A controlled experiment on software clones

Most software systems contain sections of duplicated source code - clones - that are believed to make maintenance more difficult. Recent studies tested this assumption by retrospective analyses of software archives. While giving important insights, the analysis of historical data relies only on snapshots and misses the human interaction in between. We conducted a controlled experiment to investigate how clones affect the programmer´s performance in common bug-fixing tasks. While our results do not exhibit a decisive difference in the time needed to correct cloned bugs, we observed many cases in which cloned bugs were not corrected completely.