Analysis of code reading to gain more insight in program comprehension

Code reading, although an integral part of program comprehension, is rarely reflected. In this paper, we want to argue for a research approach and direction exploiting the potential that lies in the analysis of reading processes. Based on the vast experience compiled in psychology and some studies in computing, eye tracking and think aloud were elaborated as a viable research instrument for code reading studies. We conducted a feasibility study, designed to examine the actual process of code reading as the sensory starting point of comprehension. Computational and statistical tools were developed to facilitate data capture and analysis for eye tracking experiments. Results do not just provide proof of concept but already emphasize differences between reading natural language text and source code, as well as a distinct attention allocation within different code elements like keywords and operators. In conclusion we suggest a combination of theory-driven selected stimuli material, a carefully designed procedure of eye tracking, complemented with suitable post-tests on comprehension as well as retrospective think aloud in order to obtain additional information on the linking process between perception and comprehension. As an addition to other research approaches this should most certainly help us to improve our knowledge of comprehension within an educational research framework.