Previous research has shown that older listeners have greater difficulty understanding fast speech than younger listeners. They may thus benefit from a reduced speech rate or additional pauses during fast speech, as this provides them with extra processing time. The present study aimed to examine how inserting pauses in linguistically appropriate positions, under varying speech rates, affects speech processing in normal-hearing older listeners. We examined this in terms of neural tracking of speech, specifically to understand how the underlying neural processing of speech relates to differences in perceived listening effort and speech comprehension performance.

To this end, electroencephalograms (EEG) were recorded from 18 native Korean older adults (mean age: 64.9), with normal hearing and cognitive abilities, while they listened to Korean sentences that varied in speech rate (artificially compressed vs. uncompressed) and pausing (with vs. without pauses). The rate for the fast condition was individually determined using an adaptive staircase procedure, to equalise speech comprehension accuracy across subjects. All naturally produced pauses were removed in the no-pause condition, whereas 500ms silences were inserted at prosodically appropriate positions in sentences used for the pause condition. After each sentence, subjects answered a comprehension question and rated their perceived listening effort. Neural tracking of the speech envelope was measured in EEG using multivariate Temporal Response Functions (mTRF). We also computed the predictability of each word in a sentence using a Korean large language model (Mi:dm 2.0) and examined how EEG responses varied as a function of the semantic predictability using mTRF.

The results found that speech rate did not affect speech comprehension accuracy in the main experiment, whereas subjective effort was higher when listening to fast speech than normal speech. In contrast, theta-band tracking was stronger at normal than fast rates. Pauses had no effect on any of the behavioural responses or neural tracking. When the degree of neural tracking was included in mixed-effects models predicting listening effort ratings, the results demonstrated that perceived effort increased as tracking increased, but only in the fast condition. In contrast, stronger tracking was significantly associated with poorer comprehension accuracy in the fast condition. These results suggest that speech tracking outside of the normal modulation rates was difficult for older listeners. However, speech tracking was enhanced when understanding fast speech was more difficult, likely because listeners expended greater listening effort. The semantic analysis is in progress to understand how lexical and semantic processing vary under these manipulations.