The ecology of patch reef fishes was studied to quantify the main factors that affect the natural variability of the fish community and to determine the effects produced on the community by experimental removal of predators. Initially, a year-long baseline description was completed of the physical, biological and ecological characteristics of 8 pristine patch reefs at Midway lagoon. For over 3 subsequent years, piscivorous predators were spearfished at least monthly, often for days at a time, on 4 of the 8 reefs. Fish populations were visually censused throughout the experiment. In all seasons and years of the project, daily recruitment rate of postlarval fishes to natural patch reefs was compared to that measured on standardized, artificial reefs of various sizes and degrees of inter-reef isolation. Finally, all baseline measurements were replicated and complete collections were made of all fishes, to validate the visual census method. Visual censusing was found to be of adequate precision and accuracy for most resident, non-cryptic species (highest for small patch reefs). Fishes could be assigned to size classes underwater by visual estimate with high accuracy. Rotenone collections were highly effective in quantifying many species commonly missed or underestimated in visual censuses. Only a few species composed the bulk of all recruits, while most species were rare or not seen at all. Variation between species was related to life history strategies or behavioral requirements. High temporal variability was found at the following scales: 1) Annuallywhere variability increased with the magnitude of recruitment, and different species recruited heavily in different years, suggesting that species specific factors in the plankton are more important than general oceanographic conditions; 2) Seasonally- pulsing strongly in summer, and occasionally late fall, when favorable environmental conditions may maximize growth and survival; and 3) Daily- with 1 or 2 strong peaks (each only a few days long) over a period of several weeks of low, variable recruitment. Small-scale spatial variability between replicate attractors (standardized artificial reefs) and between attractor types (coral and wire) were both high for a few species recruiting abundantly, although most recruits are probably substrate generalists. Rigorous visual fish censuses can adequately document moderate- to long-term temporal variation in the abundances of recently recruited juveniles on patch reefs (i.e., based on similar temporal patterns assessed by daily attractors). Daily total recruitment rate increased, although at diminishing densities, with (attractor) reef size, and with degree of inter-reef isolation. Abundances of recently recruited fish censused on neighboring, natural patch reefs (much larger than attractors) increased with reef size. The effect of isolation on these natural reefs was confounded by the stronger effect of reef size. These results suggest that if optimum size and spacing of reefs is provided, either by proper design of artificial reefs or selection of marine reserves, managers may enhance fish recruitment and ultimately improve local fisheries: Of the 135 fishes censused on the patch reefs studied, only 6 species together accounted for 70% of the total number of all fish, mainly due to heavy seasonal recruitment pulses. Strong seasonal and annual variability in recruitment was responsible for most of the temporal variation in fish abundance. The structure of patch reef fish communities at Midway was characterized by high unpredictability (e.g., great seasonal and/or annual variability in recruitment by common species, recruitment limitation for most species, and a high turnover rate detected by frequent sampling). Some predictions of the theory of island biogeography were also met by these fish communities (e.g., species richness correlated strongly with patch reef area, volume and relief). and total fish abundance. Some populations also exhibited a degree of long-term stability. Species diversity [H'] was similar among different size reefs. The experimental fishing on piscivores produced a catch composed mainly of lizardfish, due largely to immigration following the removal of other, competitively superior, highly resident piscivores. Scorpionfish and moray eels were also dominant predators. The expected decreases in catch-per-unit-effort were not realized, except for a quantitatively insignificant family (hawkfish). Conversely, the catch of the highly migratory lizardfish actually increased as fishing progressed. Changes in the catch composition for other piscivores related mainly to major changes in reef size or to patterns of large, inter-year recruitment fluctuations. Census data confirmed the major trends indicted by catch results. Sharks and jacks were attracted to the experimental reefs by spearfishing; the study was unable to determine whether their piscivorous effect was different between reef treatments. Patch reef fish communities at Midway were relatively resilient to long-term, intense fishing pressure on piscivores. However, enhanced survival of a large, annual, summer recruitment pulse of a common cardinalfish, synchronized with a temporary but significant reduction of lizardfish (the most prevalent piscivore) by fishing, suggested that an effect of predation on reef fish populations is experimentally detectable and considerable. However, temporal and spatial variability in recruitment, and reef size differences and changes in size were the primary factors responsible for the observed temporal patterns in fish abundance. COlnmunity analysis involves numerous confounding effects and requires the most careful interpretation for valid conclusions. / xvi, 321 leaves, bound : ill. ; 29 cm.