Eye-movement indices of reading while debugging Python source code

作     者：Dempsey, Jack Tsiola, Anna Bosch, Nigel Christianson, Kiel Stites, Mallory 

作者机构：Univ Illinois Dept Educ Psychol Champaign IL USA Univ Illinois Dept Linguist Champaign IL USA Univ Illinois Sch Informat Sci Champaign IL USA Univ Illinois Beckman Inst Adv Sci & Technol Urbana IL USA Sandia Natl Labs Albuquerque NM USA 

出 版 物：《JOURNAL OF COGNITIVE PSYCHOLOGY》 (J. Cogn. Psychol.)

年 卷 期：2025年第37卷第2期

页      面：89-107页

核心收录：

学科分类：0402[教育学-心理学(可授教育学、理学学位）] 04[教育学] 

基　　金：Sandia Laboratory Directed Research and Development program at Sandia National Laboratories [DE-NA0003525] U.S. Department of Energy's National Nuclear Security Administration DOE Public Access Plan 

主　　题：Eye-tracking python reading programming languages debugging 

摘      要：Unlike text reading, the eye-movement behaviours associated with reading Python, a computer programming language, are largely understudied through a psycholinguistic lens. A general understanding of the eye movements involved in reading while troubleshooting Python, and how these behaviours compare to proofreading text, is critical for developing educational interventions and interactive tools for helping programmers debug their code. These data may also highlight to what extent humans use their underlying text reading ability when reading source code. The current work provides a profile of global reading behaviours associated with reading Python source code for debugging purposes. To this end, we recorded experienced programmers eye movements while they determined whether 21 different Python functions would produce the desired output, an incorrect output, or an error message. Some reading behaviours seem to mirror those found in text reading (e.g. effects of stimulus complexity), while others may be specific to reading code. Results suggest that semantic errors that produce undesired outputs in programming source code may influence early stages of processing, likely due to the largely top-down strategy employed by experienced programmers when reading source code. The findings are framed to invigorate discussion and further exploration into psycholinguistic analysis of human source code reading.