AbstractSubjects performed a nonword detection task, in which they responded to occasional nonwords embedded in a series of words. The stimuli were equally likely to be presented in the visual or auditory modality. Some of the words were repetitions of items that had occurred six items previously. Repetitions were in either the same or in the alternative modality.Compared to event-related potentials (ERPs) evoked by unrepeated, visually presented words, visual-visual repetitions gave rise to a sustained positive shift, which onset around 250 msec. Auditory-visual repetition also gave rise to a positive shift. This onset some 100 msec later than that associated with within-modality repetition. For auditory ERPs, within- and across-modality repetition gave rise to almost identical effects, consisting of a sustained positive-going shift onsetting around 400 msec.The findings were interpreted as reflecting the different representations generated by visually and auditorily presented words. Whereas visually presented words lead to the generation of both orthographic and phonological representations, auditory input leads solely to the generation of phonological representations.