The patterns of reciprocity between retrogradely labeled thalamocortical cells of origin and anterogradely projecting corticothalamic axon terminals were studied in the subdivisions of the adult rat medial geniculate body following auditory cortical injections of mixtures of horseradish peroxidase and [3H]leucine. The labeling produced by each method was examined independently, both qualitatively and quantitatively, in adjacent series of tetramethylbenzidine-processed sections and in autoradiographs after 24-96 hour survivals. The distribution and number of labeled cells and axon terminals were assessed separately for each method and compared systematically throughout the rostro-caudal extent of the medial geniculate complex. The principal finding was that zones containing many retrogradely labeled neuronal somata are not completely coextensive with areas of heavy terminal labeling within the medial geniculate body, although there is a gross congruence of thalamocortical-corticothalamic projections. Conversely, we found many zones of autoradiographic silver grains without retrogradely labeled somata in the adjacent sections; in general, the autoradiographic zones of non-reciprocity were more extensive and marked than were retrograde zones of non-reciprocity. The rat medial geniculate complex could be subdivided on the basis of its neuronal organization, cytoarchitecture, fiber architecture, and thalamocortical and corticothalamic connections into three major parts: the ventral, dorsal, and medial divisions. This pattern of organization was comparable, though not identical, to that of the corresponding subdivisions in the cat medial geniculate body (Winer: Adv. Anat. Embryol. Cell Biol. 86:1-98, '85). While the retrograde labeling appeared to mark many of the different types of neurons in each of the three divisions, there were distinct local and quantitative and qualitative differences in the distribution of autoradiographic terminal labeling. The ventral division received the heaviest cortical input, the medial division the least labeling, while the dorsal division was intermediate. Thus, corticogeniculate projections to the ventral division often produced values 20-100 times above background (absolute values: 2,001-10,000 silver grains/14,400 micron2; background: less than 100 silver grains/14,400 micron2); the same projection to the dorsal division usually resulted in grain counts no more than 5-20 times above background (501-2,000/14,400 micron2), while in the medial division the number of silver grains rarely exceeded two to five times the background (201-500/14,400 micron2).(ABSTRACT TRUNCATED AT 400 WORDS)