Reward omission was used to produce retroactive interference in a serial delayed-alternation task. Rats were given forced choices in 3 T mazes presented in succession followed by free choices in which the alternate arm in each maze was rewarded. Nonreward during the delay between forced-and free-choice runs produced a decrement in performance (i.e., retroactive interference) in comparison to reward during the delay. Interference in retention of the first maze in a list occurred when the second maze contained surprising reward or omission (i.e., an outcome contrary to previous training in that maze) in comparison to expected reward or omission. In addition to producing retroactive interference, omission sometimes increased alternation in the maze in which it occurred. These results support a differential rehearsal hypothesis of spatial memory.The purpose of the present research was to study the short-term effects of reward and reward omission on the immediate retention of recently presented maze locations. Differential processing evoked by reinforcers is sometimes studied using a retroactive interference design, in which a reinforcing stimulus is presented during the retention interval after a to-be-remembered event. Such a comparison of reward presence or absence has produced mixed results. Harper, McLean, and Dalrymple-Alford (1993) found that food given during the delay impaired radial maze performance, particularly reducing retention of the most recently entered arms. They attributed this to the disruptive effect of consummatory activity during the delay interval on memory. In contrast, Kasprow (1987) found that retentioninterval reward enhanced performance of rats in a delayedalternation task, which he attributed to a reminder effect. Terry (1987), who also studied delayed alternation, did not note any difference between reward and omission presented during the retention interval.This discrepancy can be found in other memory tasks. Jans and Catania (1980) found that food access reduced delayed-response performance by pigeons. Medin, Reynolds, and Parkinson (1980), testing monkeys in delayed matching-to-sample tasks, obtained either retroactive facilitation or interference, depending on the similarity of the