Do artificial reefs serve as productive marine fish habitats (sources) or do fish assemblages of such reefs contribute little to the gene pool of succeeding generations (sinks)? Using data from a 24-year study of the breakwater at King Harbor (California, USA), annual densities of reef fish larvae were compared with densities observed elsewhere in the Southern California Bight. Larval production at King Harbor has decreased over the study period, as has the size of the fish assemblage. Both declines may be related to recorded increases in water temperature. Larval densities from similar water depths (0–15 m) throughout the bight vary widely between sites and years. However, the mean percentage represented by reef fish among all larvae at King Harbor was 51% during those 5 years for which bight-wide samples were available (35% for all years) compared to 5% in the bight-wide samples. Utilizing larvae of 12 genera of reef fish species common to the samples, five were significantly more abundant in King Harbor samples. Results indicate that the breakwater represents a mature artificial reef and contributes to the reef fish larval pool of the bight, acting as a source rather than a sink. Copyright 2002 International Council for the Exploration of the Sea. Published by Elsevier Science Ltd. All rights reserved.