Establishing a causal role between the activity of specific individual nerve cells and the behaviors they produce (or the neural computations they execute) is made difficult in vertebrate animals because of the large numbers of neurons involved. Traditional techniques for establishing causal roles, such as tract cutting and electrolytic lesions, are limited because they produce damage that affects a variety of different cell types, invariably intermingled, and often of uncertain identity. We propose here an alternative lesioning technique in which large numbers of neurons are lesioned, but the lesioned neurons are specifically identified by fluorescent labeling. We use the locomotor control system of the larval zebrafish to illustrate this approach. In this example, the technique involves injection of fluorescent dextrans into far-rostral spinal cord to label descending nerve fibers. Such injections appear to interrupt the descending nerve fibers, and therefore their accompanying locomotor control signals. This protocol is shown to produce significant behavioral deficits. Because the CNS of the larval zebrafish is transparent, the entire population of lesioned cells can be imaged in vivo and reconstructed using confocal microscopy. This large-scale lesioning technique is important, even in this relatively 'simple' vertebrate animal, because the ablation of smaller numbers of neurons, using more precise laser-ablation techniques, often fails to produce observable behavioral deficits. While this technique is most readily applied in simpler and transparent vertebrate animals, the approach is general in nature and might, in principle, be applied to any vertebrate nerve tract.