A major objective of the use of marine reserves in management of coral reef fisheries is protection of a critical spawning stock biomass to ensure recruitment supply to fished areas via larval dispersal. However, very little empirical evidence exists on the rates and patterns of increase of density and biomass of target species following the closure of a coral reef or part of a reef, nor on how quickly any gains potentially useful to fisheries can be lost if reserves are subsequently opened to fishing. This paper presents empirical evidence derived from the visual monitoring of density and biomass of large predatory coral reef fish (Serranidae (Epinephelinae), Lutjanidae, Lethrinidae and Carangidae as a group) in two small marine reserves and at two control sites in the Philippines from 1983 to 1993. At one reserve (Sumilon) a complex history of management allowed seven measurements of density and biomass at durations of reserve protection ranging from ‐2 (i.e., fished for 2 yr) to 9 yr. At the second reserve (Apo), seven measurements were taken at durations of reserve protection ranging from 1 to 11 yr. Density of large predators provided an excellent indicator of the effects of marine reserve protection and fishing. Density decreased significantly twice when the Sumilon reserve was opened to fishing (1985, 1993), and increased significantly three times following durations of marine reserve protection of 5 yr (Sumilon reserve and nonreserve sites, 1987‐1991) and >6 yr (Apo reserve beyond 1988). Density of large predators at the Apo nonreserve site (open to fishing) did not change significantly, remaining low throughout the study. Significant positive linear correlations of mean density of large predators with years of reserve protection were observed at both reserves. The rates of increase were 1.15 and 0.72 fish·1000 m‐2·yr‐1 at Sumilon and Apo reserves, respectively, with mean density ranges of ≈ 4‐17 fish/1000 m2 (‐2 to 9 yr of protection) at Sumilon and 0.5‐9.5 fish/1000 m2 (1‐11 yr of protection) at Apo. The pattern of increase of mean biomass with years of reserve protection was more curvilinear than that of mean density, particularly at Sumilon, where a slow increase was observed in the first 3‐5 yr (reflecting delayed recruitment and a natural delay to the period of maximum individual mass growth), followed by an increasing rate over the next 4 yr. Mean biomass ranges were ≈ 1.5‐18 kg/1000 m2 (‐2 to 9 yr of protection) at Sumilon and 1‐10.5 kg/1000 m2 (1‐11 yr of protection) at Apo. At Sumilon reserve, 2 and 1.5 yr of unregulated fishing, respectively, eliminated density and biomass gains accumulated over 5 and 9 yr of marine reserve protection.