The generation effect is moderated by experimental design, as are a number of other encoding variables, such that the generation effect recall is typically larger in mixed-list than pure-list designs. In typical experiments on design effects, each study list is followed by its own recall test. Rowland, Littrell-Baez, Sensenig, and DeLosh (2014) found that the testing effect was not moderated by experimental design using a procedures in which multiple study lists were followed by a single, end-of-session recall test over all lists. This may indicate an important difference between the testing effect and the set of effects (including generation) moderated by experimental design, or it may signal a more general limitation of design effects. Three experiments used the procedures of Rowland et al. and found that the generation effect was likewise unmoderated by experimental design: The generation effect was robust for pure lists and of equivalent size as with mixed lists. This was found for both a perceptual (letter transposition) and semantic (antonym) generation task. Along with prior research, these results constitute similarities between the generation and testing manipulations with respect to design effects and their limitations. This, in turn, implies a parallel between the mnemonic effects of retrieving information from semantic (generation effect) and episodic (testing effect) memory.