This article presents a model of prosodic structure development that takes account of the fundamental continuity between child and adult systems, the surface level divergence of child forms from their adult target forms, and the overall developmental paths of prosodic structure. The main empirical base for the study comes from longitudinal data collected from three Japanese-speaking children (1; 0-2; 6). Evidence for word-internal prosodic constituents including the mora and the foot is found in compensatory lengthening phenomena, syllable size restrictions and word size restrictions in early word production. By implementing the representational principles that organize these prosodic categories as rankable and violable constraints, Optimality Theory can provide a systematic account of the differences in the prosodic structure of child and adult Japanese while assuming representational continuity between the two. A constraint-based model of prosodic structure acquisition is also shown to demarcate the learning paths in a way that is consistent with the data.