This paper addresses "the 2-chart problem," with respect to diagrammatic representations of complex event-driven reactive systems. In particular we focus on behavioral modeling with statecharts. A taxonomy of cues and a heuristic for using them is proposed. This facilitates consideration of behavioral issues of each chart individually. The more complex issues that arise from consideration of both of them jointly are then considered. We show how this heuristic may be applied iteratively with a top-down traversal of the system model. In particular it is shown how to use this approach to address the issues first addressed in Lehman's work on software evolution: the unbounded number of features in any application domain and the need to identify tacit assumptions that give rise to these features that become obsolete as the application and its world of operation evolve.