Recent researches have been investigating how different manipulations on reinforcement contingencies can affect response variability. One possible manipulation on reinforcement delivery is the omission of an expected reinforcer. The objective of Experiment 1 was to investigate if reinforcement omission can affect the behavior of organisms reinforced for presenting repetition or different levels of variability of operant responses. 32 male Wistar rats were subjects. It was made a training of operant variability and repetition and the response considered was a sequence of four presses to any of two levers. During this condition the animals that were reinforced for varying were reinforced according to the relative frequence schedule. There was a group of high variability requirement, a group of low variability requirement and a group of intermediate variability requirement. A fourth group was reinforced continuously for repeating the sequence RLLL and a fifth group received reinforcement according to the distribution of reinforcements of the subjects of the group with high variability requirement, but without having to vary responses for that. After reaching response stability in the training condition, subjects went through two tests conditions in which 25% and 50% of the responses that reached criteria for reinforcement delivery were not reinforced (reinforcement omission). It was made measures of the response rate and the response latency on the intertrial interval after reinforcement delivery, on the intertrial interval after not delivering reinforcement and on the intertrial interval after reinforcement omission. It was also made measures of the U index and of the frequency of each of the 16 possible responses. The results show that: 1) the schedule of reinforcement of variability used was able to produce different levels of response variability in each group; 2) the reinforcement omission effect occurred for all groups in both tests conditions; and 3) only the subjects of the repetition group showed significant changes in the U index comparing the training condition to both tests conditions. In Experiment 2 the same number of subjects formed the same five groups. The procedure used was very similar to the prior one, except that a discriminative stimulus was used to differentiate the trial (moment in which responses counted for reinforcement) from the intertrial interval (moment in which responses did not count for reinforcement). Furthermore, the sequence chosen for reinforcement in the repetition group was LLLL and between the tests with 25% and 50% of reinforcement omission it was made five sessions of training with 100% of reinforcement. The results were very similar to those found in Experiment 1, except for the difference in response rate between post-omission condition and post-error condition in the test with 25% of reinforcement omission. Therefore, the present research sustains the hypothesis that reinforcing response variability can make the organisms more resistant to changes than reinforcing repeti...