Summary Three experiments investigated free recall and recognition performance as a function of whether the S expected a free-recall or recognition test. One, two, or five lists of 10 unrelated meaningful words were learned. The lists were always followed by the expected test, except on the final list. On the final list (i.e., List 1, 2, or 5), half of the Ss received the expected test, while the remainder received the unexpected test. Recognition accuracy never changed as a function of the test expectations. With regard to the free-recall data, the expectation of a free-recall test produced a larger primacy effect than the expectation of a recognition set, except when four lists preceded the final list. In list 1, the recency effect was larger when a recognition test was expected than when a free-recall test was expected; the opposite emerged on List 2, while no difference appeared on List 5. The recency effects were explained by the different output orders the Ss used, revealing different retrieval stra...