When top-down meets bottom-up: the interaction of semantic context and clear speech in real-time word segmentation

Real-time segmentation of continuous speech into words is essential for spoken language comprehension. Hyperarticulated clear speech is known to aid this segmentation process but how this benefit interacts with top-down semantic context remains unclear. We investigated this question using a visual-world eye-tracking experiment in which native English listeners heard conversationally and clearly produced sentences containing temporary ambiguity across a word boundary (e.g. She saw the ham starting … ), preceded by context sentences that were semantically more compatible with the intended target (e.g. ham) or a competitor (e.g. hamster). Eye fixation analyses showed that clear speech improved target-word segmentation only in contexts semantically incongruent with the target, but not in congruent contexts. Moreover, listeners’ fixations were less influenced by semantic context in clear speech than in conversational speech. These findings suggest that bottom-up, signal-dependent cues to word boundaries gain more weight under clear speech enhancement, reducing the influence of contextual-semantic information.