The difference between using context to predict versus using context to repair: a study of listening effort

Abstract

Semantic context in sentences has been shown to increase intelligibility and reduce listening effort. However, context has typically been tested in situations where it is used to predict upcoming words rather than repair earlier words. The current study attempts to directly compare predictive versus retroactive context for the first time for the same listeners in the same study. We hypothesized an increase in effort for retroactive context compared to the elevated effort of predictive context. Cochlear implant and Normal Hearing listeners heard sentences where a single content word was masked by noise, either early in the sentence (requiring retroactive repair) or later in the sentence (where it was predicted by earlier words). Changes in pupil dilation were measured as an index of listening effort linked with the moment of each word and its context. Effort increases when context is used to repair an earlier missing word, even if intelligibility is perfect. There is a smaller increase of effort when the missing word follows context that enabled prediction of the word. This study provides a novel approach in directly comparing the mental cost of using context to repair versus predict misperceived words in a sentence.