An Empirical Investigation of the Defect Detection Capabilities of Requirements Specification Languages

It is a frequently reported effect of applying requirements specification languages that the formalization of informal requirements leads to the detection of defects such as omissions, conflicts, and ambiguities. However, there is little quantitative data available on this effect. This paper presents an empirical study with requirements specification languages, which addresses two research questions. First, which types of defects are detected by a requirements engineer during the development of a requirements model, and second, what happen to those defects that are not detected? The results indicate that ambiguities require special care during formalization, because they are less frequently reported than other types of defects. Instead, ambiguities tend to become often disambiguated unconsciously, which is a serious problem, because implicit assumptions are more likely than in our study to be wrong when the system is more complex. Moreover, ambiguities are misinterpreted more often than other types of defects. Finally, ambiguities, if noticed, require immediate clarification.