Christiane Penk, Dirk Richter

• Since the turn of the century, an increasing number of low-stakes assessments (i.e., assessments without direct consequences for the test-takers) are being used to evaluate the quality of educational systems. Internationally, research has shown that low-stakes test results can be biased due to students’ low test-taking motivation and that students’ effort levels can vary throughout a testing session involving both cognitive and noncognitive tests. Thus, it is possible that students’ motivation varies throughout a single cognitive test and in turn affects test performance. This study examines the change in test-taking motivation within a 2-h cognitive low-stakes test and its association with test performance. Based on expectancy-value theory, we assessed three components of test-taking motivation (expectancy for success, value, and effort) and investigated its change. Using data from a large-scale student achievement study of German ninth-graders, we employed second-order latent growth modeling and structural equation modeling toSince the turn of the century, an increasing number of low-stakes assessments (i.e., assessments without direct consequences for the test-takers) are being used to evaluate the quality of educational systems. Internationally, research has shown that low-stakes test results can be biased due to students’ low test-taking motivation and that students’ effort levels can vary throughout a testing session involving both cognitive and noncognitive tests. Thus, it is possible that students’ motivation varies throughout a single cognitive test and in turn affects test performance. This study examines the change in test-taking motivation within a 2-h cognitive low-stakes test and its association with test performance. Based on expectancy-value theory, we assessed three components of test-taking motivation (expectancy for success, value, and effort) and investigated its change. Using data from a large-scale student achievement study of German ninth-graders, we employed second-order latent growth modeling and structural equation modeling to predict test performance in mathematics. On average, students’ effort and perceived value of the test decreased, whereas expectancy for success remained stable. Overall, initial test-taking motivation was a better predictor of test performance than change in motivation. Only the variability of change in the expectancy component was positively related to test performance. The theoretical and practical implications for test practitioners are discussed.…