We examined primary afferent depolarization (PAD) in the anesthetized cat elicited in 109 pairs of intraspinal collaterals of single group I afferents from the gastrocnemius nerve, one of the pair ending in the L3 segment, around the Clarke's column nuclei, and the other in the L6 segment within the intermediate zone. Tests for refractoriness were made to assess whether the responses produced by intraspinal stimulation in the L3 and L6 segments were due to activation of collaterals of the same afferent fiber. PAD in each collateral was estimated by independent computer-controlled measurement of the intraspinal current required to maintain a constant probability of antidromic firing. In most fibers, stimulation of the ipsilateral posterior biceps and semitendinosus (PBSt) nerve with trains of pulses maximal for group I afferents had a qualitatively similar effect but produced a larger PAD in the L6 than in the L3 collaterals. Stimulation of cutaneous nerves (sural and superficial peroneus) with single pulses and of the posterior articular nerve, the ipsilateral reticular formation, nucleus raphe magnus and contralateral motor cortex with trains of pulses often had qualitatively different effects. They could produce PAD and/or facilitate the PBSt-induced PAD in one collateral, and produce PAH and/or inhibit the PAD in the other collateral. These patterns could be changed in a differential manner by sensory or supraspinal conditioning stimulation. In summary, the present investigation suggests that the segmental and ascending collaterals of individual afferents are not fixed routes for information transmission, but parts of dynamic systems in which information transmitted to segmental reflex pathways and to Clarke's column neurons by common sources can be decoupled by sensory and descending inputs and funneled to specific targets according to the motor tasks to be performed.