The present experiments were designed to test the hypothesis that context effects in recognition memory are due to the Ss' semantic interpretations of to-be-remembered words which appear in different contexts from the presentation to the test periods. In Exp. I, Ss were not informed of the kind of retention test, and the presentation rate was 1sec per item. In Exp. II, the presentation rate was changed to 2sec per item. In Exp. III, Ss were forewarned of the recognition test. Context effects were found only in Exp. II; the results of Exp. I and III did not support the hypothesis. It was suggested that the context effects occur only when the experimental conditions enabled Ss to process the stimulus words even at the semantic level.