Neurons containing glutamatic acid decarboxylase (GAD) are known to exist in the spinal dorsal horn, dorsal column nuclei (DCN), n. ventralis posterior (VP), and somatosensory cortex of cats. Recent work suggested that species differences exist concerning the presence and/or density of GAD-positive neurons in VP. The present experiments demonstrate that, in contrast with carnivores and primates, the rat's VP contains virtually no GAD-positive neurons and that virtually all neurons in it project to the cortex. This conclusion is supported by the failure to find, in Golgi-impregnated material, neurons with characteristics commonly attributed to Golgi type II neurons in VP of cats. The lack of GAD-positive neurons in VP of rats contrasts also with the presence of such neurons in the DCN in the same species. As in cats, about one third of the neurons in the cuneate n. are GAD-positive; these have mostly small perikarya and they are present throughout the nucleus. It is likely that these are intrinsic neurons, i.e. non-projecting beyond the limits of the DCN since a comparable percentage of neurons are unlabeled by simultaneous injections of horseradish peroxidase in multiple targets of the DCN. Like GAD-positive neurons, neurons unlabeled by the retrograde transport of HRP have, for the most part, small perikarya. It is possible that inhibitory mechanisms necessary for basic transfer functions in VP of rats are sustained through projections to this nucleus from the n. reticularis thalami. Extrinsic source of GABAergic input to the DCN seem to be absent or very weak. From this and previous evidence it may be proposed that intrinsic inhibitory interneurons have gradually developed in VP of rabbits, carnivores, and primates in parallel with more elaborate levels of thalamic integration of somatosensation.