A construct definition study of a standardized ESL vocabulary test

This paper reports a construct definition study of a standardized English as a second language (ESL) vocabulary test. The test, which contained 20 stimulus word items and 17 context word items, was administered to two different samples of ESL students at different proficiency levels and to a sample of native speakers of English. The item responses for the 40 items were used as the dependent variable in a stepwise multiple regression, construct specification equation and 12 variables (measures of the stimulus words' and keyed responses' fluency, distribution, abstractness, length, affective impact, and prototypicality) were used as predictors. Only six of the predictors accounted for a significant amount of variance in the dependent variable, and the number and arrangement of significant predictors varied across proficiency strata, indicating the utility of incorporating the examinees' stages of learning into construct validation studies. Construct definition studies such as the current one permit the statement of explicit predictions and a means to falsify construct theory, which is a perceived need in educational and psychological measurement. The paper concludes with an agenda for future research based on the results of the present study.