There is considerable converging evidence that by the end of grade school, children achieve success at mastering basic skills of reading, writing, and arithmetic. There are disturbing signs, however, that many students lack a firm conceptual grasp of the goal of the activities in which they engage in school. It seems that they can perform the necessary subskills or algorithms on demand, but they do not grasp their significance. We have argued elsewhere (Campione, Brown, & Connell, 1988) 1 that several features of grade school education could be responsible for this state of affairs. First, there is a clear emphasis on direct instruction with strong teacher control. Lower-level skills are taught before higher-level understanding, causing predictable problems of metacognition (Brown, 1975, 1978). Students fundamentally misunderstand the goal of early education; they come to believe that reading is decoding and that math consists only of quickly running off well-practiced algorithms without error. This emphasis on skill training is stressed to an even greater degree for low-achieving students, those for whom explicit instruction in understanding is particularly necessary. Strategies are rarely taught. When practice in understanding is provided, frequently it, too, is treated as consisting of decomposable skills. Such activities are presented as ends in themselves, rather than as a means to a more meaningful end. Little attention is paid to the flexible or opportunistic use of strategies in appropriate contexts. This common approach leads to the acquisition of "inert knowledge" (Whitehead, 1916) that cannot be applied broadly or flexibly. We argue that one major reason for this state of affairs is that traditional educational practice rarely incorporates metacognitive and contextual factors in learning (Campione, et al., 1988). In this chapter we will consider traditional methods of teaching science and mathematics in grade school and then look at some innovative alternatives that we are currently developing.