1. Conventional electrophysiological techniques were used to record from isolated rat phrenic nerve-hemidiaphragm preparations. After periods of rest (20 min) or nerve stimulation (7/sec for 20 min) the bathing medium of the preparation was removed and assayed for adenosine triphosphate (ATP) and adenosine diphosphate (ADP) using a sensitive modification of the firefly luciferase method (Silinsky, 1974). 2. In the presence of tubocurarine and normal (2 mM) calcium, fourteen periods of nerve stimulation (eight preparations) caused the appearance of ATP and/or ADP in amounts ranging from 28 to 641 p-mole. Experiments using carbachol (30 muM or 1 mM) suggested that this nucleotide efflux was not produced by a secondary action of released acetylcholine (ACh). 3. Stimulation of isolate phrenic nerve trunks at 7/sec for 20 min did not cause the efflux of ATP or ADP. 4. In solutions of normal osmotic pressure and reduced calcium concentrations (0-1 mM or 'calcium-free'), stimulation failed to release adenine nucleotide from non-contracting preparations. 5. Diaphragms were bathed in normal calcium and indirectly stimulated at 11/sec for 80-90 min in the presence of 5 times 10-minus 5 M hemicholinium-3. After all detectable signs of ACh release were eliminated, nerve stimulation failed to release ATP or ADP. 6. These results in conjunction with experiments on the hydrolysis of exogenous ATP suggest that ATP is released from the motor nerve ending and is subsequently degraded by enzymatic activity. It is also suggested that the released nucleotide may be derived from the cholinergic vesicle.