1. The equilibrium relationship between acetylcholine (ACh) concentration and response (fraction of channels open), corrected for the effects of desensitization, has been estimated by single-ion-channel recording at the adult frog skeletal neuromuscular junction. At high ACh concentration channel openings occur in well-defined clusters separated by long desensitized intervals. The response, po, was estimated as the proportion of time for which a single channel was open during a cluster. 2. At negative membrane potential (-120 mV) po reached a maximum value of 0.9 at 100 microM-ACh and was half-maximum at 15 microM with a Hill slope of 1.6 at this point. At concentrations higher than 200 microM-ACh, po declined as a result of open-channel block by free ACh itself. 3. At positive membrane potentials (+100 mV) there was little channel block by ACh; po reached a maximum value of 0.41 at 500 microM-ACh, with half-maximum activation at 50 microM and Hill slope of 1.2 at this point. 4. Particular mechanisms for channel activation by ACh were fitted to the data by the method of least squares. Fits were fully determinate only if the two binding sites for ACh were assumed to be equivalent with no co-operativity in the ACh binding reactions. At negative potential the microscopic equilibrium constant for binding was K1 = K2 = 77 microM and the equilibrium constant for channel opening (opening/closing rates, beta/alpha) was 32. At positive potential the affinity was slightly higher, K = 32 microM, which confirms the view that the binding sites for ACh are outside the membrane electric field. The equilibrium constant for channel opening was reduced to 0.7 mainly as a result of the much shorter open lifetime (increased closing rate alpha) at positive potentials. 5. The data were also fitted well by very high values of beta/alpha together with a high degree of negative co-operativity or non-equivalence in ACh binding affinity (K2 much greater than K1). A good fit could also be obtained with moderate positive co-operativity combined with non-equivalence of the binding sites. 6. A mechanism that postulates a receptor with two independent gating subunits provided a poor fit to the data at negative potential. 7. The rate constants for channel opening and ACh dissociation were estimated by constraining the fitted parameters so that the burst length for channel opening was equal to its observed value at low concentrations of ACh.(ABSTRACT TRUNCATED AT 400 WORDS)