An experiment that uses the paired programming approach to teach beginning and advanced programming courses is discussed in this paper. Paired programming is an approach advocated by the extreme programming process [1], where a pair of programmers work side by side at a single workstation, with one person designated as the driver, and the other person as the observer. Paired programming advocates claim increased quality for this practice, with no decrease in productivity.

McDowell, Hanks, and Werner collected data from three courses in which students chose to develop programs either in pairs or individually. These two groups were compared in three areas: test scores, program quality (blind evaluation), and programming effort. The data collected seemed to show that the pairs developed higher quality programs, and that they learned more about programming. Unfortunately, the results were not statistically sound: the samples were not chosen randomly, and the sample sizes were too small to conclude that the results were significant. However, the authors have made a genuine effort to be objective, and their data is certainly more supportive of paired programming than pure anecdotal evidence.