Abstract
It has been suggested that lexical context seems to play a robust and direct role in phoneme restoration (more generally. perception) and the effect is considered to be a true feedback mechanism, whereas sentential context does not directly influence phoneme level processing and the effect is merely a post-perceptual decision bias. The experiment reported here examines how word-final ambiguous phonemes as artifacts of phonological processes such as consonant deletion and devoicing are restored by utilization of sentential context. The results show the possibility that sentential context can directly influence phoneme level processing. The data also provide evidence which supports the view that there is a level of phoneme representation in spoken word recognition as has been claimed in TRACE model, a connectionist model of speech perception and word recognition. I suggest that the nature of the role of context is not necessarily dichotomous between the purely true feedback and the post-perceptual decision bias, and is dependent on the degree of the constraint intrinsic to context.
