SUMMARY1. Previous electrophysiological, pharmacological and anatomical evidence had suggested a possible participation of rostral ventromedial medulla (RVM) in the supraspinal part of the loop underlying diffuse noxious inhibitory controls (DNICs). In order to test this hypothesis, two experimental series were performed during which DNICs were compared in control sham-operated rats and rats with lesions of the RVM and of an adjacent candidate for such a role, the nucleus gigantocellularis reticularis (Gi).2. In the first experimental series, lesions were induced, in anaesthetized animals, by injections of quinolinic acid (0-3-0-8 ,1 of a 360 nmol/,l solution) into the RVM or Gi. In the control animals (n = 10), the vehicle alone (artificial cerebrospinal fluid) was injected. Histological lesion reconstructions were performed after each electrophysiological experiment. Three groups of animals were considered: in the first group (n = 5), the lesion was centred on the RVM, including the two caudal thirds of nucleus raphe magnus (NRM) and adjacent reticular areas; in the second group (n = 5), the lesions extended more rostrally and involved the rostral pole of NRM; in the third group (n = 5), the lesion extended more laterally and dorsally and included nucleus reticularis gigantocellularis pars alpha (GiA), nucleus reticularis paragigantocellularis lateralis (LPGi) and the Gi. In each case, all the neuronal cell bodies within the lesioned area were destroyed.3. In the second experimental series, electrolytic lesions of the total rostrocaudal extent of the NRM (n = 5) were induced, in anaesthetized animals, by passing cathodal current (5 mA, 8 s). In the control animals (n = 5), the electrode was lowered but current was not applied. 4. One week after lesioning, the animals were anaesthetized, paralysed, artificially ventilated and recordings were made from convergent neurones in trigeminal nucleus caudalis. These neurones gave responses due to activation of A and C fibres when percutaneous electrical stimuli were applied to their receptive fields. DNICs were triggered by immersion of each paw in a 50°C water-bath. Both the general properties of the convergent neurones and the inhibitions of the C fibre-evoked responses produced by these heterotopic noxious stimuli were compared in the different groups of animals.5. The sizes of receptive field, spontaneous activities, thresholds for C fibre-evoked MS 1059 D. BOUHASSIRA, Z. BING AND D. LE BARS responses and responses to C fibre activation were not different in the control and lesioned animals. The percentage inhibitions of the C fibre-evoked responses both during and in the 44 s following the conditioning periods were also very similar in the different groups of animals.6. These results indicate that neither the RVM nor Gi are involved, at least directly, in DNIC. Taken together with the results of previous studies using midbrain lesions it is concluded that the midbrain and medullary structures which are known to be involved in the modulation of pain, do not play a ...