Consecutive sections of rat cerebella were incubated with antisera raised against glycine or gamma-aminobutyric acid (GABA) conjugated to protein by glutaraldehyde. The sections were subsequently processed according to the peroxidase-antiperoxidase technique (semithin sections) or treated with a secondary antibody coupled to colloidal gold particles (ultrathin sections). Corroborating previous light microscopic observations based on pre-embedding immunocytochemistry, a major proportion (about 70%) of the Golgi cell bodies showed immunoreactivity for both glycine and GABA. Analyses of semithin sections further suggested that the two immunoreactivities were colocalized in the same glomeruli and even in the same Golgi cell terminals. This was confirmed by electron microscopy. Quantification of the immunogold labelling for glycine (which is assumed to play metabolic roles in addition to its presumed role as a transmitter) showed that the net gold particle density was an order of magnitude higher over Golgi cell terminals than over the other constituents of the cerebellar glomeruli (mossy fibre terminals and granule cell dendrites). The total particle density over the latter was only slightly higher than the background level (over empty resin), suggesting that the concentration of 'metabolic' glycine is generally low compared to the concentration of glycine in Golgi cells. The stellate and basket cell terminals (which similarly to the Golgi cells are thought to release GABA as transmitter) were immunoreactive for GABA, but (with very few exceptions) virtually unlabelled for glycine, suggesting that our results were not confounded by any crossreactivity of the glycine antiserum with fixed GABA. Direct evidence that the sera reacted selectively with fixed glycine or GABA under the conditions used was obtained by incubating the tissue sections together with test sections containing a series of different amino acid-glutaraldehyde-brain macromolecule conjugates. Adsorption tests with soluble amino acid-glutaraldehyde complexes similarly suggested that the double-labelling of the Golgi terminals indeed reflected a colocalization of glycine and GABA. The results show that two 'classical' transmitters, both being inhibitory and acting on Cl- channels, may coexist in the same nerve terminals.