Functional Connectivity of Inner Speech in Schizophrenia

We explored fronto-temporal connectivity using fMRI of patients and controls generating inner speech. Method The subjects were eight schizophrenic patients in remission recruited from the wards and clinics of the Maudsley Hospital, and 8 controls matched for age, sex and IQ. Diagnosis was made by clinical interview and case-note review, using DSM IV criteria. All subjects were trained to overtly say the word 'rest' once per second and once every four seconds. When they had demonstrated stability of this, they were asked to do the same covertly. Before and after the scan, their silent speech rates were monitored by asking them to tap their finger at each covert speech act. They were then scanned for 5 minutes in a 1.5T GE Signa MRI while covertly repeating the word for alternating 30 second blocks at iHz and 0.25Hz in an AB design, without finger tapping. The desired rate was shown to subjects on a screen visible from the scanner. Image analysis was performed using XBAM. All individual maps were then entered into a conjunction analysis that selected only those clusters that were both significant, and whose significance was not due to the effects of the patient group or control group alone. Time series for these clusters were extracted at the individual level, and averaged over the sub-groups of patients and controls. Pearson correlations were calculated between averaged time-series for suprathreshold clusters in the areas of interest. Results The conjunction activation map showed greater left inferior and medial frontal cortex, left and right superior and middle temporal gyrus, at the faster rate, compared with the slower rate, of speech generation. The left inferior frontal lobe activation was significantly correlated with the left superior temporal lobe (0.264, p=0.008) and middle temporal lobe (0.315, p--0.001 ) in controls. In patients, the left inferior frontal lobe was more weakly correlated with the middle temporal lobe (0.268, p=0.007), but was not correlated with the superior temporal lobe when corrected for multiple comparisons. Discussion This method allows comparison of the same activation areas, selected from patients and controls as groups. Both groups showed widespread activation of the expected areas. Though frontal/superior temporal correlation was weak in the controls, it was not significant in the patients, supporting differential fronto-temporal disconnectivity in schizophrenia.