1. The effects of exogenous ATP or adenosine on end-plate currents (e.p.cs; evoked by simultaneous action of a few hundred quanta of ACh) or on miniature e.p.cs (m.e.p.cs) were studied under voltage clamp conditions on frog sartorius muscle fibres. 2. ATP or adenosine (100 microM(-1) mM) reduced the e.p.c. amplitude but did not affect m.e.p.c. amplitude, decay time constant and voltage-dependence of m.e.p.c., suggesting that e.p.c. depression induced by these purines had presynaptic origin only. 3. The action of ATP, unlike that of adenosine, was prevented by the P2-purinoceptor antagonist suramin (100 microM). The stable ATP analogue alpha,beta-methylene ATP (100 microM), known to be desensitizing agent on P2X receptors, also abolished the depressant effect of ATP while sparing the action of adenosine. Concanavalin A, an inhibitor of ecto-5'-nucleotidase, did not affect the presynaptic action of exogenously applied ATP. 4. The presynaptic action of adenosine was prevented by theophylline (1 mM), a blocker of adenosine receptors, while the effect of ATP was not changed under these conditions. The selective blocker of A1 adenosine receptors, 8-cyclopentyl-1,3,dipropylxanthine (DPCPX; 0.1 microM), abolished the presynaptic action of adenosine but did not prevent the depressant effect of ATP. 5. The effects of ATP and adenosine (at nearly saturating concentration) were additive suggesting that these purines activated not only distinct receptors but also different intracellular signalling mechanisms. 6. In contrast to the hypothesis that at the neuromuscular junction ATP reduces transmitter release via enzymatic degradation to presynaptically active adenosine, our data suggest that ATP (through its own presynaptic receptors) directly inhibits ACh release. Thus, ATP and adenosine might be almost equipotent as endogenous prejunctional neuromodulators at the neuromuscular junction.