I studied effects of three risk factors, predator proximity, persistence in attacking, and speed, on latency to emerge after entering a refuge in the scincid lizard Eumeces laticeps. I simulated a predator by directly approaching a lizard until it took refuge, usually a tree hole. Costs per unit time of lost opportunities to forage or engage in other activities enhancing fitness did not vary with risk factors. Skinks remained in refuges longer when I stayed near the refuges than when I withdrew to a greater distance. At both distances, skinks monitored my presence visually from positions at or near refuge entrances. Skinks remained in refuges longer after the second of two successive similar approaches at the same speed than after the first, suggesting that they perceived increased risk due to persistence by an individual predator, but the assessment might have been based on attack rate without individual recognition. Latency to emergence increased with predator approach speed, giving another indication of response to higher risk. If lizards were approached twice in succession, latency to emergence was much greater when predator speed was faster during the second approach. When the second approach was slower, there was no significant difference in latency between trials. Risks associated with speed and persistence thus simultaneously affect risk assessment. A recent model (Sih, 1992) of emergence from refuge supposes that decisions affecting latency to emerge are based on a balance of costs of remaining in refuge (often predominantly costs of lost opportunities for feeding or mating) against risk of predation. All data are consistent with a generalized view of Sih's (1992) model, but studies are needed to assess effects of cost and limited information about predator presence.