An Empirical Eye-Tracking Study of Feature Model Comprehension
Software Product Lines (SPLs) are families of related software systems which are distinguished by the set of features each system provides. Feature Models are the de facto standard for modelling the variability of SPLs because they describe the features, their relations, and all the combinations of features that constitute a SPL. Because of their key role, feature models are at the core of many tasks in SPL engineering. Our work presents an empirical study on the comprehension of feature models for the task of checking the validity of configurations. Our study explored the relation between the number of features and the number of cross-tree constraints with the accuracy of participants' answers to validity checking questions, and used eye fixations for analyzing the difficulty in interpreting fixated information and the amount of cognitive processing of the different parts of the feature model stimuli. We found that answer accuracy does not relate individually to the number of features or to the number of cross-tree constrains of a feature model, but both factors do show an interaction on accuracy. Additionally, our study identified differences in feature models with cross-tree constraints in both number of fixations and fixation time, but no differences in those models without cross-tree constraints.