1. We have investigated the facilitation of extracellularly recorded responses of ON-OFF directionally selective (DS) ganglion cells of the rabbit retina to two-slit preferred-direction apparent motion produced by both prolonged light steps, which simulate movement of an edge past two apertures, and light flashes, which simulate movement of a spot or slit. 2. Within the excitatory receptive-field center of DS ganglion cells, apparent motion with prolonged light steps elicits preferred-direction facilitation whose rise time (220 ± 150 ms, average rise to 90% of maximum for 6 cells) is typically longer than the rise time of the excitatory response elicited by each slit. The decay time to baseline of facilitation during prolonged light steps is generally longer than 500 ms and hence greatly exceeds the typical duration of the excitatory response elicited by the slits. 3. Prolonged light steps are generally effective for facilitating any given excitatory receptive- field locus from a roughly ovoid area that typically extends on the order of 100-200 μm in the preferred direction, which is less than one-half the size of the excitatory receptive-field center. Within 100 μm, facilitation can occur for motion diagonal to the preferred-null axis as long as the projection of the motion on the preferred-null axis points in the preferred direction. 4. The time course of preferred-direction facilitation between two slits does not appear to have a strong systematical dependence on the interslit distance over the range in which facilitation is effective. 5. Short light flashes are ineffective for eliciting facilitation and, at sufficiently long interslit delays, elicit inhibition all around the test slit. This inhibition may be due to the antagonistic surround mechanism within the receptive-field center, which is effectively elicited by short- duration stimuli. 6. The effect of preferred-direction facilitation is addition-like, rather than multiplication-like. That is, the facilitatory effect of the first slit appears as the addition of a fixed value to the response-versus-contrast curve of the second slit, rather than a multiplication of the curve by a constant factor. The functional relationship between strength of facilitation and contrast of the first slit is sigmoidal, however, and thus nonlinear. 7. Experiments with long light steps show that the interaction between excitation and preferred-direction facilitation is largely segregated between the ON and OFF pathways. Thus, for preferred- direction apparent motion, light onset of the first slit strongly facilitates the response to the onset, but generally not the offset, of the second slit. Light offset of the first slit similarly tends to facilitate only the response to light offset of the second slit. However, at short distances, where like-sign interactions are significantly facilitatory, opposite sign interactions tend to be inhibitory. 8. We discuss several models for facilitation from the perspective of the above data and propose a new class of models, called 'gated-enhancer' models, that can account for the data. The gated-enhancer model postulates that an enhancer (such as a molecule) that is produced by a previous light stimulus enhances the response to subsequent stimuli along trajectories corresponding to preferred-direction movement. This enhancer cannot produce a response by itself, however, because it is only effective when gated by a subsequent, active input.