The "Pair" as a Problematic Unit of Analysis for Pair Programming

This paper explores the problematic nature of using an isolated pair as the unit of analysis in studies and evaluations of pair programming. Using empirical data from an observational case study within a software development organization, we show pairs spending 20% of their pairing time interacting with people outside the pair. These interactions, which are encouraged by this organization as part of its highly collaborative system, represent important value exchanges with people outside the pair. This suggests that research on pairs in isolation may not be indicative of how pair programming works in situ when enacted by teams accomplished in the practice, and may misrepresent the net value proposition of pair programming.