Some neurones in the dorsal horn of the spinal cord are strongly inhibited when a nociceptive stimulus is applied to any part of the body, distinct from their excitatory receptive fields. This phenomenon was termed "Diffuse Noxious Inhibitory Controls" (DNIC). DNIC influence only convergent neurones, and these inhibitions can be triggered only by conditioning stimuli which are nociceptive. The inhibitions are extremely potent, affect all the activities of the convergent neurones and persist after the removal of the conditioning stimulus. Only activity of A delta- or A delta- and C- peripheral fibres can trigger DNIC. DNIC are sustained by a complex loop which involves supraspinal structures since, unlike segmental inhibitions, they are not observed in animals in which the cord has previously been transected at the cervical level. The ascending and descending limbs of this loop travel respectively through the ventro-lateral and dorso-lateral funiculi, respectively. We proposed that DNIC result from the physiological activation of some brain structures putatively involved in descending inhibition. However, lesions of the mesencephalon, including the periaqueductal grey (PAG) and the rostral ventromedial medulla (RVM), including nucleus raphé magnus, did not modify DNIC. By contrast, lesions of subnucleus reticularis dorsalis (SRD) in the caudal medulla strongly reduced DNIC. Both electrophysiological and anatomical data support the involvement of SRD neurones in spino-bulbo-spinal loop(s). In man, very similar results have been obtained by means of combined psychophysical measurements and recordings of nociceptive reflexes (RIII reflex). Painful heterotopic conditioning stimuli depress both the reflex and the associated painful sensation, with stronger effects being observed with more intense conditioning stimuli. By contrast, in tetraplegic patients, heterotopic nociceptive stimulation did not produce any depression of the RIII reflex. Observations were also made on patients with cerebral lesions causing contralateral hemi-analgesia, either a unilateral thalamic lesion or a lesion of the retro-olivary part of the medulla (Wallenberg's syndrome). In the patients with Wallenberg's syndrome, no inhibitions were observed when the nociceptive conditioning stimuli were applied to the affected side whereas if these stimuli were applied to the normal side they triggered inhibitory effects and post-effects very similar to those seen in normal subjects. These results show that in humans, brainstem--probably reticular--structures seem to play a key role in these phenomena. The data suggest that nociceptive stimuli, even though there are unquestionably perceived as being painful activate certain inhibitory controls which originate in the brainstem. Since all convergent neurones are subject to DNIC, one can make the assertion that the transmission of nociceptive signals towards higher centres is under the influence of these controls. In other words, the descending inhibitory controls may play a physiological role in the detection of nociceptive signals. It is proposed that DNIC constitute both a filter which allows the extraction of the signal for pain and an amplifier in the transmission system which increases the potential alarm function of the nociceptive signals. This hypothesis is supported by the finding that DNIC are blocked by low doses of morphine in both rat and man.