NMDA (N-methyl-D-aspartate) receptors are one class of ionotropic receptor for the ubiquitous excitatory neurotransmitter L-glutamate. The receptor is made up of four protein subunits combined from a larger library of proteins, which gives this receptor a great deal of variability. This explains the large number of modulatory sites, a variety of sites at which antagonists can interact, and therefore a number of potential drug targets. Sensitivity of the NMDA ion channel to ambient levels of Mg++ gives it a voltage dependence that suits a function of responding to intense synaptic activation; the ability of the channel to admit Ca++ tends to trigger long-term processes. The receptor is thereby involved in long-term physiological processes such as learning and memory as well as in pathological processes such as neuropathic pain. Separating these functions therapeutically with NMDA antagonists has been a major difficulty, and has not yet been achieved with currently-available agents. This review summarises the preclinical rationale, based on animal models, and the clinical evidence on the use of NMDA antagonists in pain states. It also summarises the details of the receptor so as to explain the rationale for targeting either specific sites on the receptor, or exploiting anatomical differences in subtype expression, so as to provide the beneficial effects of NMDA receptor block with an improved side effect profile. In particular, agents that are selective for receptors that include the NR2B subunit preclinically have a substantially better profile for treating neuropathic pain than do current NMDA antagonists; some emerging clinical evidence supports this view.