Williams, L., E. Wiebe, K.Yang, M. Ferzli, and C. Miller (in press). In support of paired programming in the introductory computer science course. Journal of Computer Science Education.

Prior research results indicates that pair programming, whereby two programmers working collaboratively on the same design, algorithm, code, or test, produce higher quality code in essentially half the time as solo programmers.  Educators utilizing the pair programming model in the classroom have also observed educational benefits including superior results on graded assignments, increased satisfaction/reduced frustration from the students, increased confidence from the students on their project results, and reduced workload of the teaching staff.  A formal experiment was run at North Carolina to empirically assess the educational efficacy of the technique in a CS1 course.  Results indicate that students who practice pair programming perform better on programming projects and are more likely to succeed by completing the class with a C or better.  Student pairs are more self-sufficient which reduces their reliance on the teaching staff.  Qualitatively, paired students demonstrate higher-order thinking skills than students who worked alone.  These results are supportive of pair programming as a collaborative learning technique.