Palaniyappan suggests using the Thought and Language Index (TLI) instead of the TLC in order to quantify formal thought disorder. The TLI, he argues, would have the advantage over the TLC of a standardised method of eliciting thought flow. Unfortunately, the TLI was established after the start of our study, ^{Reference Liddle, Ngan, Caissie, Anderson, Bates and Quested1} whereas the TLC was an established instrument that has been successfully used in numerous studies as a reliable instrument to quantify formal thought disorder. ^{Reference Andreasen and Grove2} We agree that future studies might benefit from the application of the more standardised TLI. However, the distribution of the severity of formal thought disorder in the patient group should not change substantially just by changing from the TLC to the TLI. Therefore, we regard our results as reliable.

Furthermore, Palaniyappan criticises the factor analysis of the PANSS in the small patient samples used in neuroimaging studies. He is right that a factor analytic approach in such small samples is critical if the patients group is randomly selected. In order to be able to study formal thought disorder with neuroimaging techniques, we recruited a specifically selected patient sample, which mainly differs in formal thought disorder severity and barely in other psychopathological categories. To obtain such a selected sample, the recruitment took years. Patients were matched as closely as possible for all other psychopathology traits in addition to the items of formal thought disorder of PANSS. The factor analysis and the correlation of the factors with the severity of formal thought disorder were only used to document this special patient selection. Therefore, it is no surprise that the factors of our factor analysis do not match the factors of unselected samples of patients with schizophrenia. ^{Reference Fitzgerald, de Castella, Brewer, Filia, Collins and Davey3} To study specific psychopathological phenomena like formal thought disorder by means of neuroimaging, such a specific selection of patients is necessary to extract reliable results. In our patient group we did not observe any significant correlation between formal thought disorder and (a) positive symptoms (without PANSS item P2): r = 0.39, P = 0.19; and (b) negative symptoms (without PANSS item N5): r = 0.04, P = 0.90. We therefore can conclude that our results are due to positive or negative symptoms in general.

Concerning the issue of the sample size in neuroimaging studies, our results were corrected for multiple comparisons. This approach is a common and accepted way to handle this problem in neuroimaging studies in general.