Cooperation Is socially desirable when users obey the rules of cooperation. This is the case in networks controlled by a single entity, such as military networks. On the other hand, in commercial wireless networks wherein users may behave for selfish intentions, it may be difficult to achieve a socially efficient cooperation without a mechanism that ensures users adhere to the rules of cooperation. In the absence of such mechanism, cooperation will be dominated by social-dilemma where each user is tempted to deviate from cooperation. In this paper we examine performance of Amplify-and-Forward (AF) In the presence a selfish user. We model selfish behavior as On-Off channel where the user randomly flip-flops between cooperative and selfish states. Based on such random characterization, we show both analytically and by simulation the physical layer consequences of selfish users In cooperative diversity.