For contemporary legal theory, law is essentially an interpretative and hermeneutics practice (Ackerman (1991), Horwitz (1992)). A straightforward consequence is that legal disputes between parties are motivated by their divergent interpretations regarding what the law says on their case. This point of view fits well with the growing evidence showing that litigants’ cognitive performances display optimistic bias or self-serving bias (Babcock and Lowenstein (1997)). This paper provides a theoretical analysis of the influence of such a cognitive bias on pretrial negotiations. However, we also consider that this effect is mitigated because of the litigants’ confidence in their own ability to predict the verdict; we model this issue assuming that litigants are risk averse in the sense of Yaari (1987), i.e. they display a kind of (rational) probability distortion which is also well documented in experimental economics. In a model à la Bebcuck (1984), we show that the consequences of self-serving bias are partially consistent with the "optimistic model", but that parties’ risk aversion has more ambiguous/unpredictable effects. These results contribute to explaining that the beliefs in the result of the trial are not sufficient in themselves to understand the behaviors of litigants. As suggested by legal theory, the confidence the parties have in their beliefs is probably more important.