8 pairs of Ss were run in each of 4 different dyadic reward systems using a continuous response task. The 2 Ss within each pair were made interdependent in their rewards by imposing a functional relationship between the responses made by one S and the outcomes delivered to the other S, and vice versa. Each pair was given 100 consecutive trials. 3 influence responses were appended to the task for 4 of the 8 pairs in each reward system. These appended responses permitted an S to communicate a request to the other S concerning the latter's continuous response on the next trial. Results from systems with no communication confirm the predictions of reward stability based on a stochastic model. Without communication, 1 of the 4 systems stabilize; with communication, 2 systems stabilize. (PsycINFO Database Record (c) 2006 APA, all rights reserved)