Multilevel Discourse Processing Analyses in Adults with a Cognitive Impairment

Abstract

Participants with a cognitive impairment associated with Alzheimer's disease (AD) present with discourse impairments early within the disorder. These discourse impairment are associated with declines in episodic, semantic, and working memory. These impairments provide researchers with the opportunity to examine the linguistic and other cognitive systems responsible for discourse, as well as determine how impairments to other cognitive systems impact discourse. Moreover, since these linguistic impairments are often qualitatively and quantitatively different from healthy aging, researchers may be able to use a multilevel discourse analysis to improve screening methods for cognitive impairment. The purpose of this study, then, is two-fold: (1) determine how participants with AD fit into the construction-integration model of discourse processing with special attention given to the role semantic memory declines have on discourse and (2) determine if a multilevel discourse analysis can discriminate between the participants with AD and healthy controls. Participants include 12 participants with AD and 12 healthy controls matched for age and education. To assess cognitive and linguistic abilities, participants complete three tasks: cognitive tasks, semantic tasks, and a discourse tasks. The cognitive task will include measures of episodic and working memory from the Wechsler Memory Scale - III. The semantic tasks involved the Pyramid and Palm Tree Tests, Boston Naming Tests, and a Category Flunecy Test. The discourse task will require participants to tell a story from two wordless pictures books. The discourse samples were analyzed for micro- and macrolinguistic errors; percentage of living things; percentage of light verbs; and thematic elements and actions. For study aim one, a MANOVA determined that participants with a cognitive impairment associated with AD produced more micro- and macrolinguistic errors, as well as fewer thematic actions compared with healthy controls. For aim two, a binary logistic regression model correctly grouped 87.5% of the participants into their correct group. While this is promising, more research is needed to understand the impact AD has on discourse and whether or not discourse can be used to improve screening method.