In the evaluation of implemented software architectures, metrics can be used to provide an indication of the degree of encapsulation within a system and to serve as a basis for an informed discussion about how well-suited the system is for expected changes. Current literature shows that over 40 different architecture-level metrics are available to quantify encapsulation, but empirical validation of these metrics against changes in a system is not available. In this paper we survey existing architecture metrics for their suitability to be used in a late software evaluation context. For 12 metrics that were found suitable we correlate the values of the metric, which are calculated on a single point in time, against the ratio of local change over time using the history of 10 open-source systems. In the design of our experiment we ensure that the value of the existing metrics are representative for the time period which is analyzed. Our study shows that one of the suitable architecture metrics can be considered a valid indicator for the degree of encapsulation of systems. We discuss the implications of these findings both for the research into architecture-level metrics and for software architecture evaluations in industry.