EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS:   EFFECTS
OF ISLAND SIZE1
Daniel Simberloff
Department of Biological Science, Florida State University, Tallahassee, Florida 32306 USA
Abstract. A controlled experiment performed on 8 small mangrove islands constituted an exact test of several biogeographic hypotheses which had rested largely on unsatisfying statistical treatments. The islands were censused for arboreal arthropods; each contained a different subset of a species pool of ^ 500. Sections of the islands were then removed and censuses retaken after a waiting period; this procedure was repeated on four islands a second time. The results were consistent with a model which posits the islands as originally in a state of dynamic equilibrium between immigration and extinction, then re-equilibrating quickly when forced into an oversaturated condition. It was specifically demonstrated that: (1) species number increases with island size alone, independent of habitat diversity; (2) the increase with area is faster on separate islands than on increasing subsections of one island; (3) the area effect and the underlying dynamic equilibrium are not due only to an increased sample of transients and vagrants on larger islands, though there are a number of species which are particularly prone to quick extinction and which contribute disproportionately to the observed high turnover rates; and (4) predictions of the effects of decreased area on species composition can be stochastic at best, not deterministic. The equilibrium theory prediction of higher turnover rate on smaller islands could not be conclusively tested because of small sample size. Several species interactions suggested by the distributional data are so subtle that it is apparent that much more intensive work will be required to demonstrate even the existence of interactions, let alone whether they are important. It is clear that much of the dynamic equilibrium and its associated turnover in this system can be ascribed to individual species characteristics and the effects of a rigorous physical environment.
Key words: Area] equilibrium bio geography; Florida; mangrove islands; species interaction; species number; species turnover.
Introduction
The effects of island size on biotic richness have been exhaustively discussed over the last two decades; in recent reviews of the relevant literature (Simberloff 1972, 1974) I have depicted two independent hypotheses, and supporting evidence for each, which dominate current thought on this phenomenon. First is the notion, associated with the more general and much older consideration of the effects of area on species number, that larger islands tend to have more habitats than smaller ones. Because each habitat contains its own set of species, larger islands have more species.
Restricted habitat preferences are so fundamental a part of any biologist's working view of nature that habitat diversity was widely presumed to account for the entire area effect until the last decade. At that time Mac Arthur and Wilson (1963, 1967) suggested that island biotas are dynamic equilibrium entities, with species composition at any instant the result of continuing island-wide extinction balanced by immigration from the species pool of the mainland and/or surrounding islands. This hypothesis has attracted such wide attention that further  general
1 Manuscript received 11 October 1975; accepted 13 January  1976.
discussion might be likened to flogging a dead horse; suffice it to say that some direct experimental evidence supports the concept of a dynamic equilibrium biota and much observation appears consistent with it (Simberloff 1974). On the other hand, Lynch and Johnson (1974) attack this interpretation of some observations, particularly the inference that local extinction is a common event for birds.
With respect to the effects of area, the equilibrium view suggests that larger islands should have more species independent of habitat diversity, simply as a consequence of decreased extinction rates (Fig. 1). If, in addition, larger islands have increased immigration rates, this would also contribute to higher species numbers. MacArthur (1972) gives a probabilistic interpretation of the immigration and extinction rates of Fig. 1, partially responding to the observation (Simberloff 1969) that these rates are not ordinary single-valued functions of species number. So a direct effect of area on species number independent of habitat diversity, but possibly acting simultaneously with it, is posited: all species on an island have finite extinction ("turnover") probabilities. The smaller the island, the smaller are population sizes, therefore, the higher these "species-extinction" rates (Simberloff 1969), which, of course, comprise the island-wide extinction rate of Fig.  1.
630
DANIEL SIMBERLOFF
Ecology, Vol. 57, No. 4
z 3E	\ \(,large> l\	/ ^small	
smallx*			/^Harge
largeA*		,_______■-------                   1 _____L	
Table 1.    Parameters of experimental islands
s*small                                    s* large
NUMBER    OF    SPECIES
Fig. 1. Relationship of equilibrium species number (S*) and turnover rate (AT*) to island size. Dotted line depicts possible effect of area on immigration rate; large island may have higher immigration as well as lower extinction rate.
If the increase in species-extinction rates with decreasing population size were not linear, but rather reflected a critical population size below which extinction would be exceedingly likely (MacArthur and Wilson 1967), the independent effect of area on species number would be exaggerated. Direct experimental evidence on this proposition is lacking (Simberloff 1974), but so long as species-extinction rates were monotonie decreasing functions of population size, area would affect species number directly. Depending on the shape and distribution of the species-extinction vs. population size curves, one might even propose species-area curves resembling the allometric ones which are dictated (Johnson and Raven 1973) by the probability statistics of habitat addition with increased area. For this reason, as well as the limited information on local extinction in the general literature, we may question how we could detect the independent effect of area on species number and, even if it must exist, whether it would be nearly as important for most islands as the effect of habitat diversify.
Equilibrium theory also suggests that, of two islands identical except for area, the smaller should have the higher equilibrium turnover rate AT* (Fig. 1); this is what produces the lower equilibrium number of species S*. If immigration, as well as extinction rates are affected by area, no clear prediction can be made about turnover rates; perhaps one could surmise in the absence of exact data on the shapes of the curves that if extinction is more sensitive than immigration to area, smaller islands would probably still have higher extinction rates.
			D2 = distance
		Dl = distance	(m) from
		(m) from	nearest large
Island	Area	nearest	(> 1,000m2)
name	(m2)	island	island
CR1	343	2	2
Gl	519	5	120
INI	264	110	110
Jl	1,263	432	432
MUD1	990	38	82
MUD2	942	5	5
Rl	721	336	336
SQ1	1,082	139	139
WH1	380	84	84
Species and area data for island archipelagoes are automatically confounded by habitat diversity differences, many of which may be difficult to measure or even too subtle to recognize. For this reason multiple regression techniques which tend to emphasize the importance of habitat diversity as opposed to population sizes are not too compelling. Furthermore, the high correlation of habitat diversity (however it is measured) and area makes stepwise regression interpretations subjective; we could as well ascribe most of the variability in species number to area as to habitat diversity in many published studies (Simberloff 1974). Because of the high intercorre-lation of the independent variables, an experiment is needed that changes one variable, enabling us to distinguish between the effects of area and habitat diversity. This paper reports an experiment designed to separate the intercorrelated variables and to examine whether faunal turnover is significant in sizable communities or whether it is a quaint, mathematically tractable, but usually minor effect observable only in systems smaller than those of interest to most ecologists.
Materials and Methods
Nine red mangrove (Rhizophora mangle) islands were chosen in Florida Bay near Sugarloaf Key (Figs. 2 and 3). Areas, distances from nearest island and distances from nearest large island (> 1,000 m2), are listed in Table 1. Islands Rl and SQ1 had small satellite islands (diam < 4 m, distance < 5 m) not listed; each satellite consisted of a single Rhizophora mangle tree, which in no instance contained animal species not present on the larger island. A general description of mangrove islands is provided by Wilson and Simberloff (1969), who fumigated several of them to ascertain that each island did, in fact, have an equilibrium number of species. The islands of this experiment differ physically from those primarily in their greater area, up to 5 X that of the largest fumigated island. A consequence is the greater frequency of black mangrove (Avicennia germinans)
Summer 1976
EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS
631
gtv* MUD 2
^MUD I

Key West
Fig. 2.    Experimental and control islands in Florida Bay.
and even white mangrove {Lagune id aria race f nosa) on these larger islands. The nine islands were chosen partly to minimize the presence of these other trees, but small Avicennia bushes were present on INI, SQl, Gl, and MUDl; the latter two islands each had one small Laguncularia bush, while Rl had a larger dead Laguncularia tree. In no instance was an animal found on these other plants which was not also recorded from Rhizophora on the same island. Few animals generally were found on these small plants; neither species supports an entomo-fauna as large as that of red mangrove.
The nonmarine fauna of small Rhizophora islands consists almost entirely of arboreal arthropods (Wilson and Simberloff 1969); the intertidal substrate is covered by seawater twice daily, effectively elimi-
nating a ground fauna. A few insects are found on the intertidal mud when the water recedes, and retreat onto mangrove roots when the tides come in (Table 2). Those which are never found on the island proper and do not appear to feed on the mangrove or to interact with its other inhabitants are excluded from further consideration. The intertidal algae around the roots contain several arthropods (Table 2); again, most of these are never found above the algae and so will be discounted. The tree snail Littorina angulifcra and tree crab Aratus pisonii are ubiquitous but are omitted from subsequent census data. The snail rasps lichens from the bark and its fortunes are probably unrelated to those of other colonists. The crab has been observed to rasp and thereby   severely   to  damage  mangrove   leaves   and
632
DANIEL SIMBERLOFF
Ecology, Vol. 57, No. 4


b*N#
Fič   3.    Upper:    Control   mangrove   island   INI.    Loner:    Experimental mangrove island SQ1   (arrow), near Squirrel Key.   Tiny fumigated island El is in foreground.
also to catch and consume a large cricket (Tafalisca lurida): it may be an important faunal component. Even if extinction (disappearance of every individual) should occur for either of these species, recoloniza-tion would be qualitatively different from that of other colonists inasmuch as both species have plank-tonic larvae and the adult crab swims well.
I omitted dipterans because censusing was too inaccurate; larvae or pupae were discovered only 14 times in this experiment, anyway. There are many species of mites on mangrove islands (Simberloff and Wilson 1969, 1970), but the size of these islands precluded certain species counts, so these too were omitted. Finally, I did not count vertebrates, although breeding cormorants {Phalacrocorax auritus) were present on Rl and SQ1, breeding green herons (Bútor id c s v ire sc ens) at one time or another on most islands, nesting white-crowned pigeons (Columba leu-cocephala) on SQ1, breeding water snakes (Natrix si-pedon) on Rl, and a transient cotton rat (Sigmodon hispidus) on Jl. None of these animals are insectivorous, so it is unlikely that they contribute to population decline of other island colonists. Simberloff and Wil son (1969) showed that birds may carry small arboreal arthropods to mangrove islands in nesting material, while phoresy must occasionally occur.   A number of small insectivorous birds fre-
quent small mangrove islands, although only the gray kingbird (Tyrannus dominicensis) commonly nests there. Available data do not permit an assessment of how important these birds are in the population dynamics of resident insects, but it is suggestive that more than half the time that investigators were on the islands of this experiment, none of these birds were present. Abandoned birds' nests were found on all islands; their significance will be discussed.
This leaves a large community of arboreal insects, spiders, centipedes, millipedes, isopods, pseudo-scorpions, and scorpions as potential colonists of these islands. In this study 254 species were encountered, while 351 species have been collected on small Rhizophora islands I have visited. There are probably about 500 species routinely found in large Florida Keys Rhizophora swamps, and perhaps 4,000 in all the Keys.
Species were considered present on an island if individuals were observed that were potentially capable of population increase (reproduction); presumably this would be a pair or a fertilized female for sexual species, or just a female for species capable of parthenogenesis. If only males were observed (as occasionally happened for ants, for example) the species was considered absent, while an immature individual was considered evidence of reproduction.
Summer 1976
EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS
633
40 number of man-hours
Fig. 4. Cumulative counts of species vs. time spent collecting on island Gl in 3 successive yr. Each year an asymptote was reached.
With this criterion for presence or absence, at least 99% of all individual organisms encountered during this study could be either unambiguously identified in the field or collected and subsequently identified by specialists. In rare instances collection was impossible, and in others an immature form could not be certainly referred to a species (though often it could be placed to family and all but once to order). In such situations the most conservative possible accounting was made: if the doubtful individual could conceivably belong to any species recorded from its island, it was placed in that species.
The basic technique for censusing these islands, described by Wilson and Simberloff (1969), involves exhaustive examination of all microhabitats which comprise the mangrove: dead hollow twigs and branches, dead bark and tree holes, leaves, flowers, fruits, and green shoots. Because these islands are larger than the fumigated ones it was impossible to examine as great a fraction of most of these micro-habitats as in the earlier experiment, but the islands are still sufficiently small and simple that it was possible to examine virtually the entire island. In addition, entire abandoned birds' nests, small parts of active nests, bracket fungi, and litter and duff from tree holes and under bark were collected and subjected to light-powered Berlese funnels. Collections were continued until an asymptote was reached in cumulative number of species collected vs. time: 40-82 man-hours of collection not counting Berlese
funnels (which were always operated within the first 20 man-hours). Figure 4 depicts these curves of cumulative numbers of species vs. time spent sampling for island Gl during 3 successive yr. Of course the islands did not all yield identical curves: weather, condition of collectors, and size of island all caused differences, but the shape of the curve, with negative second derivative and a clear asymptote, was universal.
In the summer of 1969 all nine islands were cen-sused. Island INI served as a control throughout the experiment. In November 1969 fractions of the other islands were removed. In the summer of 1970 all islands were censused again, after which four of them (CR1, Gl, MUD1, and WH1) had further fractions removed. In autumn of 1971 these four islands plus INI and MUD2 were censused once again. A crew of workmen, under the author's supervision, using gasoline-powered saws on the canopy and hand-tools on the roots removed fractions of islands (Fig. 5). All removed areas were leveled to below the highest algal growths on the roots, so that they were totally submerged by high tides twice daily. Repeated subsequent examination of the remaining dead algae-encrusted roots (as many as possible were simply pulled out) revealed no colonists. The removed section of each island was loaded onto a barge, towed at least 300 m from the island to deep water, and dumped overboard to sink (green Rhizophora and most dead wood do not float). Areas of all islands were determined by planimetry of aerial photographs. Throughout the remainder of the experiment no regrowth of the removed sections was observed, and the few newly established Rhizophora seedlings were continually pulled out. However, by 1975 significant growth from new seedlings was observed on islands JI, MUD2, and Rl, with lesser establishment on all other experimental islands except for SQL It should be emphasized that these islands were chosen from among thousands of similar islands in the Keys, that such islands frequently are devastated naturally by hurricanes and high seas, and that the animals on these islands are all distributed at least throughout the Keys and are not in any sense rare or endangered.
Because the validity of all results and conclusions rests on the censuses, evidence for their completeness is presented first; it consists of seven kinds:
1)   That the cumulative species vs. sampling effort curves each reached an asymptote suggests that if there is a faunal component which these censuses fail to record, it must be forever hidden and not subject to discovery by greater effort (more time). Furthermore, one may infer that if such species exist, they would be missed on all islands and so at least the error is uniform.
2)   In no instance did the Berlese extraction re-
634
DANIEL SIMBERLOFF
Ecology, Vol. 57, No. 4
j*ř
Fig. 5.    Upper:   Island Rl  as tide comes in when  removal of right-hand section is nearly complete.    Lower: Removal of roots on island Rl; process was continued until water covered tops.   Note satellite island in background.
veal more than four species which had not been recorded by the normal census procedure; the maximum addition to any census was 5%. Even small, agile, secretive species like collembolans, psocop-terans,  and thrips are generally discovered in situ.
3)   As in the fumigation experiment (Wilson and Simberloff 1969), several dusk and night censuses were performed, and in no instance was a species recorded that had not been counted already. Crepuscular and nocturnal species are discovered in their daytime microhabitats in a system as simple physically as this one.
4)   When one island (Jl) was cut, ~ 5% of the cut vegetation (< 2% of the total size of the island) was placed on a stage constructed inside a covered canvas raft. The raft was then sealed and the contents fumigated with chloropicrin (tear gas), which works as a noxious agent for insects, driving out even woodborers (Wilson and Simberloff 1969); in effect, a giant floating Berlese funnel was created. The vegetation and raft floor were then examined for insects; 49 species (68% of the total census) were recorded from this small fraction, but none that had not been collected already. Woodborers can be discovered easily by hand in mangrove, but they must be one of the more difficult groups to census with certainty.
5)   On three islands (MUD2-1969, Rl-1969, and
WH1-1969), the island fraction which was to be removed was initially uncensused; only after an asymptote was reached for the remainder of the island was examination extended to the whole island. One goal of this censusing procedure was to test whether differential effort in different island sections, which I tried to avoid but which must nonetheless have occurred to some extent, could bias the results. Major differences would have dictated even more intensive collecting. But the censuses of the three islands minus the fractions to be cut revealed 91%, 96%, and 95%, respectively, of the species in the full censuses, while the censused fractions comprised only 35%, 66%, and 69%, respectively, of the island areas. These results imply that precisely equal sampling effort over all parts of an island is unnecessary and that most species, although they may have stringent microhabitat needs, are not particularly localized geographically within an island. These conclusions apply a fortiori to the census results from the tiny fumigated fragment of Jl.
6) On all islands except for the 1969 versions of CR1, MUD1, MUD2, Rl, SQ1, and WH1, records were kept of the number of species found more than once vs. those found but once. The rationale behind these data was that a high percentage of species recorded from one individual would indicate the existence  of  a  correspondingly  high  percentage  of
Summer 1976
EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS
635
Table 2. Animals of intertidal mud and algae. 1 = Found on mud, 2 = found in algae, X = not counted in censuses
Table 3.   Changes in parameters on experimental islands
Order
Species
Collembola
Orthoptera Dermaptera Hemiptera Coleoptera
Lepidoptera Araneida
Isopoda
Anurida maritimer'2*x Axelsonia littoralis1'2'x Seira bipunctata1
Hygronemobius alleni1, x
Labidura riparia1
Pentacora sphacelata1'x
Actinopteryx fucicola1'x Anisomeristes sp.1, x Bembidion cf. contractum1' Micronaspis floridana2 T achy s occulaíoŕ Gen. sp. (Staphylinidae)1,x
Gen. sp. (Unknown)2,x
Clubiona littoralis2'x Piráta arenicola2'x Stenoonops minutuŕ
Lige a exotica1, x
the true species complement which were unrecorded at all (cf. Preston's [1962] discussion of the "veil line" in lognormal species abundance curves). The range for fraction of species recorded more than once was 0.645-0.836, with a mean of 0.742. Only one major group, the small parasitic wasps, contributed disproportionately to the fraction of records consisting of just one individual, and this is the group for which, by nature of its small size, position in the trophic web, and active habits, the censuses are most likely to be deficient. But even for this group the majority of records are for more than one insect.
7) Finally, on island SQ1-1970 all individuals up to 200 per species were counted and then killed. For the 80% of species represented by more than one individual, usually many more than one were seen (x = 52). This supports the claim that if a species is present at all, it is likely to be present in number and to have been recorded.
A second question which may be raised is whether the time allotted for re-equilibration (7 mo, then 1 yr) was sufficient. The results of the fumigated island experiment (Simberloff and Wilson 1969, 1970) imply that for smaller mangrove islands with this degree of isolation 4-8 mo is sufficient for re-establishment of the equilibrium number of species, although the composition may still be peculiar (e.g., lacking millipedes, centipedes, and certain ants). In the experiment reported here, island MUD2 served as a control in the matter of whether enough time elapsed for re-equilibration; it was reduced only in 1969. That the number of species fell by 17 in the 7 months following area reduction, then by only 1 in the following year supports the claim that suf-
Island		Area	Total	Special11
name	Year	(m2)	species	species
CR1	1969 1970 1971	343 104 54	74 65 62	0 5 2
Gl	1969 1970 1971	519 327 169	86 77 69	0 0 0
INI	1969 1970 1971	264 264 264	63 63 68	0 0 0
Jl	1969 1970	1,263 779	75 71	4 3
MUD1	1969 1970 1971	990 565 320	79 76 71	0 0 0
MUD2	1969 1970 1971	942 327 327	79 62 61	4 0 0
Rl	1969 1970	721 478	103 85	0 4
SQ1	1969 1970	1,082 731	88 82	0 5
WH1	1969 1970 1971	380 261 123	86 73 72	0 0 1
a In Berlesates of birds' nests or bracket fungi.
ficient time was allowed. However, it may be noted that MUD2 was the second most severely reduced of the original group, down to a third of its original area, which ought to have exaggerated the difference between the initial species number decline and the subsequent one.
Results
Areas and species numbers of all islands are listed in Table 3; compositions are in the Appendix. To begin to assess the effects of area on species number, an attempt was made to regress species number of the original nine islands on the variables listed in Table 1. Because "stepping stone" islands were present for Gl and MUD1 only, variables Dl and D2 were highly correlated (r = 0.97) and the stepping stone data could as well have been omitted. In any event, no significant simple or multiple regression was produced with all possible combinations of log-transformed and untransformed variables.
MacArthur and Wilson (1967) suggest that on very small islands high area-independent extinction rates may eliminate an area effect, but mangrove islands clearly do not fit this pattern. The smaller fumigated islands always had fewer species than did these islands (Simberloff and Wilson 1969), and tiny Rhizophora islets consisting of a single bush, 1 m or so high, typically have even fewer species. However, a much larger island (Bill Finds Key), which
636
DANIEL SIMBERLOFF
Ecology, Vol. 57, No. 4
is a virtual Rhizophora monoculture, has yielded 120 species in only 24 man-hours of collection.
One reason why these nine islands do not show a clear area effect upon regression is their narrow span of areas; the largest is < 5 X the size of the smallest. Darlington (1957) suggested that for island herpetofaunas, a tenfold area increase is usually associated with only a doubling of species number; a similar rule can be extracted from Preston's bird data (1962). Particularly with such a small sample of islands, species number variance not associated with area variance would likely obscure any statistical dependence of species number on area over this narrow range of areas.
It was impossible to increase sample size for these regressions beyond nine islands by including the decreased versions of the eight experimental islands because the successive species sets on these islands were not independent, even though the numbers may have re-equilibrated. If each fauna were viewed as an independent draw from a collection of 254 equi-probable colonists, the number of ways in which two islands with n and m (^ n) species, respectively, could share exactly (m - Í) species (with i ^ 254 - n)
is (254ř"n)(mn_i). The expected number of shared species under this hypothesis is then
ě(254f")U"-r)(™-0/(2£4) = W254.
'     t = 0
Because species are not equally likely to colonize (discussed below), the observed number of species held in common between two islands is always much greater than this expected number; because the expected number is heavily dependent on the sizes of the faunas (m and n), the probability of a number as large as or larger than the actual number is a better measure of the nonrandomness of two faunas than is the deviation of actual number from expected number. For all 36 possible pairs of the original nine islands, the mean excess of actual number of shared species over expected was 22.2, and the median probability of a deviation as large as observed was 4 X 10-1(). For all 15 possible pairs of islands where one island was simply the other (modified physically or not) during the next census period, the mean excess of actual number of shared species over expected was 31.8, and none of these deviations was as probable as 10~13. As time passes an island fauna becomes successively more different from (and independent of) the original composition: for the six possible pairs of islands where the first differs from the second by two census periods (19 mo), the mean excess was 27.4 species (always less than the deviation between successive censuses on the same island), and two of the deviations were as probable as  100"18.   But over the course of this experiment
225
area (m2)
Fig. 6.    Individual   species   number   vs.   area   curves for the experimental islands in a log-log plot.
the composition revealed at each census period was largely determined by the original composition.
The second reason, in addition to the combination of narrow area range and small sample size, that these nine islands do not show a clear area effect upon regression is the reason why regression is not quite suitable for elucidating area effects generally: the presence of other, possibly unrecognized, independent variables affecting species number. On an island-by-island basis (Fig. 6) the importance of area is clear: there were 12 decreases and no increase. Such an extreme result, if the null hypothesis of no area effect were correct, would occur just (0.5)V2 = .0002 of the time. The rise in species number on control island INI is probably no more than random variation about an equilibrium, particulars in light of the concurrent decrease in species number of the partial control island MUD2. But the control changes in no way suggest that the observed effect of area is an artifact.
It was not possible to modify just the one parameter of area. For example, perimeter could not be held constant without also changing shape. But, as has been said earlier, no colonists have been observed to prefer a geographic location like the edge (as opposed to a microhabitat like the canopy of leaves). Mangrove islands are remarkably homogeneous; a fraction of one is truly just like an entire one except for area because all Rhizophora microhabitats are present in all sections, probably in similar proportions with the possible exception of dead bark. However, the effects of microhabitat diversity may be very subtle; on CRl, JI, MUD2,
Summer 1976
EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS
637
Rl, SQ1, and WH1 a few species were found only in abandoned birds' nests, while Berlesates of bracket fungi on Rl and SQ1 produced species not observed elsewhere during the appropriate census. Numbers of these special colonists are given in Table 3. In no instance is the independent effect of area brought into question.
That these islands are behaving like islands for this biota should be emphasized. The data indicate that most or all of the species do not treat these islands as part of a fine-grained foraging system; invasion of one of these islands from without is much more difficult than dispersal within a large island. If this were not so, the higher equilibrium number of species on larger islands could be viewed simply as a consequence of a higher immigration curve, rather than of a lower extinction curve. Osman (1975) has been able to separate, for small marine rocks, the contributions of decreased extinction and increased immigration, respectively, to an observed area effect and has demonstrated that the latter predominates. In effect, the rocks are acting as different-sized sampling devices collecting settling marine organisms in proportion to their sizes. As stated earlier, for three islands in 1969 (MUD2, Rl, and WH1) censusing was first restricted (until an asymptote was reached) to the section which was* to be left after the subsequent area-reduction. The censuses of the three islands minus the fractions to be cut yielded 116%, 116%, and 112%, respectively, of the subsequent censuses 7* mo later, when the same territories comprised entire islands instead of parts of larger islands. This suggests that the mangrove islands, unlike marine rocks, are not just sampling, proportionally to their sizes, a constantly active pool of propagules.
Looked at another way, this result shows that species number increases more slowly with area when successively larger quadrats within a large system are examined than when successively larger discrete islands are taken. This is the same phenomenon depicted by MacArthur and Wilson (1967) when they pointed out that the exponent in the allometric species-area relationship is lower for mainland sections than for an archipelago of true islands; this experiment has produced the first completely appropriate test of this proposition, for precisely the same territory was first a section of a larger island, then a discrete island. MacArthur and Wilson suggest that the reason for the lower exponent for quadrats within a large island or mainland is that species which cannot maintain themselves in a small area may be frequent transients if the small area is embedded in a larger one which can support breeding populations. Such transients would be rarer if the area were a separate island with no nearby source for these species.  Whether this is the true explanation for the
result on MUD2, Rl, and WH1 cannot be determined from the numerical information alone. That > 90% of the species present were found in 35%-69% of each island is certainly consistent with MacArthur and Wilson's hypothesis. There is no reason to think that all observed species were actually breeding in the censused fragment, and in this sense at least some were likely transients from the un-censused part of the island.
As explained below, the best one can conceivably do to predict the compositional nature of the species number increase on larger and larger islands is to state probabilities that given species will be present given a certain area, and the mangrove insect data do not yet permit even this kind of stochastic prediction. I can summarize the numerical species-area data by saying that they are in accord with the hypothesis of higher probabilities of extinction on smaller islands for all species, resulting in fewer species present at any given time.
All other things being equal, and assuming that immigration rate is much less affected by area than is extinction rate, a small island ought to have a higher turnover rate than a large one (Fig. 1). In this experiment the number of immigrations may serve as an estimate of the equilibrium turnover rate; the excess of extinctions over immigrations may be viewed as the additional extinction wrought by the decreased area. Of course the observed immigrations and extinctions underestimate the actual numbers, because species which immigrate and go extinct (or vice-versa) between monitorings will not contribute. For the fumigated mangrove islands with 18-day intervals between monitorings I suggested (Simberloff 1969) that perhaps half of all turnover would go unrecorded for this reason; Gilroy (1975) feels that this is unduly pessimistic and that 10% might be a more realistic figure. The much longer intervals between censuses on the area-modified islands ought to increase the unobserved turnover at the same time as the generally larger island size ought to decrease actual turnover rates. Which one will affect observed turnover more cannot be guessed, but it is the observed rates with which we must work, and we must assume they are correlated with the actual rates.
In Fig. 7 the turnover rates thus defined are plotted against number of species, with single island values connected. Three rates rose upon decrease in area, in accord with the prediction, and two declined. For all points, the correlation coefficient is -0.252 (n = 15, NS). That the control (INI) turnover rate changed (increased) more than any other contributes to the insignificance of this result. If the same data are examined as percent turnover vs. species number (as in Diamond 1969) four rates rise and one falls on the modified islands; only 19%
638
DANIEL SIMBERLOFF
Ecology, Vol. 57, No. 4
initial number of species
Fig. 7.    Turnover rate vs. area curves for the experimental islands.
of the time would a result this extreme be observed if the null hypothesis of no effect of area were true. Furthermore, for these data r = -0.687 (P < 0.01), but this nearly monotonie decrease in percent turnover vs. S, similar to that published by Diamond (1969) for Channel Island birds, must surely be an artifact. Species number, the denominator of percent turnover, decreases when area is reduced, so there is a bias for percent turnover to increase on smaller islands. It is the actual value of the turnover, not the percentage, which is important.
Discussion
I have interpreted this experiment as a disturbance of islands from an equilibrium condition, followed by a re-equilibration similar to that of the fumigated mangrove islands (Simberloff and Wilson 1969). Here, of course, equilibration will be occurring from an oversaturated rather than undersaturated condition. If one assumes constant coefficients of immigration and extinction (Diamond 1972), the curve for fall from oversaturation to equilibrium would be St = S* + (S0-S*)<?-(A'i + A'()ř, where St = number of species at time ŕ, S0 = number at time of area change, S* = equilibrium number for new area, and Kt and K, are the coefficients of immigration and extinction, respectively. This asymptotic curve is a mirror image about S* of the expected curve for an undersaturated (e.g., fumigated) island (Fig. 8). The time to fall any fraction x of the way to S* for the oversaturated island (or to rise a fraction x of the way to S* for an undersaturated one) would be given for this model by tx = -ln(l -x)/(Ki + Kc). The presence in the denominator of Kc, which one might expect to be a decreasing function of area, suggests the intuitively obvious: that large islands should take longer to re-equilibrate from over-saturation (where extinction is the rate-limiting event).
.St = S++(S0-SJe"(K' + KeU
time = t
Fig. 8. Predicted "relaxation" curve of species number vs. time for an island out of equilibrium, given constant coefficients of immigration and extinction (Ki and Kc). Dotted lines represent more realistic model where coefficients are functions of S, the number of species present.
{Kt in the denominator implies that distant islands should take longer to re-equilibrate from under-saturation, where immigration is the rate-limiting event.) Diamond (1972) found exactly this trend— larger islands remaining oversaturated for longer—for birds on islands off New Guinea. I have already noted that immigration rates might be higher for larger islands, which would increase Kt and therefore the denominator of the equation for tx, lowering the "relaxation" time required for re-equilibration to be substantially complete. There is no corresponding expected effect of distance on Kc.
The Is and Es curves are likely concave in nature, as indicated in Fig. 1 (MacArthur and Wilson 1967); the reason is that certain species will be good disperses and others (possibly some of the same ones) will be good at persisting on islands. As the former invade, they lower the immigration curve disproportionately; as the latter are added to the biota, they raise the extinction curve disproportionately (i.e., subsequent species will have higher species-extinction rates). So K{ and Kc are functions of S, not constants; in the same vein, Terborgh (1975) treats Ke as a constant but uses a factor of S2. This restricts the shape of the extinction curve in Fig. 1, but the effect on return to equilibrium from perturbation is similar: to make the return quicker than it otherwise would be (Fig. 8). The precise equation for re-equilibration analogous to the equation already given for td, depends on the functional relationship of Ki and Kc to 5.
Summer 1976
EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS
639
The interpretation oí the area effect in this experiment as predominantly an increase in the extinction rate which lowers a dynamic equilibrium deserves further comment. Lynch and Johnson (1974) point out that much claimed avian turnover could actually be construed as kkpseudoturnover,,: either an artifact of incomplete censuses or passage of transients, migrants, and vagrants which, if studied carefully, would be found never to breed on an island and so ought not to be regarded as contributing to immigration or extinction rates. Their second concern is that turnover attributable to human modification of the environment cannot tell us much about the workings of nature. In the Keys mangrove, aside from the area modification, it is hard to conceive oi another secular environmental change, man-induced or otherwise.
With respect to their first contention, I have already listed evidence for completeness of my censuses. Whether or not actual breeding occurred usually cannot be proven, although the frequent presence of immature individuals of most of the species suggests that we are dealing here with breeding populations. Certainly there must be some species in the Appendix (such as herbivores which cannot reproduce on Rhizophora) which, although they fulfill the technical definition of "propagule," are unlikely ever to breed on such islands or to constitute a major component of the trophic web. The infinitesimal probabilities described in the previous discussion of similarities of island faunas are largely due to the presence in the species pool of such obligate transients. But the majority of species recur again and again in this system; 171 of 254 species encountered in this study (67%) were counted in more than one census, and I have recorded at least another 45 (18%) on other Rhizophora islands. Furthermore, 48 (58%) of the species recorded just once in this study are either parasitic wasps or predators, which ought to have low population densities on thermodynamic grounds alone and might be expected to have high species-extinction rates. A few of the larger ones, such as the vespid Polistes and braconid Iphiaulax, may not be treating these islands as islands, but rather as parts of larger "home ranges" or fine-grained foraging areas. Although sample sizes are too small for a chi-squared contingency test and differences often rested on just one or two species, it may be noted that in 10 of 12 situations where area was decreased the proportion of predators and parasites decreased; this happened in two of three control situations.
On the other hand, there were a number of extinctions recorded of species whose individuals are largely restricted to one island where it seems highly unlikely that the species could have been missed in 40  or   more   man-hours   of  observation.    The   ants
Camponotus abdominalis floridanus, Crematogaster ashmead'u and Paratrechina bourbonica all have large colonies of active foragers and became extinct 4 times in this experiment. The adult spiders Nephila c lav i pes, Gasteracantha ellipsoides, Allepeira lem-niscata, and Tetragnatha an til lana are large and have highly visible webs; even the spiderlings are easily discovered, particularly by their webs. Together they accounted for 16 extinctions. Micronaspis floridana, a lampyrid beetle, is large and usually appears to be present in number if at all; it disappeared five times. Because of its microhabitat requirements and sizable colonies the termite Cryptotermes cavifrons is easily discovered"; six extinctions occurred. These are just some of the more striking species recorded as going extinct; in the final analysis assessment of the amount of turnover must rest on faith in the census methods, and for the reasons already stated I feel that the great majority of listed extinctions are real. Furthermore, because most of them involve "valid1' mangrove colonists, the area effect described here is not simply the uninteresting phenomenon of larger sample sizes turning up more transients and vagrants. This is not to say that even among species which regularly colonize Rhizophora islands there are not some species which are particularly good dispersers and others which are especially persistent. The de-faunation experiment (Simberloff and Wilson 1969) provided evidence particularly on differences in dispersal capabilities; this experiment sheds additional light on extinction. A measure of a species1 ability to persist is the ratio of its observed number of extinctions to the potential number of extinctions. (Presence in 1970 on Jl, Rl, and SQl and 1971 on the other islands of this experiment could not have been followed by observed extinction.) There were 1,148 extinctions which could conceivably have been observed, but only 349 actual extinctions. One group of 96 species (38% of all those observed) contributed 111 of these extinctions out of only 121 possible; these species are poorly adapted to survival in this habitat and it is among them that transients in the sense of Lynch and Johnson (1974) must be sought. At the other end of the gradient, a group of 22 species were present at all censuses or absent only once, and produced only 11 extinctions of a potential 320. Among these are four ant species which are typically common and among the ecological dominants; two weevils, a termite, and a moth which contribute a large fraction of the woodboring which produces the favored microhabitat of many mangrove species (including ants); and two crickets and two roaches which are ubiquitous and numerous. Finally, three of these species, the nonballooning spider Ariadna arthuri, the polyxenid millipede Lophoproctinus bar t sc hi, and the oniscid isopod Rhyscotus sp., combine poor dispersal ability (Sim-
640
DANIEL SIMBERLOFF
Ecology, Vol. 57, No. 4
berloff and Wilson 1969) with extremely good persistence once present. This contrasts with several spiders (such as the orb-weavers already mentioned) and bugs which seem to be adept at reaching islands but are quite likely to be extinguished once there.
Extinction is a stochastic rather than deterministic event in this system, and the increase in extinction caused by decreased area did not have precisely predictable components. Certain species (e.g., the ant Camponotus abdominalis floridanus and the scorpion Centruroides key si) are much more frequently found on large islands than small ones, and among the ants, in particular, there is a quite ordered succession of nested subsets on larger and larger islands. But there are exceptions to all these rules (e.g., the presence of scorpions on the small remaining fraction of CRl in 1971), and aside from the weak rule that predators and parasites are more likely to suffer extinction from decreased area, no generalizations can be made. Any species is more likely to disappear on a smaller island.
Finally, a word is in order about the nature of the biological processes or phenomena that organize mangrove island insect communities and allow numerical, and rarely compositional, predictions to be made. Heatwole and Levins (1972) suggest that the trophic web of these communities is such an organizing force and that a species' trophic position and interactions with other species determine in large part whether or not it will be found at some time on some mangrove island. I have questioned (Sim-berloff 1976) whether the data from the fumigated islands support this claim and proposed that interactions determining presence or absence, if they exist, are likely to be subtle and complex and to require detailed observation to confirm their importance. I would extend this principle to the area-modified islands as well; the presence-or-absence data do not permit many inferences to be made about interactions. Four suggestive patterns are now described which indicate the sort of additional information required.
First, the nearly omnipresent roach Latiblattella sp. nov. has black nymphs with two thin white bands crossing the abdomen. These bands appear to give the roach a waist, and these roaches are occasionally seen running in files of the black ant Crematogaster ashmeadi. Furthermore, they have been observed to elevate the rear of their abdomen not unlike the characteristic way in which Crematogaster raises its gaster when disturbed. Crematogaster dominates red mangrove islands, except when large colonies of Camponotus abdominalis floridanus are present, everywhere except the Mud Keys, where it is rare or absent. In this experiment Crematogaster was present and abundant everywhere but INI (it was absent in 1970, rare the other 2 yr)   and the two
Mud Keys, MUD1, and MUD2, which are in an archipelago distant from all other islands of this experiment. Latiblattella sp. nov. was recorded everywhere except the Mud Keys, where it was never observed. The similarity between MUD1 and MUD2 transcends this pair of species but is subtle enough that it could not be detected by cursory inspection of the Appendix. The deviation between expected and actual overlap, given the assumptions described above, is larger (31.2 species) and less probable (< 10-13) than for any other pair of islands.
Second, Camponotus abdominalis floridanus is a very large, active ant whose size precludes it from nesting in the thin hollow twigs used by all other mangrove ants. It is forced to use special micro-habitats: large, low, hollow aerial roots, the space between two large, appressed branches, or tree holes. The roots are labile, frequently falling into the water, and probably are too small to support large colonies. The other two microhabitats are rarer, but when a Camponotus colony establishes itself in one of them it may grow very large, the workers swarming all over the branches and even being visible from a boat. This situation occurred once in this experiment, on INI, and on a nearby fumigated island, ST2. On the former Crematogaster has always been absent or present just as tiny new colonies consisting of a queen and a few undersized workers, probably new colonies (E. O. Wilson, personal communication). On the latter Crematogaster was originally extremely abundant but has become progressively rarer. The nature of the interaction, if one exists, cannot yet be guessed.
Third, the waxy scale insect Ceroplastes rubens is not seen on many mangrove islands. In this experiment it occurred very frequently on INI, tended by Camponotus abdominalis floridanus, and in 1971 on CRl, also tended by Camponotus. One may imagine that Ceroplastes has a much higher species-extinction rate in the absence of the ant.
Fourth, the centipede Orphnaeus brasilianus is a common mangrove island inhabitant and quite possibly its predation on crickets helps to control cricket population sizes (Simberloff and Wilson 1969). It was found in great abundance on two of the original fumigated islands, seven of the islands in this experiment, and five other small mangrove islands. A second centipede, Cryptops sp., which also lives in hollow twigs and under dead bark, is about the same size and is also probably a generalized predator (R. Crabill, personal communication). It was observed only on CRl, where Orphnaeus was absent, and WH1, where Orphnaeus disappeared after 1969. For Orphnaeus, at least, dispersal to small mangrove islands is a problem; it never recolonized the two fumigated islands E7 and E9.   There is no reason
Summer 1976
EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS
641
to think that Cryptops is more adept at overwater dispersal. From the distributions and the disappearance of Orphnaeus on WH1 one might tentatively postulate that these centipedes exclude one another competitively or by predation. (Crabill suspects that they are even cannibalistic.) The wider distribution of Orphnaeus may be due either to better dispersal powers or to a stochastic competitive superiority.
Conclusions and Summary
This experiment demonstrates conclusively that a smaller island will support fewer species than a larger one, all other things being equal. In particular, if we equate microhabitats of the mangrove island system with habitats of larger archipelagoes, it is clear that habitat diversity differences are not a sufficient explanation of the widely observed effect of area on species number. That population sizes of most species on these islands are large suggests that the independent effect of area on species number is not restricted to microcosms of little interest to most ecologists. The area effect in this experiment was a consequence of the dynamic equilibrium on any island between immigration of new species and extinction of those already present. It supports the notion that any species is more likely to become extinct on a smaller island. Because of unobserved immigration and extinction it is impossible to measure turnover rates exactly, but they must have a minimum of about 15 species per year for an island with a standing equilibrium fauna of about 75 species and are likely much larger. Some species are exceptionally prone to extinction, but some of this turnover is not generated by the fluke arrival and swift subsequent extinction of species unsuited for mangrove island colonization.
The results of this experiment are consistent with an hypothesis of fast asymptotic approach to equilibrium from an oversaturated condition, the precise curve resting on the functional relationship between immigration and extinction rates and number of species. Equilibrium theory predicts that turnover rates should be higher on a smaller island; this is neither confirmed nor refuted by these data, probably because sample size was too small and turnover rates on untreated islands of this type are too high and variable. A comparison of percent turnover with island size has an inherent statistical bias which makes it uninteresting.
The processes underlying the observed turnover and its relation to area remain obscure. It is clear not only that there are distributions of dispersal and colonization abilities among potentially colonizing species but also that there is a large historical component in the determination of an island biota. In this experiment  19 mo was not a sufficient in-
terval to render two successive censuses of the same island independent of one another, or at least as independent as contemporaneous censuses of two different islands. It is my hypothesis that the characteristics of individual species—invasion and species-extinction rates—adequately explain most of the observed pattern of species number vs. area. Proof of this hypothesis will be difficult, resting on the demonstration that interactions between species are less important factors in population dynamics in this system than are aspects of the physical environment. Even the interior of these islands is exposed to wind-driven waves, and violent rainstorms characterize the Keys. Even ordinary rains and moderate wind and seas frequently cause damage such as broken branches and leaves killed by mud-spattering. Drowned individuals are often seen, but to prove that such physical factors are dominant causes of population decline would be a formidable task. The distributional data alone cannot be used either to prove or to disprove the hypothesis, and the paucity of evidence on four subtle interactions suggested by the data imply that even a description of most interactions between species, let alone an assessment of their importance, will be a monumental undertaking.
Acknowledgments
I thank Jack Watson (Great White Heron National Wildlife Refuge) for assisting me to find safe areas where this experiment would not cause ecological damage, and Steve Tendrich (National Exterminators, Ft. Lauderdale) for imaginatively and assiduously coping with unique problems encountered in changing island areas. Assistance in collection was provided by T. Kelley, R. Swain, and D. C. K. Young, and in curatorial chores by T. Ray. Joe Beatty and Edward Mockford generously contributed their time to identifying colonists, and systematic assistance, without which this experiment would have been impossible, was provided by B. D. Burks, K. A. Christiansen, R. E. Crabill, G. W. Dekle, H. E. Evans, R. C. Froeschner, R. Gordon, A. B. Gurney, J. F. Lawrence, P. M. Marsh, L. Masner, R. W. Matthews, E. M. Miller, W. B. Muchmore, C. W. O'Brien, E. S. Ross, L. M. Russell, L. J. Stannard, H. H. Tippins, B. D. Valentine, T. J. Walker, F. G. Werner, R. E. White, S. L. Wood, and P. W. Wygodzinsky. The manuscript profited from a careful reading by E. Connor, K. Heck, and E. McCoy. Finally, Edward O. Wilson provided both encouragement and inspiration throughout. This work was supported by NSF grants GB-8668 and GB-28082.
Literature Cited
Darlington, P. J. 1957. Zoogeography: The geographical distribution of animals. John Wiley & Sons, New York.    675 p.
Diamond, J. M. 1969. Avifaunal equilibria and species turnover rates on the Channel Islands of California. Proc. Natl. Acad. Sei. U.S.A. 64:57-63.
---------.    1972.    Biogeographic kinetics:   Estimation of
relaxation times for avifaunas of Southwest Pacific islands.    Proc. Natl. Acad. Sei. U.S.A. 69:3199-3203.
642
DANIEL SIMBERLOFF
Ecology, Vol. 57, No. 4
Gilroy, D. 1975. The determination of the rate constants of island colonization.    Ecology 56:915-923.
Heatwole, H., and R. Levins. 1972. Trophic structure stability and faunal change during recolonization. Ecology 53:531-534.
Johnson, M. P., and P. H. Raven. 1973. Species number and endemism: The Galapagos Archipelago revisited.    Science 179:893-895.
Lynch, J. F., and N. K. Johnson. 1974. Turnover and equilibria in insular avifaunas, with special reference to the California Channel Islands. Condor 76:370-384.
MacArthur, R. H. 1972. Geographical ecology. Harper & Row, New York.    269 p.
MacArthur, R. H., and E. O. Wilson. 1963. An equilibrium theory of insular zoogeography. Evolution 17:373-387.
MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Princeton Univ. Press, Princeton, N.J.    203 p.
Preston, F. W. 1962. The canonical distribution of commonness and rarity. Ecology 43:185-215, 410-432.
Simberloff, D. S. 1969. Experimental zoogeography of islands: A model for insular colonization. Ecology 50:296-314.
----------.    1972.    Models in biogeography.   In T. J.  M.
Schopf [ed.] Models in paleobiology. Freeman, Cooper & Co., San Francisco.    250 p.
----------. 1974. Equilibrium theory of island biogeography and ecology.    Annu. Rev. Ecol. Syst. 5:161-182.
----------.    1976.    Trophic   structure   determination   and
equilibrium in an arthropod community. Ecology 57: 395-398.
Simberloff, D. S., and E. O. Wilson. 1969. Experimental zoogeography of islands: The colonization of empty islands.    Ecology 50:278-296.
Simberloff, D. S., and E. O. Wilson. 1970. Experimental zoogeography of islands: A two-year record of colonization.    Ecology 51:934-937.
Terborgh, J. 1975. Faunal equilibria and the design of wildlife preserves. In F. B. Golley and E. Medina [ed.] Tropical ecological systems. Springer-Verlag, New York.    443 p.
Wilson, E. O., and D. S. Simberloff. 1969. Experimental zoogeography of islands: Defaunation and monitoring techniques.    Ecology 50:267-278.
Appe	ndix:    Colonists of	the experimental islands.   An "X"	in column			U 2,	or	3	indicates		species observed				in	1969,	1970,		or	1971	, respectively.						3 3
	Family	Genus and Species	(	CR1			Gl		INI			Jl 1    2		MUDl			MUD2			Rl 1   2		SQ1 1   2		WH1			(T>
Order			1	2	3	1	2	3	1	2	3			1	2	3	1	2	3					1	2	3	-J ON
Thysanura	Lepismatidae	Lepisma sp.					X	X	X	X	X		X	X	X	X						X	X		X	X	
Collembola	Entomobryidae Hypogastruridae	Seira bipunctata Brachystomella Villa-lobosa Xenylla grisea	X	X	X		X	X	X	X	X X	X X	X	X	X	X	X	X	X			X	X X			X	
		Xenylla sp. nov. cf. stachi	X	X	X	X	X		X	X		X	X	X	X	X	X	X	X	X	X		X	X	X	X	
	Sminthuridae	Sminthurides bixidus				X																					
Orthoptera	Blattidae	Aglaopteryx gemma	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X			X	X	X	X		
		Latiblattella rehni	X	X		X	X	X	X	X	X	X	X	X	X	X	X	X	X			X	X	X	X	X	w X
		Latiblattella sp. nov.	X	X	X	X		X	X	X	X	X	X							X	X	X	X	X	X	X	
		Pycnoscelis surinamensis																						X	X	X	w
	Gryllidae	Cycloptilum antillarum	X	X	X	X	X	X				X	X	X			X		X			X	X		X	X	FÖ
		Cycloptilum spectabile	X	X		X		X	X	X		X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	s
		Cycloptilum trigonipalpum															X										
		Cyrtoxipha confusa	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X		X	X	X	X	X	
		Cyrtoxipha sp. nov.												X													>
		Orocharis gryllodes	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X			X			r
		Tafalisca lurida	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	N
Isoptera	Kalotermitidae	Cryptotermes cavifrons Neotermes castaneus	X			X			X	X		X		X X	X X	X	X			X	X			X	X		o o 0
		Neotermes jouteli	X			X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	m
Dermaptera	Labiduridae	Labidura riparia										X															o 0
Embioptera	Oligembiidae	Oligembia hubbardi					X															X					*>
	Teratembiidae	Diradius caribbeana	X	X	X	X	X	X	X	X	X	X	X	X			X	X	X	X	X	X	X	X	X	X	>
		Diradius vandykei										X			X	X	X					X	X				
Psocoptera	Caeciliidae Lepidopsocidae	Caecilius incoloratus Echmepteryx youngi	X	X X	X	X	X			X	X	X	X X	X X	X X	X X	X X	X X	X X	X	X	X	X	X	X	X	o
	Liposcelidae	Belaphotroctes ghesquierei Belaphotroctes simberloffi			X	X				X X	X X			X	X	X	X	X		X X	X X	X X	X X	X	X	X	"Ti GO r
		Liposcelis bostrychophilus	X	X	X	X	X	X	X		X	X	X	X					X		X	X	X	X	X	X	
		Liposcelis entomophilus												X	X	X		X		X			X				>
	Peripsocidae	Ectopsocus maindroni																			X						Ö CO
		Ectopsocus sp. cf. vilhenai		X																							
		Ectopsocus sp. C		X							X		X							X							
		Peripsocus pauliani																X		X		X					
	Pseudocaeciliidae	Pseudocaecilius citricola																		X	X	X	X	X			
	Psocidae	Indiopsocus texanus			X	X	X	X			X			X	X		X	X			X	X				X	
	Psoquillidae	Rhyopsocus sp. cf. bentonae		X								X															
	Psyllipsocidae	Psocatropos micro p s	X																								
Thysanoptera	Idolothripidae Phlaeothripidae	Allothrips watsoni Barythrips sculpticauda Hoplothrips sp. cf. fungosus Karnyothrips flavipes	X	X		X				X			X	X	X X		X	X			X X		X X X	X X	X	X	
		Li o t h rips i si an die a	X	X		X	X	X	X		X	X	X	X	X	X		X	X	X	X	X	X	X			
		Neurothrips magna]emoralis													X												
		Neurothrips punanus																X									ON
			Appendix:	(Continued).																			
	Family	Genus and Species	ČŘÍ		Gl	INI		Jl 1   2		MUDl			MUD2			Ř1		SQ1 1   2		WH1 1   2   3			
Order			1   2   3	1	2   3	1	2   3			1	2	3	1	2	3	1	2						
Thysanoptera	Thripidae	Pseudothrips inequalis	X   X	X	X   X	X	X   X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	X	
Hemiptera	Anthocoridae	Cardiastethus assimilis Dufouriellus afer Xylocoris galactinus	X				X						X					X					
	Miridae	Campylomma sp.	XXX	X	X   X		X    X	X	X	X	X	X	X	X	X	X	X	X	X		X	X	
	Nabidae	Carthasis decoratus	XXX	X	X   X					X	X	X	X		X	X	X	X	X	X			
	Pentatomidae	Gen. sp.										X											
	Reduviidae	Ploiaria sp.															X						
	Unknown	Gen. sp.	X																				
Homoptera	Aleyrodidae	Aleurothrixus sp.	X		X   X		X    X	X	X							X	X	X	X	X	X	X	
		Paraleyrodes sp.		X	X			X	X	X	X	X		X	X	X	X	X	X	X	X		
		Tetraleurodes sp.								X					X								
	Cicadellidae	Coelidia melanota		X	X   X	X	X   X	X	X	X	X	X	X	X	X	X			X	X	X	X	
		Scaphytopius nigrinotus	XXX	X	X   X	X	X    X	X	X	X	X	X	X	X	X	X	X		X	X	X	X	
	Cicadidae	Gen. sp.											X										> 2
	Cixiidae	Nymphocixia vanduzeei	X		X		X			X	X	X	X			X							
	Coccidae	Cataenococcus oliv ace u s		X	X   X											X		X		X			
		Ceroplastes rubens	X			X	X   X																s
		Dysmicoccus alazon		X		X												X		X			C/5
		Dysmicoccus bispinosus				X										X	X						1—H w
		Dysmicoccus brevipes	X   X				X   X	X	X	X	X	X		X	X				X		X	X	
		Ferrisia virgata																		X			
		Pseudococcus sp.																X					*>
	Flatidae	Flatoidinus acutus				X							X										o "TJ
	Psyllidae	Gen. sp.	X																				
	Tropiduchidae	Neurotmeta breviceps	X	X	X		X	X	X	X			X		X	X	X	X	X	X	X		■TJ
	Unknown	Gen. sp.					X																
Neuroptera	Berothidae	Lomamyia sp.								X													
	Chrysopidae	Chrysopa collaris Chrysopa rufilabris	X   X	X	X	X	X	X	X		X	X		X	X	X X	X X	X X	X X	X	X	X	
Coleoptera	Aderidae Anobiidae	Cnopus impressus Ganascus ventricosus Cryptorama holosericeum	X	X		X					X X					X	X						
Anthicidae Anthribidae
Buprestidae
Cryptorama minu tum
P e talium sp.
Tricorynus abbreviatus                 X   X
Tricorynus inflatus                                 X
Sapintus fulvipes
Acaromimus americanus
Cisanthribus sp. nov.
Gen. nov., sp. nov.
Actenodes auronotata                        X   X
Chrysobothris sexfasciatus            X
Chrysobothris tranquebarica         X
Mastogenius sp.
X x
XXX   x   x
XXX
x x
x x
XX   XXX   XXX
XXX
x
XX   XXX   x   x x
x x x
x x
x x
XXX
x x x x
x
x x
M o
o o
<
o
2
o
Summer 1976
3
.9 c
ô
TS
e
If	CM
CZ!	CM
2	CM
CM S	CM T—1
	CM
	CM
	m
	CM
	m
O	CM
	^
2 u	CM
	c« "o Si C/5
	-d G cd
	CO 3 C O
	1 cd
	S-H •s o
EXPERIMENTAL ZOOGEOGRAPHY OF ISLANDS
x
645
x
x x
x x
x x
x x
x
x x x
x x x x
x x
XX  XX
x x x
x	x      XXX      XX
x	XXX     x
x	XXX     x     x
	xxxxxxxx
	x x x x x x
x x
X    XXX
XX x x x x x
x x x x
x x x
x	x		x x	x x
x x		x	x x	x x
	x	x x	x x	x x
x	x	x x	x x	x
x		x	x x	XXX
x		x	x x	x
x		x	XXX	x x
x		x	x x	x
x		x	x x	
x		x	x x	x
	x		x x x x	x
	x		XXX	x
	x		XXX	x
	x		x x	x
x	x	x	x	x    x
x	XXX		XXX	x    x
XX   xxxxx x x    xxxxxxx
XXX          x
x    x x    x
XXX      x XXX          x
xxxxx    x x x    x    x
xxxxx    x
XX          x
xxxxx xxxxx        x x           x
x x
x    x x
x    x
x x	x x
xxxxx	x x
xxxxx	x    x
x x	x
x	x
x x	x
x x
x x
x
x
x    x
fc.  ^  o
.§1

8
a >>
TD^O
"■s s
Oh O
U
O} < ^^
TD d
'-C TD
SS
co .3
>> O
»-. o X3 O
. ^ *i s
c o a 2 &*£> =
§1
s s á
-C)
S^ž   co         6,   *>
^
c c s
^  co    q
^   03    S
££  2
?<.   s*.     O

*> 53 o ^ ^ c os
°§-l
S3.** O                     -*•
O  v.             ^CM^
í^            "^  o-,  S                   »C
cxi cL,^ c^^ ci, d, d, o áiái
CO  CO   ŕ! "^ O   co   co  co  O  to   to
c c S» & ^ c c c E c c
.3 .g
.0
O^3
co -q
9-5 o^SŠ
£ S. ^^> c *> £ 5
•3 * \£
C'"*
C** C»    CO    CO    CO    CO    CO
3 S.....
c c c c

^   S   <L>   <L>   <L>   O   <L>
Si?
^ÖikJeqO^OOOÖO ^0 ^ (2 c^
TD
TD
žš
5(3

^ cd .a >> G
co TD b »C  Sž  Sž
«-^s >> a s J2
STD cii n S^3 O tí 2 ^:3 <U  <t)  Í3  «>  <u •T3
^ 'TT'
cd O
cd
TD

^ CM m Tt >o
C  C  C  C
>>    2S o o  m ä ^ SJí-l-g S o o o o o  S?
lil '■SililflSJJJJJsá'
Q o 00 o  g-
■  ^1
c c c c c
HDD  WOKOEíiSwDDDDD <0Q
o
TD
O
Appendix:    (Continued).
ON ON
Order
Family
Genus and Species
CR1
1   2   3
Gl
1   2   3
INI 1   2   3
Jl
1   2
MUDl
1   2   3
MUD2
1   2   3
Rl
1   2
SQ1
1   2
WH1
1   2   3
Hymenoptera      Braconidae
Chalcidae
Chalcidoidea
Colletidae Diapriidae Encyrtidae
Eulophidae
Eumenidae Eupelmidae
Evaniidae Formicidae
Gasteruptiidae
x  x
X   X
Acrophasmus sp. cf. lycti
A g at his sp.
Apanteles hemileucae
Callihormius bifasciatus
Heterospilus sp.                                X
Iphiaulax sp.                                     X
Macrocentrus sp.                                   X
Orgilus sp.
Gen. sp.  1
Gen. sp. 2
Arachnophaga sp.                            X                  X
Brachymeria psyche
Trigonura insularis
Gen. sp. 1                                         XX
Gen. sp. 2
Hylaeus sp.                                                  X                  X
Paramesius sp.
Encyrtus sp.
Euryrhopalus schwärzt                                              X   X
Ooencyrtus submetallicus
Plagiomerus diaspidis
Pseudaphycus sp.                                                       X
Encarsia var ie gat a
Eretmocerus sp.
Syntomosphyrum ischnopterae
Tetrastichus sp.
Pachodynerus nasidens                    XXX
Anastatus sp.
Metapelma schwarzi
Poly moria sp.
Gen. sp.
B rachy my r mex obscurior                X
Camponotus abdominalis                X        X
Camponotus tortuganus                             X
Camponotus (Colobopsis)  sp.                            XXX
Camponotus sp.
Crematogaster ashmeadi                 XXX
Monomorium floricola                    XXX
Par aery ptocerus varians                  XXX
Paratrechina bourbonica                 XXX
Pseudomyrmex elongatus                XXX
Pseudomyrmex "flavidula"             XXX
Tapinoma littorale                           XXX
Xenomyrmex floridanus                 XXX
Gen. sp
X   X
X   X
X   X
X   X
X   X
X X
X X
X
X X
X X
X
X X
X X
XXX
X X
X
X X XXX
XXX XXX
X X
X X
X X X
X X
X X
X X
XXX XXX XXX
X X X X
X X X
XXX XXX XXX
XXX XXX
X
X
X X X
X X X X X
X X
X X X
X
X
X
X
X   X
X
X   X
X   X
X   X
X   X
X   X X
X   X
X   X
X
X
X X
X X X
X X X
X X X
X   X
X
X   X
X   X
X   X
X   X
X   X
X   X
XXX
X    X X X
X    X
X
X X
X X
X
X X
X X X
X XXX XXX X X XXX XXX XXX
XXX X
XXX XXX
> z 5
2 do
w
r o ►n 41
o
o*
CTQ
<
o
o
			Appendix:	(Continued).										3 3
	Family	Genus and Species	CR1	Gl		INI	Jl 1   2	MUDl	MUD2	Rl	SQ1 1   2	WH1		
Order			1   2   3	1   2	3	1   2   3		1   2   3	1   2   3	1   2		1   2	3	
Hymenoptera	Ichneumonidae Pteromalidae	Calliephialtes ferrugineus Casinaria texana Spalangia drosophilae		X   X X   X		X	X	X	X	X    X X X	X    X X X	X	X	
	Scelionidae	Parabaeus sp.		X   X		XXX	X	X    X	X    X	X    X	X			
		Probaryconus striatus	X	X	X	X				X	X	X	X	
		Probaryconus sp.								X				
		Telenomus sp.		X				X	X	X	X			
		Trimorus sp.		X		X          X								w
	Sphecidae	N i tela sp.		X					X					X
		Trypargilum johannis			X	X   X	X		X    X	X	X	X   X		
	Vespidae	Poli st e s sp.				X				X		X   X		*
	Unknown 1	Gen. sp.			X									1—H w
	Unknown 2	Gen. sp.					X							
	Unknown 3	Gen. sp.						X						2
	Unknown 4	Gen. sp.											X	2 r N o
	Unknown 5	Gen. sp.											X	
	Unknown 6	Gen. sp.											X	
Unknown	Unknown	Gen. sp.						X						
Araneida	Anyphaenidae	Aysha velox	XXX	X   X	X	XXX	X   X	XXX	XXX	X    X	X   X	X   X	X	o o M
		Gen. sp.	X			X								
	Araneidae	Allepeira lemniscata Araneus juniperi		X			X X   X		X   X	X   X	X	X		o o > -0
		Argiope argentata.		X						X				
		Eustala sp. A	X   X	X   X	X	X          X	X   X	XXX		X    X	X   X	X   X	X	
		Eustala sp. B					X							x
		Eustala sp. C							XXX					<
		Gasteracantha ellipsoides	X   X	X		X          X	X   X	X          X	X   X	X	X   X	X		o
		Neoscona sp.					X							►n
		Nephila clavipes	X   X	X   X		XXX	X	X          X		X    X	X	X		C/5
	Clubionidae	Gen. sp.	X	X	X	X   X		X   X	X		X			r
	Dictynidae	Dictyna sp.										X		> Ü
	Gnaphosidae	Gnaphosa sp.	X											
		Sergiolus sp. A	XXX			X		XXX	XXX		X	X    X	X	
		Sergiolus sp. B					X							
	Hersiliidae	Tama habanensis		X   X	X			XXX	XXX					
	Oonopidae	Stenoonops minu t us					X			X				
	Piratidae	Piráta sp.							X					
	Salticidae	Admestina sp.	XXX	X   X	X	X	X   X	XXX	X    X	X    X				
		Hentzia grenada	XXX	X   X	X	XXX	X   X	XXX	XXX	X   X	X   X	X   X	X	
		Mae via sp.	X									X		
		Metacyrba taeniola		X   X	X		X    X	XXX	XXX		X   X			
		Metacyrba undata	X   X									X	X	
		Metaphidippus sp.										X		
		Gen. sp.  1										X		
		Gen. sp. 2										X		
														ON 4^
Appendix:     (Continued ).
	Family	Genus and Species	i	CR1		Gl		INI	Jl	MUDl	MUD2	Rl		SQ1 1   2	WH1			
Order			1	2	3	1   2	3	1   2   3	1   2	1   2   3	1   2   3	1	2		1	2	3	
Araneida	Segestriidae	Ariadna arthuri	x	x	x	x  x	x	x	x x	XXX	XXX	x	x	x x	x	x	x	
	Tetragnathidae	Leucauge venusta Plesiometa argyra		x		x	x	x	x	x x	x	x	x	x x				
		Tetragnatha antillana		x		x  x	x	x	x x	XXX	x       x	x		x	x	x	x	
		Tetragnatha sp.				x			x									
	Theridiidae	A r gy r ode s sp. Coleosoma acutiventer Euryopis sp. Theridion adamsoni Theridion sp. Gen. sp.	x x					x	x	x		x x						> s r
	Unknown 1	Gen. sp.			x													C/3 I—H
	Unknown 2	Gen. sp.						x										2
	Unknown 3	Gen. sp.		x														öd
	Unknown 4	Gen. sp.											x					
	Unknown 5	Gen. sp.		x												x		r
Chelonethida	Cheiridiidae	Neocheiridium sp.									x							o
	Cheliferidae	Tyrannochelifer sp.	x	x	x	x  x	x	XXX	x x	XXX	XXX	x	x					►n
	Chernetidae	Americhernes reductus Parachernes bisetus		x	x	x  x	x	x XXX				x			x		x	
	Olpiidae	Serianus sp. Solinus sp.			x	x	x			x       x						x		
Scorpionida	Buthidae	Centruroides key si			x						XXX							
Chilopoda	Cryptopidae	Cry p to p s sp.	x	x											x	x	x	
	Oryidae	Orphnaeus brasilianus				x	x		x	x  x	x	x	x	x x	x			
Diplopoda	Polyxenidae	Lophoproctinus bartschi	x	x	x	x  x	x	XXX	x x	XXX	XXX	x	x	x x	x	x	x	
Isopoda	Armadilliidae	Gen. sp.				x		x	x					x	x	x	x	
	Oniscidae	Rhyscotus sp.	x	x	x	x  x	x	XXX	x x	XXX	XXX	x	x	x x	x	x	x	M o
o
CľQ V-