An experiment to assess different defect detection methods for software requirements inspections

Software requirements specifications (SRS) are usually validated by inspections, in which several reviewers read all or part of the specification and search for defects. We hypothesize that different methods for conducting these searches may have significantly different rates of success. Using a controlled experiment, we show that a scenario-based detection method, in which each reviewer executes a specific procedure to discover a particular class of defects has a higher defect detection rate than either ad hoc or checklist methods. We describe the design, execution and analysis of the experiment so others may reproduce it and test our results for different kinds of software developments and different populations of software engineers.<>