The NMDA subtype of glutamate receptor is hypothesized to mediate synaptic competition in the developing brain by stabilizing converging synapses that have correlated activity patterns. Disruption of NMDA receptor function during development interferes with synapse elimination and sensory map formation. Moreover, many studies indicate that NMDA receptor function is high during times of synaptic rearrangement. In this review, a corollary of the NMDA receptor hypothesis for activity-dependent synapse stabilization is proposed. As developing inputs increase in number and strength, the increasing excitatory synaptic activity in young neurons should lead to increases in postsynaptic Ca2+ influx through NMDA receptors. This Ca2+ flux is postulated to trigger a feedback system that changes the subunit composition of the NMDA receptor complex so that less Ca2+ enters postsynaptic cells upon NMDA receptor activation. Changes in NMDA receptor effectiveness resulting from manipulations of activity are consistent with the idea that NMDA receptor function is under the control of activity. This postulate of activity-dependent control of NMDA receptor expression has implications for the control of brain plasticity. If particular combinations of NMDA receptor subunits typically expressed in young animals are better than adult receptor types at maintaining synapses in regions where they are not well correlated with other inputs, then expression of these juvenile subunit combinations could facilitate synaptic rearrangements in the mature brain after the normal end of synaptic plasticity. Thus, understanding the regulation of NMDA receptor function during development could provide a novel approach to restructuring circuitry in the adult brain to compensate for damage produced by trauma or disease.