We previously demonstrated using PET in normal subjects (Demonet et ai, Brain 1992; 115: 1753–68) that, by comparison to a reference task of monitoring for pure tones, a phoneme monitoring task involving two factors of complexity (sequence processing and perceptual ambiguity) activated Wernicke's area and Broca's area. In the present experiment, we explored the respective influence of these factors on brain activation. In addition to the reference task, four phoneme monitoring tasks in non-words were used with stimuli presented binaurally. These included an easy, nonsequential, unambiguous task; a perceptually ambiguous, but non-sequential task; a sequential, but perceptually unambiguous task; a sequential and ambiguous task identical to the one we used in our previous study. Changes in regional cerebral blood flow were measured in nine right-handed volunteers using PET and infusion of H215O. The sequential unambiguous task was associated with the same pattern of activation as that observed for the equivalent non-sequential task, i.e. activation of Wernicke's area. The two tasks with perceptual ambiguity gave rise to activations in the left fusiform gyrus and in Broca's area, respectively. Activation of the left fusiform gyrus in the perceptually ambiguous but non-sequential task suggests a prominent role for a strategy involving phoneme-to-grapheme transcoding in subjects, who thereby attempted to visualize non-words that they could not decipher auditorily. Activation of Broca's area in the sequential and ambiguous task reproduces our previous result and suggests that subjects resorted to a predominantly verbal rehearsal strategy when performing this task.