A Comparison of Pair Programming to Inspections for Software Defect Reduction

Even though convinced of their ef®cacy, software development managers are looking for ways to reduce the cost of inspections. Recent work shows that inspections by two engineers have as good results as by using a larger team of ®ve. This indicates that pair programming might be effective as an inspection technique, as other research shows that a meeting is not necessarily better than not meeting. An experiment was conducted to see if pair programming reduces defects more than formal inspections. Results indicate that pair programming is more effective. A defect rate of 9.6 per thousand lines of code, much lower than that of a heavier method, were achieved. The implications for teaching are explored. One of the principal practices of the agile method eXtreme Programming (XP) is pair programming (Beck, 2000). Advocates of XP and pair programming claim that it helps ®nd defects sooner, as a different set of eyes examine all products as they are made (Williams & Kessler, 2000). They liken the pairs to an inspection team (Fagan, 1976). Earlier research seems to support this contention. A study by Porter and Johnson of inspection teams meeting versus those that do not meet indicates that the results are the same, so it would seem that it is good to have inspections, but these can be by individuals (Porter & Johnson, 1997). Further experimentation by Porter et al. of different sized teams of two, three, and ®ve that did meet indicated that the team of two had similar effectiveness to the team of ®ve, so, apparently, `̀ three's a crowd'' in terms of inspections as well (Porter et al., 1997). This is potentially good news for managers trying to reduce the cost of inspections, since teams of two can be suf®cient. This can also support the claims of the effectiveness of pair programming. We tried an experiment to ®nd out if this was true. We wanted to know whether to teach Address correspondence to: James E. Tomayko, School of Computer Science, Carnegie Mellon, 5000 Forbes Avenue, Pittsburgh, PA 15213, USA. E-mail: [email protected] full formal inspections or simply to emphasize defect reduction as a side product of pair programming.