on this study, a trigeminal model was used in which high threshold C-fibre-evoked activities of convergent neurones located in the subnucleus oralis of the trigeminal complex are modulated through the superficial part, the substantia gelatinosa, of the subnucleus caudalis. The two subnuclei are located 3 mm apart, therefore, it was possible to inject dizocilpine, a non-competitive N-methyl-D-aspartate (NMDA) receptor antagonist, into either the super¢cial or the deep parts of subnucleus caudalis without interfering with ongoing recording of convergent neurones in subnucleus oralis. A di¡erential NMDA-dependent modulation of wind-up was observed according to the dose and the injection target. (1) The injections of small non-di¡usible doses (0.12 Wg) of dizocilpine into the super¢cial part of subnucleus caudalis facilitated wind-up. The e¡ect peaked at 25 min with a mean increase above control of 173 þ 31%. Injection (0.5 Wg) of either the less active enantiomer dizocilpine or saline into superficial subnucleus caudalis had no signi¢cant e¡ect on subnucleus oralis convergent neurones. This suggests that NMDA-dependent interneurones, probably located in substantia gelatinosa of subnucleus caudalis, exert, in normal conditions, an inhibitory control on wind-up of convergent subnucleus oralis neurones. (2) The injection of larger doses (0.5 Wg) into the superficial part of subnucleus caudalis induced a predominant inhibitory ffect on wind-up. The mean peak effect at 15 min was 46 þ 7% compared to control (100%). Small and large doses of dizocilpine injected into the deep part of subnucleus caudalis had a predominant inhibitory e¡ect. The inhibition of wind-up of subnucleus oralis neurones after injection of NMDA receptor antagonists in super¢cial or deep subnucleus caudalis indicates that wind-up may be due, at least in part, to NMDA activation at synapses that do not involve the recorded convergent neurones. ß