Identifying Common Code Reading Patterns using Scanpath Trend Analysis with a Tolerance

Eye tracking data, particularly scanpath, provides valuable insights about code reading patterns which could describe actual comprehension strategies used. However, aggregating multiple scanpaths into one representative path is challenging since individual scanpaths tend to be different and are highly individualistic. The differences may affect the identification of a representative path which could decrease its similarity to individual scanpaths. Thus, we aim to identify a trending scanpath using Scanpath Trend Analysis (STA) with a tolerance to reveal common code reading patterns of high and low performing students while finding bugs in a static source code. Results show that variance exists in the scanpaths of high performing students which suggests that they follow varied code reading patterns while low performing students follow similar code reading patterns. Further, high performing students read code in a logical manner and a somewhat linear code reading pattern along with chunking of program code was employed which makes it possible to perceive the program better and hence, error regions are fixated. In contrast, low performing students jump directly to certain statements without following program's logic. This study addresses the challenge of identifying common code reading patterns that could help us determine effective strategies to be explicitly taught to students and develop learning materials to help improve their code reading and code comprehension skills.