Restructuring efforts in the U.S. electric power sector have tried to encourage transmission investment by independent (non-utility) transmission companies, and have promoted various levels of market-based transmission investment. Underlying this shift to "merchant" transmission investment is an assumption that new transmission infrastructure can be classified as providing a congestion-relief benefit or a reliability benefit. In this paper, we demonstrate that this assumption is largely incorrect for meshed interconnections such as electric power networks. We focus on a particular network topology known as the Wheatstone network to show how congestion and reliability can represent tradeoffs. Lines that cause congestion may be justified on reliability grounds. We decompose the congestion and reliability effects of a given network alteration, and demonstrate their dependence through simulations on a 118-bus test network. The true relationship between congestion and reliability depends critically on identifying the relevant range of demand for evaluating any network externalities.