Models of lexical knowledge assessment of second language learners of English at higher levels of language proficiency (2005)

Zareva, A. (2005). Models of lexical knowledge assessment of second language learners of English at higher levels of language proficiency. System, 33(4), 547-562.


Abstract

The study presented in this paper was conducted within the theoretical framework of the three-dimensional global-trait model of lexical knowledge proposed by [Henrikson, B. 1999. Three dimensions of vocabulary development. Studies in Second Language Acquisition 21, pp. 303–317], consisting of breadth, depth, and receptive-productive dimensions. The experiment empirically tested the practical effectiveness of the framework to account for differences in the vocabulary knowledge of three groups of participants: NSs of English (n = 30), L2 advanced (n = 17) and intermediate learners (n = 17). In addition, it was of interest to find out the degree to which five commonly used predictors of lexical competence, i.e., self-reported knowledge of words, vocabulary size, knowledge of words from various frequency bands, native-like commonality of associations, and number of associations, could help explain the greatest amount of variance in participants’ actual knowledge of words. The expectation was that such an analysis would also allow for identifying the smallest and practically most efficient set of predictors of both NSs’ and L2 learners’ overall state of the lexicon and, respectively, of their lexical development with an increase in proficiency. To address the research questions, two statistical procedures were used: multiple regression analysis, which is frequently employed for predictive purposes, and all possible regressions, which is one of several procedures used for identifying the most efficient predictors from a pool of variables. The regression analysis pointed to a statistically significant relationship between participants’ actual knowledge of words and the five predictors. All possible regressions revealed that the smallest “best” set of predictors of lexical knowledge across the three groups was a two-predictor model consisting of verifiable self-report and vocabulary size. The implications of the research findings are discussed from the practical perspective of using discrete item tests for vocabulary knowledge assessment of L2 learners at higher levels of proficiency.