Despite converging agreement about the definition of persistent vegetative state, recent reports have raised concerns about the accuracy of diagnosis in some patients, and the extent to which, in a selection of cases, residual cognitive functions may remain undetected. Objective assessment of residual cognitive function can be extremely difficult as motor responses may be minimal, inconsistent, and difficult to document in many patients, or may be undetectable in others because no cognitive output is possible. Here we describe strategies for using H(2)(15)O positron emission tomography activation studies to study covert cognitive processing in patients with a clinical diagnosis of persistent vegetative state. Three cases are described in detail. Of these, two exhibited clear and predicted regional cerebral blood flow responses during well-documented activation paradigms (face recognition and speech perception) which have been shown to produce specific, robust and reproducible activation patterns in normal volunteers. Some months after scanning, both patients made a significant recovery. In a third case, blood flow data were acquired during a speech perception task, although methodological difficulties precluded any systematic interpretation of the results. In spite of the multiple logistic and procedural problems involved, these results have major clinical and scientific implications and provide a strong basis for the systematic study of possible residual cognitive function in patients diagnosed as being in a persistent vegetative state.