An event-related fMRI study of overt and covert word stem completion

In fMRI studies of language processing, it would be extremely useful to obtain high-quality images during tasks requiring spoken output. Recent studies have suggested that this may be possible, particularly if event-related fMRI methods are used. This study assesses the feasibility of acquiring interpretable images during speech by applying event-related methods to visual word stem completion, a task that has been studied extensively. On each trial, a different three-letter word stem (e.g., COU) was presented visually and subjects were required to generate a word beginning with that stem (e.g., COUSIN). In covert runs, subjects were instructed to say the word once to themselves, without moving their lips. In overt runs, subjects were instructed to say the word once aloud. Ten subjects were scanned during six overt runs and six covert runs at three presentation rates. Data were analyzed using an implementation of the general linear model making no assumptions about response shape. Images were relatively free of artifacts, and regions demonstrating task-related activation were similar to those reported in previous imaging studies. Regions active during overt task performance were similar to those active during covert task performance, with the addition of several regions commonly associated with motor aspects of speech production. Consistent with other studies, magnitude of activation was greater in the overt condition than in the covert condition, and there was a modest decrease in magnitude at the fastest presentation rate. Together, these results help to validate the use of event-related fMRI during tasks that require spoken output.