This dissertation studies several problems centered around developing a better understanding of the energy efficiency of cooperative wireless communication systems. Cooperative communication is a technique where two or more nodes in a wireless network pool their antenna resources to form a "virtual antenna array". Over the last decade, researchers have shown that many of the benefits of real antenna arrays, e.g. spatial diversity, increased range, and/or decreased transmission energy, can be achieved by nodes using cooperative transmission. This dissertation extends the current body of knowledge by providing a comprehensive study of the energy efficiency of two-source cooperative transmission under differing assumptions about channel state knowledge, cooperative protocol, and node selfishness. The first part of this dissertation analyzes the effect of channel state information on the optimum energy allocation and energy efficiency of a simple cooperative transmission protocol called "orthogonal amplify-and-forward" (OAF). The source nodes are required to achieve a quality-of service (QoS) constraint, e.g. signal to noise ratio or outage probability, at the destination. Since a QoS constraint does not specify a unique transmit energy allocation when the nodes use OAF cooperative transmission, minimum total energy strategies are provided for both short-term and long-term QoS constraints. For independent Rayleigh fading channels, full knowledge of the channel state at both of the sources and at the destination is shown to significantly improve the energy efficiency of OAF cooperative transmission as well as direct (non-cooperative) transmission. The results also demonstrate how channel state knowledge affects the minimum total energy allocation strategy. Under identical channel state knowledge assumptions, the results demonstrate that OAF cooperative transmission tends to have better energy efficiency than direct transmission over a wide range of channel conditions. The second part of this dissertation focuses on the development of an opportunistic hybrid cooperative transmission protocol that achieves increased energy efficiency by not only optimizing the resource allocation but also by selecting the most energy efficient cooperative transmission protocol from a set of available protocols according to the current channel state. The protocols considered in the development of the hybrid cooperative transmission protocol include compress-and-forward (CF), estimate-and-forward (EF), non-orthogonal amplify-and-forward (NAF), and decode-and-forward (DF). Instantaneous capacity results are analyzed under the assumption of full channel state knowledge at both of the sources and the destination node. Numerical results are presented showing that the delay limited capacity and outage probability of the hybrid cooperative transmission protocol are superior to that of any single protocol and are also close to the cut-set bound over a wide range of channel conditions. The final part of this dissertation focuses on the issue of node selfishness in cooperative transmission. It is common to assume in networks with a central authority, e.g. military networks, that nodes will always be willing to offer help to other nodes when requested to do so. This assumption may not be valid in ad hoc networks operating without a central authority. This section of the dissertation considers the effect selfish behavior on the energy efficiency of cooperative communication systems. Using tools from non-cooperative game theory, a two-player relaying game is formulated and analyzed in non-fading and fading channel scenarios. In non-fading channels, it is shown that a cooperative equilibrium can exist between two self-interested sources given that the end of the cooperative interaction is uncertain, that the sources can achieve mutual benefit through cooperation, and that the sources are sufficiently patient in the sense that they value future payoffs. In fading channels, a cooperative conditional trigger strategy is proposed and shown to be an equilibrium of the two-player game. Sources following this strategy are shown to achieve an energy efficiency very close to that of a centrally-controlled system when they are sufficiently patient. The results in this section show that cooperation can often be established between two purely self-interested sources without the development of extrinsic incentive mechanisms like virtual currency.