Objective: Comparison of program defects caused by programmer pairs and solo developers. Design: Analysis of programs developed during two counter balanced experiments. Setting: Programming lab at University. Experimental Units: 42 programs developed by computer science students participating in an extreme programming lab course. Main Outcome Measures: Programmer pairs make as many algorithmic mistakes but fewer expression mistakes than solo programmers. Results: The second result is significant on the 5 percent level. Conclusions: For simple problems, pair programming seems to lead to fewer mistakes than solo programming .
Matthias M. Müller