Working memory was designed to explain four benchmark memory effects: the word length effect, the irrelevant speech effect, the acoustic confusion effect, and the concurrent articulation effect. However, almost all research thus far has used tests that emphasize forward recall. In four experiments, we examine whether each effect is observable when the items are recalled in reverse order. Subjects did not know which recall direction would be required until the time of test, ensuring that encoding processes would be identical for both recall directions. Contrary to predictions of both the primacy model and the feature model, the benchmark memory effect was either absent or greatly attenuated with backward recall, despite being present with forward recall. Direction of recall had no effect on the more difficult conditions (e.g., long words, similar-sounding items, items presented with irrelevant speech, and items studied with concurrent articulation). Several factors not considered by the primacy and feature models are noted, and a possible explanation within the framework of the SIMPLE model is briefly presented.