Component mode synthesis is a popular reduced-order modeling technique with a substructuring strategy in structural dynamics. The ideal goal of component mode synthesis is to construct compact reduced models with adequate accuracy for specific intended applications. In this paper, five a priori mode selection methods plus an a posteriori mode selection method have been evaluated as applied to several representative structural dynamics problems to assess their relative merits. The present evaluation offers a guideline for the informed use of these methods, and it is hoped that it will provide a basis for the further development and/or unification of several mode selection methods. Although there is no clear winner for all the problem classes studied, a slight edge of an a posteriori method over the a priori methods is demonstrated for a certain class of problems, provided the analyst is willing to allow extra computational overhead.