Neural Mechanisms Underlying Linguistic Context Use for Speech Processing in Aging - Project Summary/Abstract A frequent complaint in older adults is difficulty with speech comprehension, especially in the presence of background noise, such as when at a busy restaurant. A common strategy that listeners use in such situations is to use the linguistic context to fill in gaps in comprehension. Older adults tend to do well with this, at least in laboratory situations: given a meaningful linguistic context, their word recognition is much more like young adults’ than when speech is meaningless. An open question is what cognitive mechanisms underly this ability. Based on research in young adults, it is widely believed that prediction plays a central role in language processing – listeners could predict upcoming words from a meaningful context, and thus compensate for noisy input. Yet, a line of research using an electroencephalography (EEG) signature of predictive language processing, the N400, suggests that predictive processing is impaired in older adults. This raises the question of how, then, older adults can successfully use context to facilitate speech recognition. This proposal uses cutting-edge neuroimaging techniques to examine a wider range of predictive speech processing than previous work, as well as non- predictive compensatory mechanisms, in younger and older adults. To maximize relevance to older adults’ real- world experience, brain responses are measured while listening to continuous narrative speech, in quiet and in multi-talker babble background noise, approximating conditions in a busy restaurant. EEG is used to model the temporal dynamics of predictive language processing at multiple levels (acoustic-phonetic, sublexical, lexical, sentential). This will test several hypothesis that older adults use predictions during speech processing, but at different levels than young adults. Functional magnetic resonance imaging (fMRI) is used to characterize language networks related to speech processing in quiet and in noise. Whereas previous research has mostly relied on small sample sizes, obscuring individual differences, this work will be based on a large dataset, adequately powered for individual difference analysis, and will collect a range of individual variables that may be related to speech comprehension, such as hearing ability, working memory span and inhibitory functioning. This work will relate individual differences in prediction at multiple levels of the processing hierarchy to language, cognitive, and functional outcomes. Together, this work aims to uncover which cognitive strategies (predictive or non-predictive), neural representations and network are related to successful speech comprehension in older adults. By relating neural measures to comprehension, it will distinguish between successful compensation and maladaptive changes in brain activation. The dataset will be shared publicly, providing a unique testbed for future researchers interested in speech and aging.
