The differential effects of four task selection methods on training efficiency and transfer in computer-based training for Air Traffic Control were investigated. A non-dynamic condition, in which the learning tasks were presented to the participants in a fixed, predetermined sequence, was compared to three dynamic conditions, in which learning tasks were selected on the basis of performance, mental effort, and a combination of both (i.e., mental efficiency). Using the 3-factor mental efficiency formula of Tuovinen and Paas (2004, this issue), the hypothesis that dynamic task selection leads to more efficient training than non-dynamic task selection was confirmed. However, the hypothesis that dynamic task selection based on mental efficiency leads to more efficient training than dynamic task selection based on performance or mental effort alone was not supported. The results are discussed in light of the theoretical framework and suggestions are given for future research.