Measuring Lexical Diversity for Discourse Analysis in Aphasia: Moving-Average Type–Token Ratio and Word Information Measure

Abstract

Purpose The purpose of this study was to compare the utility of two automated indices of lexical diversity, the Moving-Average Type–Token Ratio (MATTR) and the Word Information Measure (WIM), in predicting aphasia diagnosis and responding to differences in severity and aphasia subtype. Method Transcripts of a single discourse task were analyzed for 478 speakers, 225 of whom had aphasia per an aphasia battery. We calculated the MATTR and the WIM for each participant. We compared the group means among speakers with aphasia, neurotypical controls, and left-hemisphere stroke survivors with mild aphasia not detected by an aphasia battery. We examined whether each measure distinguished levels of aphasia severity and subtypes of aphasia. We used each measure to classify aphasia versus neurotypical control and compared the areas under the curve. Results The WIM and the MATTR differentiated among people with aphasia, neurotypical controls, and people with mild aphasia. Both measures demonstrated moderately high predictive accuracy in classifying aphasia. The WIM demonstrated greater sensitivity to aphasia severity and subtype compared to the MATTR. Conclusions The WIM and the MATTR are promising measures that quantify lexical diversity in different and complementary ways. The WIM may be more useful for quantifying the effect of treatment or disease progression, whereas the MATTR may be more useful for discriminating discourse produced by people with very mild aphasia from discourse produced by neurotypical controls. Further validation is required.