Lukijat sanaston monimuotoisuutta määrittämässä: Leksikaalisen diversiteetin tarkastelua määrällisen ja laadullisen tutkimuksen rajapinnassa

This article examines the construct of lexical diversity while focusing on research methodology and the potential for lexical diversity to be used as an index of language proficiency. The study gives attention to questions of inter-rater reliability, the effects of texts' lexical characteristics on raters' lexical diversity ratings, and which set of features best accounts for raters' perceptions of lexical diversity. The present study focuses on lexical diversity in Finnish, and it does this by comparing the lexical characteristics of texts with how they are perceived. The data consists of narrative texts written by school children, as well as lexical diversity ratings assigned to the same texts by adult raters. Each text (n = 60) was rated by 23 raters, whose ratings were tested statistically for inter-rater reliability. A regression model was then used to investigate which lexical features the raters relied on while assessing the texts' levels of lexical diversity. The results show that the raters' lexical diversity ratings were highly consistent with one another (Cronbach's alpha = 0.959). Despite the large number of texts they were asked to rate, all raters rated each of the 60 texts, and their intra-rater consistency remained high from the beginning to the end of the rating task. The results have important implications for the construct definition of lexical diversity: of all the lexical features examined in the present study, four alone suffice to account for nearly three quarters (roughly 73%) of the variance in the ratings. The results suggest that raters' perceptions of lexical diversity are strongly associated with 1) the number of different words in a text (abundance), 2) the intervals between occurrences of the same word (dispersion), 3) the semantic quality of individual words in the text (specialness) and 4) the overall degree of repetitiveness in the text (variety).