Processing of acoustically degraded speech in primary progressive aphasia and Alzheimer’s disease

Abstract

AbstractBackgroundIn daily life, spoken messages are often “degraded” by competing sounds and vocal idiosyncrasies. Comprehension of degraded speech demands intense computations across distributed neural networks; it is likely to present a particular challenge for patients with neurodegenerative pathologies. However, this issue has not been studied systematically.MethodHere, we studied the processing of degraded speech signals in a cohort of patients representing all major variant syndromes of primary progressive aphasia (N = 32), in relation to patients with typical Alzheimer’s disease (N = 19) and healthy age‐matched controls (N = 25). As a model paradigm for the degraded ‘noisy’ speech signals of daily life, we used noise‐vocoding: digital division of the speech signal into a variable number of frequency channels constituted from amplitude‐modulated white noise, such that fewer channels convey less spectrotemporal detail and reduce intelligibility. We assessed the impact of noise‐vocoding on the recognition of spoken three‐digit numbers (e.g., ‘five‐hundred‐and‐eighty‐seven’) and used psychometric modelling to ascertain the threshold number of noise‐vocoding ‘channels’ required for intelligibility by each participant.ResultCompared with healthy controls, all patient groups showed normal comprehension of natural speech; for vocoded speech, patients with semantic variant primary progressive aphasia performed normally whereas other patient groups had a raised noise‐vocoding threshold and this was particularly marked in logopenic aphasia.ConclusionOur findings suggest that degraded speech is a promising paradigm for probing real‐world communication difficulties in people with dementia, and for stratifying syndromes and pathologies. Future work should assess the potential of this paradigm for developing new biomarkers and interventions to assess and improve communication function in these diseases.