Reliability assessment for two versions of Vocabulary Levels Tests

This article reports a reliability study of two versions of the Vocabulary Levels Test at the 5000 word level. This study was motivated by a finding from an ongoing longitudinal study of vocabulary acquisition that Version A and Version B of Vocabulary Levels Test at the 5000 word level were not parallel. In order to investigate this issue, Versions A and B were combined to create a single instrument. This was administered at one time to discover whether score differences found in the longitudinal study were present once the variable of time was removed. The data was analysed using correlation, and in order to discover if there was a significant difference between the two means of Version A and Version B, a t-test was used. Following that, a further examination of item facility values was conducted. The data analysis showed that Version A and Version B at the 5000 were highly correlated and highly reliable. However, the item analysis shows that the facility values of Version B contain a number of more difficult items. While versions of the Vocabulary Levels Tests at the 2000, 3000 and Academic levels may be treated as parallel for longitudinal studies, this does not hold at the 5000 word level. We suggest changes that need to be made to the test before it is used in future longitudinal vocabulary growth studies.