This study assesses longitudinal effects of different training designs on joint negotiation performance. In so doing, the study experimentally compares (a) bilateral training of both the seller and the buyer within a dyad with both (b) a no-training control condition and 2 conditions with unilateral training of either (c) the buyer or (d) the seller. Moreover, underlying psychological mechanisms of the training effect are assessed. Results of the study with 360 participants reveal a significant overall training effect on negotiation outcomes that remains stable over time. Consistent with our hypotheses, unilateral negotiation training is only effective if the trained party is the seller, and it fails if the trained party is the buyer. Additional mediation analyses reveal exchange of priority-related information as a causal mechanism underlying these effects.