This review article examines a recent study (Jarvis and Hashimoto, 2021) investigating three lexical diversity (LD) measures, each using five different word unit operationalizations. Jarvis and Hashimoto’s aim is to determine the most effective LD measures and demonstrate the potential influences of the different word units on each LD index. The LD measures include the measure of textual lexical diversity (MTLD), moving average MTLD with wrap-around measurement (MTLD-W), and moving-average type-token ratio (MATTR). Each measure is investigated with types operationalized as orthographic forms, lemmas using automated part-of-speech (POS) tags, lemmas with manually corrected POS tags, flemmas, and word families. These measures are used to examine 60 narrative essays written by English, Finnish, and Swedish first-language speakers; correlations with the LD ratings of 55 human raters are investigated. Jarvis and Hashimoto conclude that while the three LD measures are comparable, two of the word unit operationalizations produce mixed results.
In the review, following a summary of the paper and explanation of important concepts, the strengths and weaknesses of the study are evaluated with a view to assessing its importance in the field. Jarvis and Hashimoto’s research undoubtedly advances understanding of lexical diversity and its measurement. However, there are problems concerning the corpus of texts and the human raters used in the study. Also, the question of which word units work best remains unanswered.