Cognition and language are intimately connected. How dementia is manifested in language is a matter of intense research around the world by neuropsychologists, linguists and more recently, by specialists in artificial intelligence who comb through language data for telltale signs of cognitive impairment. Formulaic expressions, word recall, propositional density are among the linguistic features that have been investigated in relation to dementia, with conflicting results.

We assume that changes in language are early signs of cognitive decline, and take a structural approach to study the language of early cognitive impairment, as manifested in language use. We know that in the normal population structurally complex sentences require more cognitive resources to process. The onset of cognitive decline may cause a parallel decline in the use of sentences with complex structures, resulting in language deficits that solidify as the disease progresses. Using natural speech data collected from volunteers who participate in the Community Health Intergenerational (CHI) study, we search for linguistic markers of dementia. Early results from a small-scale pilot study are promising. We analyzed the speech data from 36 subjects, 28 healthy and 8 showing signs of mild cognitive impairment (MCI). Some broad grammatical features, such as type-token ratio, do not show a difference between the two groups of subjects. Other features, such as the use of complex noun phrases, exhibit a significant decline in usage rate among people with MCI.

This talk presents a usage profile of select grammatical features in the language of MCI sufferers.