Retrieving information by testing improves subsequent retention more than restudy, a phenomenon known as the retrieval practice effect. According to the retrieval effort hypothesis (REH), difficult items require more retrieval effort than easier items and, consequently, should benefit more from retrieval practice. In two experiments, we tested this prediction. Participants learned sets of easy and difficult Swahili–Portuguese word pairs (study phase) and repeatedly restudied half of these items and repeatedly retrieval practiced the other half (practice phase). Forty-eight hours later, they took a cued-recall test (final test phase). In both experiments, we replicated both the retrieval practice and the item difficulty effects. In Experiment 1 (N = 51), we found a greater retrieval practice effect for easy items, MDifference = .26, SD = .17, than for difficult items, MDifference = .19, SD = .19, t(50) = 2.01, p = .05, d = 0.28. In Experiment 2 (N = 28), we found a nonsignificant trend—F(1, 27) = 2.86, p = .10,
$$ {\upeta}_{\mathrm{p}}^2 $$
= .10—toward a greater retrieval practice effect for difficult items, MDifference = .28, SD = .22, than for easy items, MDifference = .18, SD = .21. This was especially true for individuals who benefit from retrieval practice (difficult: MDifference = .32, SD = .18; easy: MDifference = .20, SD = .20), t(24) = –2.08, p = .05, d = –0.42. The results provide no clear evidence for the REH and are discussed in relation to current accounts of the retrieval practice effect.