With regard to Professor Crow's stimulating commentary (Reference CrowCrow, 2000) on our recent paper (Reference Spence, Liddle and StefanSpence et al, 2000), we write to correct some errors of fact and interpretation which appeared in his article.

In our paper we demonstrated that during word generation people with schizophrenia exhibited a ‘functional disconnection’ between neural activity in dorsolateral prefrontal and anterior cingulate cortices relative to controls, as stated by Crow. People with schizophrenia also exhibited relative overactivity of the precuneus compared to controls and those at genetic risk (not only to controls, as stated by Crow).

Crow criticised our choice of an a priori hypothesis, based as it was on previous studies of word generation (summarised in Table 1 of our paper). We examined the hypothesis that focal or distributed brain dysfunction might provide a trait marker for schizophrenia, specifically implicating the left superior temporal gyrus, and a proposed frontotemporal ‘disconnection’ (Reference Friston, Herold, Fletcher and WatsonFriston et al, 1995). Crow suggested that we should have specified a dysfunction of right prefrontal cortex, despite the absence of such a finding from previous studies or statistical confirmation of such ‘dysfunction’ in our data-set. Indeed, despite many papers on the relevance of language to schizophrenia, Crow has never previously hypothesised such a specific region of dysfunction. In fact, although he has provided diagrams of how language ‘must’ be organised in the brain (e.g. Reference CrowCrow, 1998), it is notable that they exist in isolation from contemporary cognitive neurobiological accounts of the functional anatomy of language; and despite his emphasis on modelling the ‘first rank symptoms’ of schizophrenia, he has ignored those studies which have specifically addressed the neural correlates of these phenomena (e.g. Reference Spence, Brooks and HirschSpence et al, 1997).

Failure to address contemporary developments exposes Professor Crow's theories to the risk of appearing increasingly anachronistic.

When neuroimaging studies are published in psychiatric journals there is a particular responsibility incumbent on referees to be cautious in their interpretation of these data (Reference BrodieBrodie, 1996) and on authors to be rigorous in their application of statistics, lest a false impression be given to clinicians not used to examining such analyses (Reference SpenceSpence, 1999). Such responsibilities should also extend to those invited to comment on others' work.