The extent to which individual neural networks can produce phase-constant motor patterns as cycle frequency is altered has not been studied extensively. I investigated this issue in the well-defined, rhythmic pyloric neural network. When pyloric cycle frequency is altered three- to fivefold, pyloric inter-neuronal delays shift by hundreds to thousands of msec, and all pyloric pattern elements show strong phase maintenance. The experimental paradigm used is unlikely to activate exogenous inputs to the network, and these delay changes are thus likely to arise from phase-compensatory mechanisms intrinsic to the network. Pyloric inter-neuronal delays depend on the time constants of the network's synapses and of the membrane properties of its neurons. The observed delay shifts thus suggest that, in response to changes in overall cycle frequency, these constants vary so as to maintain pattern phasing.