A modularization is a partitioning of a software system into components based on a variety of criteria, each depending on the clustering approach and desired level of abstraction. Source-header dependency graphs are bipartite graphs that are formed by flattening include file dependencies and enumerating source-file-to-header-file dependencies. In this paper, we describe an approach for identifying candidate modularizations of software systems by analyzing connectivity properties of source-header dependency graphs. In addition, we apply the approach to a large software system to demonstrate its applicability