PURPOSE: A microlinguistic content analysis for assessing lexical semantics in people with aphasia (PWA) is lexical diversity (LD). Sophisticated techniques have been developed to measure LD. However, validity evidence for these methodologies when applied to the discourse of PWA is lacking. The purpose of this study was to evaluate four measures of LD to determine how effective they were at measuring LD in PWA. METHOD: Four measures of LD were applied to short discourse samples produced by 101 PWA: (a) the Measure of Textual Lexical Diversity (MTLD; McCarthy, 2005), (b) the Moving-Average Type-Token Ratio (MATTR; Covington, 2007), (c) D (McKee, Malvern, & Richards, 2000), and (d) the Hypergeometric Distribution (HD-D; McCarthy & Jarvis, 2007). LD was estimated using each method, and the scores were subjected to a series of analyses (e.g., curve-fitting, analysis of variance, confirmatory factor analysis). RESULTS: Results from the confirmatory factor analysis suggested that MTLD and MATTR reflect LD and little of anything else. Further, two indices (HD-D and D) were found to be equivalent, suggesting that either one can be used when samples are >50 tokens. CONCLUSION: MTLD and MATTR yielded the strongest evidence for producing unbiased LD scores, suggesting that they may be the best measures for capturing LD in PWA.
PURPOSE: A microlinguistic content analysis for assessing lexical semantics in people with aphasia (PWA) is lexical diversity (LD). Sophisticated techniques have been developed to measure LD. However, validity evidence for these methodologies when applied to the discourse of PWA is lacking. The purpose of this study was to evaluate four measures of LD to determine how effective they were at measuring LD in PWA. METHOD: Four measures of LD were applied to short discourse samples produced by 101 PWA: (a) the Measure of Textual Lexical Diversity (MTLD; McCarthy, 2005), (b) the Moving-Average Type-Token Ratio (MATTR; Covington, 2007), (c) D (McKee, Malvern, & Richards, 2000), and (d) the Hypergeometric Distribution (HD-D; McCarthy & Jarvis, 2007). LD was estimated using each method, and the scores were subjected to a series of analyses (e.g., curve-fitting, analysis of variance, confirmatory factor analysis). RESULTS: Results from the confirmatory factor analysis suggested that MTLD and MATTR reflect LD and little of anything else. Further, two indices (HD-D and D) were found to be equivalent, suggesting that either one can be used when samples are >50 tokens. CONCLUSION: MTLD and MATTR yielded the strongest evidence for producing unbiased LD scores, suggesting that they may be the best measures for capturing LD in PWA.
Authors: Davida Fromm; Joel Greenhouse; Kaiyue Hou; G Austin Russell; Xizhen Cai; Margaret Forbes; Audrey Holland; Brian MacWhinney Journal: J Speech Lang Hear Res Date: 2016-10-01 Impact factor: 2.297
Authors: Davida Fromm; Margaret Forbes; Audrey Holland; Sarah Grace Dalton; Jessica Richardson; Brian MacWhinney Journal: Am J Speech Lang Pathol Date: 2017-08-15 Impact factor: 2.408