In the preceding companion article (Berkowitz and Stein, 1994b), we showed that many descending propriospinal neurons in the turtle were rhythmically activated during two different motor patterns, fictive rostral scratching and fictive pocket scratching. In this article, we present phase analyses of the activity of each such neuron during fictive scratching. Each neuron's activity was concentrated in a particular phase of the ipsilateral hip flexor muscle nerve (VP-HP) activity cycle; each had a distinct "preferred phase." Each neuron's preferred phase during fictive rostral scratching was similar to its preferred phase during fictive pocket scratching. This result is consistent with the idea that some descending propriospinal neurons may contribute to the generation of both rostral scratching and pocket scratching. Many descending propriospinal neurons were rhythmically activated during fictive scratching evoked on either side of the body. This activity may contribute to production of bilateral hindlimb movements during scratching. It is also possible that synaptic interactions between the two sides of the spinal cord may be important in generating the motor patterns for movement of a single hindlimb. In addition, we present a model which illustrates that a population of propriospinal neurons, each of which is broadly tuned to a region of the body surface and is rhythmically activated in a constant phase of the hip control cycle, could mediate the selection and generation of rostral scratching and pocket scratching. Thus, the selection of an appropriate motor pattern and the production of the required knee-hip synergy may each be distributed over a diverse population of spinal cord neurons. This model requires that each such neuron project to both knee muscle and hip muscle motoneurons. According to this model, the process of selecting a motor pattern would not be completed until knee muscle motoneurons integrate overlapping excitatory and inhibitory inputs.