1. Mechanisms underlying the selective expression of different motor patterns in vertebrates are poorly understood. Immobilized, spinal Xenopus embryos are used here to examine the motor patterns evoked by various concentrations of excitatory amino acids. 2. Relatively low concentrations of N-methyl-D-aspartate (NMDA) (40-60 microM), kainate (7-8 microM), and alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) (5 microM) evoked motor root discharge characteristic of swimming. Brief applications of higher concentrations of kainate (20-40 microM), AMPA (25-30 microM), quisqualate (5 microM), and glutamate (1-4 mM) evoked sequences of a different motor pattern: struggling. This is characterized by a longer cycle period, increased burst duration, and a reversed longitudinal pattern of motor root discharge. The struggling pattern was never evoked by higher concentrations of NMDA (300-500 microM), but was evoked by 30 microM AMPA or 5 microM quisqualate in the presence of 50 microM D-2-amino-5-phosphonopentanoic acid. 3. Intracellular recordings from presumed spinal motoneurons showed different patterns of activity during agonist-evoked swimming and struggling. The patterns were like those described previously during sensory-evoked behavior. 4. Caudal applications of excitatory amino acids that produced struggling discharge did so only at caudal motor roots, whereas caudal applications of NMDA evoked swimming activity throughout the spinal cord. 5. During excitatory-amino-acid-evoked struggling, sensory Rohon-Beard neurons depolarized up to 7 mV, but did not fire. 6. The results show that expression of the struggling pattern, like swimming, is not critically dependent on sensory discharge. The results are also consistent with the idea that expression of the two very different motor patterns for swimming or struggling in this simple vertebrate preparation can be controlled by the level of excitation within the spinal motor circuitry, and need not involve the activity of a specific external neuromodulator.