Test Improvement toward Criterion-referenced Evaluation and Education : Based on Classical Test Theory and Item Response Theory

This study reports test data analyzed by both Classical Test Theory and Item Response Theory as fundamental language testing research. Since Criterion-referenced Evaluation has been chosen for introduction into Japanese elementary and junior high schools in 2002, Japanese high schools in 2003, and Hiroshima University in 2004, measurements such as paper tests should be consolidated for the purposes. Since norm-referenced evaluation has assumed a main role in evaluation, language testing research has focused on pre-test stages of test administration. This has occurred because a test is inferred to be sufficient when it can show ranking of test takers with validity and reliability at the time the test is conducted. However, in criterion-referenced evaluation, tests are reasonably inferred to measure longitudinal transitions of learners' proficiency. Item Response Theory is recommended to analyze test data because it can estimate item characteristics regardless of test takers' proficiency. Classical Test Theory, which is comparatively simple, but limited by test takers, is used to show fundamental data. Results obtained by these analyses represent highly reliable open data for Japanese EFL learners. Those data can be used for further test development or test improvement