The present study has revealed that OC axons gave rise to a number of thin collaterals. Due to the abundance of these non-CF thin collaterals, it seems better to make a distinction between the terms CFs and OC axons, as was done in the present paper. The present findings on the innervation of PC dendrites by CFs are basically similar to those in previous reports (Ramón y Cajal, 1911; Palay and Chan-Palay, 1974). The number of swellings on a single CF in the present study (n = 250) is comparable to a previously measured value in the rat (n = 288; Rossi et al., 1993) and larger than a value in the frog (n = about 100 beads; Llinás et al., 1969). The average number of CFs per OC axon in this study was close to the number (n = about 7) inferred in the rat by counting the total number of IO neurons and PCs (Schild, 1970). Contact of interneurons by some swellings of CFs in the molecular layer was emphasized by Scheibel and Scheibel (1954) in their study with Golgi staining. Despite the contact of CF terminals on interneurons, the formation of a synaptic structure between them has been excluded in an electron-microscopic study (Hámori and Szentàothai, 1980). On the other hand, electrophysiological studies have demonstrated a weak excitatory effect of CFs on some interneurons (Eccles et al., 1966). Terminals in the granular layer were originated either from thin collaterals of OC axons or from retrograde collaterals of CF terminal arborizations. The former was the main source of swellings in the granular layer. The morphology of the thin collaterals in the present study was consistent with "globose varicosities connected by a fine thread" as described in Golgi preparations and electron micrograms (Chan-Palay and Palay, 1971). Swellings of thin collaterals (about 1.7% of the total number of swellings per OC axon) were most abundant in the upper portion of the granular layer just underneath the PC layer, in which Golgi cells are usually located. Furthermore, some of these swellings were observed to touch presumed Golgi cells in the present study, which is consistent with electron-microscopic findings on the innervation of somata of Golgi cells by thin collaterals (Hámori and Szentàothai, 1980; Chan-Palay and Palay, 1971). Inferior olive stimulation has been shown electrophysiologically to have a weak direct excitatory effect on Golgi cells (Eccles et al., 1966). Ninety-one percent of the OC axons examined had nuclear collaterals; since the possibility of insufficient staining could not be excluded, this percentage may be an underestimation. The ratio of swellings in the cerebellar nuclei versus those of CF terminal arborizations was about 0.036 in individual OC axons in the present study. However, since the volume of the cerebellar nuclei is much smaller than that of the cerebellar cortex, and significant convergence of input from OC axons to cerebellar nucleus neurons is present (Sugihara et al., 1996), cerebellar nucleus projection of OC fibers can still be functionally important. Some swellings seemed to make contact with the soma and the proximal portions of dendrites of large neurons in the present study, which is consistent with the steep rising phase of postsynaptic excitatory potentials in cerebellar nucleus neurons following IO stimulation (Kitai et al., 1977; Shinoda et al., 1987). Although intracellular potentials were presumably recorded only from large output neurons in the cerebellar nuclei, the present study suggested that small neurons were also innervated by OC axons. The present study revealed that virtually all reconstructed LRN axons projected not only to the Cx as mossy fibers, but also to the DCN including the VN by their axon collaterals. None of the LRN neurons specifically projected to the DCN without projecting to the Cx, namely all axon terminals of LRN neurons in the DCN and VN belonged to axon collaterals of mossy fibers projecting to the Cx. (ABSTRACT TRUNCATED)