Antisera raised against the fixation products of l-glutamate and l-aspartate were used, singly or in combination, to study the ultrastructural localization of the amino acids in the rat dorsal horn, with post-embedding immunogold techniques. Immunostaining for each of the amino acids was also combined with immunolocalization of GABA, an important inhibitory neurotransmitter in the spinal cord, or synaptophysin, a synaptic vesicle glycoprotein. In addition, we examined the localization of glutamate immunoreactivity in relation to that of calcitonin-gene related peptide and substance P, two neuropeptides present in high concentrations in the dorsal horn. Glutamate- and aspartate-immunoreactive neuronal cell bodies, dendrites, axons and terminals were apparent in the first three laminae of the dorsal horn. In somatic and dendritic profiles, the immunolabel was present over the general cytoplasm and mitochondria; in the terminals, it was found over small, agranular vesicles, mitochondria and, at times, synaptic densities. Quantitative estimation indicated that the colloidal gold density in the glutamate-immunoreactive terminals was five-fold more than in any other neuronal profile. Both glutamate- and aspartate-immunopositive terminals made asymmetric synaptic contacts onto unlabelled dendrites; glutamate-positive terminals often formed the core of type I and II glomeruli. After double labelling of the same sections, glutamate and aspartate immunoreactivities consistently occurred in different axonal and terminal profiles. In these preparations, it was clearly seen that glutamate-immunoreactive terminals were far more numerous than (more than 10-fold) those immunoreactive for aspartate. Double labelling for glutamate or aspartate and GABA also revealed distinct staining of different terminals. Simultaneous immunolocalization of each of the amino acids and synaptophysin showed the amino acid and glycoprotein immunoreactivities co-localized in small, agranular vesicles in immunoreactive terminals. Finally, triple labelling of the same sections for glutamate, calcitonin gene-related peptide and substance P revealed that glutamate was often co-localized with either of the two neuropeptides in the same axonal boutons; terminals that showed simultaneous labelling for glutamate, calcitonin gene-related peptide and substance P were also noted. In all cases, the glutamate immunoreactivity was restricted to small, clear vesicles whereas the neuropeptide immunoreactivities were present in larger, dense-cored vesicles. Our observations demonstrate that there is an abundant glutamate immunoreactivity in the superficial layers of the rat dorsal horn, localized in neuronal profiles distinct from those containing aspartate or GABA. Most glutamate-positive synaptic terminals are probably of primary sensory origin since they were particularly numerous in the spinal area where most small and medium-sized sensory afferents terminate and since they formed the core of type I and II glomeruli. In addition, a number of glutamate-positive terminals were seen to contain immunoreactivity for calcitonin gene-related peptide and/or substance P, two neuropeptides contained in many primary sensory afferents. That the amino acid and neuropeptide immunoreactivities were present in different subcellular organelles raises the possibility of selective mechanisms of release and thus of differential effects on postsynaptic structures.