This study tested the common assumption that, to be most effective, working memory (WM) training should be adaptive (i.e., task difficulty is adjusted to individual performance). Indirect evidence for this assumption stems from studies comparing adaptive training to a condition in which tasks are practiced on the easiest level of difficulty only [cf. Klingberg (Trends Cogn Sci 14:317–324, 2010)], thereby, however, confounding adaptivity and exposure to varying task difficulty. For a more direct test of this hypothesis, we randomly assigned 130 young adults to one of the three WM training procedures (adaptive, randomized, or self-selected change in training task difficulty) or to an active control group. Despite large performance increases in the trained WM tasks, we observed neither transfer to untrained structurally dissimilar WM tasks nor far transfer to reasoning. Surprisingly, neither training nor transfer effects were modulated by training procedure, indicating that exposure to varying levels of task difficulty is sufficient for inducing training gains.