1. I have used several methods to estimate the conduction velocities of retinal afferents innervating the upper 50-100 micron of the stratum griseum superficiale (the upper SGS). The measurements were based on a unitary extracellular potential unique to this sublamina, which was first described by McIlwain (28). He termed it the juxtazonal potential (JZP), and showed that it results when a single spike invades the terminal arbor of a single retinal afferent to the upper SGS, triggering synchronous excitatory postsynaptic potentials in postsynaptic collicular cells. 2. Individual unitary JZPs were evoked at fixed latencies by weak shocks to the optic disk, chiasm, or tract. When the same JZP could be evoked in isolation from two stimulus sites, the conduction velocity of the axon triggering the JZP was estimated by dividing the conduction time between the stimulating electrodes (i.e., the "latency difference") into the distance separating these electrodes. This "latency-difference method" lacked general utility, however, since the same JZP could only rarely be evoked in isolation from two stimulus sites. 3. This limitation was circumvented by means of a collision method. When a stimulus that evoked a JZP in isolation was preceded by a sufficiently intense conditioning shock to a second, more central stimulus site, the conditioning stimulus caused the JZP to fail in an all-or-none fashion. It was assumed that when the JZP failed, the conditioning stimulus had exceeded the spike threshold of the axon mediating the JZP and that an antidromic action potential had collided with the orthodromic spike initiated at the peripheral stimulus site. Assessment of the critical interstimulus interval for producing such a collision, together with measurements of the axon's refractory period and the interelectrode conduction distance, permitted an estimate of the conduction velocity of the JZP-triggering axon. Conduction-velocity estimates generated in this way closely matched those based on the latency-difference technique when both methods could be applied. 4. Conduction velocities of 31 JZP-triggering axons analyzed by the collision method ranged from 2.9 to 6.8 m/s [4.6 +/- 1.0 (mean +/- SD)]. Comparable estimates were obtained for such axons by alternative methods based on the absolute latencies of electrically evoked JZPs or of the field potential to which they contribute. The conduction velocities of JZP-triggering axons fell within the range reported for retinal W-cells and entirely outside those of X- and Y-cells, confirming earlier evidence for W-cell input to the upper SGS (7, 15, 18, 28).(ABSTRACT TRUNCATED AT 400 WORDS)