Semantic retrieval of spoken words with an obliterated initial phoneme in a sentence context.

Perception of a phoneme may occur even when the speech sound is missing (e.g., when an extraneous noise replaces the sound). This phenomenon, called phonemic restoration, has been observed to depend on the type of distortion. It requires a replacing sound that provides acoustic input to the auditory system, since the restoration has not been found when a speech sound was replaced by silence. We examined the brain activation underlying speech processing when the word's initial phoneme was completely replaced by a silent gap. Event-related potentials (ERPs) and reaction times (RTs) were measured as indicators of semantic processing of sentence final words. Slower voice onset times during repetition of the manipulated words as compared to normal words indicated increased difficulty in retrieving their meaning. The N400, which is related to the increased demands of the semantic integration of words, was elicited by less expected words as compared to highly expected ones. For manipulated words, the N400 was elicited at the same latency than for normal words, with respect to the onset of the remaining word fragment. The amplitude of the N400 was not increased, nor did it last longer, thereby indicating successful retrieval of the word's meaning based on the semantic context and remaining phonetic information. Thus, semantic retrieval does not seem to require the word's initial phoneme to be present in a sentence context. The results suggest that both context-driven expectancy (top-down) and stimulus-driven processes (bottom-up) are utilized in word processing and contribute to the overall N400 response.