Two experiments were carried out to investigate the effects of semantic constraints of sentence contexts on the elaborative learning. In the first experiment 40 subjects (adults) were asked to rate whether the target word would fit into sentence frame with five point scales followed by an unexpected recalland recognition tests. Three types of sentence frame were used: Congruous-Interchangeable(CI), Congruous-Noninterchangeable(CN), and Incongruous(I). The target words fitted into CI and CN sentence frames, but did not into I sentence frames. In a CI sentence frame, both the target word and its associate fitted sensibly, whereas in a CN sentence frame the target word fitted sensibly but its associate did not. Recall performance was highest in a CN sentence frame followed a CI and a I sentence frame. False recognition scores to the associates were higher in CI sentence frames than in CN ones. These results were interpreted as showing that effective elaboration seemed to depend on the semantic constraints of sentence contexts embedding the target words. In the second experiment 29 second and 33 sixth graders were asked to answer orally whether the target word would fit into sentence frame followed by an unexpected recall and reccognition tests. CI and CN sentence frames led to a better recall than I ones, but the difference between the former two sentence frames was not observed. The difference of frequency of false recognitions to the asociates between CI and CN sentence frames was not observed. These results showed that school children would be unable to utilize a CN sentence frame for effective elaboration of target word. It was suggested that the ability of encoding target words with sentence contexts was critical to effective elaboration.