Often the code regions that are assigned for a maintenance task do not follow the modularization of the original application program, but instead include parts of code from many different units scattered throughout the application. In this paper, we investigate an approach to testing which we call concern-based testing, which leverages existing tools to help software maintainers identify the relevant code for their assigned task, their concern. The main contribution is a demonstration of the possible savings in test suite execution overhead and the increased precision in coverage information that can be obtained for a software maintainer if testing tasks are performed with respect to concerns. Based on a concern graph representation of the concern, a framework for guiding selective instrumentation for scalable coverage analysis is also presented.