This paper aims to provide a comprehensive model of litigation procedure as a grounding work for empirical research. Firstly, it extensively examines the signaling nature of litigation selection under asymmetric information. For the robustness of the analysis, we first separate the litigation selection process, where the signaling from the informed party plays the key role, from the actual settlement where, to avoid ad-hoc first-mover advantage, a more neutral bargaining method than the usual `take-it-orleave-it' is adopted. With full characterization of an equilibrium of the signaling game both under the defendant's and the plaintiff's private information, our model is believed to provide richer testable hypotheses for the future comparative static analyses. To be sure, these include the fairly famous hypotheses in law and economics such as the litigation puzzle and the 50% win rate hypothesis.