Do small-scale experiments showing spatial density dependence in marine fishes scale-up to temporal density dependence and regulation of relatively large local populations? If so, what are the causative mechanisms and their implications? We conducted an 8-year multi-generation study of population dynamics of bicolor damselfish (Stegastes partitus) inhabiting four large coral reefs in the Bahamas. After a 4-year baseline period, it was clear that two populations naturally received very few settlement-stage larvae, so recruitment of recently settled fish was artificially enhanced at one low-settlement reef and reduced at one high-settlement reef to ensure a broad range of population sizes over which to test for regulation. Over all 8 years, populations on the two naturally high-settlement reefs experienced temporal density dependence in multiple per capita demographic rates mortality, survival to adulthood, and fecundity. These local populations also displayed components of regulation persistence, boundedness, and return tendency. Reefs supporting regulated populations were structurally complex, providing sufficient prey refuges that ensured high survival at low densities. In contrast, populations at low-settlement reefs experienced either density-independent or slightly inversely density-dependent demographic rates, even though recruitment was artificially augmented to high levels at one reef. There was no evidence of regulation at these reefs, and indeed, one local population suffered temporary extirpation. Here, habitat complexity was relatively low, increasing the risk of predation, especially at low population densities when fish would have to travel longer distances when finding mates or home sites inhabited by conspecifics. Among all demographic parameters, density dependence in individual growth (an indicator of within-species competition for food) was least correlated with the presence or absence of local population regulation. We conclude that, for systems like these, the environmental context of a local population -- especially predation risk and the distribution and abundance of spatial refuges -- is more important than the magnitude of larval supply alone in determining the existence of regulating density dependence. At the broader metapopulation scale, density dependence in both survival and fecundity exogenously caused by predation may provide regulation for the entire stock, even when endogenous within-species competition (assumed to be important in most fisheries models) is weak.