barnacle dropped greatly and have since recovered little. False mussels and barnacles disappeared in oiled rivers and were not recorded again until 15 months after the spill. The greatest cover on roots in all oiled habitats after the spill was oil (Fig. 4). In addition, most roots sampled for epibiota (28) were dead, broken, or rotting by August 1987 (54% on open coasts, 56% in channels, and 66% in rivers at oiled sites versus 8%, 1%, and 3%, respectively, at unoiled sites). Thus the mangrove-fringe root habitat has been largely destroyed at oiled sites and will not be restored unless new trees grow. Seagrasses. Extensive meadows of the seagrass Thalassia testudinum cover much of the intertidal reef platforms and subtidal floors of shallow embayments and lagoons along this coast (24). Entire beds of intertidal Thalassia were killed on some heavily oiled reef flats, as shown by abundant oil-covered dead leaves washed ashore and dead but intact root-rhizome mats (for example, the site at Largo Remo North) (Table 2). In contrast, subtidal Thalassia survived every. where after the spill, although leaves became brown and heavily fouled by algae for several months in heavily oiled areas. Thalassia provides food, habitat, and refuge for dense populations of invertebrates and fishes (24, 29). Before the spill, amphipods, small decapods, ophiuroids and other echinoderms, bivalves, gastropods, polychaetes, and small fishes were all abundant in seagrass within the study area (24) and still are in unoiled beds (Fig. 5). Sampling of infauna began September 1986 at three oiled and four unoiled sites (30) (Fig. 2). Four taxa of invertebrates were significantly less abundant in oiled grassbeds after the spill, but four others showed no such difference (Fig. 5). Abundances of most taxa increased between September and January; preliminary analyses of body size indicate that recruitment caused these increases. Only the abundance of hermit crabs increased in oiled relative to unoiled areas (7.3 versus 3.4 crabs per sample, respectively, in January 1987, P < 0.05, repeated measures analysis of variance), perhaps because of a surplus of shells of recently dead snails (31). Intertidal reef flats. Platforms of fringing reefs from extensive shallow to intertidal flats throughout the study region (Fig. 3) (7, 9, 17, 32). Populations on the reef flat at Punta Galeta have been monitored for up to 15 years (9) (Fig. 2). Damage was most extensive at the seaward border, where the oil accumulated at low tide. Immediately after the spill a bloom of microalgae covered recently vacated substratum (33) (Fig. 6). This area had been dominated by perennial macroalgae, particularly the fleshy red alga Laurencia papillosa, crustose corallines, and the articulated calcareous green alga Halimede opuntia. Cover of all these plants was reduced to. levels well below those observed previously, but had regained or exceeded typical abundance within 12 to 18 months, concurrent with a reduction of the microalgae. The most common sessile animals before the spill were zoanthids (Zoanthus sociatus and Palythoa spp.), hydrocorals (Millepora spp.), and scleractinian corals (Ponites spp.). At the seaward border of the reef flat, populations of all these animals were severely reduced, and only Zoanthus had returned to typical abundance after 18 months (Fig. 6). Densities of sea urchins on the reef flat varied over three orders of magnitude before the spill (Fig. 6) (9). At the seaward edge, the most abundant Percent cover Urchins per 10 m2 10 Fig. 6. Pre- and postspill comparisons of the abundance of algae and sessile animals (top seven panels) and the sea urchin Echinometra hucunter (bottom panel) on the seaward portion of the Galeta reef flat (33). Filled symbols denotes the heavily oiled zone closest to the water's edge at low tide (Fig. 3). Open symbols and the smooth line in the E. cuter graph indicate less oiled (more landward) habitats; and + mark postspill surveys of percent cover that were significantly higher or lower than any prespill survey at P ≤ 0.05 and P = 0.01, respectively. In all cases, differences were significant only in the heavily oiled zone (). Cover was compared survey by survey with the use of paired t tests matched by individual transects within surveys. Abundances of E. lucunter were compared with the use of residuals from regressions (34). 1 83 84 85 87 88 Oil Spill Year 6 JANUARY 1989 ARTICLES 41 urchin Echinometra lucunter was reduced by about 80% within a few days of oiling, and the reef flat was littered with its skeletons. Although this decrease was small in an absolute sense relative to changes in abundance among years, it nevertheless departed significantly from the average trend in abundance between months of May and June over the last 8 years (34). Farther inshore, no such reductions occurred. Gonodactylid stomatopods (mantis shrimp) are abundant in intertidal Thalassia beds from Isla Margarita to Isla Largo Remo, where they live in and aggressively defend cavities in coral rubble at densities up to 20 per square meter (35). They are prey for shorebirds, fish, and crabs, and in turn consume large numbers of hermit crabs, snails, and other animals (36). Population data are available for stomatopods from intertidal seagrass beds on four reef flats, including certain types of data collected at various times before the spill (Table 2 and Fig. 2). Two of the sites were heavily oiled, and large amounts of oil were still present on mangrove roots and in sandy sediment when sampling first began 3 months after the spill; the other two flats were at most lightly oiled. Densities of stomatopods were less on the heavily oiled flats, particularly for animals greater than 40 mm long (37). Loss of these large animals apparently made larger cavities available to the survivors, which consequently fought less and suffered fewer injuries (38) compared to the same sites before the spill and compared to lightly oiled flats afterwards (Table 2). Growth of larger survivors also increased on heavily oiled flats (39), probably because of decreased competition for cavities and an apparent population explosion of hermit crabs for food (35, 40). These effects have persisted in diminished form through our last census in September 1987. Subtidal reefs. Populations of subtidal sessile organisms wer surveyed on six fringing reefs between Isla Margarita and ka Grande within 1 year before and 4 months after the oil spill (Fig. 2. (41). These include the heavily oiled reef at Punta Galera, a lightly oiled reef at Isla Margarita, and four unoiled reefs east of Portobelo Abundance of most common scleractinian coral genera in depths ≤3 m decreased at Galeta by 51 to 96%, and total coral cover decreased by 76%; even at 9 to 12 m the drop was 45% (Fig Reductions were less at Isla Margarita and generally absent on the unoiled reefs, except at one site northeast of Portobelo (Fig. 2) for no apparent reason. This relation between amount of oiling on the six reefs and decrease in coral cover was significant for 0 to 3 m but not deeper (42). Sublethal effects were also substantial. Within Bahia las Munas. most of the scleractinians still alive in depths less than 3 m showed signs of recent stress including bleaching or swelling of tours. conspicuous production of mucus, recently dead areas devoid of coral tissue, and globules of oil (Fig. 3) (43). In some cases, bleached or dead areas were surrounded by a black halo characteristic of bacterial infection (44). Both the frequency and size of recendy dead lesions on the commonest massive corals increased markedly with the amount of oiling at each reef and decreased with water depth (Fig. 8) (45, 46). These effects were also species specific. In the case of S. siderea, which suffered most, new partial mortality was still disproportionately common on heavily oiled reefs 1 year after the spill (46). The oil spill also affected other organisms, including snails on the reef flat and intertidal zone at Punta Galeta and mobile epifauna. particularly shrimps above subtidal seagrasses. In summary, the spil harmed prominent organisms in all intertidal and suboidal enviKEY ments examined, infauna and epifauna, and members of all trophie levels including primary producers, herbivores, carnivores, and detritivores. Fig. 7. Percent cover of corals before the oil spill plotted against their cover afterward. The diagonal lines represents no change in coral cover, points below this line indicate a decrease after the spill, and points above the line show an increase. The points are means ± 1 SE; some error bars lie within the plotting symbol. Oiling significantly affected coral cover only at the shallowest depth (42). Abbreviations: 3, 0- to 3-m depth; 6. 3- to 6-m depth; and 12. 6 to 12-m depth. JUG and PALW are located just west of Isla Grande, DMA and DÖNR are northeast of Portobelo. GAL is at Punta Galera, and MAR3 is at Isla Margarita (Fig. 2, half-filled circles). 42 Patterns of Responses and Their Significance The severe damage to intertidal biotas and their response, such as the flush of ephemeral microalgae just after the spill, are summular in those documented previously for other oiled intertidal commuarates (47-50), including those inferred for Galeta after the Witwater spill (4). Likewise, biological effects of spills are usually greatest in low energy environments, where oil tends to accumulate and be retained in fine sediments, than in high energy environments where it is soon washed away (50). This pattern matches the greater and more persistent disturbance to riverine and channel mangrove root communities than to those along open coasts and the rapid recovery a abundance of many organisms at the seaward edge of the reef flat (Fig. 6). In contrast, other results were not expected, and in some cases contradict widely held views about the effects of oil spills and the ways they are studied. First, extensive mortality of subtidal corals and infauna of seagrasses had not been demonstrated before (1. A and contradicts undocumented assertions that these orgasms are not affected by oil spills. In part, subtidal effects may have bee caused by the use of dispersant (51) but it seems unlikely that the was the only reason, given the regional breadth of the impact and the relatively small amount of dispersant employed (14). Second, the magnitude of subtidal coral mortality and unparv within Bahia las Minas is in striking contrast to findings of no lasting change in coral condition or growth after exposure to od with or without chemical dispersant, in small-scale experiments (51 52). Such discrepancies underlie the importance of detailed long term ecological studies, and the dangers inherent in extrapolation to SCIENCE, VOL. 341 Fig. 8. Frequency of injury for the three most common species of massive corals in relation to the amount of oiling at 12 reefs (4 unoiled, 3 lightly to moderately oiled, and 5 heavily oiled). Open bars represent coral colonics with injuries that did not exceed 10% of the surface area of the coral; filled bars represent injuries greater than 10% (maximum observed, 100%). Sample sizes are shown above each category of oiling. Abbreviations; Dip, Diplonia clivosa, Por, Parites astreoides, Sid, Siderastrea siderea; s, shallow (0 to 1 m), and d, deep (1 to 2 m). natural populations from laboratory-based physiological data or small-scale, short-term press perturbation experiments in the field (2, 3, 48, 52, 53). Third, sublethal effects are extensive and may be more important in the long term than initial mortality (3, 48). Changes in stomatopod behavior and population structure were more pronounced than might be expected from a simple reduction in numbers. Likewise, injury of corals has allowed colonization of their bare skeleton by algae and other sessile organisms that may overgrow parts of colonies that survived the initial effects of oil (3, 54). Corals stressed by oil are probably also more susceptible to epidemic disease and likely to grow and reproduce more slowly than unaffected colonies (3, 5, 52). Any combination of these effects may further reduce overall abundance of corals as much as the initial spill (54). Such a chain reaction could continue long after any petroleum hydrocarbons are present in the environment or coral tissues, just as staghorn coral continued to decline precipitously following a severe hurricane in Jamaica (55). The response of organisms to an oil spill, or any other major disturbance, will depend on the conditions in which they normally Live (15, 48, 56). Moreover, the suite of organisms able to survive under conditions of chronic pollution, and their resistance to further stress, is typically different from that in similar unpolluted habitats (48, 57). Much of Bahia las Minas has been subjected to human disturbance, beginning with decades of excavation, dredging, and landfilling for the construction of the Panama Canal and the City of Colon, drainage and spraying of mangroves for mosquito control, construction of the refinery and a large cement plant on landfill, a major oil spill in 1968, and unknown amounts of chronic oil pollution from the refinery and ships passing to and from the Canal (4, 5, 17, 58). It is a measure of the severity of the 1986 oil spill that the biological consequences were so detectable despite this history of environmental abuse. REFERENCES AND NOTES 1 National Research Council, Oil in the Sea: Inputs, Fates, & Effects (National Academy Press, Washington, DC, 1985). 2. R. B. Clark, Philos. Trans. R. Soc. London Ser. B 297, 185 (1982); R. S. Carney, Am Sa 74, 298 (1986). 3. Y. Loys and B. Rinkevich, Mar Ecol. Prog. Ser. 3, 167 (1980). 6 JANUARY 1989 4. K. Ruzier and W. Sterrer, BioScience 20, 222 (1970). 5. C. Birkeland, A. A. Reimer, J. R. Young. Survey of Marine Communities in Panama and Experments with Oil (Publ. EPA-600/3-76-028, Environmental Protection Agency Ecological Research Series, Narragansett, RJ, 1976). 6. J. D. Cubit et al., in 1987 Oil Spill Conference (Publ. 4452, American Petroleum Institute. Washington, DC, 1987), pp. 401-406. 7. J. Cubit and S. Williams, Atoll Res. Bull. 269, 1 (1983). & J. D. Cubit, R. C. Thompson, H. M. Caffey, D. M. Windsor, Smithsonian Contrib Mar. Sci. 32, 1 (1988); L. G. Macintyre and P. W. Glynn, Am. Assoc. Petrol Geol. Bull 60, 1054 (1976). 9. G. L. Hendler, in Proceedings of the Third Intemational Coral Reef Symposium, D. L Tavlor, Ed. (University of Miami, Miami, FL, 1977), vol. 1, pp. 217-223, 1. D Cubit, D. M. Windsor, R. C. Thompson, J. M. Burgett, Estuarine Coastal Shelf Ser 22, 719 (1986). 10. J. C. Ogden and E. H. Gladfelter, UNESCO Tech. Pap. Mar Sci. 23, 1 (1983). Important species in common throughout the tropical western Atlantic are red mangrove R. mangle, seagrass T testudinum, most reef corals and algae, and commercially important lobsters (Paulinus spp.) and oysters (C. rhizophorae). 11. Ongoing drilling and refining off Yucatan and possible production along the west coast of Florida 12. The oil type was 70% Venezuelan crude, 30% Mexican Isthmus crude, specific gravity 27 (American Petroleum Institute). Percent composition determined at the Bermuda Biological Station by column chromatography with the use of alumina and silica gels: 47.4% sanurates (fraction 1, hexane); 6.4% light aromatics (fraction 2, 10% ether/90% hexane); 4.8% heavy aromatics (fraction 3, 20% methylene chloride/80% hexane); 18.9% polar fraction (fraction 4, methylene chloride); and 22.5% unrecovered. 13. C. D. Getter, G. I. Scort, J. Michel, in 1981 Oil Spill Conference (Publ. 75-4161. American Petroleum Institute, Washington, DC. 1981), pp. 535-540. 14. Refinery officials estimated that <21.000 liters of Coreut 9527 (Exxons Chemicals America) were spraved from aircraft (6) STRI personnel observed this spraving between Asana Chaquita and Punta Galeta. 15. J. D. Woodley et al., Science 214, 749 (1981), J. H. Connell, ibid. 199, 1302 (1978). 16. Surveys bevond Punta Galeta began the first week of June 1986 and lasted 1 month Videotapes were made of the coast between Rio Chagres and Islas Naranjos, and still photographs were made all along the coast. Visual assessments were made of degree of oiling (heavy, moderate, light, or absent) and the habitats and types of organisms obvioush affected. 17. J. D. Cubit, G. Barista de Yec, A. Roman, V. Batista, in Agonia de la Naturaleza, S. Heckadon Moreno and J. Espinosa Gonzalez, Eds. (Impretex, Panama, 1985), pp. 183-199, Revista Medica Panama 9, 56 (1984). 18. Oil was observed commonly until the ame of manuscript revision (July 1988) around Isla Pavardi. where it emerged directly from the coral landfill on which the refinery is built, and Punta Muerto, Bahia Cativa, eastern Isla Largo Remo, and Punta Galeta 19. Within this area oil commonly escaped from sediments when we walked among mangroves or cored into shallow subudal seagrass beds. Intertidal surfaces were heavily coated with oil. 20. Most of the oil that escaped from Bahia Cativa was driven west. Coasts facing north to northeast were heavily oiled, whereas adjacent areas facing west or south received little oil. The best studied example is at the northwestern side of Isla Largo Remo (Table 2). 21. These low tides occur annually at this time of year (8, 9). Sea levels during most of this time were as much as 10 cm below the 9-year average for thus period of normally low sea level. 22. Corals were collected by divers on days when on surface slicks were visible, placed in solvent-rinsed aluminum foil, and frozen. Tissues were removed from skeletons with an air pick and homogenized with a tissue grinder. To minimize variations due to water or carbonate inclusions, standard Lowrey protein determinations were used as one measure of tissue mass. After solvent extraction, total lipid weights were determined microgravimetrically Extracts were fractionated by adsorption chromatography. Hydrocarbons were analyzed with a capillary gas chromatograph equipped with a flame ionization detector and 25-m SE 52 fused silica columns. 23. R. H. Green, Sampling Design and Statistical Methods for Environmental Biologists (Wiley, New York, 1979) 24. J. B. C. Jackson, Bull Mar Sa 23, (1973); K. L. Heck, Mar. Biol. 41, 335 (1977); R. Vasquez-Montoya, An. Inst. Cien Mar Limnol Univ. Nac. Auton Mex. 10, 1 (1983). 25. This estimate was extrapolated from aerial photographs of 36 km of the 82-km shoreline, of which 76 km contains mangroves Of the photographed shoreline, 36% conrained a band of dead mangroves. 26. Sprouting success was measured for seedlings collected from an unoiled site, and planted in one unoiled and two oiled sites. Of 25 seedlings transplanted into each site, 16 (64%) sprouted at the unouled site and 1 (2%) sprouted at the oiled sites. 27. J. P. Sutherland, Mar Biol 58, 75 (1980), W. E. Odum, C. C. Mclvor, T. J. Smith III, The Ecology of the Mangroves of South Florida A Community Profile (Publ. 81/24, U.S. Fish Wildlife Service, Washington, DC, 1982). 28. Roots were sampled three times before the spill in rivers and channels (September to October 1981, January 1982, and June 1982) and twice in open habitats (September to October 1981 and January 1982). Several areas of shore within each habitat were chosen haphazardly Fifteen to 25 intertidal roots 20 cm long that had not become firmly attached to the mud were lifted from the water, and cover of attached organisms was measured for each root by 100 random point counts after the oil spill or was estimated visually before the spill. Afte the spill, previously sampled riverine and channel habitats included both oiled and unoiled sites but all open sites were oiled. Thus it was necessary to choose sites near Portobelo (Fig. 2), ARTICLES 43 the nearest equivalent unoiled shoreline, for unoiled open habitat similar to oiled open habitat. One hundred roots were sampled in each oiled and unoiled habitat after the spill. 29. A. K. O'Gower and J. W. Wacasey, Bull. Mar. Sci. 17, 175 (1967); J. B. C. Jackson, Mar. Biol: 14, 304 (1972); D. K. Young and M. W. Young, in Ecology of Manne Benthes, B. C. Coull, Ed. (University of South Carolina Press, Columbia, 1977), pp. 359-382; J. C. Ogden, in Handbook of Seagrass Biology: An Ecosystem Perspective, R. C. Phillips and C. P. McRoy, Eds. (Garland, New York, 1980), pp. 173-198, A. W. Stoner, Bull. Mar. Sci. 30, 537 (1980). 30. All seagrass beds sampled lie adjacent to mangrove shoreline and are 1 m deep; unoiled beds meeting these criteria occur near Portobelo and Isla Grande (Fig. 2). Infauna were collected with 10.2-cm inside diameter polyvinyl chloride (PVC) tubes pushed into the sediment to a depth of at least 10 cm. Each sample consists of three pooled cores (240 cm2). At each census eight samples were collected haphazardly at each site. Samples were sieved over 500-um mesh and preserved in 10% formalin with rose bengal. Animals were sorted to major taxa and counted or weighed wet. 31. M. B. Bertness; Ecology 62, 751 (1981). 32. J. D. Cubit, in Proceedings of the Fifth International Coral Reef Congress, C. Gabrie, J. L Toffart, B. Salvat, Eds. (Antenne Museum-Ephe, Moorea, French Polynesia, 1985), vol. 3, pp. 111-118 33. Percent cover of sessile species attached to the substratum was estimated by point sampling along ten permanent transects 9 to 22 m long, perpendicular to the reef edge and spaced at random intervals within a 20-m strata. The number of points sampled per survey was between 1510 and 1560. The seaward 6 m of these transects were more directiv exposed to oil than the landward portions. This 6-m band forms the "heavily oiled zone" of Fig. 6. Sea urchins were counted in two 1 x 20 m transects; one at the reef edge, the other 26 m farther landward. 34. For the reef edge transect only, residuals from regressions of changes in abundance from May to June were highly negative in 1986 compared with all other transects in all other years (n=41.0.88, P < 0.001). 35. R. L. Caldwell and H. Dingle, Sci. Am 234, 80 (January 1976); R. Steger, Ecology 68, 1520 (1987). 36. M. L. Reaka, in Proceedings of the Fifth International Coral Reef Congress, M. Harmelin Vivien and B. Salvat, Eds. (Antenne Museum-Ephe, Moorea, French Polynesia, 1985), vol. 5, pp. 439-444: R. L. Caldwell G. Rodenck, S. Shuster, la! Zool Union Mone, in press; R. L. Caldwell, Bull Mar Sa. 41, 135 (1987). 37. The four collection sites ranged from 100 to 125 m x 25 to 40 m (all larger than 3000 m2) and were selected before the spill to have intertidal Thalassia beds with firm, sandy substratum. and moderate amounts of rubble and coralline algal nodules. Densines of stomatopods were estimated from 30 quadrats (0.5 m2) chosen at random within 400 m2 plots at each site. All hard substrata were removed from each quadrat and broken, and all stomatopods were retained. A minimum of 300 individuals was collected, including ones from pieces of coral rubble collected haphazardly within each site, but outside the plot used for quadrats. These animals were included in the analysis of injuries and growth. 38. Cuticular injuries were recorded for all animals as described by I. K. Berzins and R. L. Caldwell Mar Behav Physiol. 10, 83 (1983)]. 39. We estimated growth by comparing the length of the carapace of arumals newly molted in the laboratory with its length before molting, using only animals that molted within 3 days of capture and were longer than 35 mm. 40. Almost every piece of rubble in the heavily owed plots was covered by hundreds of hermut crabs. Such densites were not observed in unoued plots or in any years previoush 41. See P. W. Glynn (Bull Biol Soc Wash 2, 13 (1972)] and J W Porter (ibid., p. 89) for general descriptions of Panamanian reets We selected fringing reefs that extended at least 100 m along the shoreline and to a depth of at least 12 m. At each reef four or five line transects were extended from haphazardly chosen points at the shoreward edge of the reef to the deepest point of the reef Along each line, 1-m2 quadrats were placed contiguously or up to 3 m apart. depending on transect length 18 to 81 quadrats per transect). Quadrats were divided into 100-cm2 cells, and total cover was estimated by eve within each cell. 42. We determined differences in coral cover before and after the spill (July to October 1985 and July to August 1986) by all quadrats at each of theme dep (0 to 3, 3 to 6, and 6 to 12 m; prespill n = 306, 219, and 125 quadram. respectively; postspill = 294, 200, and 187) at each suse and then calculang (mean cover before the spill/mean cover afterward). The effect of cil was crassaved by one-way analysis of variance of these values for one heavily clad, one lighty oiled, and four unoiled reefs; df = 2. For 0 to 3 m depth, F 13.3, P = 0.032, 3 to 6 m, F = 4.48; P = 0.13; 6 ∞ 12 m, F = 0.90, P = 0.49. 43. Dead areas were bare or colonized by thin growth of filamentous sight. Ten colonics of S. siders were sealed in plastic bags underwater, brought to the laboratory, and placed in clean tanks and seawater. A film of oil appeared on the surface of each tank when the filamentous algae growing over the recandy dead areas were squeezed by hand. 44. A. Antonius, in Proceedings of the Fourth International Coral Reef Sympo E D Gomez et al., Eds. (Marine Sciences Center, University of the Philippines, Quezon City, Philippines, 1981), vol. 2, pp. 3-14, ibid., vol 2 7-14. 45. Recent injuries were counted on 12 reefs for all corals in 2 x 50 m 46. Significance was tested with hierarchical log-linear models (SPSS/PC+, 47. A. J. Southward and E. C. Southward, J. Fish. Res. Board Can 35, 682 (1978) 48. A. J. Southward, Philos. Trans. R. Soc. London Ser. B 297, 241 (1982) 49. R. J. Nadeau and E. T. Bergquist, in 1977 Od Spill Conferem (Pab. €23€, American Petroleum Institute, Washington, DC, 1977), pp. 535–538. E. 1 Chan sid, pp. 539-542; L. E. G. Cintron and Y. Schaeffer-Novelli, UNESCO TP Mar Sci. 23, 87 (1982). 50. J. H. Vandermeulen, Philos. Trans. R. Soc. London Ser. B 297, 335 (1982). 51. A. H. Knap et al., Oil Petrochem. Poll. 1, 157 (1983); R. E. Dodge et al., Canal Regi 3, 191 (1984); A. H. Knap, Mar. Pollest. Bull. 18, 119 (1987). 52. B. E. Brown and L. S. Howard, Adv Mar. Biol. 22, 1 (1985) 53. P. Yodzis, Ecology 69, 508 (1988). 54. S. R. Palumbi and J. B. C. Jackson, J. Exp. Mar Biol. Ecol. 64, 103 (1982), T. P Hughes and J. C. Jackson, Ecol. Monogr. 55, 141 (1985) 55. N. Knowlton, J. C. Lang, M. C. Rooney, P. Clifford, Nater 294, 251 (1981), N Knowlton, J. C. Lang, B. D. Keller, Smithsonian Comer. Mar. Sc., in p 56. J. H. Connell, in Population Dynamic, R. M. Anderson, B. D. Tumer, L. R. Tanke, Eds. (Blackwell Scientific, Oxford, 1979), pp. 141-163, R. T. Pane and SA Levin, Ecol. Monogr. 51, 145 (1981); S. T. A Pickett and P. S. White. The Euningy of Natural Disturbance and Patch Dynamics (Academic Press, Orlando, FL, 1985, 57. R. P. M. Bak, Mar. Polba. Bull. 18, 534 (1987) 58. H. H. Rousseau, in The Panama Canal International Engineering Congress (Nical. San Francisco, 1915), pp. 374-432; W. G. Comber, in The Panama Coma A Engineering Treatise, G. W. Goethals, Ed. (McGraw-Hill, New York, 1916), pp. 459-493, D. P. Curry, Am. J. Trop. Med 5,1 (1925), 1. G. Claybou, Digung on the Panama Canal (A. Kroch, Chicago, 1931), EC Webster, La Defense Prom (Editorial Universitaria Panama, Panama, 1973). 59. 1. Brawn and R. Green helped with statistics, and G. Jacome, D. Matas, and K. Moss provided technical assistance. B. Carey, D Dodge, R. Green; Kar N Knowlton, Y. Loya, and A. Smith reviewed the manuscript and helped se mate other ways. This work was supported by the U.S. Munerals Management Service under contract 14-12-0001-30393 and by the Smuchaouan Tropical Research Institute and the Smithsonian Institution Environmental Sciences Program 1 43-269 (824) О SCIENCE, VOL. 243 |