iMeasuring and Characterizing Eye Movements

Software review is a technique to improve the quality of software documents and to detect defects by reading the documents. To increase review performance (the number of defect detections and/or defect detection efficiency), many review techniques and support environments have been proposed. However, the difference between individuals is more dominant than the review techniques and the other factors. Hence, understanding the fac-tor of individual differences between a high-performance reviewer and a low-performance reviewer is necessary to develop practical support and training methods. This thesis re-veals the factors affecting review performance from an analysis of the reading procedure in software review. To analyze the reviewers ’ reading procedure quantitatively, this thesis proposes to use the eye movements of the reviewer. Measuring eye movements on each line and in a docu-ment allow us a correlation analysis between reading procedure and review performance. In this thesis, eye movements are classified into two types: Eye movements between lines and eye movements between documents. Two experiments analyzed the relationship be