The ability of various 5-HT(3) receptor antagonists to potentiate spinal glycine responses was investigated. Whereas (3-alpha-tropanyl)-1H-indole-3-carboxylate (ICS 205930), (3-alpha-tropanyl)-3,5-dichlorobenzoate (MDL 72222) and 1-methyl-N-(3-alpha-tropanyl)-1H-indazole-3-carboxamide (LY 278584) exhibited this property, even in identified motoneurones, several other chemically similar 5-HT(3) receptor antagonists did not. Introducing a methyl group on the nitrogen of the azabicyclo moiety of ICS 205930 greatly reduced the ability to potentiate glycine responses. Neither endo-1-methyl-N-(9-methyl-9-azabicyclo[3.3. 1]non-3-yl)-indazole-3-carboxamide (granisetron), differing from LY 278584 by an additional carbon in this cycle, nor 2beta-carbomethoxy-3beta-benzoyloxytropane (cocaine), 1,2,3, 9-tetrahydro-9-methyl-3-[(2-methyl-1H-imidazol-1-yl)-methyl]-4H-carba zol-4-one (ondansetron) and (S)-4-amino-N-(1-azabicyclo[2.2. 2]oct-3-yl)-5-chloro-2-methoxy-benzamide ((S)-zacopride) could potentiate glycine responses. A pharmacophore model of the glycinergic potentiators was generated by molecular modelling using MDL 72222 as a template. According to this model, an aromatic ring, a carbonyl group and a tropane nitrogen atom are required for glycinergic potentiation, as previously described for 5-HT(3) receptor antagonism. However, the steric allowance at the glycine receptor site and the tridimensional arrangement of the pharmacophoric elements appear to be more restricted.