Coman et al. describe two studies of pair programming. They define the term backup behavior as the extent to which team members help each other perform their roles, that is, how much they act as backup for their peers.

The two studies, a four-month study of student projects and a nine-month study of professional programmers, measured programmers’ activities along with the interactions of their peers. Pair programming was but one of many different options available to the teams. The authors found a wide variety of interaction patterns, but in the large majority of the time, programmers chose to work alone. Peers were occasionally brought in to help, either to resolve a problem or to help with the initial effort. No cases of pure pair programming from start to finish were observed.

I particularly liked this paper because it resonated with my own work experiences. Pair programming has a place in development activities, at least in these real-world efforts, as a support mechanism for a developer’s natural programming processes. I believe that this paper is worth reading for anyone contemplating introducing formal pair-programming methodologies into their development portfolio.