Abstract
In this paper, we present final results of our long-term investigation whose goal was to study the contribution of software models on source-code comprehensibility. In this investigation we considered unified modeling language (UML) models produced in the analysis and design phases, and we conducted 12 controlled experiments in different contexts with different kinds of participants (e.g., PhD students and software practitioners). The total number of observations from these experiments was 333. We considered both source-code comprehensibility and the time to complete comprehension tasks. We use a meta-analysis to integrate gathered data and to obtain a global effect of analysis and design models on source-code comprehensibility. Results suggest that the use of UML models affects source-code comprehensibility, but in two opposite directions. In particular, models produced in the analysis phase reduce sourcecode comprehensibility and increase the time to complete comprehension tasks, while models produced in the design phase improve source-code comprehensibility and reduce the time to complete comprehension tasks.