We have studied, in the salamander (Ambystoma tigrinum), an apparent hindrance to the sprouting of certain segmental nerves into adjacent skin supplied by neighbouring segmental nerves. The mechanosensory fields of the dorsal skin of the hindlimb were examined by recording the afferent impulses evoked by a fine bristle applied to the skin. Three segmental nerves normally supply the handlimb, and together divide the dorsal skin into two major mechanosensory fields, an anterior one innervated by both the 15th spinal nerve (N15) and the anterior division of N16 (N16A), and a posterior one innervated by N17 and N16P. Denervations were done so as to leave only one of the two major fields intact, that of N15 (with or without an intact N16A) or of N17 (with or without N16P). The results were surprising: after a small (but significant) expansion of the remaining N15 or N17 field the sprouting of either of these nerves into the adjacent denervated skin ceased for a total post-operative period of at least two months; however, when N16A or N16P was present the N16 axons sprouted apparently without hindrance into the adjacent skin (showing that there was no mechanical barrier to sprouting at the frontier zone). The failure of N15 or N17 to invade the adjacent territory was not because their axons had reached an upper limit in their capacity to enlarge the size of their terminal fields: when a proportion of the axons of a remaining N17 was eliminated, the remainder sprouted progressively over the following weeks. Significantly they invaded denervated skin that was almost entirely confined to the territory of the `parent' N17; the immediately adjacent skin formerly supplied by N15 was largely ignored. These results, and similar ones for N14, N16 and N18, suggest that the segmental cutaneous axons are hindered from sprouting to any great extent into skin normally occupied by a neighbouring segmental nerve, even when that nerve is removed. The axons will sprout readily within the territory of the parent nerve, and we call these territories `domains'; their borders usually correspond fairly closely to those of the segmental dermatomes. Regenerating nerves freely invaded `foreign' skin, and thus were not susceptible to these spatial constraints. After a two month period the intact axons of a remaining N17 began freely to invade the adjacent domain, as though the constraint had suddenly disappeared. To test for skin `specificity', skin flaps were excised, and reimplanted after rotation by 0-180 degrees, in limbs that were either fully innervated or partially denervated. In about half of the instances the segmental nerves sprouted to re-establish their former spatial territories in the flaps with respect to the coordinates of the limb, without regard for the type of skin that they innervated (i.e. their former skin or `foreign' skin). In the remaining cases the invasion of the flaps was often indiscriminate, as though by regenerating nerves. The spatial constraint thus seemed to operate at the borders of domains, and to relate to the limb as a whole, rather than to skin itself. The possible genesis of domains, and their significance, are discussed.