The proposals concerning working memory outlined in this paper involve the architecture of working memory, the reasoning mechanisms that draw on it, and the ways in which working memory may develop with age. Ways of assessing task demands and children's working memory capacities are also considered. It is noted that there is long-standing evidence that much human reasoning is analogical. Recently developed models have shown that analogies can be conceptualized as a mapping from a base structure to a target structure. This paper begins with a review of the concepts of short-term and primary memory, followed by a literature review. Discussion then turns to the question of whether working memory increases or remains constant with age. It is maintained that the lack of adequate measures of working memory capacity makes it impossible to resolve the question. However, when the question is analyzed in terms of neural net architecture, the hypothesis that overall capacity remains constant but representations become more differentiated with age appears theoretically and empirically reasonable. If the hypothesis were valid, it would mean that the ability to map structures increases with age, and the development of reasoning capacity has a maturational component. Contains 35 references. (LB)