Proc. Nat. Acad. Sci. USA Vol. 73, No. 1, pp. 262-266, January 1976 Zoology Species-area relation for birds of the Solomon Archipelago (biogeography/islands/) JARED M. DIAMOND** AND ERNST MAYRt * Physiology Department, UCLA Medical Center, Los Angeles, California 90024; t Museum of Comparative Zoology, Harvard University, Cambridge, Massachusetts 02138 Contributed by Ernst Mayr, October 23, 1975

ABSTRACT Accurate values of number of breeding bird Third, some detailed ecological information about Solomon species have been obtained for 50 islands of the Solomon Ar- bird species is available from ref. 20 and from unpublished chipelago. From information about species altitudinal distri- butions on each island, the values are apportioned into num- observations by the Whitney Expedition and by Diamond. ber of montane species (Smt) and of species present at - This information will be used below in partialing out the ef- level (SIO,). Sio, increases linearly with the logarithm of fect of elevation on species number. Finally, bird distribu- island area A over a million-fold range of areas (correlation tions in the Solomons are now well known. Between 1927 coefficient 0.99) and with a comparatively low slope, while and 1930 the Whitney Expedition collected on all major and the log S-log A relation is markedly curved. With increasing many minor islands, searching especially for geographically isolation of an archipelago, the species-area relation de- species creases in slope and may shift in form from a power function variable species, taxonomic novelties, and montane to an exponential. Comparison of Pacific archipelagoes at [distributional results summarized by Mayr (21)]. Whitney different distances from the colonization source of New surveys of the larger islands were exhaustive, but wide-rang- Guinea shows that the decrease in slope is due to increasing ing "tramp" species were often ignored on the smaller intra-archipelago immigration rates, arising from overrep- islands. More recently Diamond surveyed 42 Solomon resentation of the most vagile inter-archipe ago immigrants islands, including 15 very small ones, reached the summits in more distant archipelagoes. When colonists are sorted into documenting species sets correlated with their dispersal abilities, the slope of the of many islands, and concentrated on species-area relation for the most vagile set is close to zero, distributions, altitudinal ranges, and interisland variation in but for the least vagile set is close to the value predicted by ecology and behavior. Additional information was obtained Preston for "isolated universes." from older resident men, many of whom, in the Solomons as in (22, 23), proved to be walking encyclopedias A major goal of biogeography is to explain how the number of natural history. The best informants were capable of reli- of locally coexisting species varies over the earth's surface. ably distinguishing sibling species in silhouette at consider- Islands with different species numbers have provided favor- able distances by behavioral cues, and accurately described ite study objects, because island communities have sharp, species seen or heard on a single occasion several decades unambiguous limits. After the recognition that island species previously under circumstances permitting independent number S is related to island physical properties such as area confirmation of the account. For these reasons we believe A, isolation d, and elevation L (1-3), attempts were made by that the list of island species numbers in Table 1, extracted multiple regression analysis to partition the variation of S from detailed tabulations by Mayr and Diamond (24), is sub- into its dependence on A, d, and L (e.g., refs. 4-6). Numer- stantially correct. ous subsequent studies have sought to interpret the resulting In this paper we discuss the dependence of species num- patterns in terms of immigration and extinction rates, ber on island area, while the following paper of this series species-abundance relations, the biology of individual (25) will discuss the dependence on isolation. species, environmental productivity or stability, and effects of history (7-17). Effect of area The present paper, which is the first of a series on the evo- lution and ecology of Northern Melanesian birds, continues Table 1 summarizes for each island the area A, elevation L, these studies by analyzing variation in breeding land and distance d of "isolated" islands (see table legend for defini- fresh-water bird species number among islands of the Solo- tion) from nearest species-rich island, number of breeding mon Archipelago in the tropical southwest Pacific. Each of land and fresh-water bird species normally resident in the us has carried out surveys of Solomon island avifaunas, Ernst lowlands S0ow, and number of montane species Smt. Since A, Mayr as a member of the Whitney South Sea Expedition in L, and d may all affect species number on islands, analyses 1929 and 1930, Jared Diamond in three expeditions between often begin by multiple regression of Stotal = Slow + Smt on 1969 and 1974. Solomon birds offer several advantages for A, L, and d (e.g., 4-6, 18). When exploring expeditions have understanding variation in species number. First, the islands provided many specimens but few field notes, so that Stotal is of this archipelago share similar climate, so that interisland known but species altitudinal ranges are not, such multiple variation in habitat is correlated mainly with island area A regression may be the only way to estimate the effect of ele- and elevation L. Were this not true, e.g., were rainfall a vation on Stotal. This method has the disadvantage of poor major independent variable, then habitat variation indepen- sensitivity to elevational effects, because the fraction Smt/ dent of variation in A and L would yield much unexplained Stotal is generally small. Since information about altitudinal variance in regressions of species number on A and L (cf. distributions is available for Solomon birds, we instead begin Fig. 6 of ref. 9; 18). Second, the evolutionary history of the by considering the relation between Slow and A, excluding Solomon avifauna has been intensively studied by Mayr (19). the isolated islands. We consider elsewhere (25) the depen- dence of Slow on d and A for isolated islands. We also show t Address reprint requests to J.M.D. elsewhere (24) that: (a) Smt can be analyzed separately from 262 Downloaded by guest on September 28, 2021 Zo-ology: Diamond and Mayr Proc. Nat. Acad. Sci. USA 73 (1976) 263 Slow, and is determined by A, L, and the altitudinal distribtl- cause of statistical fluctuations in numbers of species and in- tion of island area§; (b) effects of history on S are negligible dividuals on these islets (cf. also Fig. 1 of ref. 28). (see also legend to Fig. 1) because S on all is The interpretation of the second and third conclusions at or close to equilibrium, in contrast to the situation in cer- may be sought in the theoretical relation between species- tain other archipelagoes (12, 26). abundance relations and species-area relations, as developed Figs. -1 and 2 plot Slow and log Slow, respectively, against by Preston (7) and comprehensively reexamined by May log A. If the Slow-A relation were an exponential function (14). Suppose that islands were isolated universes, containing [exp (Slow) cc A], Fig. 1 would be linear; if a power function large, heterogeneous assemblies of species whose abundances (Slow a Az), Fig. 2 would be linear. Four conclusions are - depended on many more-or-less independent factors. Then vious: the central limit theorem of statistics leads to the prediction (i) Larger islands have more species. The biological fac- that the distribution of individuals a'mong species (the so- tors underlying this familiar finding are multiple and are called species-abundance relation) should be lognormal (p. analyzed elsewhere in the light of single-species distribution 89 of ref. 14). Assuming further that the total number of in- patterns (pp. 364-371 of ref. 15). dividuals is approximately proportional to island area, then (ii) The exponential function two consequences follow: (a) the species-area relation should be only slightly sublinear on a log-log plot (Fig. 10 of ref. Slow =134.7 + 12.08 log A [i] 14). (b) The asymptotic slope of this relation should be around 0.25 (likely extreme values, 0.15-0.39). If, on the other hand, one constructed the species-area relation for a provides an excellent fit (correlation coefficient 0.99) as A nested set of areas on a single land mass, then the species varies over a range of six orders of magnitude. Naturally, present on a given area would include not only the species this equation applies to bird species number only in the Solo- capable of persisting on the same area if it were an isolated mons, not in other archipelagoes. Below A about 1 square universe, but also species whose populations were "subsi- mile the log Sliw-log A relation is notably curved and de- dized" by immigration from adjacent areas. Such nested viates from a power function. The curvature is also much sample areas often yield the following properties: (a) the greater than predicted theoretically from a lognormal species-abundance relation may be of logseries or geometric species-abundance relation (Fig. 10 of ref. 14). rather than lognormal form. (b) The slope of the species- (iii) The slopes of Figs. 1 and 2 are very low in compari- area relation may be much less than for isolated universes. son with most published values for other island studies. That (c) The relation may be very sublinear on a log-log plot, and is, Slow increases relatively slowly with A in the Solomons. A approximately linear on a semilog plot. In these terms, both power function fitted through the points of Fig. 2 for A > 1 the slope and the form of the species-area relation suggest square mile yields a slope z = 0.087, comparable to or even that birds of the Solomon Archipelago do not see these lower than the value often observed for species-area rela- islands as isolated universes but as more like nested sample tions of nested continental sample areas, and much lower areas. Why is this so? than the values (typically 0.20-0.35) calculated in studies of The explanation is probably to be sought in the magnitude most other tropical island archipelagoes (7, 9, 14)1. of interisland immigration rates (17). For a fauna composed (iv) Slow values show relatively greater scatter (Fig. 2) for of species' with poor dispersal ability, immigration rates will islands smaller than 1 square mile than for larger islands, be- be low, and the species-abundance relation on an island will approach that of an isolated § Since for even the most mountainous Solomon island 59% of the universe. For a fauna of highly total area is below 2000 feet, and since the relation Silw against vagile species, the interisland water gaps become decreas- log A has a very low slope, the decrease in Slow with increasing ingly significant as barriers to immigration, and dispersal proportion of total island area above the 2000-foot contour is neg- among islands may eventually approach that among sample ligible. areas on the same land mass. Thus, for faunas of increasingly We suspect that somelz values reported in the literature for island vagile species the species-area relation should decrease in birds significantly overestimate the true values, for the following slope and may shift reasons. The species numbers on which island biogeographic anal- progressively in form from a power yses are based have often not been gathered by the biogeograph- function towards an exponential function. In the extreme of ers interested in the species-area relation, but have been extracted infinitely high immigration rates, all islands would share all from records of biological exploring expeditions seeking popula- species, and z would be 0. Recall that most bird species of tions of taxonomic interest. Like the Whitney Expedition, these the Solomons and of other tropical southwest Pacific islands expeditions tended to concentrate on the larger islands, because are derived from the source island of New Guinea. New they harbor taxonomically interesting endemic forms, and made more cursory surveys of smaller islands harboring mainly wide- Guinea bird species differ greatly in dispersal ability, and spread tramps. Undersurvey of small islands inflates the slope of S the most vagile of these species will become progressively versus A relations. Thus, Greenslade (27), using information then overrepresented among immigrant populations on islands at available to him on Solomon bird distributions, obtained an expo- increasing distances from New Guinea (15, 30). Thus, with nential-function slope of 22.2 compared to our value (Fig. 1) of distance from New Guinea the immigration rates of these 12.1, because his underestimate of S is now known to approach a vagile colonists between islands of an archipelago should in- factor of 2 or 3 for the smaller islands, as shown by Diamond's re- crease, even though immigration rates from New Guinea it- cent surveys. In addition, the same motivation often led biological exploring expeditions entirely to ignore islands smaller than 1 self decrease. Assuming that intra-archipelago immigration square mile, yet comparison of Figs. 1 and 2 shows that fits of the contributes far more colonists than inter-archipelago immi- species-area relation to a power function and to an exponential gration for an island in a remote archipelago, we therefore function could scarcely be distinguished if the points for islands smaller than 1 square mile were unavailable. Thus, before one 11However, not all sampling distributions have a logseries can assess either the slope or the form of a species-area relation, form, and not all observed logseries or geometric species-abun- one must consider whether these two common systematic biases dance relations arise as sampling distributions. For instance, early (undersurvey or complete neglect of small islands) have distorted successional communities often yield geometric species-abun- the data base. dance relations. See refs. 14 and 29 for discussion. Downloaded by guest on September 28, 2021 264 Zoology: Diamond and Mayr Proc. Nat. Acad. Sci. USA 73 (1976) Table 1. Well-surveyed Solomon Islands and their bird species number A (square L d A (square L Island S1ow Smt miles) (feet) (miles) Island Slow Smt miles) (feet) d (miles) 1. Bougainville 82 16 3317 8500 26. Wana Wana 56 0 26.5 262 2. 79 23 2039 8028 27. Nusalavata 9 0 0.032 -0 3. Ysabel 71 3 1581 4100 28. Kundu Kundu 12 0 0.0187 -0 4. Choiseul 70 3 1145 3180 29. 65 1 247 2600 5. Buka 63 0 236 1318 30. Ganonga 53 3 55 2800 6. Nggela (Florida) 61 0 142 1312 31. Rendova 61 4 147 348t 7. Shortland (Alu) 58 0 89.5 676 32. Tetipari 56 0 47 1330 8. Fauro 51 0 27.4 1926 33. 69 7 1663 4200 9. Vatilau (Buena 34. San Cristobal .Vista) 41 0 5.4 990 () 69 6 1193 3410 10. Nusave 35 0 0.206 230 35. Pavuvu & Banika 11. Bagora 22 0 0.1263 200 (Russells) 45 0 68 1780 28.8 (2) 12. Nugu 9 0 0.0577 -0 6.7 (6) 36. Mono (Treasury) 43 0 28.1 1165 17.8 (7) 13. 19 0 0.0351 -0 37. Ugi 50 0 16.3 670 14. New 15 0 0.0273 -0 38. Nissan 29 0 14.3 110 38.9 (5) 15. Dalakalonga 10 0 0.0257 128 39. Gizo 56 0 13.6 654 16. Elo 14 0 0.0211 50 40. Savo 37 0 11.8 1588 8.4 (2) 17. Kosha 13 0 0.0199 30 41. Santa Anna 46 0 5.66 520 18. Kukuvulu 11 0 0.0156 100 42. 43 0 5.0 1100 19. Tapanu 13 0 0.0062 -0 43. Three Sisters 32 0 4.28 250 12.2 (34) 20. Kanasata 6 0 0.0035 50 44. Santa Catalina 38 0 1.91 320 21. Nameless 8 0 0.0027 -0 45. Borokua (Mur- 22. NewGeorgia 65 1 789 3300 ray) 13 0 1.56 1181 36.4 (25) 23. 65 15 272 5800 46. Kicha 15 0 0.044 -0 24. Vang-unu 62 5 210 3686 47. Rennell 42 0 264 504 104.4 (34) 25. Gatukai 55 3 42 2912 48. Bellona 20 0 7.62 250 97.6 (2) 49. Ongtong Java 9 0 3.69 -0 147.0 (3) 50. 6 0 0.502 -0 108.5 (33) For each island, SiO is the number of resident land and fresh-water bird species normally occurring near sea-level; Smt, the number of species confined to mountains; A, area in square miles; L, elevation in feet; and d, distance in miles from the nearest island with Stotai > 50, identified by the island number in parentheses. d is not tabulated for islands with Stotai > 50, or for islands within 6 miles of such an island; such islands are considered not isolated (see ref. 25 for discussion). Islands 1-21 were joined during late-Pleistocene periods of low sea-level into a single expanded island, Greater Bukida. Islands 22-28, 29-30, and 31-32 were similarly joined into other Pleistocene islands, termed Greater Gatumbangara, Greater Vellonga, and Greater Rendipari, respectively. Islands 33-50 have had no recent connections.

expect the slope of an archipelago's species/area relation to tropical archipelagoes, as suggested by these considerations decrease with isolation**. This prediction is confirmed for of intra-archipelago immigration. four well-surveyed archipelagoes: the power-function expo- Additional evidence that the slope of the species-area rela- nent z for islands larger than 1 square mile is 0.22 for satel- tion varies inversely with interisland immigration rates lite islands of New Guinea itself (12), 0.18 for satellite islands comes from detailed examination of the species actually es- of 55-315 miles from New Guinea (28), 0.09 tablished in each archipelago. Diamond (15) assigned the for the Solomon Archipelago 390-840 miles from New bird species of the Bismarck Archipelago to several catego- Guinea (this paper), and 0.05 for the New Hebrides Archi- ries according to their distributional strategies (e.g., so-called pelago 11h1430 miles from New Guinea (Diamond and high-S species, A-tramps, B-tramps, C-tramps, D-tramps). Marshall, unpublished). Schoener's (17) analyses of variation Dispersal ability or interisland immigration rates, as gauged in bird species numbers within archipelagoes agree in show- by several types of evidence, increased from high-S species ing that z decreases with distance from the source, and dem- towards D-tramps. We have assigned species of the Solo- onstrate further that z is lower for temperate archipelagoes mons to distributional categories, and Diamond and Mar- than for tropical archipelagoes, correlated with the greater shall (unpublished) have done similarly for New Hebridean dispersal of temperate bird species. Schoener has also shown bird species, by the same criteria as previously applied to the that the species-area relation does shift from a power func- Bismarcks. Comparison of results for the three archipelagoes tion towards an exponential function as one proceeds from shows that the proportion of the archipelago's species that tropical to temperate archipelagoes, or from near to remote belongs to the least vagile categories decreases with archipel- ago distance from New Guinea (e.g., high-S species repre- ** MacArthur and Wilson (pp. 23-27 and Figs. 10 and 12 of ref. 9) sent 37%, 27%, and 11% of Bismarck, Solomon, and New stated that the slope of the species-area relation should increase Hebridean species, respectively). Conversely, the proportion with increasing distance, apparently opposite to the present re- of species in the most vagile categories increases with dis- sults. However, their discussion is based on scattered islands for tance (C- plus D-tramps represent 25, 46, and 57% of Bis- which immigration is assumed to be predominantly inter- rather marck, Solomon, and New Hebridean species, respectively). intra-archipelago. Their prediction is confirmed for the scattered A- isolated Solomon islands (25), which do fit their assumption. They If one considers only the poorly vagile high-S species and correctly predicted (their pp. 29-30) that island "clumping," by and B-tramps, the power-function exponent z for Solomon favoring intra-archipelago immigration, should reduce the slope. islands larger than 1 square mile is 0.28, a typical "isolated- Downloaded by guest on September 28, 2021 Zo-ology: Diamond and Mayr Proc. Nat. Acad. Sci. USA 73 (1976) 265

910- of the Gulf of Guinea islands [z = 0.49, or 0.44 after removal 0 810- o of distance effects by multiple regression analysis (6)]. In 7110- contrast to many other islands, some islands of these archi- 6io- pelagoes are not at species equilibrium, because they were 50 Slow 5 connected to species-rich sources by land bridges (or bridges 4S'0io- of montane vegetation in ref. 31) until the late Pleistocene. So- 31 Especially on the larger islands, species numbers have not 2i yet relaxed to the eventual lower values at which extinction AAAA40 AA 10_ and immigration rates will become equal. The high zs there- 0I fore reflect the dependence of extinction rates or so-called 0.001 0.01 0.1 10 100 1,000 1 0,000 Area (square miles) relaxation times (12) on island area, rather than reflecting FIG. 1. Species-area relation for birds of nonisolated islands of properties of islands at equilibrium. The high z of the four the Solomon Archipelago. Siow (ordinate) against log A (abscissa). Gulf of Guinea islands has been considered puzzling [(6), Symbols: 0, islands derived from the expanded Pleistocene island and p. 9 of ref. 9], but previous analyses did not consider the

of Greater Bukida (islands 1-21 of Table 1); A, islands similarly Pleistocene land bridge that joined the largest island, Fer- derived from Greater Gatumbangara, Greater Vellonga, and nando Po, to , leaving its species number presently still Greater Rendipari (islands 22-32); 0, islands with no recent con- supersaturated and the archipelago's z high. Omission of nections (islands 33-50). The line is the linear regression equation Fernando Po and multiple regression of S on A and d for the Si., = 34.7 + 12.08 log A fitted through all the points. Note the good fit of this line (correlation coefficient 0.99) over a 106-fold remaining islands, which lacked land bridges and should be = range in island area. Note also that the sets of points 0, A, and at equilibrium, yields z 0.33, within the range expected do not systematically differ from each other, meaning that histori- for "isolated universes." cal effects of Pleistocene landbridges on present-day bird species number are negligible in the Solomons. Outlook We conclude by drawing attention to two caveats and two universe" value. Conversely, the highly vagile D-tramps of unsolved problems. the Solomons yield z = 0.025: immigration rates are so high If species-area relations are constructed from island sur- that all islands of any size above 1 square mile share almost veys that were performed by other workers for other pur- all species. The progressive overrepresentation of vagile poses, we stress the need to evaluate whether the thorough- colonists with z nearly zero, and the progressive underre- ness of surveys varied systematically with island size; e.g., presentation of poorly vagile colonists with z of ca. 0.28, in whether smaller islands were considered "less interesting" increasingly remote archipelagoes cause the previously-dis- and were examined more cursorily. If surveys are carried cussed differences in z values averaged over all the archipel- out de novo, we stress the need to include really small ago's species. islands, so that fits of species-area values to an exponential We have now seen how z for an archipelago can be con- function and to a power function can be distinguished. siderably lower than the values around 0.20-0.35 expected Unsolved problems in the control of bird species number for a set of "isolated universes," correlated with high intra- on islands include the following: archipelago immigration. Some other explanation must be Is it true that the species-abundance relation fits a lognor- sought, however, in four cases where z is considerably great- mal or logseries distribution, for islands belonging to archi- er than 0.20-0.35: birds of New Guinea land-bridge islands pelagoes whose species-area relation fits a power function or (z = 0.39, calculated from ref. 12), mammals of Great Basin an exponential function, respectively?. mountaintops [z = 0.43 (31)], birds of neotropical land- How can the high intra-archipelago immigration rates of bridge islands (z = 0.44: calculated from ref. 24), and birds the Solomons, deduced from the form and slope of the species-area relation, be reconciled with the interisland vari- ation in bird populations at the subspecies and semi-species 100- level, for which the Solomons are famous (19, 32)? To what 80- , extent does the explanation for this seeming paradox lie in the ability of coselected or coevolved sets of local popula- tions to exclude invaders indefinitely? To what extent is the 40- / paradox an artifact of an analysis that averages immigration low ' rates for the whole species pool, and that thereby conceals the enormous differences in colonizing ability among differ- 20- 0.00 ent species, among different populations of the same species, and within the same population at different times in a cycle 10- 0/ of colonization? ,A 8 °,, In further analyses of these two problems, we suggest refs. 7 and 14, and 15 and 30, respectively, as starting points. 6- 01/ 0.001 0.01 0.1 1.0 I10 100 1,000 10,000 We thank R. May and T. Schoener for valuable clarifying sugges- Area(square miles) tions; the governments and numerous residents of the British Solo- mon and New for the field work islands of Islands Guinea, making FIG. 2. Species-area relation for birds of nonisolated possible; and the National Geographic Society for support. the Solomon Archipelago. As Fig. 1 , except that the ordinate is log instead of Slow. The solid straight line is the linear regression Slow Proc. 6th Pac. Sci. 191-216. equation log Siow = 1.59 + 0.087 log A fitted through the points for 1. Mayr, E. (1941) Congr., A > 1 mile2. Note that the relation actually defined by the points 2. Ripley, S. D. (1944) Bull. Mus. Comp. Zool. 94,305-430. is markedly curved. The dashed curve is the regression equation 3. Darlington, P. J., Jr. (1957) Zoogeography (Wiley, New York). Slow = 34.7 + 12.08 log A depicted in Fig. 1. 4. Hamilton, T. H. & Rubinoff, I. (1963) Evolution 17,388-403. Downloaded by guest on September 28, 2021 266 Zoology: Diamond and Mayr Proc. Nat. Acad. Sci. USA 73 (1976)

5. Hamilton, T. H., Barth, R. H., Jr. & Rubinoff, I. (1964) Proc. 18. Power, D. (1972) Evolution 26,451-463. Nat. Acad. Sci. USA 52,132-140. 19. Mayr, E. (1931-1957) numerous papers in Am. Mus. Novit. 6. Hamilton, T. H. & Armstrong, N. E. (1965) Nature 207, 20. Cain, A. J. & Galbraith, I. C. J. (1956) Ibis 98, 100-134 and 148-151. 262-295. 7. Preston, F. W. (1962) Ecology 43, 185-215, 410-432. 21. Mayr, E. (1945) Birds of the Southwest Pacific (MacMillan, 8. MacArthur, R. H. & Wilson, E. 0. (1963) Evolution 17,373- New York). 387. 22. Mayr, E. (1932) Nat. Hist. 32,83-97. 9. MacArthur, R. H. & Wilson E. 0. (1967) The Theory of 23. Diamond, J. M. (1966) Science 151, 1102-1104. Island Biogeography (Princeton University Press, Princeton, 24. Mayr, E. & Diamond, J. M. (1976) Bull. Mus. Comp. Zool., in N.J.). press. 10. Mayr, E. (1965) Science 150, 1587-1588. 25. Diamond, J. M. & Mayr, E. (1976) Proc. Nat. Acad. Sci. USA, 11. Pianka, E. (1966) Am. Nat. 100, 33-46. in press. 12. Diamond, J. M. (1972) Proc. Nat. Acad. Sci. USA 69, 3199- 26. Terborgh, J. W. (1975) in Tropical Ecological Systems: 3203. Trends in Terrestrial and Aquatic Research, eds. Golley, F. & 13. Simberloff, D. S. (1974) Annu. Rev. Ecol. Syst. 5, 161-182. Medina, E. (Springer, New York), pp. 369-80. 14. May, R. M. (1975) in Ecology and Evolution of Communi- ties, eds. Cody, M. L. & Diamond, J. M. (Harvard University 27. Greenslade, P. J. M. (1968) Evolution 22,751-761. Press, Cambridge), pp. 81-120. 28. Diamond, J. M. (1974) Science 184,803-80. 15. Diamond, J. M. (1975) in Ecology and Evolution of Com- 29. Pielou, E. C. (1969) An Introduction to Mathematical Ecolo- munities, eds. Cody, M. L. & Diamond, J. M. (Harvard Uni- gy (Wiley, New York), chap. 17. versity Press, Cambridge), pp. 342-444. 30. Mayr, E. (1965) in The Genetics of Colonizing Species, eds. 16. Lack, D. (1975) Island Birds (University of California Press, Baker, H. G. & Stebbins, G. L. (Academic Press, New York), Berkeley, Calif.). pp. 29-47. 17. Schoener, T. W. (1976) Proc. 16th Intern. Ornith. Congr., in 31. Brown, J. H. (1971) Am. Nat. 105,467-478. press. 32. Mayr, E. (1969) Biol. J. Linn. Soc. 1, 1-17. Downloaded by guest on September 28, 2021