Local item dependence on the vocabulary levels test revisited

Abstract

The purpose of this study was to address the question of local item dependence (LID) on the Vocabulary Levels Test (VLT). The test format adopts a matching format consisting of six words and three definitions. A review of the literature suggested that items presented in such a format cannot be considered independent. However, Schmitt, Schmitt, and Clapham reported that according to Rasch analysis, most of the items performed independently. The present study examined effects of LID from a non-Rasch approach. A set of three clusters was collapsed into one large cluster of 18 words and 9 definitions on the assumption that such a treatment would make items practically independent. Both the collapsed VLT and the intact VLT were given to 114 Japanese English as a Foreign Language (EFL) students at an interval of one week. Results showed that scores were 15% higher on the original VLT than on the collapsed form. Furthermore, an LID index based on the correct/wrong response types between the two tests indicated that scores on the original VLT were about 19% inflated. Implications were drawn and discussed.