Test-language effects in bilingual education: Evidence from CLIL classes in Germany

In the last decades, many bilingual programs such as content and language integrated learning (CLIL) have emerged. A crucial question in CLIL is whether to assess students' content-subject knowledge in the common language or the dominant language of instruction. In two experimental field studies with N = 361 6th-graders and N = 703 8th-graders, we investigated test-language effects and their moderation by CLIL-experience, language background, instructional language use and item format. In both studies, students' achievement was higher when items were presented in the common language. This effect was less pronounced in experienced CLILlearners and for multiple-choice than cloze-test items. Moderation effects of language background were only partially and for instructional language use not statistically significant. The results show that using the dominant language of instruction to assess content-subject knowledge in bilingual education might underestimate students' learning and that test-language and item-format effects need to be considered.