1. In the motor system of the jellyfish, Aglantha digitale, there are eight giant axons connected by chemical synapses to a muscle epithelium. The simplicity of this structure makes it possible to assess the contribution of different ion conductances in the axon membrane to the two forms of swimming that provide the behavioral output of the system. In situ recordings from large clusters of ion channels provide a means of studying these membrane conductances in isolation so that the features that permit them to perform their behavioral function may be identified. 2. In Aglantha motor axons, low-amplitude, low-threshold spikes are associated with slow swimming, whereas escape swimming depends on a higher-threshold, overshooting action potential. The action potential was abolished by a sodium-free (choline-containing) bathing medium but was resistant to tetrodotoxin (0.09 mM; 3 x 10(-5) g/ml). It was prolonged by tetraethylammonium (TEA) ions (50 mM) but little affected by changes in holding potential in the range of -51 to -82 mV. The low-threshold spikes were unaffected by sodium-free saline containing TEA (30 mM). They were inactivated by holding the membrane potential at -51 mV. Average axon resting potentials were -63 +/- 6 (SD) mV (n = 17). 3. Shortened axons studied with the two-electrode voltage-clamp technique had a transient inward current with a low threshold for activation (about -60 mV). The inward current was fully inactivated at -51 mV; it was present in sodium-free saline and abolished by Mg2+ (120 mM) just like the low-threshold spike. 4. Calcium-dependent low-threshold spikes and sodium action potentials coexist in the same axons but may be elicited separately because an outward current limits the peak of the low-threshold spike to a level below the threshold of the action potential (about -20 mV). 5. Analysis of ensemble currents showed that axon-attached membrane patches contained clusters of different voltage-dependent potassium channels. Three channel classes were distinguished by prepulse inactivation experiments. All three channels were found to inactivate, but they had different voltage-dependencies and different inactivation kinetics (fast, intermediate, or slow). Recovery from inactivation was slow in each case (time constant 2-10 s). 6. All axon-attached membrane patches were found to contain one or two of the three classes of potassium channel. Channels with intermediate kinetics were found less frequently and may have been present at lower density.(ABSTRACT TRUNCATED AT 400 WORDS)