DOI: 10.3389/fnagi.2020.607449 One-Liner oral lexical retrieval works better than qualitative narrative analysis to classify dementia; and semantic fluency + Disfluency features chucked on an SVM returns pretty good results. Novelty Tried two different assays of measuring linguistic ability: oral lexical retrieval metrics, and qualitative discourse features analysis of speech. Notable Methods Subjects divided into three groups Great cog. decline Impaired but stable Healthy controls Administered BNT and SVF tests as baseline Key Figs Table 3 This figure tells us that the percentages of unrelated utterances was a statistically significant metric to figure differences between the three experimental groups. (CD, CS, HC: cognitive decline, cognitively stable (but declining normally), healthy control) (no other items are bolded) Table 4 This figure tells us the disfluency features analyzed. None of them were independently statistically significant. Table 5 This figure tells us that analyzing Semantic Verbal Fluency, plus the information of disfluency, trained on an SVM, actually shows >90% recall value? New Concepts Discourse-Completion Task oral lexical retrieval discourse features modalization Semantic Verbal Fluency Boston Naming Test