In many real world negotiations, from wage contract bargaining to product liability disputes, the bargaining parties often interact repeatedly and have the option of seeking outside judgement. This paper studies a model of repeated bargaining with a third party to analyze how and why bargaining postures endogenously evolve over time. A privately informed long-lived player bargains with a sequence of short-lived players, one at a time. Should the players fail to reach an agreement, an unbiased yet imperfect third party is called upon to make a judgement. The uninformed short-lived players learn through two channels: observed behavior of the informed player (\soft" information) and, if any, verdicts of the third party (\hard" information).  The long-lived player wants to guard his private information by bargaining tough but at the expense of more information disclosure from the third party. As a result of the strategic use of these two sources of information, the players' bargaining postures change as the uninformed players' beliefs evolve. Interestingly, as third party information becomes more precise, the players adopt tough bargaining postures for a wider range of beliefs. Many repeated bargaining problems can be analyzed in this framework. In particular, the equilibrium dynamics provide an explanation for the puzzling contrast between the bargaining postures of Merck and Pfizer in their recent high-profile product liability litigations. The results also help us understand several other phenomena documented in the related literature.