Effects of defects in UML models: an experimental investigation

Author(s): Christian F. J. Lange, Michel R. V. Chaudron
Venue: International Conference on Software Engineering
Date: 2006

Type of Experiement: Controlled Experiment
Sample Size: 159
Class/Experience Level: Undergraduate Student, Professional


This paper investigates the effect that UML defects have on the way UML is interpreted. It also investigates how well the implementers of the UML diagrams can detect defects.

The authors conducted two experiments using computer science students (111), and industry practitioners (48). Participants were given UML fragments and prompted to answer questions based on the model fragments. The questions about model fragments containing an injected defect were paired with a similar control question that focused on the same aspect but using a UML fragment without the defect.

The experiment showed that defects often remain undetected and cause misinterpretation. However the experiments show that is some cases "the results show that some defect types are detected by almost all subjects (e.g. 96% of the subjects detect Class not in Sequence Diagram) whereas other defect types are hardly detected (e.g. Multiple Definitions of the same Class is detected by 10% only)."