Global coherence analysis of discourse in amnestic mild cognitive impairment and associated cognitive function

Abstract

Objectives: Global coherence measures include a mean score of utterances according to the degree of theme maintenance, a method of estimating the ratio of utterances that inhibit coherence, and a subjective rating score given whole discourse (i.e., a global coherence index, an error rate, a subjective rating score). The purposes of this study are as follows: to compare scores of three measures between patients with amnestic mild cognitive impairment (aMCI) and the normal elderly; to find out adequate global coherence measure among them; and to investigate cognitive functions which are associated with global coherence. Methods: Analysis of covariance was used for the group comparison between 20 patients with aMCI and 20 normal elderly. Measures were investigated through logistic regression analysis and receiver operating characteristic curve, and then cut-points were computed with sensitivity and specificity. Partial correlation analysis was used for finding out cognitive functions related with global coherence. Results: Significant group differences were found in a global coherence index and an error rate. The logistic regression analysis resulted in significantly lower global coherence index and significantly higher error rate of aMCI compared to the normal elderly. Cognitive functions which were associated with three measures were immediate recall and semantic verbal fluency. Conclusion: Discourses of aMCI are liable to be irrelevant, egocentric, and repetitive, or have more than 30% error utterances. Scores under the cut-points imply communicative and semantic-pragmatic declines of language use, and reduced memory and executive functions. These linguistic and cognitive impairments are reflected in global coherence of aMCI.