This article presents three studies, two of them longitudinal, which investigated the relations between age, processing speed, working memory (WM), and fluid intelligence ("g[subscript f]") from 4 to 16 years of age. Structural equation modeling showed that speed was a powerful covariate of age ([approximately] - 0.6 to - 0.7) from 4 to 13 years, declining thereafter (to [approximately] - 0.2). WM was stably related to speed through the whole age-span studied ([approximately] - 0.4 to - 0.5). A large part (59%) of age-related changes in "g[subscript f]" (83%) from 4 to 7 years and a lower but significant part later on, especially in adolescence ([approximately] 10-20% out of [approximately] 40-50%), were mediated by WM. However, with speed and age controlled, WM was almost fully commensurate with "g[subscript f]" ([approximately] 0.9), from about the age of 8-9 years onwards. A series of models suggested an ever present efficiency level defined by speed and control and a representational level defined by WM and "g[subscript f]", which are increasingly differentiated with development. All processes develop in cycles concerted by a dynamic G. Change in each process over time originated from within the processes themselves and G, in proportions varying with developmental phase. Overall, speed signified age-associated changes in processing capabilities, partly expressed in WM expansions and "g[subscript f]" reconstructions. An overarching model is proposed integrating differential with developmental theories of human intelligence. (Contains 2 tables and 5 figures.)