Event-related brain potentials (ERPs) were recorded to terminal words of visually presented sentences that were identical to those used in an auditory modality study examining the effects of phonological and semantic characteristics of words on ERPs (Connolly and Phillips, 1994). Phonological and/or semantic characteristics of terminal words were manipulated in 4 conditions in which: (1) the terminal word was the highest cloze probability ending for the sentence and was thus phonologically and semantically appropriate to the sentence context (e.g., Ray fell down and skinned his knee(s).; (2) the terminal word had the initial phonemic sound of the highest cloze probability sentence-ending word but was semantically anomalous to the context of the sentence (e.g., They sat together without saying a single worm. [word]); (3) the initial phoneme of the terminal word was phonologically unexpected but the word was semantically correct (e.g., The dough was put in the hot pan. [oven]); or (4) the initial phoneme of the terminal word was phonologically unexpected and the word was semantically inappropriate to the sentence context (e.g., Bill jumped into the lake and made a big farm. [splash]). A left fronto-temporally distributed negative peak was observed in the 250-300 msec range to varying degrees in all conditions but was largest in the condition in which terminal words were both phonologically unexpected and semantically inappropriate. A second, parietally distributed, symmetrical negativity (the N400) was found which peaked around 365 msec and was largest in the two conditions involving semantically anomalous terminal words. Results are discussed in terms of modality factors in ERP manifestations of receptive language functions and the implications of ERP modality differences for theories of word recognition.