The application of the Rasch model to Yes/No vocabularly tests
Yes/No tests have become well-established as a method for assessing the vocabulary size of first and second language learners (Anderson & Freebody, 1983; Meara & Buxton, 1987). However, despite a general acceptance of their usefulness, no evaluation studies had ever been conducted of the tes...
| Main Author: | |
|---|---|
| Published: |
Swansea University
1999
|
| Subjects: | |
| Online Access: | http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.594177 |
| Summary: | Yes/No tests have become well-established as a method for assessing the vocabulary size of first and second language learners (Anderson & Freebody, 1983; Meara & Buxton, 1987). However, despite a general acceptance of their usefulness, no evaluation studies had ever been conducted of the tests. The results from two experiments showed quite conclusively that the corrected scores from five Yes/No tests were unreliable as measures of vocabulary size. The results also revealed that responses to real words alone were the single best predictor of vocabulary knowledge. Three further experiments were therefore conducted using Yes/No tests constructed from real words which had been selected from different corpora. The results from the Rasch analysis showed that all the test scores were extremely reliable and the application of the Rasch model to the data was valid. The results also indicated there is a close relationship between the frequency of a word and its difficulty, but only a moderate one between Yes/No test scores and scores from tests of vocabulary knowledge and general proficiency. It is concluded that using the revised Yes/No format and Rasch analysis together will result in very accurate estimates of vocabulary size. It is suggested that if Yes/No tests are used in conjunction with other measures of vocabulary knowledge, they may help to define more clearly how the lexicon develops. |
|---|