic patients with abnormal frequency disparity exhibited no significant cingulate/supplementary motor area response to changing behavioral demands (Figure 4). Unable to perform tone recognition at low-disparity frequencies, this group may have perceptual abnormalities that are due to severely impaired attention and memory resources. This subgroup of patients with schizophrenia exhibited significant differences from the healthy comparison subjects and schizophrenic patients with normal frequency disparity in motor and premotor regions; this suggests that the circuits associated with stimulus choice, impulse inhibition, and set maintenance are potentially defective. Not clinically distinct, the six schizophrenic patient volunteers with abnormal frequency disparity were not remarkable with respect to symptoms or course of illness. Furthermore, this group of six volunteers showed no capacity to improve in their auditory memory performance in spite of extensive training; the other 12 schizophrenic patient volunteers significantly improved in tone recognition over the training period (31). Unable to use the frontal, cingulate, and supplementary motor area cortices to manipulate incoming information quickly, this subgroup may have relied more heavily on secondary auditory cortex and motor regions for sensory integration and recent stimulus representation. This group’s largest change clusters (seen in the sensory motor control minus rest analyses) were found in the left auditory and motor/hand regions, not in the association cortex. Although distinctly different from the schizophrenic patients with normal frequency disparity in this study, this group may appear more continuous with the other patients with schizophrenia in tasks requiring a different set of cognitive resources.

The blood flow contrast patterns (decision minus sensory motor control) between the comparison subjects and schizophrenic patients with normal frequency disparity were not significantly different in the frontal lobes. Yet, it is widely appreciated that schizophrenic patient volunteers perform particularly poorly on working memory and long-term memory tasks, which are highly dependent on frontal lobe function. The comparison subjects exhibited a marked tendency to have increased rCBF in the frontal lobes that was associated with progressively longer response times. During the first 300–500 msec of a trial, the comparison subjects and patients with schizophrenia may have relied primarily on the auditory cortex for making a match between the percept and its correct representation. Behavioral data (15, 25) suggest that healthy subjects become less and less accurate the longer they suppress a response. Responses made 1500 msec after stimulus presentation are far less likely to be accurate than those made at 400 msec. Response inhibition may be associated with a search for the correct match in the association cortex. Then, the longer the response is suppressed, the noisier becomes the percept, and the less accurate becomes the decision. The response of the comparison subjects to prolonged (greater than 1,000 msec) response inhibition was to facilitate greater activation of the frontal lobe; the response to prolonged response inhibition of patients with schizophrenia was to enhance primary auditory cortex rCBF. Lesion studies (18, 19) have suggested that frontal lobe pathology, especially the right side, strongly compromises auditory memory. When patients with schizophrenia fail to execute a decision after 1 second they may not continue to search for the correct match in the frontal association cortex; they may be required to rely on the primary auditory cortex, a site less equipped to provide accurate long-delay matches. Although attentive to the task, the group of patients with schizophrenia may be unable to engage the optimal search strategy. Attention is not sufficient for successful recognition when searching for ambiguous signals.

This model is consistent with work by investigators who have studied patients with schizophrenia making memory decisions (32). When asked to make automatic decisions, the group of patients with schizophrenia often appeared similar to the comparison subjects. When, however, a time delay is imposed (e.g., 2-Back Working Memory Task) that exceeds 1 second, then marked differences appear between groups physiologically and behaviorally. Blood flow apparently diminishes in frontal lobes of patients with schizophrenia attempting to make long memory decisions; their accuracy also tends to drop precipitously (16). These studies support a strong time component for the frontal lobe failure of patients with schizophrenia. As long as the schizophrenic patient volunteer is able to make most of his or her decisions quickly (e.g., under 800 msec), frontal lobe activity is likely to appear similar to that of comparison subjects. When, however, patients with schizophrenia are either compelled to wait, or choose to wait, for over 1,200 msec, then they are likely to exhibit marked differences in frontal flow from that of comparison subjects.

When the identity of a stimulus becomes extremely ambiguous, schizophrenic patient volunteers became markedly impaired in their recognition abilities. This may be partly due to the increased time needed to search out and confirm an accurate match. The longer the search is maintained, the less available is the frontal lobe. Consequently, the group of patients with schizophrenia will exhibit progressive failure with greater time delays whether the information is either 1) unambiguous and the motor response delay is externally enforced or 2) it requires an extended internal search and greater time delays because of the subject’s uncertainty. In either case, schizophrenic patient volunteers appear unable to use the frontal lobe to sustain a search over a prolonged time period. The cingulate cortex may play an important role in modulating the frontal lobe’s activities (30). If the cingulate cannot shift the frontal lobe’s state from one level of activity to another, then the frontal lobe may prematurely terminate or diminish its probe for the correct answer (31, 33, 34).

Future studies that use single-event functional MRI methods may be better suited to untangle these relationships. Schizophrenic patient volunteers engaged in a task similar to one performed by comparison subjects exhibited similar activity change patterns. They did, however, exhibit exactly opposite blood flow/response time correlations in sensory cortex and frontal lobes. Functional MRI may be able to assess to what extent the cingulate/supplementary motor area region becomes “disconnected” from the frontal lobe during responses requiring particularly long time intervals.