Background: Engaging students in computer-assisted guided inquiry learning has great potential to scaffold their scientific understanding: Students are expected to improve their scientific problem-solving skills, and at the same time gain a deep conceptual understanding of the subject-matter. Additional generative activities such as creating video explanations subsequent to inquiry learning activities can also further deepen students' knowledge. Objectives: In this experiment, we therefore compared the effectiveness of computer-based direct instruction (to mimic traditional classroom teaching) versus computer-based inquiry learning. Methods: University students (N = 118) either received video-based direct instruction (direct instruction), including the demonstration of a virtual experiment, or conducted the virtual experiment themselves supported by prompts (inquiry learning). A third group of students additionally generated a video explanation as consolidation activity subsequent to conducting the virtual experiment (inquiry learning + generative activity). Results and Conclusions: Contrarily to our hypotheses, the direct instruction condition outperformed the inquiry learning conditions. There were no significant differences between the inquiry learning and the inquiry learning + generative activity condition. Moderation analyses revealed that the effectiveness of direct instruction predominantly held true for students with low levels of domain-specific self-concept. All in all, our present study contributes to a better understanding of effects of direct instruction versus guided inquiry learning in computer-based science education settings. Importantly, our findings show that the effectiveness of instructional approaches may depend on students' domain-specific self-concept as a motivational prerequisite. As such it is up for further research in science education to identify motivating instructional strategies to enhance students' learning.