1. Gradients of excitation in the swimmeret system were created by applying either pilocarpine or carbachol to selected ganglia in isolated abdominal nerve cords. The state of the system was monitored in each segment with extracellular electrodes on nerves that innervated swimmerets. In preparations that were quiescent before drugs were applied, these cholinergic agonists elicited well-coordinated swimmeret motor patterns from the entire system, including ganglia that were not directly treated with pilocarpine or carbachol. 2. The periods of these patterns depended on the number of ganglia that were directly excited. As this number increased, period decreased. When the same numbers of ganglia were excited by direct application of a drug, the mean period of the swimmeret activity elicited by pilocarpine was greater than that elicited by carbachol. 3. Selective excitation of anterior or posterior ganglia caused significant changes in intersegmental phase at the boundary between excited and nonexcited regions of the nerve cord. When only anterior ganglia were excited directly, the phases of their power-stroke activity relative to the most posterior ganglion were advanced. When only posterior ganglia were excited directly, the phases of power-stroke activity in more anterior ganglia were retarded. Neither pilocarpine nor carbachol caused a complete reversal of the normal phase relations of the swimmeret motor patterns. 4. These results are consistent with an asymmetric-coupling model of the intersegmental coordinating circuit of the swimmeret system but contradict an alternative excitability-gradient model.