Research-based learning (RBL) is regarded as a panacea when it comes to effective instructional formats in higher education settings. It is said to improve a wide set of research-related skills and is a recommended learning experience for students. However, whether RBL in the social sciences is indeed as effective as has been postulated for other disciplines has not yet been systematically examined. We thus administered a pre-post-test study to N = 952 students enrolled in 70 RBL courses at 10 German universities and examined potential changes in cognitive and affective-motivational research dispositions. Latent change score modelling indicated that students increased their cognitive research dispositions, whereas most affective-motivational research dispositions decreased. The instructors' interest in the students' work served as a significant predictor of changes in research interest and joy. Practical implications for designing RBL environments can be inferred from the results.