Novice programmers, who have yet to form effective mental models of the domain, often experience high cognitive load, low confidence, and high anxiety, negatively affecting learning and retention rates. These cognitive and affective limitations pose an instructional challenge. This study aimed to investigate the effectiveness of a whole-task instructional approach compared with a part-task instructional approach for novices learning to program from a cognitive and affective perspective. A fully randomized between-subjects controlled experiment was designed, including two online instructional conditions (whole-task vs. part-task). The "whole-task condition" followed the Task-Centered Instructional Strategy and included explicit instruction in the context of whole tasks. The "part-task condition" followed a part-task instructional strategy and included the same explicit instruction, yet in the context of objectives and topic-related tasks. Based on Bandura's triadic model (Bandura, Social foundations of thought and action: A social cognitive theory, Prentice-Hall, 1986), we propose a conceptual model, which we used to hypothesize that the Task-Centered Instructional Strategy may be more effective for novices learning to program. Sixty-five students with no programming experience volunteered to participate in the study and were randomly assigned to one of the conditions. Participants in the whole-task condition performed significantly better on the near and far transfer posttests. In accordance with our model, confidence and cognitive load during learning were found to be significant partial mediators of the effect of instructional strategy on performance. Overall, we found that the task-centered instructional strategy, combining explicit instruction with whole-tasks, is effective for addressing the cognitive and affective considerations relevant to novices in computer science education.