Strategy and Instruction for Code Comprehension: From Reading Behavior Analysis to Instructional Evaluation

Ziyi Zhang

Program comprehension is a foundational activity during software development. It is central to tasks such as maintenance, testing, and code review, which account for a significant portion of developers' time and effort. Given its importance, researchers have investigated program comprehension from multiple angles: some focus on measuring the cognitive processes involved through techniques such as eye tracking and neuroimaging, while others seek to understand the underlying procedures and actions through interviews, cognitive models, and behavioral analysis. These efforts have substantially advanced our understanding of how developers, particularly experienced ones, comprehend code, and have begun to examine the sub-steps that students follow during comprehension. However, much less is known about how to systematically support students in developing these skills in educational settings. In typical computer science curricula, early courses emphasize writing code, while advanced courses assume students can already read and understand code independently. As a result, code comprehension often remains an implicit skill that students are expected to acquire on their own. In this dissertation, we contribute empirical perspectives and evidence-based approaches to code comprehension education. First, to understand how students approach code comprehension in practice, we identify and characterize the reading strategies that students employ during comprehension tasks. We link the observable reading behaviors to high-level cognitive goals and to performance outcomes such as correctness, time, and cognitive load. Results show that certain reading patterns are associated with improved task performance. Having established what reading behaviors students employ, we then turn to the question of how to support students in developing effective comprehension skills. We introduce and evaluate a context-specific instruction approach that integrates concrete, problem-specific worked examples with procedural guidance. We demonstrate that this approach leads to more effective skill acquisition and more robust long-term retention when compared to traditional guidelines. In particular, concrete problem-specific examples prove to be a critical component: Participants receiving such examples not only achieve higher correctness and shorter completion times, but also report reduced cognitive load and greater confidence in their comprehension process.

Strategy and Instruction for Code Comprehension: From Reading Behavior Analysis to Instructional Evaluation

Files and links (1)

Abstract

Metrics

Details