The purpose of the present study was twofold: first, to investigate whether the part-to-whole transfer effect originally observed in free-recall learning of unstructured word list can be generalized to the learning of cohesive sentences; and second, to evaluate the organizational hypothesis for this effect. There were two experimental groups (Sequential and Random groups) and a control group. Each subject learned a common whole list which consisted of a poem of 16 lines. Prior to the whole-list learning, each subject learned a part list. The part list for the experimental subjects consisted of 8 lines which were randomly selected from the whole-list poem. The subjects in the Sequential group learned the part list in a sequential arrangement, while the subjects in the Random group learned the same part list in a random arrangement. The part list for the control group consisted of another poem of 8 lines. The results indicated that the negative part-to-whole tranfer effect occurred only in the Random group. These results were interpreted as supporting the organizational hypothesis of the part-to-whole transfer effect, and the theoretical implications for the models of sentence memory were discussed.