Pair Programming vs. Solo Programming: What Do We Know After 15 Years of Research?

Author(s): Carolina Alves de Lima Salge, Nicholas Berente
Venue: 2016 49th Hawaii International Conference on System Sciences (HICSS)
Date: 2016

Type of Experiement: Other
Sample Size: 22


In this study, the aim was to find if we discovered the benefits of Pair Programming over the course of the last 15 years of it being a practice. They collected the results of 22 other studies and conduct a meta-analysis of 15 years of empirical tests of these relationships, and find that: pairs generally (a) produce higher quality software, (b) learn more, and (c) program faster, supporting Beck's arguments.

They iterated over Beck's potential pros and cons of pair programming and then used the results of their study to determine if these were true or not. Their analysis shows that the data supports Beck's arguments about the effect of pair programming on quality, learning, and duration. Quality will improve due to pair programming for the fundamental reason that two people are more likely to be able to solve complex problems more effectively than just one, and a dialog between two developers is likely to generate better outcomes than either acting alone. In learning, in participating in pair programming, developers learn from each other. And finally, overall calendar time (i.e., duration) for developing a bit of code could shrink. All three of these are supported by their data in the results section.