Blocking activity with intraocular tetrodotoxin prevents the sharpening of the retinotectal map formed during regeneration of the optic nerve. If (under normal conditions) the initially diffuse map sharpens because of correlated activity in neighboring but not distant ganglion cells, then sharpening should also be prevented merely by disrupting the spatiotemporal correlation in the pattern of activity. To test this idea, fish were exposed during regeneration to stroboscopic illumination in a featureless environment, or were maintained in complete darkness. The regenerating cells remained visually responsive after axotomy, and the xenon strobe effectively drove each ganglion cell at a constant latency. The maps formed in the strobe-reared fish were normally oriented, but the multiunit receptive fields were greatly enlarged, averaging 32 degrees. In control regenerates, multiunit receptive fields averaged only 11-12 degrees, nearly the same as for single units. Dark rearing, which allows only spontaneous activity, also resulted in enlarged multiunit receptive fields, averaging more than 28 degrees. Both effects parallel those reported previously with tetrodotoxin block. The mature projection did not become diffuse as a result of the strobe rearing, and the sensitive period corresponded to the early stage of synaptogenesis (20-34 days). Periods of normal visual exposure after 35 days produced very little sharpening of the diffuse maps produced during either strobe or dark rearing. The results are attributed to an activity-dependent stabilization of developing synapses. The correlated firing of neighboring ganglion cells could allow postsynaptic summation of their responses, and the retention of those more effective, retinotopically placed synapses might then occur via a Hebbian mechanism.