Factors in sensory processing of prosody in schizotypal personality disorder: an fMRI experiment

Abstract

Introduction: Persons diagnosed with schizophrenia demonstrate deficits in prosody recognition. To examine prosody along the schizophrenia spectrum, antipsychotic-naïve schizotypal personality disorder (SPD) subjects and healthy control subjects were compared. It was hypothesized that SPD subjects would perform more poorly; with cognitive and demographic factors contributing to the poor performance. The superior temporal gyrus (STG) was selected as the region-of-interest (ROI) given its known abnormalities in SPD and its important role in the processing of prosody.

Methods: SPD and healthy comparison (HC) subjects were matched on age, IQ, and parental social-economic status (PSES). Cognitive measures included the Speech Sound Perception Test (SSPT) to examine phonological processing (SPD=68, HC=74) and the Verbal Fluency task to examine executive functioning (SPD=129, HC=138). The main experiment was a novel fMRI task of prosody identification using semantically neutral sentences spoken with emotional prosody (SPD=16, HC=13). Finally, volumetric measurement of the superior temporal sulcus (STS), a key region for processing prosody, and partially overlapping with the STG, was performed (SPD=30, HC=30).

Results: Phonological processing and executive functioning were both impaired in SPD subjects compared with HC subjects. Contrary to the prediction, SPD subjects, as a group, were similar to HC subjects in terms of correctly indentifying the emotion conveyed and reaction time. Within the SPD group, prosody identification accuracy was influenced by executive functioning, IQ and perhaps PSES, relationships not found with HC subjects. Phonological perception aided prosody identification in both diagnostic groups. As expected, both groups activated the STG while performing the prosody identification task. However, SPD subjects may have been less "efficient" in their recruitment of STG neurons. Finally, SPD subjects demonstrated a trend toward smaller STS volumes on the left, particularly the lower bank.

Conclusions: These data suggest that subtle differences between SPD and controls in phonological processing, executive functioning, IQ, and possibly PSES, contributed to difficulty in processing prosody for some SPD subjects.

Figure 1. fMRI Experimental Design

The onset vector begins with the first phoneme of the sentence (arrow) and the epoch continues to the onset of the next sentence in the series. SOA = stimulus onset asynchrony.

Figure 2. Manual Tracing of STS

(a). Vertical lines on the sagittal image demarcate the anterior (green) and posterior (magenta) boundaries of STS drawing. (b). The coronal image shows the segmented right STS upper bank (salmon) and the lower bank (yellow).

Figure 3. Speech Sound Perception Test (SSPT)

SPD subjects demonstrated deficits in phonemic processing when the two subjects who were outliers on this measure were removed from the analysis (one SPD and one HC, data points circled). When their data was included, the difference in performance was statistically significant at the trend level (F(1,139) = 2.930, p < 0.09). Bars and numbers represent mean values.

Figure 4. Verbal Fluency

(a). Semantic Fluency. (b). Phonemic Fluency. SPD subjects have executive functioning deficits as measured by the Verbal Fluency task in both the semantic and phonemic domains. Bars and numbers represent mean values.

Figure 5. fMRI Behavioral Measure of Number Correctly Identified Prosodic Emotions

(a). Total number of sentences correctly identified. There was no difference between groups on this measure. Some SPD subjects, however, appeared to have difficulty. To examine those subjects more closely, both groups were examined with split-half comparisons. The mean number of correctly identified sentences for the HC subjects (154.8) was used to separate each diagnostic group into two groups, higher performers and lower performers. (b). Number of correctly identified sentences by prosodic emotion condition. There were 50 sentences per condition. There were no statistically significant differences between groups although there was a trend toward SPD subjects having more difficulty with “sad”.

Figure 6. Correlations among Cognitive and Behavioral Measures: Factors Influencing Prosody Processing

(a). Role of Phonological Processing. Phonological processing was significantly correlated with prosody processing for both SPD and HC subjects. (b). Role of Executive Functioning. Executive function significantly affected SPD subjects only. (c). Role of IQ. IQ significantly affected SPD subjects only. (d). Role of Social Class of Origin. There was a positive correlation for SPD subjects but a negative correlation for HC subjects. Neither correlation in this case reached the significance level, but there was a statistically significant difference in the correlation coefficient between groups.

Figure 7. fMRI Parametric Maps for ROI Analysis across All Conditions, Threshold = 0.003

Using the one-sample t test both groups demonstrated activation in the (a) left STG and (b) right STG. When the two groups were directly compared with a two-sample t test there were areas of both the left and the right STG seen in the HC > SPD analysis (c). However, with the two-sample t test there were no super-threshold voxels in the SPD>HC comparison (d). MNI x, y, z coordinates given for voxel of peak magnitude activation as well as color bar for t score significance.

Figure 8. Correlations between the Extent of Activation in STG and Behavioral Measure of Number Correct Across All Conditions

For the healthy control subjects there was a tight coupling between ROI activation bilaterally and accuracy of prosodic emotion identification (orange graphs). SPD subjects did not display such tight coupling (blue graphs). Top: right STG data. Bottom: left STG data.

Figure 9. fMRI Parametric Maps for Exploratory ROI Analysis for Each Separate Emotion, Threshold = 0.003

Using one-sample t tests, the left (left-sided images) and the right STG (right-sided images) parametric maps are displayed for each emotion separately for HC and SPD subjects. Note that these analyses are post-hoc as there were no emotion-specific hypotheses. MNI x, y, z coordinates given as well as color bar for t score significance.

Figure 10. fMRI Parametric Maps for Secondary, Exploratory Whole Brain Analysis across All Conditions, Threshold 0.001

These whole brain parametric maps were produced using one sample t tests and are included for further hypothesis generation. The results suggest that SPD subjects have large hemodynamic responses in frontal regions while the HC have little. The areas in red are those where the significance threshold of 0.001 was exceeded. Specific regions are noted in Table 2.