We read with considerable interest the paper by Shergill et al (Reference Shergill, Brammer and Amaro2004) about the temporal course of brain activity associated with auditory verbal hallucinations. The researchers used functional magnetic resonance imaging to reveal those brain regions activated before, during and after such hallucinations (the occurrence of which was indicated by patients pressing a button). They concluded that activation of the left inferior frontal gyrus some 9 seconds prior to button pressing supports the theory that hallucinations originate in brain areas involved in the generation of 'inner speech'. Given the importance of this question for future paradigm development, we wish to offer constructive comment.

There is a difficulty associated with the experimental method as described. Because no control condition was included (in which, for example, subjects might self-initiate button presses, unrelated to the timing of hallucinations) we cannot ascertain whether the frontal activation was attributable to the auditory verbal hallucinations or the procedure of button pressing itself; this problem emerged in the interpretation of an earlier, similar study (Reference McGuire, Shah and MurrayMcGuire et al, 1993; Reference Krams, Deiber and FrackowiakKrams et al, 1996). In healthy individuals we have observed that the left frontal cortex also activates 9 seconds prior to simple, self-initiated button pressing (Reference Hunter, Green and WilkinsonHunter et al, 2004). Obviously, in healthy individuals this has no relationship to auditory verbal hallucinations (it is a feature of the temporal evolution of normal voluntary motor behaviour). During such behaviour, maximal frontal activity is seen in the middle and inferior frontal gyri (9 s prior to button pressing). The temporal sequence of frontal activation observed by Shergill et al (Reference Shergill, Brammer and Amaro2004) could be related to the hallucinations or be attributed to the self-initiation of motor action (button pressing). This methodological consideration radically constrains the authors' conclusions. The techniques of functional neuroimaging are complex and unfamiliar to most general readers. We hope that the concern we raise is helpful in elucidating the methodological issues inherent in studies such as these.