The temporal dynamics of evoked brain responses are normally characterized using electrophysiological techniques but the positron emission tomography study presented here revealed a temporal aspect of reading by correlating the duration a word remained in the visual ¢eld with evoked haemodynamic response. Three distinct types of e¡ects were observed: in visual processing areas, there were linear increases in activity with duration suggesting that visual processing endures throughout the time the stimulus remains in the visual ¢eld. In right hemisphere areas, there were monotonic decreases in activity with increased duration which we relate to decreased attention for longer stimulus durations. In left hemisphere word processing areas there were inverted U-shaped dependencies between activity and word duration indicating that, after 400^600 ms, activity in word processing areas is progressively reduced if the word remains in the visual ¢eld. We conclude that these inverted U e¡ects in left hemisphere language areas re£ect the temporal dynamics of visual word processing and we highlight the implication of these e¡ects for the design of activation studies involving reading.
I N T RO DUC T IONThe aim of this work was to characterize the haemodynamic responses evoked by visually presented words as a function of the duration they remain in the visual ¢eld. These e¡ects of stimulus duration are distinct from those of presentation rate (frequency) where the number of words presented per unit time is changed but the duration of each word is held constant. Previous studies manipulating stimulus duration on words and falsefonts (Price et al. 1994(Price et al. , 1996a have demonstrated that when stimuli remain in the visual ¢eld for 1000 ms relative to 150 ms, there is enhanced activation in visual processing areas and, for words only, there is decreased activation in temporal and frontal word processing areas. The contrasting e¡ects of stimulus duration suggest functional specialization in the di¡ering cortical regions. Further, the counter-intuitive observation that responses in temporal and frontal word processing regions are greater when word presentation is brief (150 ms) relative to when the same words remain in the visual ¢eld for longer periods of time (1000 ms), may be indicative of the time course of word processing. This hypothesis motivated a more detailed investigation of the relationship between evoked activity and the length of time a word remains in the visual ¢eld.Three di¡erent explanations for the decreased activity in word processing regions with longer word durations have been considered. One specious argument is that there are additional e¡ects with short durations which re£ect cognitive processes other than those related to word processing. Such`irrelevant processing' would occur during the longer interstimulus intervals associated with short relative to long durations (the rate of presentation being constant). This explanation can be discounted because it predicts similar e¡ects when the interstimulus...