Eight pigeons responded in a three-component concurrent-chains procedure, with either independent or dependent initial links. Relative probability and immediacy of reinforcement in the terminal links were both varied, and outcomes on individual trials (reinforcement or nonreinforcement) were either signaled or unsignaled. Terminal-link fixed-time schedules were varied across components within conditions to yield immediacy ratios of 1:2, 1:1 and 2:1. The probabilities of reinforcement were varied across conditions to yield reinforcer ratios of 1:5, 1:2, 2:1 and 5:1. Results showed that a model based on the generalized matching law provided a good description of response allocation, accounting for 92% of the variance overall. As expected, sensitivity to probability was greater in the unsignaled conditions. However, sensitivity to immediacy was also greater in the unsignaled conditions, suggesting that the effect of signaling terminal-link outcomes may not be limited to probability but apply to reinforcer variables in general. The effects of signaling can be explained in terms of conditioned reinforcement added to each alternative's outcomes in the matching law. There was some evidence for an interaction between reinforcer probability and immediacy, particularly for the dependent-schedules group, such that sensitivity to immediacy was greater at moderate rather than extreme reinforcer ratios. However, further analysis suggested that this could have been due to a ceiling effect on response allocation imposed by dependent scheduling. Overall, the present results show that the generalized matching law can provide a useful account of choice between outcomes that vary in both probability and immediacy of reinforcement. (Contains 4 tables and 7 figures.)