Longitudinal Analysis of Spoken Language Characteristics in the Nun Study

DESCRIPTION (provided by applicant): Verbal communication is one of the most complex and vital human behaviors negatively affected by neurodegenerative disease, and previous research strongly suggests that linguistic characteristics are promising as early clinical indicators. The Nun Study data set offers a rare opportunity to investigate the application of linguistic methods to the assessment of late life cognitive deficits and the development of neurodegenerative disease. We propose to investigate rate of decline in linguistic ability using previously unanalyzed spoken autobiography audio samples repeated over four waves of assessment. Aim 1 will capitalize on work already done to digitize audio recordings of the Nun Study longitudinal spoken autobiography samples. Verbatim transcripts will be produced and will be time-aligned with the audio from each speech sample. Aim 2 will use computational linguistic methods to calculate measures of syntactic complexity and propositional content, and will evaluate the rate of change on these measures over the 41/2 year follow-up period. Mean rates of change will be estimated in those sisters with repeated speech samples available, and will be compared between those sisters with and without dementia to determine if the rate of decline in linguistic ability is associated with diagnostic status. This project leverages the extensive NIH resources already invested in the Nun Study to 1) analyze an under-utilized aspect of this valuable data set, 2) expand upon the study's early findings in the application of linguistic analysis to the study of aging and disease-development, and 3) update the analysis methodology applied to the autobiography samples by making use of automated methods from the field of computational linguistics to combine information about speech content with information from the audio recording.