Effects of Hearing Loss on Semantic Prediction: Delayed Prediction for Intelligible Speech When Listening Is Demanding.

Abstract

Objectives: Linguistic context can be used during speech listening to predict what a talker will say next. These predictions may be particularly useful in adverse listening conditions, since they can facilitate speech processing. In this study, we investigated the impact of postlingual hearing loss on prediction processes. Because hearing loss leads to a perceptual deficit (i.e., degraded auditory input), that can also have cognitive impacts (i.e., increased competition for cognitive resources due to increased listening effort), it is a naturalistic test case of how different sorts of challenge affect prediction.

Design: We report a visual world eye-tracking study run with 3 participant groups: older adults (range: 53 to 80 years old) with normal hearing (n = 30), older adults with hearing loss listening under low demand (n = 32), and older adults with hearing loss listening under high demand (n = 31). Using highly semantically constraining predictable sentences, we analyzed the timecourse of simple associative predictions based on the agent of the sentence (sub-experiment 1), and the timecourse by which these predictions were narrowed with additional constraint provided by the verb (sub-experiment 2).

Results: Although there was no effect of group on early agent-based predictions, we saw that the buildup and tailoring of verb-based prediction was delayed with hearing loss and exacerbated by listening demand. As there was no comparable group difference for semantically unconstraining neutral sentences, this cannot be explained as a result of delayed lexical access in the hearing loss groups. We also assessed the cost of incorrect predictions but did not see any group differences.

Conclusion: These findings indicate two separable stages of prediction that are differently affected by hearing loss and listening demand (potentially due to changes in listening effort), and reveal delayed prediction as a cognitive impact of hearing loss that could compound simple audibility effects.