Gender DIF in reading tests: A synthesis of research

Abstract

The purpose of this study is to synthesize gender DIF in reading tests and the possible reasons behind it. After a thorough search of studies, 1,210 items from 18 articles were included in the final analysis. It is found that 23.3% of the items show gender DIF. There is, however, a large variation in the percentage of items determined as showing DIF across studies ranging from 0 to 77%. Among the DIF items, about half favor males and half favor females, and this pattern holds true for studies using the item response theory-likelihood ratio (IRT-LR) test and for those using the Mantel–Haenszel (M–H). Furthermore, items from shorter tests are more likely to be determined as having DIF than are items from longer tests. Other DIF patterns appear to depend on the specific DIF methods and the tests. The implications and limitations of this synthesis are also discussed.