This article concerns the structure of a double object construction. It is argued that Larson's arguments for the structure of double objects raise some empirical problems concerning anaphor binding and a semantic relation between the two objects. Next, it is pointed out that the structure corresponding to the present-day double object construction was extremely common in Old English and the structure similar to the dative construction of today began to appear at the beginning of the Middle English period. To accommodate the facts and at the same time to solve the problems with Larson's analysis, it is proposed that the structure of a double object construction is a basic form and the structure of a dative construction is derived from it by passivization within VP.