According to Epstein (1985), extinction is one of the main processes for creative problem solving. To test this assertion, we present an experiment in which we manipulated extinction in a problem situation using the computer game Portal 2®. Three groups of participants were exposed to the same problem (i.e., get out of a chamber) under two different conditions. Two groups of participants learned the prerequisite behavior to solve the problem (NOEXT and EXT), but one group did not receive this training (BL). Under one condition, the problem could be solved in two ways: either by emitting a new flexible behavior (innovative) or by doing all of the previously learned behaviors (not innovative). Under the second condition, the problem had to be solved by emitting the innovative behavior, and the learned behaviors were under operational extinction. NOEXT was exposed to the first condition and EXT and LB groups to the second. Results showed that all participants from EXT and NOEXT, but fewer than half of LB groups, solved the problem, corroborating previous data that showed that creative problem solving is a function of previous learning. All but two participants from the NOEXT group solved the problem by emitting the learned behaviors; thus, extinction increased the likelihood of emitting new and flexible problem-solving behaviors. The results also showed that the EXT group spent more time solving the problem, emitted a higher frequency of behaviors other than E1, and interacted more often with objects than the NOEXT group. Therefore, even if the problem-solving behavior of EXT and NOEXT groups was equally effective for obtaining the solution, it was not equally efficient. We discuss that the constraints (i.e., extinction vs. no extinction) of each problem imposed different learning conditions during problem solving. New learning was necessary to produce the new and flexible problem solving of the EXT group, and this process demands more time and effort than solving the problem by just performing what was learned in the past, as the NOEXT group did. Finally, results also suggested a modulatory effect of gaming experience of the effects of experimental training on creative problem solving, because the participants from LB who solved the problem were those with more experience with different types of games.