Preparations of the nearly isolated leech nerve cord containing as few as two ganglia are sufficient to generate intersegmentally coordinated swim oscillations, provided that they receive tonic excitation from other segments via the median connective (Faivre's nerve). Due to their greatly reduced complexity, these preparations should provide useful experimental models of neuronal coordination. As a step in the development of such models, we have characterized the intersegmental coordination of nerve-cord chains ranging from 2 to 18 ganglia in length. We found that increases in swim-cycle period give rise to increases in intersegmental delay between homologous motoneuron bursts. Thus the intersegmental phase relationships are nearly independent of period. The relationship between intersegmental delay and period is approximately linear and extrapolates to intersect the period axis at approximately 0.3 s. This value is in close agreement with the analogous measure derived from tension measurements in the intact swimming leech. Chain length (number of connected ganglia in a preparation) has a pronounced influence on the magnitude of intersegmental phase lag. The longest chains (18 ganglia) exhibited phase lags of approximately 8 degrees per segment, whereas for pairs of ganglia the phase lag was approximately 40 degrees per segment. This dependence of phase lags on chain length was apparent at both the motor and oscillator levels. The intersegmental phase lag is not the same in all parts of the nerve cord. Rather, it increases steadily toward the posterior end of the chain, providing a deceleration in the rearward progression of the metachronal activity. The rearward increase in intersegmental phase lag is paralleled by a propensity of chains taken from more posterior sections of the nerve cord to exhibit larger phase lags. That is, there appears to be a phase-lag gradient intrinsic to the nerve cord to account for the deceleration of activity. The anterior and posterior ends of an isolated nerve cord continue to exhibit phase-locked bursting when an intervening section of five ganglia is bathed in elevated Mg2+ saline. Thus, information sufficient to coordinate oscillations in separate ganglia travels at least six segments. The phase lag across the blocked section is reduced but within each unblocked section is increased so that the phase lag between extreme ends is nearly unchanged. This altered burst pattern is due to a combination of synaptic block in segmental ganglia and conduction block in through-fibers.