The regenerating retinotectal projection in goldfish goes through an activity-driven refinement that appears to involve the elimination of inappropriate branches from early arbors. Retinotopically appropriate branches may be stabilized because the normally correlated firing of neighboring ganglion cells causes summation of their postsynaptic responses and increases their effectiveness by a Hebbian mechanism. In this study, I report that the regenerating projection has an increased capacity for long-term potentiation (LTP) that may be related to the activity-driven sharpening. In the normal projection, field potentials, reflecting currents from EPSPs elicited by optic nerve shock, are large (greater than 4 mV) and very stable. In newly regenerated projections, field potentials are initially small (less than 1 mV), but a train of 20 stimuli at 0.1 Hz results in a large (100–200%) increase in amplitude that is stable for at least 8 hr, and in 3 cases overnight. The capacity for potentiation is greatest from 20 to 40 d postcrush, the time just after arrival of the optic fibers, and during the period of retinotopic sharpening. A greater-than-normal capacity for potentiation persists for many months. Topical application of NMDA receptor blockers AP5 or AP7 at 25 microM prevents potentiation without decrementing ongoing responses. The closely related agent AP6, which is not an NMDA receptor blocker, does not prevent potentiation. In addition, infusion of the NMDA receptor blockers AP5 or AP7 into the tectal ventricle (4 microliters/d of 500 microM solution) for 2–3 weeks during regeneration prevented retinotopic sharpening, as assessed by electrophysiological mapping. At each tectal point, responsive areas in the visual field were enlarged to 28 degrees vs 11–12 degrees in control regenerates and normals. This was comparable to data from fish regenerating with activity blocked with intraocular tetrodotoxin or synchronized by stroboscopic illumination and indicates uncorrected errors in targeting of regenerated arbors (Schmidt, 1985). The results support the involvement of NMDA receptors in sharpening and suggest that the initial step in stabilizing appropriate branches may be a long- lasting increase in synaptic gain.