We studied the relationship between the receptive fields of directionally selective retinal ganglion cells and the dendritic arbors of the same cells. The cells were recorded from extracellularly under visual control and then injected with Lucifer yellow. The arbor of Lucifer-filled dendrites could then be directly compared with the properties of the receptive field. A large population of on-off directionally selective cells was injected and drawn. The directionally selective ganglion cells had bistratified receptive fields similar to those previously described by others in the central retina. In the periphery, the dendritic fields became larger, rounder, and sparser than centrally. The diameters of the dendrites were measured in living or lightly fixed retinas; they were found to be somewhat larger than previously estimated by electron microscopy. The local structure of the dendritic arbor bore no obvious relation to the directional properties of the cell. The receptive fields of most cells were centered symmetrically around their dendritic fields. For about 10% of the cells, however, the receptive field was displaced. The displacement was always toward the preferred direction, relative to the dendritic field. The meaning of these shifts is not clear. In both cases, the diameter of the receptive field exceeded the diameter of the dendritic field only slightly; in our sample, the diameters of the receptive fields averaged 6% larger than the dendritic fields. This means that the neurons afferent to the directionally selective ganglion cells must either have narrow dendritic fields or, if they are wide spreading, have dendrites that do not conduct effectively along their length. It also means that the observed spread of neurobiotin between DS ganglion cells (Vaney, 1991) must be due to a very few gap junctions, or to some mechanism other than a gap junction.