Many theoretical positions on memory development anticipate that forgetting rates should vary substantially with age. The nature of these age variations is also relevant to many applied questions about child development that have major social policy implications, such as the veracity of children's eyewitness testimony and the long-term effectiveness of classroom instruction. Surprisingly, developmental studies of long-term retention have repeatedly produced the puzzling finding that forgetting rates are age invariant. It now seems, however, that these null age trends may have been artifacts of variables such as measurement insensitivity, floor effects, and stages-of-learning confounds. Assuming, as some later studies suggest, that forgetting rates vary with age when these factors are controlled, there are three overriding questions that must be dealt with in the developmental analysis of forgetting: the relative importance of storage failure versus retrieval failure, the relative importance of true forgetting processes versus test-induced processes, and the relative importance of storage-based reminiscence versus retrieval-based reminiscence. We describe a framework (disintegration/redintegation theory) that provides a conceptual environment within which research on these questions can progress. This framework, which evolved from fuzzy-trace theory, reinterprets processes such as storage failure, retrieval failure, restorage, and retrieval relearning in terms of levels of featural integration in traces (i.e., the extent to which contextual information is integrated with core semantic gist to produce a coherent representation). The theory is implemented in a mathematical model (the trace-integrity model) whose parameters deliver measurements of relevant memory processes on a common ratio scale. In a series of experiments, the model was used to study the theory's predictions about the contributions of these memory processes to long-term retention in subjects between the ages of 7 and 70. All the experiments were standard long-term retention designs (an initial acquisition session, followed by a 1-2-week forgetting interval, followed by a series of retention tests).(ABSTRACT TRUNCATED AT 400 WORDS)