The ultrastructure of normal and neurotoxin-treated human detrusor muscle was studied to define the bases for myogenic and neurogenic control and to determine the extent and selectivity of effects of the toxins. No gap junctions were observed in the 30 strips studied but there were other cell-to-cell junctions which may allow cell-to-cell coupling. Nerve varicosities in the muscle bundles were present in high frequency (9 to 15 profiles per 100 muscle cell profiles); they sometimes formed close contacts with muscle and most were bare or partly bare of glial covering. They were often present as single axons or in bundles of only a few nerves. No myelinated nerves or nerve cell bodies were seen in or near muscle bundles. This muscle seemed to be designed primarily for neural control of contractile function with single nerves innervating discrete muscle regions. Most of the nerve varicosities contained a large majority of small agranular vesicles along with some large granular vesicles, but about 20% contained a few small granular vesicles as well. The proportion of the small granular vesicles was not increased but the percentage of varicosities containing them increased slightly after treatment with 5- or 6-hydroxydopamine. Scorpion venom seriously damaged 80% of nerve varicosities, sparing no particular class, but spared axons from structural damage. Beta-bungarotoxin had similar effects. 4-Aminopyridine did not induce gap junction formation but had neurotoxic effects. These data suggest that scorpion venom is a useful tool to eliminate selectively physiological functions dependent on synaptic vesicles, that all nerve types contain scorpion venom binding Na channels and that the persistent atropine- and tetrodotoxin-insensitive responses of this muscle involve mechanisms independent of synaptic vesicles observable by electron microscopy. The common occurrence of multilamellar membranous structures in bladder muscle cells was reported and their origin suggested.