1. The action of tubocurarine on voltage-clamped frog muscle end-plates has been re-examined by means (a) equilibrium dose-ratio measurements, (b) current fluctuation measurements and (c) voltage-jump relaxation measurements. 2. The equilibrium measurements can be interpreted as implying that tubocurarine has (a) a competitive blocking action, with a dissociation constant of 0.34 microM, which is not dependent on membrane potential, and (b) an additional voltage-dependent blocking action. 3. In the presence of tubocurarine two kinetic components can be seen. The faster one is similar to, but rather faster than, the normal ion channel closing rate. The other is much slower (1--3 sec), and, in relaxation experiments it is in the opposite direction to the fast relaxation. 4. A number of alternative explanations for the results are discussed. The mechanism that fits them best appears to be a combination of competitive block (or block of shut channels), with a strongly voltage-dependent block of open ion channels by tubocurarine. Estimates of the rate constants for channel blocking (and their voltage dependence) are derived. From these estimates the dissociation constant for the binding of tubocurarine to open channels appears to be roughly 0.12 microM at --70 mV and 0.02 microM at --12 mV. 5. Several potential sources of error in the experiments, and in their interpretation, are discussed. The most serious of these are problems associated with diffusion in the small volume of the synaptic cleft, viz. (a) changes in cleft concentration consequent on changes in binding, and (b) ionophoretic flux of antagonist and agonist into the synaptic cleft.