The problem of how to partition a software system and thus capture its overall architecture and its constituent components has become a research focus in the community of software engineering. In the literature, many methods have been proposed for solving this problem. For example, both top-down and bottom-up methods based on analyzing the graph representation of software systems have been proposed. We report an experimental study of a top-down method and a bottom-up method. In our study, we focus on the capability of component capture, the capability of architecture recovery and the time complexity for the two methods. According to our results on two real world systems, the studied bottom-up method is superior to the studied top-down method in both aspects, although the time complexity of the bottom-up method remains a big concern for large systems.