1. Simultaneous electromyographic (EMG) records were obtained from a single-joint extensor muscle of each of the four limbs of intact cats during repeated overground stepping trials. 2. In each limb, the temporal spacing of step cycles was determined by measurements of the intervals between consecutive terminations of EMG activity, since this occurs in a consistent relationship to the removal of the limb from the ground. By measuring the latencies between step cycles so determined, the temporal spacing of step cycles between limbs was determined. Each latency was expressed as a function of step duration or as a phase interval. 3. Analysis of the cooordination of step cycles of both homologous limb pairs (the forelimbs and hindlimbs), both homolateral limb pairs (the fore- and hindlimb on the right and left sides), and both sets of diagonal limbs suggest that the step cycles of the four limbs are coordinated according to a few frequently occurring patterns. However, the representation of a large number of phase intervals between these preferred patterns indicates a substantial amount of variability in interlimb coupling. 4. Analysis of the interaction of different interlimb-coupling patterns indicates that during alternate coordination of hindlimbs, coupling of the other limbs is fairly predictable. The step cycles of the forelimbs and hindlimbs are spaced according to a trotting form of coupling. During in-phase coordination of hindlimbs, the patterns of coordination of the other limbs are more diffuse. Forelimbs step cycles are coupled via a number of different modes, as are those of the forelimbs and hindlimbs. 5. It is concluded that the step cycles of different limbs are coordinated, but the association of observed patterns of coordination with any known neural pathways or the interaction of neural pathways should be approached with caution. The variability about the frequently occurring patterns is interpreted as an expression of the faculatative capabilities of the neural mechanisms controlling locomotion. Thus, these data favor a model of interlimb control during stepping, which recognizes preferred patterns of coordination and the variability about these patterns.