Part 1

---

Analysis of large data set (Mechanical Turk + Bloomington).

Prediction of task performance metrics from program version, code metrics, and participant experience.

Part 2

---

Analysis of eye-tracking data set (Bloomington only).

Characterization of reading behavior, identification of important code lines/elements, and presentation of specialized eyeCode library.

Part 3 (In Progress)

---

Cognitive modeling of code reading and evaluation.

Quantitative process models based on ACT-R memory and perceptual/motor components. Compared to human reading behavior and errors.