To study the organization and distribution of the inhibitory amino acid neurotransmitter GABA in the medial hypothalamus, we used a postembedding immunocytochemical approach with colloidal gold. Quantitative analysis showed that half (49%) of all synapsing boutons studied were immunoreactive for GABA, based on immunogold staining of the suprachiasmatic, arcuate, supraoptic, and paraventricular nuclei. This was corroborated with pre-embedding peroxidase immunostaining with antisera against glutamate decarboxylase, the GABA synthetic enzyme. These data suggest that GABA is the numerically dominant neurotransmitter in the hypothalamus, and emphasize the importance of inhibitory circuits in the hypothalamus. Serial ultrathin sections were used to reconstruct GABA immunoreactive boutons and axons in three dimensions. With this type of analysis we found less morphological heterogeneity between GABA immunoreactive boutons than with single ultrathin sections. Single sections sometimes showed boutons containing only small clear vesicles, and other with both clear vesicles and small dense core vesicles. However, with serial sections through individual boutons, dense core vesicles were consistently found at the periphery of the pre-synaptic GABA immunoreactive boutons, suggesting probable co-localization of GABA with unidentified peptides in most if not all boutons throughout the hypothalamus. A positive correlation was found between the density of small clear vesicles and the intensity of immunostaining with colloidal gold particles. GABA immunoreactive axons generally made symmetrical type synaptic specializations, although a small percentage made strongly asymmetrical synaptic specializations. Vesicles in GABA immunoreactive boutons were slightly smaller than those in non-reactive boutons. Synaptic efficacy is related to the position of the synapse on the post-synaptic neuron. While the majority of GABA immunoreactive axons made synaptic contact with dendrites, the distribution of GABA immunoreactive synapses on somata and dendrites was the same as would be expected from a random distribution of all boutons. No preferential innervation of cell bodies by GABA immunoreactive terminals was found. Serial ultrathin sections showed that a GABA immunoreactive axon would sometimes make repeated synaptic contacts with a single postsynaptic neuron, indicating a high degree of direct control by the presynaptic GABAergic cell. Other immunoreactive axons made synaptic contact with a number of adjacent dendrites and cells, suggesting a role for GABA in synchronizing the activity of hypothalamic neurons. Based on the density of immunogold particles per unit area, varying concentrations of immunoreactive GABA were found in different presynaptic boutons in the hypothalamus.