On the Effectiveness of the UML Object Diagrams: A Replicated experiment

Background: In the modeling of object oriented software systems, the UML object diagrams are recognized very useful to complement class diagrams. However, up to now, there exists only one experiment [Torchiano 2004] that investigates this concern. Aim: To confirm or contradict the findings of the original experiment, we have conducted a replication and the achieved results have been presented in this paper. Both the replication and the original experiment have been conducted to investigate whether the use of object diagrams to complement class diagrams affects the comprehension of software systems. Method: The replication has been conducted with a group of 24 graduated subjects in Computer Science of the University of Basilicata. The experiment adopts a counterbalanced design, thus ensuring that each subject work on two comprehension tasks, experimenting each time class and object diagrams together or class diagrams alone. The comprehension on each task has been assessed using a questionnaire-based approach. In particular, we have measured the comprehension level of each subject using an information retrieval based approach that allowed us to get a balance between correctness and completeness of the answers. Results: The results show that the subjects significantly benefit from the use of object diagrams in the comprehension of software systems, thus confirming and strengthening the findings of the original experiment. Conclusions: It is advisable to complement the usual class diagrams with object diagrams to increase the understandability of software systems. To raise the generalizability of the results, replications of this study are necessary especially with professional software engineers.