The existence of a cerebellar projection from the dorsal column nuclei (gracile and cuneate nuclei, DCN) has been proposed on electrophysiological grounds but questioned when studied with neuroanatomical techniques. The retrograde transport of horseradish peroxidase (HRP) has been used for the present study and provides anatomical evidence of a DCN-cerebellar pathway. In adult cats, 1 to 6 mul of 30% HRP were injected in pars intermedia of the anterior lobe (lobules IV-V), in paramedial lobule and in vermis of the anterior (lobules IV-V) and of the posterior lobe (lobule VII). After survival of 24 to 48 hours, all animals were perfused with a double aldehyde mixture and serial 40 mu sections through the medulla oblongata were incubated for visualization of HRP. In all cases, medullary nuclei known to project to the injected cortical regions of the cerebellum contained HRP-positive neurons mainly ipsilateral to the injection (e.g., external cuneate nucleus) or mainly contralateral to it (e.g., inferior olivary complex). Following ipsilateral injections in either the paramedian lobule or the pars intermedia, HRP-positive neurons in the cuneate nucleus were concentrated in its rostral portion where multipolar cells with radiating dendrites predominate. In contrast, none of the clusters region, in the caudal part of the cuneate nucleus, displayed HRP-positive granules. In cases in which the anterior vermis was injected a few labelled cells were present in the rostral part of the gracile nucleus but not in the clusters region of this nucleus. No labelling of DCN neurons was evident after posterior vermis injection. To compare the distribution of cells contributing to the DCN-cerebellar pathway with that of thalamic relay cells in the DCN, 0.5 to 3 mul of 30% HRP were injected in the nucleus ventralis posterolateralis of the thalamus in another series of cats. Contralateral to the thalamic injection, labelled cells were concentrated in the clusters region of the gracile and cuneate but rostrally in these nuclei they were scattered among unlabelled neurons. The preferential location in the DCN of cells which project to the cerebellum and of cells which project to the thalamus stresses the heterogeneous organization of these nuclei along the rostrocaudal axis. Further, the results indicate that regions of the DCN which have been distinguished on the basis of cytoarchitectonics (Kuypers and Tuerk, '64) and of afferents (Rustioni, '73, '74) differ also in their efferent projections.