Background. Schizophrenic patients generate fewer words than healthy controls during verbal fluency tasks. The structure of output may explain why patients generate fewer exemplars.

Methods. Twenty-four healthy controls and 24 patients with schizophrenia participated in six, 3 min semantic fluency tasks. In a subsequent session, participants were given cards, each printed with one of their own words generated from previous fluency tasks. Participants were to sort the cards into categories (e.g. subcategories of ‘animals’), thus defining their own semantic subcategories of words, and thereby eliminating experimenter assumptions about word relatedness. These clusters were matched with fluency output of each participant. The time spent searching through semantic networks within clusters and switching to other clusters when locating and producing associated words were measured.

Results. Patients produced fewer words and spent more time switching to words within clusters and to different clusters than controls, but otherwise response profiles were similar. Although controls returned more frequently to clusters and consequently made more switches between these clusters than patients, this group difference disappeared when the total number of words produced was covaried.

Conclusions. Consistent with previous literature, patients produced fewer words and made more errors than controls. The absence of a group difference in number of different clusters or mean number of items per cluster suggests that patients are similar to controls with respect to number of ideas in their semantic network. Patients' longer between-cluster switching times indicate a general slowness that may be attributed to difficulties finding new words within a semantic field.