Background: In this study, we examined the effectiveness of three instruction-production arrangements: instruction-problem solving (I-PS), problem solving-instruction (PS-I), and problem solving-instruction-problem solving (PS-I-PS)) in video tutorials for software training. Literature review: Most of the research on these arrangements stems from math and physics and has yielded equivocal outcomes. Studies of software training are scarce and have also led to varying results. In these studies, video access was prohibited once participants engaged in problem solving. Our study did not have this limitation. Research methodology: We followed an experimental approach with three conditions: I-PS, PS-I, and PS-I-PS. Research questions: 1. What is the effect of condition on video processing? 2. What is the effect of condition on motivation? 3: What is the effect of condition on procedural knowledge development? Results: In all conditions, all videos were viewed nearly in full. Replays of sections were scarce. In all conditions, self-efficacy rose substantially and more in I-PS than in PS-I. All conditions achieved very high scores on a final procedural knowledge test. I-PS did significantly better than PS-I on this test. Conclusion: The high absolute scores for self-efficacy and procedural knowledge presumably reflect the quality of the Demonstration-Based-Training videos developed for the study. Self-efficacy and procedural knowledge development was significantly higher in I-PS than in PS-I. In short, the data show that the best results were obtained for the video tutorial in which instruction preceded problem solving.