1979). However, virtually no field studies have qualitatively investigated the 

 effects of fish predators upon the teeming mobile cryptic fauna that inhabits this 

 environment. Also, while these invertebrates are the primary food source for many 

 reef fishes (Randall, 1967), no studies have experimentally examined the importance 

 of this mobile invertebrate fauna for the colonization and maintenance of popula- 

 tions of reef fishes. In July 1980 we initiated a 7h year of study that experimen- 

 tally investigated reciprocal interactions between fishes and the mobile cryptic 

 invertebrate fauna in Salt Canyon, St. Croix. A preliminary analysis of the 

 first 6 months of the data is presented here. A more thorough analysis of data 

 from the entire 2% year period is in progress. 



MATERIAL AND METHODS 



Using the Hydrolab, an underwater habitat operated by NOAA on St. Croix, U.S. 

 Virgin Islands, 15 artificial reefs were established at a depth of 20 m in Salt 

 River Canyon in July 1980. Salt River Canyon is a submarine canyon 70-100 m wide 

 with a sand floor flanked by a vertical and a sloping coral reef wall. Five reefs 

 provided habitat for fishes and invertebrates (A reefs); each consisted of 11 

 cinderblocks arranged in a pyramid 2 blocks wide and 3 blocks high, with 6 pieces 

 of sun-dried dead coral rubble placed around the pyramid base. The components were 

 tied together with nylon rope and anchored in place with iron reinforcement bars. 

 Five reefs consisted of a cinderblock pyramid without rubble, providing habitat for 

 fishes but little shelter for invertebrates (B reefs). Five reefs, each composed 

 of rubble arranged in the same pattern as in the A reefs but lacking a cinderblock 

 pyramid, provided habitat more suitable for cryptic invertebrates (C reefs). The 

 15 reefs were arranged serially (ABCABC...) 10 m apart and 10 m out from the 

 sloping east wall of the canyon. Fishes on the reefs were censused visually one 

 week after establishment, and at approximately 30-day intervals thereafter. Using 

 information from the literature (Randall, 1967; Clavijo, et al., 1980) and personal 

 observations, we assigned each individual to a feeding guild - (planktivore, herbiv- 

 ore, piscivore, piscivore-invertivore, invertivore) based on its size class (post- 

 larval, juvenile, adult) and species (Appendix Table 1). We have included fishes 

 that eat only invertebrates and those that eat both invertebrates and smaller 

 fishes here as "invertebrate-eaters", since both prey on invertebrates. The in- 

 vertebrate-eaters and planktivores (which prey upon invertebrate larvae as well as 

 holoplankton) are the fishes most likely to influence benthic prey populations; 

 hence, these taxa are emphasized in this report. After 6 months, cryptic in- 

 vertebrates in half of the rubble from each of the A and C reefs were sampled 

 quantitatively by sealing the rubble in plastic bags in situ . On shore, the rubble 

 was chiselled into small pieces and sieved (0.7 mm mesh), retaining all of the 

 resident cryptic biota. These samples were preserved in formalin and sorted, 

 counted, and measured microscopically (see Reaka, 1981, 1983, for more details). 



RESULTS 



All fishes colonized both types of cinderblock reefs (A,B) quickly, while >2 

 months passed before fishes on the rubble reefs (C) reached peak abundances (lower 

 graph, Fig. 1). On all 3 types of reefs, total numbers of individuals subsequently 

 declined (and remained low throughout the winter; numbers increased again following 

 spring recruitment; Reaka, 1981, 1983). Planktivores (mostly juvenile grunts, 

 Haemulidae) were by far the most abundant guild, so their pattern of colonization 

 followed that described above for total numbers of fishes (Fig. 1). Invertebrate- 

 eating fishes were less abundant than planktivores. The number of invertebrate- 

 eating fishes on reefs with rubble-dwelling invertebrates (A, C) did not peak until 

 2-5 months after establishment of the reefs. This period coincided with the time 



70 



