Severity of Alzheimer's disease and language features in picture descriptions

Background: Studies of connected speech of individuals with Alzheimer's disease (AD) report significant impairments relative to the language of cognitively intact participants. Considerably less research has focused on the association between dementia severity and language features. Aims: The current study examines how scores on a dementia screening test (the Mini-Mental Status Examination, MMSE) correlate with features of connected speech. Methods & Procedures: Thirty-five individuals with AD (range of MMSE scores = 3-25) and 35 cognitively intact participants provided picture descriptions. Ten language features were derived from their descriptions using an automated text analysis tool: total word number, percentage of content words, pronoun ratio, type-token ratio, mean word frequency, percentage of verbs, percentage of verbs in the most common morphological form in Hebrew, percentage of verbs in present tense, percentage of prepositions, and percentage of subordination markers. Information content was also analysed. Outcomes & Results: Group differences emerged in five language features as well as in information content, attesting for substantial lexical impairment in AD. Within the AD group, MMSE scores were correlated with type-token ratio, with mean word frequency, and with the number of information units. No equivalent correlations were found within the control group. Conclusions: Dementia severity associates with decreasing lexical diversity, increasing word frequency, and a reduction in relevant information content, but not with changes in grammatical features of language. A simple automated analysis of connected speech could be used clinically to define and track the decline in language abilities in AD.