A previous study found that task shifting and fluid intelligence, but not working memory capacity (WMC) and prior knowledge, influenced the worked example effect (Schwaighofer, Bühner, & Fischer, 2016). To increase confidence in these findings, we report a preregistered extended replication study of Schwaighofer et al.'s investigation. University students (N = 231, M[subscript age] = 22.40 [SD = 4.33], 87% women) solved statistical problems with textbook materials presented on a laptop in one of four conditions in a 2 × 2 factorial between-subjects design. We compared worked examples versus problem-solving (replication) and with versus without time pressure (extension). Time pressure was added to test whether learners in the original study were able to offload WMC demands, which would explain why the WMC moderation was not found. Results showed that the advantage of worked examples over problem-solving decreased with increasing prior knowledge, suggesting that problem-solving becomes eventually more effective than worked example study. Similarly, the benefit of worked examples over problem-solving decreased with increasing shifting ability of a learner. However, contingencies on WMC or fluid intelligence were not detected. Our extension analysis indicated that the worked example effect was also not contingent on WMC even when learners were under time pressure. These findings underline the important role that task shifting might play in scaffolded learning environments and suggest that trading in the focus on WMC for a broader perspective on cognitive architecture provides novel explanations for instructional effectiveness. Our study further highlights the importance of more customized instructional support.