We have examined the morphology and longitudinal axon projections of a population of spinal commissural interneurons in young Xenopus tadpoles. We aimed to define how the distribution of axons of the whole population constrains the longitudinal distribution of the inhibition they mediate. Forty-three neurons at different positions were filled intracellularly with biocytin and processed with avidin-conjugated horseradish peroxidase. Soma size did not vary longitudinally and only one ipsilateral axon was found. Contralateral axons ascended, descended, or usually branched to do both. Total axon length and the extent of dendritic arborisation decreased caudally. The distributions of ascending and descending axon lengths were different; there were more long ascending (mean 737 +/- standard deviation 365 microm) than long descending (447 +/- 431 microm) axons. We used the axon length distribution data with existing data on the distribution of commissural interneuron somata to calculate the overall longitudinal density of these inhibitory axons. Axon numbers showed a clear rostrocaudal gradient. Axon length distributions were then incorporated into a simple spatiotemporal model of the forms of inhibition during swimming and struggling motor patterns. The model predicts that the peak of inhibition on each cycle will decrease from head to tail in both motor patterns, a feature already confirmed physiologically for swimming. It also supports a previous proposal that ascending inhibition during struggling shortens cycle period by shortening rostral motor bursts, whereas descending inhibition could delay subsequent burst onset.