Ecology, 83(3), 2002, pp. 689-702 ? 2002 by the Ecological Society of America THE ANT FAUNA OF A TROPICAL RAIN FOREST: ESTIMATING SPECIES RICHNESS THREE DIFFERENT WAYS JOHN T. LONGINO,14 JONATHAN CODDINGTON,2 AND ROBERT K. COL WELL3 ^Evergreen State College, Olympia, Washington 98505 USA department of Entomology, National Museum of Natural History, Smithsonian Institution, Washington, D.C. 20560 USA ^Department of Ecology and Evolutionary Biology, University of Connecticut, U-43 Storrs, Connecticut 06269-3043 USA Abstract. Species richness is an important characteristic of ecological communities, but it is difficult to quantify. We report here a thorough inventory of a tropical rain forest ant fauna and use it to evaluate species richness estimators. The study was carried out in ~ 1500 ha of lowland rain forest at La Selva Biological Station, Costa Rica. Diverse methods were used, including canopy fogging, Malaise traps, Berlese samples, Winkler samples, baiting, and manual search. Workers of 437 ant species were encountered. The abundance distribution was clearly lognormal, and the distribution emerged from a veil line with each doubling of sampling effort. Three richness estimates were calculated: the area under the fitted lognormal distribution, the asymptote of the Michaelis-Menten equation fit to the species accumulation curve, and the Incidence-based Coverage Estimator (ICE). The per- formance of the estimators was evaluated with sample-based rarefaction plots. The inventory was nearly complete because the species accumulation curve approached an asymptote, the richness estimates were very close to the observed species richness, and the uniques and duplicates curves were both declining. None of the richness estimators was stable in sample- based rarefaction plots, but regions of stability of estimators occurred. The explanation of rarity is one key to understanding why richness estimates fail. Fifty-one species (12% of the total) were still uniques (known from only one sample) at the end of the inventory. The rarity of 20 of these species was explained by "edge effects": "methodological edge species" (possibly abundant at the site but difficult to sample because of their microhabitat), and "geographic edge species," known to be common in habitats or regions outside of La Selva. Rarity of 31 species remained unexplained. Most of the 51 rare species were known from additional collections outside of La Selva, either in other parts of Costa Rica or in other countries. Only six species were "global uniques," known to date from only one sample on Earth. The study demonstrates that patterns of species occurrence early in an inventory may be inadequate to estimate species richness, but that relatively complete inventories of species-rich arthropod communities are possible if multiple sampling methods and extensive effort are applied. Key words: abundance distribution; ants; biodiversity; Costa Rica; Formicidae; incidence-based coverage estimator; inventories, species; La Selva; lognormal; Michaelis-Menten; richness estimation; species accumulation curves; tropical rain forest. INTRODUCTION asked by visitors is always "How many species do you There is an abiding interest in the number of species have?" ?ne can show them a list of the sPecies ob" in local communities, because species are the funda- served and count them UP' but that does not answer mental particles of biodiversity (Bisby 1995, Gaston the question of how many species are present at the 1996). A major research agenda in ecology is to explain slte' ?bserved number of species is a biased estimate geographic patterns of species richness (Rosenzweig of community species richness, because observed rich- inn<;\ A/T ? A- JJ -.u I .-?<- J- I u ? ness can never be higher than true richness (barring 1995). Many studies address the latitudinal change in b \ t> ? , , . ., ? ? identification errors), but it can be and usually is lower, species richness and why there are so many species in ' J ., . ?., , ? ., , , Although many studies rely on taxonomists' species the tropics. Others examine change with elevation, ask- fc ' ' r , ., .. . . , , , . ,. . lists, with the assumption that the taxonomic work is ing whether there is a mid-elevation peak in diversity, . . . , .? , _ , . , . ^ . . ,, exhaustive and the lists are complete, ecologists have and it so, why. Conservation biologists examine wheth- . .. ... . . , . . . , , . also devised methods tor estimating species richness er species richness changes with land use history or degree of landscape fragmentation, and whether rich- ness changes over time. When an investigator is car- based on quantitative sampling (e.g., Soberon and Llo- rente 1993, Colwell and Coddington 1994, Palmer 1995, Peterson and Slade 1998). This study reports on rying out a taxon inventory of a site, the first question , . , , , . ? , ,TT a relatively thorough inventory of the ants (Hymenop- Manuscript received 23 March 2001; revised 4 April 2001; tera: Formicidae) in a Costa Rican rain forest, and eval- accepted 10 April 2001; final version received 16 April 2001. uates the efficacy of three richness estimation methods. 4 E-mail: longinoj@evergreen.edu There are several reasons for choosing ants as a sur- 689 690 JOHN T. LONGINO ET AL. Ecology, Vol. 83, No. 3 TABLE 1. Field sampling techniques and definitions of sampling methods used in the inventory of the ants (Formicidae) of La Selva Biological Station, Costa Rica. Sample distribution No. Method Sample unit No. samples subsites Duration Fogging 1-m2 funnel 10-40/tree 24 trees Malaise 2-wk catch 8-15/trap 7 traps 13 mo Berlese 1.7 L litter/soil core 20 sites 13 mo Winkler 6 L sifted litter 41 sites 14 yr Barger transect of 20 baits 10 sites 1 mo Thompson 1 nest 30-m2 area of forest floor 1 mo Longino Other various temporally and restricted samples various temporally and restricted samples spatially spatially multiple sites multiple sites 14 yr 23 yr vey taxon. Ants and other invertebrate taxa should be included in biodiversity studies because of their high diversity and rapid response to environmental change (Kim 1993, Kremen et al. 1993, Miller 1993, Samways 1993, Basset et al. 1998). When choosing invertebrate taxa to survey, the relative merits of different taxa are often discussed (e.g., Basset et al. 1998). Agosti et al. (2000) have made a strong case for the inclusion of ants. In most terrestrial ecosystems ants are a conspic- uous presence, and they are among the most commonly studied terrestrial invertebrates. They are mostly scav- engers and predators, and they have large impacts on soil formation, seed predation and dispersal, and in- vertebrate community structure (Folgarait 1998). They are particularly good candidates for monitoring com- munity changes that may occur due to climate change or habitat fragmentation. Ants are also an important group to monitor for the arrival and impact of invasive species. Ecological studies of ant communities are very com- mon but in most cases are the result of relatively small- scale sampling programs using one or few sampling methods. There are increasing numbers of very thor- ough inventories of leaf litter ants (Belshaw and Bolton 1994, Fisher 1996, 1998, 1999a, b, 2000, Delabie et al. 2000). Other studies involve sampling the whole community with a variety of methods, in the context of answering particular ecological questions, but they do not purport to be exhaustive inventories of a local ant fauna (e.g., Andersen and Majer 1991, Andersen 1992, Andersen and Reichel 1994, Armbrecht and Ul- loa-Chacon 1999, Gadagkar et al. 2000). Inventories involving both sustained effort and a variety of sam- pling methods, with the express purpose of capturing the entire fauna, are relatively rare. Notable examples include Van Pelt (1956) for the Welaka Reserve, Flor- ida; Talbot (1975) for the E. S. George Reserve, Mich- igan; and Deyrup and Trager (1986) for Archbold Bi- ological Station, Florida. Tropical rain forest studies include Verhaag (1990, 1991) for Panguana Reserve, Peru, and Bruhl et al. (1998) for Kinabalu National Park, Sabah, Borneo. Here we report the results from a thorough inventory of the ants of La Selva Biological Station, an active research station located in a lowland tropical rain forest in Costa Rica (McDade et al. 1993). Earlier work on the ant fauna of La Selva includes Olson (1991) and Roth et al. (1994), but these were smaller studies that did not attempt to sample the entire ant community. The results reported here are part of a larger arthro- pod inventory at La Selva, Project ALAS (Longino and Colwell 1997; also see the project ALAS web site).5 The project combines the methods of museum taxon- omists and field ecologists (Longino 1994), with the primary goal of achieving thorough inventories as ef- ficiently as possible. An additional goal is to apply quantitative sampling procedures that allow the inves- tigation of inventory efficiency and richness estima- tion. Longino and Colwell (1997) analyzed inventory efficiency; this second report addresses estimation of species richness. METHODS La Selva Biological Station, Heredia Province, Costa Rica, is approximately 1500 ha, with an elevation range of 50-150 m and ~4 m mean annual rainfall. The hab- itat is a mosaic of mature lowland rain forest, second growth forest of various ages, and abandoned pastures (McDade et al. 1993). The data set on which this paper is based is a species X sample incidence (presence-absence) matrix. The samples fall into eight categories: fogging and Malaise sample the arboreal fauna; Berlese, Winkler, Barger, and Thompson sample the soil and litter fauna; Longino and Other sample a combination of microhabitats (Ta- ble 1; also see the Appendix). Samples were processed by the four Project ALAS Parataxonomists, or by J. T. Longino (see the Appen- dix). All quantitative results in this paper were based on the worker caste; males and queens were excluded 5 URL: March 2002 ANT FAUNA OF A TROPICAL RAIN FOREST 691 (much like excluding seeds from a vegetation survey). Most ant taxonomy is based on the worker caste, and the presence of a worker is clear evidence of colony establishment. All specimens were examined and iden- tified by Longino, with the exception of one species record from the literature (Supplement). Richness estimation and evaluation of estimators Bunge and Fitzpatrick (1993) and Colwell and Cod- dington (1994) reviewed methods for estimating the number of species in a community. Three broad cate- gories of estimators are (1) fitting a lognormal abun- dance distribution and estimating the hidden or unsam- pled portion of the curve, (2) fitting asymptotic equa- tions to species accumulation curves, and (3) using nonparametric estimators that use the relative abun- dances of rare species to estimate the number of species not seen. We used these three methods to estimate spe- cies richness of the La Selva ant fauna. Abundance distributions.?Various methods have been proposed for fitting sample data to a lognormal distribution and for estimating species richness by cal- culating the total area under the fitted curve, including the portion of the curve hidden behind the "veil line" (Preston 1948, Williams 1964, Ludwig and Reynolds 1988, Magurran 1988, Lobo and Favila 1999). We fol- lowed the method of Preston, in which abundance clas- ses or "octaves" had boundaries 0.5, 1, 2, 4, 8, etc. If a species fell on a boundary, then its abundance was evenly split between the two adjacent octaves, adding 0.5 to each one. The first visible octave was 1-2, which contained one half the singletons plus one half the dou- bletons (the other half of the singletons necessarily ignored in analyses). Richness was estimated by as- suming a truncated lognormal distribution for the ob- served data, estimating the parameters of the complete distribution, and calculating the area under the curve of that complete distribution. The lognormal distribu- tion is given by S(R) = S0e<-"2R2> where S(R) is the number of species in the Rth octave, 50 is the number of species in the modal octave, and a is a parameter related to the width of the distribution. R = 0 for the modal octave; 1, 2, 3, etc. for higher octaves; ? 1, ?2, ?3, etc. for lower octaves. The parameters of the lognormal were estimated us- ing a modified version of the method in Ludwig and Reynolds (1988). Octave numbers were assigned to each observed abundance class, and the parameters S0 and a were estimated using the Nonlin module of Systat 5.2, with Quasi-Newton estimation and least-squares fit. Octave numbers were assigned such that they max- imized the corrected r2 value from the model estima- tion. Traditionally, S1*, the estimate of total species richness, is the area under the entire estimated curve. However, this contains not only the species behind the veil line, but also that portion of the estimated curve that extends beyond the highest observed abundance class. Since all the very abundant species are being observed (there are no "hidden" species with abun- dances greater than the most abundant observed spe- cies), the estimated curve should be truncated at the octave with the highest observed abundance. We fol- lowed that procedure here. Projecting species accumulation curves.?This method requires a series of samples, so that a species accumulation curve can be observed as additional sam- ples are added to the pool. An asymptotic equation is then fit to this curve, and the estimate of the asymptote parameter is the estimate of species richness. The func- tion most often used, and the one used in this report, is the Michaelis-Menten equation (Clench 1979, Sob- eron and Llorente 1993, Colwell and Coddington 1994, Chazdon et al. 1998, Keating and Quinn 1998). Param- eter values were estimated using Raaijmakers' (1987) maximum likelihood estimators for the Eadie-Hofstee transformation, as outlined in Colwell and Coddington (1994). Nonparametric estimators.?Nonparametric meth- ods show promise for richness estimation (Bunge and Fitzpatrick 1993, Colwell and Coddington 1994). Chao and colleagues developed a set of nonparametric meth- ods for estimating the number of classes in a sampling universe (Chao 1984, 1987, Chao and Lee 1992, Chao et al. 1993, Lee and Chao 1994). These methods have now been evaluated for a number of biological data sets (e.g., Colwell and Coddington 1994, Fisher 1996, 1998, 1999a, b, Chazdon et al. 1998, McKamey 1999, Anderson and Ashe 2000). An estimator that shows considerable promise is the Incidence-based Coverage Estimator (ICE). Details of its calculation may be found in Chazdon et al. (1998) and R. K. Colwell's Estimates (Version 6) web site.6 We used ICE for nonparametric richness estimation in this paper. Two parameters that are components of nonpara- metric estimators and that are often examined as in- dicators of inventory completeness are the number of "uniques" and "duplicates." Uniques are species known from a single collection; duplicates are species known from only two. Sample-based rarefaction curves and associated es- timator curves.?A key attribute of a richness estimator is that it become independent of sample size above some minimum sample size. As additional species are discovered during a survey, the species accumulation curve can be viewed as an increasingly accurate as- sessment of community species richness. As additional samples are pooled and the observed species richness curve stabilizes beyond a particular number of samples, then observed species richness is deemed an adequate estimate of community species richness for that number of samples. Pielou (1966, 1975) first suggested as- sessing diversity index measures (although not species 6 URL: 692 JOHN T. LONGINO ET AL. Ecology, Vol. 83, No. 3 richness), at progressively larger numbers of pooled samples, by adding samples to the pool in random order from an existing sample set, then computing a diversity index measure for each level of sample pooling. Pie- lou's "pooled quadrat method" (Magurran 1988) was generalized by Lloyd et al. (1968), who suggested mul- tiple re-orderings to account for variation among sam- ple re-orderings, and by Colwell and Coddington (1994), who extended this approach to richness esti- mators. If the estimate of species richness is unstable or rises with sample size, it is not a reliable estimate of total richness. (A rising Chao estimator can none- theless be regarded as a valid estimator of minimum richness, given the available data [A. Chao, personal communication]). If the estimate stabilizes and is in- dependent of sample size, it is more likely to be a reliable estimate of total species richness. Gotelli and Colwell (2001) refer to this random resampling pro- cedure as "sample-based rarefaction," since it is anal- ogous to the individual-based rarefaction of Sanders (1968; see also Simberloff 1978), but it differs from the latter in retaining the pattern of aggregation of spe- cies within samples. To investigate the effect of number of samples on richness estimates using Michaelis-Menten and ICE, R. K. Colwell's program Estimates (Version 5) was used.7 This program calculates sample-based rarefac- tion curves and associated values for a variety of rich- ness estimators, presenting the mean of a user-desig- nated number of random re-orderings of the samples. Sample-based rarefaction plots were produced for each data set. For all calculations involving multiple random orderings of the samples, 50 randomizations were used. The ICE calculations were carried out using the Estimates default "cut-point" of 10, as suggested by Lee and Chao (1994). Estimates calculates Michaelis- Menten estimates in two ways: (1) for each of the 50 sample re-orderings, then averaged ("MMRuns"), or (2) once using the smoothed species accumulation curve ("MMMeans"). MMMeans is used in this report. To compare sample-based rarefaction curves, the units of the abscissa should, in general, be number of individuals when examining community species rich- ness, whereas the units should be number of samples when examining species density (Gotelli and Colwell 2001). Because this report is concerned with species richness, the abscissa of rarefaction curves should, in principle, be number of individuals. However, ants pre- sent a problem because of their sociality, resulting in an extreme spatial clumping of individuals within sam- ples. In addition, some of the data from Project ALAS are incidence data (presence/absence data). To circum- vent this problem, all data were converted to incidence data, and number of species occurrences (not number of samples) is used throughout as a measure of abun- 7 URL: dance and sampling effort. All rarefaction curves are plotted with number of occurrences on the abscissa. We use a log scale for the abscissa of most sample- based rarefaction plots. This technique reveals patterns in estimator curves that are not easily seen in arithmetic plots. Arithmetic plots make it difficult to distinguish asymptotic curves from logarithmic curves. In a semi- log plot the former will appear sigmoidal and the latter will approach linearity. Also, curves of number of uniques that appear flat on arithmetic plots may appear more distinctly humped on log plots. Data subsets.?We applied richness estimators to a variety of data partitions, ranging from small, uniform subsets to the entire data set. Richness estimators were applied to (1) each of the eight methodological subsets, considered separately; (2) within-month, among-site subsets of the Berlese samples; (3) the quantitative data set, which is the combined data set for all the methods using structured sampling (i.e., "Longino" and "Oth- er" excluded); and (4) the full data set, consisting of the samples from all eight methods. We used the quantitative and full data sets to evaluate richness estimates based on the lognormal method and to assess the prediction that the abundance distribution would emerge from behind a veil line one octave for each doubling of sample size. Samples were pooled to yield subsets with 12.5%, 25%, 50%, and 100% of the samples. Samples were randomly selected from within each of the eight methods so that proportional repre- sentation of methods did not change as sample size increased. For each subset we observed the abundance distribution and calculated a richness estimate based on the truncated lognormal distribution. We used all data partitions and sample-based rare- faction plots to evaluate ICE and Michaelis-Menten estimates. Examination of rarity.?Richness estimates are high- ly influenced by rare species. Often an attempt is made to partition rare species into those that are low density elements of local communities and those that somehow do not belong. The latter are often referred to as "tour- ists," reflecting their transitory, nonresident status (Moran and Southwood 1982, Stork 1988, Basset and Kitching 1991, Gaston et al. 1993, Gaston 1994, No- votny and Basset 2000). To better understand rarity in the La Selva ant fauna, the unique species (those found in only one sample) were partitioned according to ad- ditional knowledge of their natural history and distri- bution. First, "edge" species were identified: species known or surmised to be common but not easily sampled with the methods reported here. Edge species were further divided into "methodological" vs. "geographic" edge species. The former are assumed to be common at La Selva but not readily sampled by the methods em- ployed. In particular, this category consisted of spe- cialized inhabitants of the soil, a subsurface community that is not easily collected. Geographic edge species March 2002 ANT FAUNA OF A TROPICAL RAIN FOREST 693 TABLE 2. Summary statistics for different types of samples used in the inventory of the ants of La Selva Biological Station, Costa Rica. No. species No. local Method No. samples occurrences! No. species uniques:): No. uniques? Fogging 459 3262 165 21 2 Malaise 62 455 103 39 0 Berlese 217 775 117 33 1 Winkler 41 1416 197 50 11 Barger 40 275 65 26 3 Thompson 222 255 52 19 0 Longino 528 894 283 96 16 Other 335 573 235 11 1 18 Total 1904 7904 437 51 f Ants are social, and entire colonies may occur in single samples. To reduce the effect of this extreme spatial aggregation, only incidence data are used in this report. $ Number of species found in only one sample for a particular method, but not necessarily for the study as a whole (following terminology of Novotny and Basset 2000). ? Number of species found in only one sample in the entire study. are known to be abundant outside of La Selva yet oc- curred in only one sample at La Selva. Geographic edge species were further subdivided into those more com- mon at higher elevation, those more common in the tropical dry forest habitats on the Pacific side of Costa Rica, and "tramp" or synanthropic species that are common in open or highly disturbed habitats (most La Selva sampling occurred in mature or second growth forest). The rarity of the remaining unique species re- mained unexplained. Finally, "global uniques" were identified: those known only from the one sample re- ported here, and as yet from nowhere else on Earth. Berlese 117 Other 235 90 60 30 Longino 1 283 mm Octave FIG. 1. Abundance distributions of ants at La Selva Bi- ological Station, based on eight different sampling methods (see the Appendix). Octaves are standard Preston abundance classes, with boundaries 1, 2, 4, 8, etc. Species falling on a boundary are split between the two adjacent octaves (see Methods: Abundance distributions). Abundance of each spe- cies is the number of samples in which a species occurred. The number below the name of each method is the total num- ber of species collected by that method. Background information on ant species was obtained from existing taxonomic literature, and from J. T. Lon- gino's ongoing survey of Costa Rican ants (available online).8 Although no other site in Costa Rica has been surveyed as intensively as La Selva, Longino has car- ried out smaller scale collecting in many different re- gions of Costa Rica and has relied on the nationwide collection at INBio. RESULTS Workers of 437 species of ants were found at La Selva (Supplement). In the data set presented here, 7904 species occurrences are distributed among 1904 samples (Table 2). Eighteen additional species were collected as queens and/or males only (discussed below but not included in the quantitative analyses). Individual methods Octave-based relative abundance plots for most of the methodological subsets, plotted independently, failed to reveal a mode (Fig. 1). Canopy fogging clearly revealed a mode, however, and produced a convincing lognormal distribution. The Winkler method had a mode in the second octave, but the distribution was irregular and not convincingly lognormal. All others showed distributions typical of tropical insect surveys: the mode was in the first octave. Most species were rare, and the distribution would fit a log series distri- bution or the right tail of a lognormal distribution equally well. The quantitatively structured samples had lower, broader distributions than the two nonquanti- tative methods (Longino and Other), as might be ex- pected, given the tendency of nonquantitative collec- tors to ignore additional specimens of species already collected. Richness estimates based on the Michaelis-Menten equation and ICE usually did not stabilize (Fig. 2). The exception was the ICE estimate for Other, which sta- bilized at ~360 species. In general, final observed rich- 8 URL: 694 JOHN T. LONGINO ET AL. Ecology, Vol. 83, No. 3 100 50 Barger 1 _ # * ~x vXxx - ' x x* x ---^ ^o? ^o? _A- 10 03 300 g o a) o. CO ?5 150 .Q E 100 Winkler x . -o ,oOv !??????? AA A A A A A A 10 500 T 250 100 Other 80 40 0 1000 10 160 80 Thompson x x . - x - o xxx-xx- o ,x- -*x ^0 ^AAAAAAAAA 160 80 Berlese ?x- o X""' o* x o o < ^o ^fiH& AAAAAA 100 Malaise X X x _ " x*xx" c 1000 10 200 100 ^^AA?A ?AAAAAAAA 100 Fogging xx*|.a* * X o? o o o ;g^?AAA]A1AggBB 1000 1000 10 -xxx o o 500 250 A A A,A <>nnDD AAAA* fin 0 100 Longino 1000 10 100 1000 10000 x ICE - MM Mean o Sobs n Uniques A Duplicates 10 100 1000' 10 100 1000 Number of species occurrences FIG. 2. Sample-based rarefaction curves and corresponding estimators, based on eight different sampling methods. The abscissa is the number of species occurrences, on a logarithmic scale. Key definitions are: ICE, incidence-based coverage estimator; MMMean, asymptote of Michaelis-Menten curve estimated from the sample-based rarefaction curve; Sobs, observed species richness; Uniques, number of species each known from only one sample; Duplicates, number of species each known from exactly two samples. ness and richness estimates were still widely separated. For canopy fogging samples, both estimates closely converged on observed richness, and in the Winkler method, the Michaelis-Menten estimate closely con- verged on the observed richness. The uniques and du- plicates curves were generally rising as the maximum number of samples was approached. For canopy fog- ging, both curves peaked and declined. For the Longino samples, the uniques curve was beginning to decline (Fig. 2). Within-month Berlese samples Richness estimation using ICE and Michaelis-Men- ten can produce stable richness estimates for uniform subsets of the data. When the Berlese samples from 16 different sites were examined within months (?=16 samples within each month), the ICE estimates stabi- lized in six of the 13 mo, Michaelis-Menten in four (Fig. 3). But the behavior of ICE and Michaelis-Menten was not correlated: no months showed stabilization of both estimators. Quantitative data set The combined data set, using only the six methods based on structured sampling, produced a lognormal distribution (Fig. 4, left column of histograms). Ran- dom subsets representing successive doublings of ef- fort showed a lognormal distribution emerging from behind a veil line. For the first two doublings of sample effort, fitted lognormal curves had similar size and shape and appeared as one curve that was revealed one octave at a time. However, the best-fit curves for the third doubling, from 50% to 100%, did not differ in position. Instead the 100% curve differed in shape, being lower and broader than the 50% curve (Fig. 4, lower left histogram). Richness estimates based on the lognormal method were thus stable for 12.5%, 25%, and 50% of sampling effort, but at a value that was clearly an underestimate of the total species richness. The estimate based on 100% of sampling effort was much higher (Fig. 5), exceeding the observed total of 437 species. The Michaelis-Menten estimates were unstable but closely approached the total observed richness (Fig. 5). The ICE estimates produced a double plateau, one at ~275 species and 100-500 species occurrences and one at ~390 species and 4000-6500 species occur- rences. The observed species accumulation curve was slightly sigmoidal, ending at 364 species. The uniques curve was broadly hump-shaped and declining at high- er sampling intensity. The duplicates curve was flat. March 2002 150 100 50 0 150 100 50 0 CD .a 10 10 CD o 150 CD ?" 100 50 0* 10 150-1 100- 50 10 150 100 50 0 March ->&*&&? X a- s:^ ANT FAUNA OF A TROPICAL RAIN FOREST April 695 100 June x * O 0?' 100 September o o 100 December 0 ib ? ?<>? 100 March x x 8 oo 1000 1000 1000 1000 150 100- 50- 0 150! 100 50 0 150 100 50 0 1 150 100 50 0 o o0< 10 100 July 1000 x oo 10 100 October 1000 -xx- n ^?? , 0 100 January 1000 . x* 10 100 - MMMean x ICE ? Sobs 1000 150 100 50 0 150 100 50 0 150 100 50 0 150 100- 50- 0 May x 10 100 August xx*x^?* X - oo 1000 10 100 November 1000 -*-* - X OOP' 10 100 February 1000 . x^mii 10 100 1000 10 100 1000 Number of species occurrences FIG. 3. Sample-based rarefaction curves and corresponding estimators, based on Berlese samples across sites within months. Abscissa, MMMean, ICE, and Sobs are defined as in Fig. 2. Full data set The entire data set, for all methods combined, pro- duced a distribution even more clearly lognormal than for the quantitative data set; the fitted curves for the subsamples approximated one underlying distribution, revealed one octave at a time (Fig. 4, right column of histograms). However, richness estimates based on the lognormal method were not stable, reflecting subtle changes in distribution size and shape with increasing sampling intensity. The richness estimate was stable for the two lowest sampling intensities, but then rose with and closely approximated observed species rich- ness (Fig. 5). The Michaelis-Menten estimates behaved very dif- ferently than the quantitative data set. They rose along with observed species richness but formed two line segments with different slopes. The transition to shal- lower slope occurred at about 340 species and 400 species occurrences (Fig. 5). The ICE estimates also appeared to have two different "phases," somewhat offset from the Michaelis-Menten estimates, acceler- ating from a shallower to a steeper curve. Neither es- timator stabilized with sample size. However, the es- timates and the observed richness closely converged at the highest observed richness: 437, 453, 463 for ob- served richness, Michaelis-Menten, and ICE, respec- tively. The species accumulation curve was slightly sigmoidal. Both uniques and duplicates curves peaked and began to decline. 696 JOHN T. LONGINO ET AL. Ecology, Vol. 83, No. 3 80 Quantitative data set Full data set 0 80 12.5% 25% .9> 80 o CD Q. CO CD E 3 50% 80 100% Octave FIG. 4. Abundance distributions of ants at La Selva. The left column of histograms is for the combined data set, con- taining all data from quantitatively structured sampling; the right column is for the full data set, pooling data from all methods, both quantitative and nonquantitative. Rows rep- resent successive doublings of sampling effort, beginning with 12.5% of samples randomly drawn from the respective data set. In each column the bottom graph shows the super- imposed curves of the lognormal distributions estimated from the four histograms; the vertical line there is the veil line. When plotted on an arithmetic abscissa (the tradi- tional representation of species accumulation curves), the gradual nature of the approach to inventory com- pletion is revealed (Fig. 6). For the majority of the inventory (between 2000 and 8000 species occurrenc- es), the Michaelis-Menten and ICE estimates are sim- ilar and outperform the estimates based on the log- normal method. Rarity Of the 437 species known from workers, 51 (12%) were uniques. Rarity of 20 of them could be explained by edge effects. Six were methodological edge species, known or suspected to be common at La Selva but not easily sampled by the methods used. These were all ants that nest and forage in the soil but apparently do not forage into the leaf litter or higher. Most of them are eyeless or have greatly reduced eyes. Geographic edge species included three species that are common in synanthropic habitats around La Selva, four species that are abundant in the wet montane forests at higher elevations above La Selva, three species that are com- mon in the seasonal dry forests of Costa Rica's Pacific lowlands, two species that are generally common in Costa Rica but inexplicably rare at La Selva, one spe- cies that is near the southern limit of its geographic range and is much more common in countries to the north, and one species that is near the northern limit of its range and is more common in countries to the south. The remaining 31 uniques have no known or suspected area of high abundance. Among the 51 uniques, 45 were known from addi- tional collections outside of La Selva. Among the re- maining six global uniques, five were morphospecies from taxonomically difficult genera, and global unique- ness cannot be stated with certainty. Only one species was certainly a global unique. This can be claimed with confidence because it is in a genus with a recent tax- onomic revision. Sixteen species in the La Selva inventory are known from queens or males only. Two species of Cremato- gaster are known from queens only. These are species known from other sites and have been associated with workers, but they are always extremely rare, and the queen morphology suggests that they are temporary or facultative social parasites. At La Selva, the queen of one species was found in the nest of another Crema- togaster species, and an alate queen of the other species was found in a canopy fogging sample. Both these species make small carton nests, and there is no ex- planation why they have not been collected, other than low density. Two species of Cylindromyrmex are known from alate queens only. Cylindromyrmex axe known to be rare in general, and nests are probably in narrow beetle galleries in solid wood. None of the col- lecting methods used so far is efficient at sampling insects inside of solid wood. One species of Discoth- yrea is known from two alate queens obtained in can- opy fogging samples. The workers of this genus are extremely small. Workers of what may be the same species have been collected in Winkler samples from other Atlantic lowland sites, but not from La Selva. The remaining 11 species are all army ant (Eciton- inae) males. Army ants have a parallel taxonomy for workers and males. Apart from a few common species that forage on the surface, workers of most species are rarely encountered. They are largely nocturnal, and many appear to be largely or entirely subterranean. In contrast, the males are large and conspicuous, and they are common at blacklights and in Malaise traps. In many cases males have never been associated with March 2002 ANT FAUNA OF A TROPICAL RAIN FOREST 697 500 400 300 200 100- co CD O 03 % ? Quantitative data set vi? w *?? x x ,K< IS^m***^ J^ 10 xx* 100 1000 ? Lognormal x ICE - MMMean o Sobs a Uniques ^ Duplicates 10000 4? 10 100 1000 Number of species occurrences 10000 FIG. 5. Sample-based rarefaction curves and associated estimators for the ants of La Selva. The top figure is for the combined data set containing all data from quantitatively structured sampling; the bottom figure is for the full data set, pooling data from all methods, both quantitative and non-quantitative. Key definitions are: Lognormal, richness estimate based on the area of an estimated lognormal distribution; ICE, incidence-based coverage estimator; MMMean, asymptote of Michaelis-Menten curve estimated from the smoothed sample-based rarefaction curve; Sobs, observed species richness; Uniques, number of species each known from only one sample; Duplicates, number of species each known from exactly two samples. The abscissa is scaled logarithmically. FIG. 6. Sample-based rarefaction curve and associated estimators for the combined data set, with the abscissa scaled arithmetically. 500 400 co (D o CD Q. CO CD E 300 200 if 100 X o X o o ICE MMMean Lognormal Sobs Uniques Duplicates * 2000 3000 4000 5000 6000 Number of species occurrences 7000 8000 698 JOHN T. LONGINO ET AL. Ecology, Vol. 83, No. 3 workers, because it is extremely rare to find them to- gether. At La Selva, 11 species of army ant are known from males only. DISCUSSION The inventory The 1500-ha La Selva Biological Station contains at least 437 resident ant species (known from workers at the site). This is one of the most thorough inventories of an entire ant community ever carried out. The ob- served abundance distribution is a truncated lognormal, with three revealed abundance classes below the mode. Richness estimates using a variety of methods are at most 6% above the observed richness. Uniques com- prise only 12% of the observed species. Similarly thorough inventories for temperate and subtropical sites show the expected lower species rich- ness outside of the tropics. Talbot (1975) reported 87 species for the 518-ha E. S. George Reserve in Mich- igan, a result of 24 summers of field work. Van Pelt (1956) found 72 species (12.5% uniques) for the 882- ha Welaka Reserve in Florida. Deyrup and Trager (1986) found 102 species (9% uniques) for the 1771- ha Archbold Biological Station in Florida. Studies in the wet tropics have reported much higher species rich- ness. Bruhl et al. (1998) found 524 species in a 400- ha area at Poring Hot Springs in Kinabaloo National Park, Borneo. Verhaagh (1990, 1991) reported 520 spe- cies for a 1000-ha area in Panguana Reserve, Peru. These tropical studies do not provide relative abun- dances or other measures of inventory completeness, but they employed diverse sampling methods and in- volved a major collecting effort. The results for La Selva are of the same order of magnitude as these other tropical sites. Richness estimation Our results for individual collection methods reveal the inadequacy of single methods for insect inventories. Most show high proportions of rare species, species accumulation curves that do not show signs of ap- proaching a plateau, uniques and duplicates curves that are rising or flat, and richness estimates that rise steeply and remain well above observed species richness. The exceptions were canopy fogging, which showed a close convergence of estimated and observed species rich- ness, and Berlese samples for some months, which showed stable richness estimates. Even for combinations of methods, the three richness estimators evaluated here showed a variety of behav- iors when plotted together with sample-based rarefac- tion curves (Fig. 5), and none showed a stable plateau. After a region of highly fluctuating or declining rich- ness at very low sample sizes, they all increased along with observed species richness as sample size in- creased. However, some showed intervals of temporary stability, and all tended to converge with observed spe- cies richness at large sample size. What causes regions of temporary stability in richness estimates is unknown and would benefit from further study. They may rep- resent subcommunities that are efficiently sampled by the combination of methods used. If true, richness es- timators might best be thought of as linked to a par- ticular set of methods. Instead of assuming that our richness estimates are of the whole community at a site, we should expect them to be of the subcommunity available to the particular combination of methods we have used. Coverage-based estimators are relatively new in the field of richness estimation, and are in a trial period. Sample-based rarefaction plots with corresponding coverage estimators have now appeared in numerous publications. In nearly all cases the estimates have shown variable behavior, stabilizing for some data sets and failing to stabilize in others. Anderson and Ashe (2000) sampled montane leaf litter beetles, McKamey (1999) sampled Homoptera at a site in Africa, and Fish- er (1996, 1998, 1999a, b) sampled ant communities in Madagascar. All of these studies involved large sam- pling efforts, with diminishing returns revealed by spe- cies accumulation curves. In these studies most cov- erage-based estimators have steadily risen with sam- pling effort. Col well and Coddington (1994) empha- sized the untested nature of coverage-based estimators. Chazdon et al. (1998) evaluated a range of nonpara- metric estimators, using a tropical tree data set. They found ICE to be one of the best, but even ICE was inconsistent, stabilizing in some cases and not in others. These estimators are valuable additions to the ecolo- gist's toolbox for richness estimation, and they are see- ing increasing usage (e.g., Feener and Schupp 1998, Sayre et al. 2000), but continued evaluation of their performance is advised, they should not be used un- critically, and they should be viewed as yielding min- imum estimates of true richness. Some of the variation in the performance of these estimators perhaps can be explained by a consideration of what is being estimated. When we set out to measure species richness, we assume that the community we are sampling is like a jar of candy, from which we draw samples and attempt to estimate the number of kinds of candy. The hard fact is that there is no jar. We attempt to define a jar by the spatial, temporal, and methodo- logical boundaries of our sampling program, but we must accept the fact that the jar leaks. The leaks are from various kinds of edge effects, which can be spa- tial, temporal, or methodological. Methodological edge effects are perhaps the most serious problem in ar- thropod inventory. Species vary in probability of cap- ture in ways that are not necessarily related to their absolute abundance. This problem is less severe for studies of large or highly visible organisms such as trees and birds (e.g., Hubbell and Foster 1986, Terborgh et al. 1990, Clark et al. 1999, Pitman et al. 1999) but is particularly vexing for small organisms that must be March 2002 ANT FAUNA OF A TROPICAL RAIN FOREST 699 extracted from the environment using a variety of traps, extraction methods, and search techniques. It is unrealistic to expect species accumulation curves and richness estimates to stabilize completely, but combinations of methods and large enough sample sizes may sample a particular "jar" in ways that reduce the effects of edge species. If observed and estimated richness closely converge, this is evidence that a rel- atively thorough inventory of a spatially, temporally, and methodologically bounded community has been achieved. Abundance distributions A persistent question is what the abundance distri- bution of organisms would be if one could collect all individuals of all species in an area and count them. The abundance distribution reported here is lognormal. The lognormal distribution has been demonstrated nu- merous times and is often assumed (Preston 1948, May 1975, Sugihara 1980, Magurran 1988). However, Lambshead and Platt (1985) and Hughes (1986) argued that there was no a priori reason that species should have a lognormal distribution in nature. They claimed that reported lognormal distributions were mostly the result of combining multiple samples and thus could be explained by the central limit theorem alone (see also Preston [1981], Ugland and Gray [1982], and Wil- son [1993] for further critiques of the lognormal). The data set reported here is certainly the result of multiple, combined samples from disparate methods and no sin- gle sample in this inventory comes even close to re- vealing a nonsingleton mode. What the abundance dis- tribution would look like if one could record and iden- tify every ant colony in La Selva's 1500-ha remains unknowable. However, we argue that even though sam- pling artifacts certainly influence the shape of the ob- served distribution, this inventory still strongly sup- ports the presence of an approximately lognormal dis- tribution in nature. The upper part of the true abundance distribution (the right hand tail) must be similar to what was ob- served here, because it contains the most abundant spe- cies that are routinely encountered in samples. They are the most conspicuous part of the whole community, and their abundances can be more accurately assessed, much like trees or birds. The distribution of rare species is the more difficult problem. Are species with few captures really less abundant, or are there features of their nesting and foraging behavior that make them harder to capture? If one could capture all the ants at La Selva, would there still be uniques? Would any spe- cies be represented by one colony (or a lone worker foraging across the border from a nest outside of La Selva)? Or would all species be represented by multiple colonies? The specialized collecting by Longino makes it very unlikely that there is a large pool of rare, unseen ant species at La Selva. In fact, the combination of non- quantitative taxonomist collecting and quantitatively structured sampling may result in biases at opposite ends of the abundance distribution that tend to cancel each other out, resulting in a relatively accurate esti- mation of the true distribution. The specialized col- lecting of a taxonomist may reveal relative abundances of rare species, because a very broad spectrum of meth- ods is used, and every sample of every rare species is kept. However, after a few collections of a species are obtained, subsequent encounters are ignored. Thus common species are underrepresented. In contrast, quantitatively structured sampling does a good job of estimating relative abundances of common species, but underrepresents rare species due to the limited scope and number of methods. These compensating tenden- cies may explain why the full data set more closely approximates a lognormal than the quantitative data set. Rarity Rarity in a sample may be caused by methodological edge effects. These are cases in which species are abun- dant at the study site but are undersampled due to the inadequacy of the sampling methods. It is easy to rec- ognize edge species within one sampling technique when multiple techniques are used in an inventory (Longino and Colwell 1997). For example, specialized plant-ants that live only in myrmecophytes are very rarely obtained in canopy fogging samples, yet manual searching can make them as abundant as one might wish. In this inventory, the combination of methods cov- ered nearly every imaginable niche for ants in the leaf litter or above. The one portion of the habitat that was poorly sampled was the subsoil. Only one method, Ber- lese samples, extended into the soil, and the volume of material sampled was relatively small. A number of species were relatively abundant in the Berlese samples and nearly or entirely absent from all other collecting methods, including Winkler samples of litter lying di- rectly on top of the soil. Subterranean species that were even moderately abundant would appear extremely rare in this survey. Six of the uniques in this inventory were species either manually excavated from soil or obtained in a Berlese sample. They are part of the undersampled soil fauna and might, in fact, be relatively abundant. Rarity at a site may also be caused by mass effects (Shmida and Wilson 1985). Species may arrive at a site by dispersal from nearby source areas where they are more abundant. Since this survey was based entirely on workers, rarity cannot be caused by a "rain" of long-distance dispersers that immediately die. For each species, at least one colony has to be established at La Selva. However, source-sink population dynamics may nonetheless apply, with a continual influx of immigrant queens necessary to maintain a low density population. Fourteen of the uniques at La Selva are known to have high population densities elsewhere, ranging from syn- 700 JOHN T. LONGINO ET AL. Ecology, Vol. 83, No. 3 anthropic species with high density immediately ad- jacent to La Selva, to species at the limits of their ranges and with high densities hundreds or thousands of kilometers north or south. An unanswered question, and one relevant to conservation biology, is whether these low-density species rely on continual immigra- tion for population persistence or are capable of per- sisting at a low density, with immigration insignificant or non existent. Rabinowitz (1981) and Rabinowitz et al. (1986) ex- amined rarity at larger spatial scales for British plants. Similar analyses of rarity have been done for Ama- zonian birds (Terborgh et al. 1990), Amazonian trees (Pitman et al. 1999), and benthic marine organisms (Carney 1997). All have found that species rare in small-scale ecological samples often have broad ranges and are frequently known to be abundant elsewhere. Species rare in ecological samples are often not rare to museum taxonomists. For insect taxonomists, rare species are often methodological edge species. The findings reported here are consistent with these obser- vations. Only seven species out of 437 are currently known from only one collection in the world. For many of the La Selva ant species we can say little about their ranges outside of Costa Rica and even less about their abundances outside of La Selva. But it was nonetheless striking how many of the La Selva uniques were known from additional collections outside of La Selva. Rarity will continue to be an important research topic in ecology, because of its obvious relevance to con- servation biology. A challenging task is to understand patterns of abundance at larger spatial scales, extending to the entire geographic range of species. Finally, an important question to address is whether rare but wide- spread species are nonetheless relatively prone to ex- tinction as the habitat becomes increasingly fragmented throughout their range. ACKNOWLEDGMENTS First and foremost we thank the ALAS staff who made the project possible: Danilo Brenes, Nelci Oconitrillo, Maylin Paniagua, and Ronald Vargas. Nichole Barger and Chris Thompson provided two of the methodological data sets while carrying out undergraduate independent research projects. Barry Bolton, Stefan Cover, and Phil Ward assisted in iden- tification of Dacetini, Pheidole, and Pseudomyrmex, respec- tively. The Directors and staff of the La Selva Biological Station have been extremely helpful. Steve Yanoviak made helpful comments on the manuscript, and James Gutholm assisted with technical aspects of manuscript preparation. This work has been supported by National Science Founda- tion grants BSR-9025024, DEB-9401069, and DEB-9706976. LITERATURE CITED Agosti, D., J. D. Majer, L. E. Alonso, and T R. Schutz. 2000. Ants: standard methods for measuring and monitoring bio- diversity. Smithsonian Institution Press, Washington, D.C., USA. Andersen, A. N. 1992. The rainforest ant fauna of the north- ern Kimberley region of Western Australia (Hymenoptera: Formicidae). Journal of the Australian Entomological So- ciety 31:187-192. Andersen, A. N., and J. D. Majer. 1991. The structure and biogeography of rainforest ant communities in the Kim- berley region of northwestern Australia. Pages 333-346 in N. L. McKenzie, R. B. Johnston, and P. J. Kendrick, editors. Kimberley rainforests of Australia. Surrey Beattie and Sons, Sydney, Australia. Andersen, A. N., and H. Reichel. 1994. The ant (Hymenop- tera: Formicidae) fauna of Holmes Jungle, a rainforest patch in the seasonal tropics of Australia's Northern Ter- ritory. Journal of the Australian Entomological Society 33: 153-158. Anderson, R. S., and J. S. Ashe. 2000. Leaf litter inhabiting beetles as surrogates for establishing priorities for conser- vation of selected tropical montane cloud forests in Hon- duras, Central America (Coleoptera: Staphylinidae, Cur- culionidae). Biodiversity and Conservation 9:617?653. Armbrecht, I., and P. Ulloa-Chacon. 1999. Rareza y diver- sidad de hormigas en fragmentos de bosque seco Colom- bianos y sus matrices. Biotropica 31:646?653. Basset, Y., and R. I. Kitching. 1991. Species number, species abundance, and body length of arboreal arthropods asso- ciated with an Australian rainforest tree. Ecological En- tomology 16:391-402. Basset, Y, V Novotny, S. E. Miller, and N. D. Springate. 1998. Assessing the impacts of forest disturbance on trop- ical invertebrates: some comments. Journal of Applied Ecology 35:461-466. Belshaw, R., and B. Bolton. 1994. A survey of the leaf litter ant fauna in Ghana, West Africa (Hymenoptera: Formici- dae). Journal of Hymenoptera Research 3:5?16. Bisby, F. A. 1995. Characterization of biodiversity. Pages 21-106 in V H. Heywood, editor. Global biodiversity as- sessment. Cambridge University Press, Cambridge, UK. Briihl, C. A., G. Gunsalam, and K. E. Linsenmair. 1998. Strat- ification of ants (Hymenoptera: Formicidae) in a primary rain forest in Sabah, Borneo. Journal of Tropical Ecology 14:285-297. Bunge, J., and M. Fitzpatrick. 1993. Estimating the number of species: a review. Journal of the American Statistical Association 88:364-373. Carney, R. S. 1997. Basing conservation policies for the deep-sea floor on current-diversity concepts: a consider- ation of rarity. Biodiversity and Conservation 6:1463? 1485. Chao, A. 1984. Non-parametric estimation of the number of classes in a population. Scandinavian Journal of Statistics 11:265-270. Chao, A. 1987. Estimating the population size for capture- recapture data with unequal catchability. Biometrics 43: 783-791. Chao, A., and S. M. Lee. 1992. Estimating the number of classes via sample coverage. Journal of the American Sta- tistical Association 87:210-217. Chao, A., M. C. Ma, and M. C. K. Yang. 1993. Stopping rules and estimation for recapture debugging with unequal failure rates. Biometrika 80:193-201. Chazdon, R. L., R. K. Colwell, J. S. Denslow, and M. R. Guariguata. 1998. Statistical methods for estimating spe- cies richness of woody regeneration in primary and sec- ondary rain forests of northeastern Costa Rica. Pages 285- 309 in F Dallmeier and J. A. Comiskey, editors. Forest biodiversity research, monitoring and modeling: conceptual background and old world case studies. Parthenon Pub- lishing, Paris, France. Clark, D. B., M. W Palmer, and D. A. Clark. 1999. Edaphic factors and the landscape-scale distributions of tropical rain forest trees. Ecology 80:2662?2675. Clench, H. K. 1979. How to make regional lists of butterflies: some thoughts. Journal of the Lepidopterists Society 33: 216-231. Colwell, R. K., and J. A. Coddington. 1994. Estimating ter- March 2002 ANT FAUNA OF A TROPICAL RAIN FOREST 701 restrial biodiversity through extrapolation. Philosophical Transactions of the Royal Society of London, Series B 345: 101-118. Delabie, J. H. C, D. Agosti, and I. C. d. Nascimento. 2000. Litter ant communities of the Brazilian Atlantic rain forest region. Pages 1?17 in D. Agosti, J. Majer, L. Alonso, and T. Schultz, editors. Sampling ground-dwelling ants: case studies from the world's rain forests. Curtin University School of Environmental Biology Bulletin No. 18, Perth, Australia. Deyrup, M., and J. Trager. 1986. Ants of the Archbold Bi- ological Station, Highlands County, Florida (Hymenoptera: Formicidae). Florida Entomologist 69:206-228. Feener, D. H., Jr., and E. W. Schupp. 1998. Effect of treefall gaps on the patchiness and species richness of Neotropical ant assemblages. Oecologia 116:191-201. Fisher, B.L. 1996. Ant diversity patterns along an elevational gradient in the Reserve Naturelle Integrale d'Andringitra, Madagascar. Fieldiana Zoology (n.s.) 85:93?108. Fisher, B.L. 1998. Ant diversity patterns along an elevational gradient in the Reserve Speciale d'Anjanaharibe-Sud and on the western Masoala Penninsula, Madagascar. Fieldiana: Zoology 90:39-67. Fisher, B. L. 1999a. Ant diversity patterns along an eleva- tional gradient in the Reserve Naturelle Integrale d'Andohahela, Madagascar. Fieldiana: Zoology 94:129? 147. Fisher, B. L. 1999b. Improving inventory efficiency: a case study of leaf litter ant diversity in Madagascar. Ecological Applications 9:714-731. Fisher, B. L. 2000. Ant inventories along elevational gradi- ents in tropical wet forests in Eastern Madagascar. Pages 41-49 in D. Agosti, J. Majer, L. Alonso, and T. Schultz, editors. Sampling ground-dwelling ants: case studies from the world's rain forests. Curtin University School of En- vironmental Biology Bulletin No. 18, Perth, Australia. Folgarait, P. J. 1998. Ant biodiversity and its relationship to ecosystem functioning: a review. Biodiversity and Con- servation 7:1221-1244. Gadagkar, R., P. Nair, K. Chandrashekara, and D. M. Bhat. 2000. Ants species diversity in the Western Ghats, India. Pages 19?30 in D. Agosti, J. Majer, L. Alonso, and T. Schultz, editors. Sampling ground-dwelling ants: case stud- ies from the world's rain forests. Curtin University School of Environmental Biology Bulletin No. 18, Perth, Australia. Gaston, K. J. 1994. Rarity. Chapman and Hall, New York, New York, USA. Gaston, K. J. 1996. Species richness: measure and measure- ment. Pages 77?113 in K. J. Gaston, editor. Biodiversity: a biology of numbers and difference. Blackwell Science, Cambridge, UK. Gaston, K. J., T. M. Blackburn, P. M. Hammond, and N. E. Stork. 1993. Relationships between abundance and body size: where do tourists fit? Ecological Entomology 18:310? 314. Gotelli, N. J., and R. K. Colwell. 2001. Quantifying biodi- versity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters 4:379? 391. Hubbell, S. P., and R. B. Foster. 1986. Commonness and rarity in a neotropical forest: implications for tropical tree con- servation. Pages 205?231 in M. E. Soule, editor. Conser- vation biology, the science of scarcity and diversity. Sin- auer Associates, Sunderland, Massachusetts, USA. Hughes, R. G. 1986. Theories and models of species abun- dance. American Naturalist 128:879-899. Keating, K. A., and J. F Quinn. 1998. Estimating species richness: the Michaelis-Menten model revisited. Oikos 81: 411-416. Kim, K. C. 1993. Biodiversity, conservation, and inventory: why insects matter. Biodiversity and Conservation 2:191- 214. Kremen, C, R. K. Colwell, T L. Erwin, D. D. Murphy, R. F Noss, and M. A. Sanjayan. 1993. Terrestrial arthropod assemblages: their use in conservation planning. Conser- vation Biology 7:796-808. Lambshead, J., and H. M. Platt. 1985. Structural patterns of marine benthic assemblages and their relationships with empirical statistical models. Pages 371?380 in P. E. Gibbs, editor. Proceedings of the 19th European Marine Biology Symposium, Plymouth, UK, 1984. Cambridge University Press, Cambridge, UK. Lee, S. M., and A. Chao. 1994. Estimating population size via sample coverage for closed capture-recapture models. Biometrics 50:88-97. Lloyd, M., R. F Inger, and F W. King. 1968. On the diversity of reptile and amphibian species in a Bornean rain forest. American Naturalist 102:497-515. Lobo, J. M., and M. E. Favila. 1999. Different ways of con- structing octaves and their consequences on the prevalence of the bimodal species abundance distribution. Oikos 87: 321-326. Longino, J. T 1994. How to measure arthropod diversity in a tropical rainforest. Biology International 28:3?13. Longino, J. T, and R. K. Colwell. 1997. Biodiversity as- sessment using structured inventory: capturing the ant fau- na of a lowland tropical rainforest. Ecological Applications 7:1263-1277. Ludwig, J. A., and J. F Reynolds. 1988. Statistical ecology: a primer on methods and computing. John Wiley and Sons, New York, New York, USA. Magurran, A. E. 1988. Ecological diversity and its measure- ment. Princeton University Press, Princeton, New Jersey, USA. May, R. M. 1975. Patterns of species abundance and diver- sity. Pages 81-120 in M. L. Cody and J. M. Diamond, editors. Ecology and evolution of communities. Belknap Press of Harvard University, Cambridge, Massachusetts, USA. McDade, L. A., K. S. Bawa, H. A. Hespenheide, and G. S. Hartshorn, editors. 1993. La Selva, ecology and natural history of a neotropical rainforest. University of Chicago Press, Chicago, Illinois, USA. McKamey, S. H. 1999. Biodiversity of tropical Homoptera, with the first data from Africa. American Entomologist 45: 213-222. Miller, J. C. 1993. Insect natural history, multi-species in- teractions and biodiversity in ecosystems. Biodiversity and Conservation 2:233?241. Moran, V. C, and T. R. E. Southwood. 1982. The guild com- position of arthropod communities in trees. Journal of An- imal Ecology 51:289-306. Novotny, V., and Y Basset. 2000. Rare species in commu- nities of tropical insect herbivores: pondering the mystery of singletons. Oikos 89:564-572. Olson, D. M. 1991. A comparison of the efficacy of litter sifting and pitfall traps for sampling leaf litter ants (Hy- menoptera: Formicidae) in a tropical wet forest. Biotropica 23:166-172. Palmer, M. W. 1995. How should one count species? Natural Areas Journal 15:124-135. Peterson, A. T, and N. A. Slade. 1998. Extrapolating inven- tory results into biodiversity estimates and the importance of stopping rules. Diversity and Distributions 4:95-105. Pielou, E. C. 1966. Species diversity and pattern diversity in the study of ecological succession. Journal of Theoretical Biology 10:370-383. Pielou, E. C. 1975. Ecological diversity. John Wiley and Sons, New York, New York, USA. Pitman, N. C. A., J. Terborgh, M. R. Silman, and P. Nunez 702 JOHN T. LONGINO ET AL. Ecology, Vol. 83, No. 3 V. 1999. Tree species distributions in an upper Amazonian forest. Ecology 80:2651-2661. Preston, F. W. 1948. The commonness, and rarity, of species. Ecology 29:254-283. Preston, F. W. 1981. Pseudo-lognormal distributions. Ecology 62:355-364. Raaijmakers, J. G. W. 1987. Statistical analysis of the Mi- chaelis-Menten equation. Biometrics 43:793-803. Rabinowitz, D. 1981. Seven forms of rarity. Pages 205?217 in H. Synge, editor. The biological aspects of rare plant conservation. Wiley, New York, New York, USA. Rabinowitz, D., S. Cairns, and T Dillon. 1986. Seven forms of rarity and their frequency in the flora of the British Isles. Pages 182?204 in M. E. Soule, editor. Conservation biol- ogy, the science of scarcity and diversity. Sinauer Asso- ciates, Sunderland, Massachusetts, USA. Rosenzweig, M. 1995. Species diversity in space and time. Cambridge University Press, Cambridge, UK. Roth, D. S., I. Perfecto, and B. Rathcke. 1994. The effects of management systems on ground-foraging ant diversity in Costa Rica. Ecological Applications 4:423?436. Samways, M. J. 1993. Insects in biodiversity conservation: some perspectives and directives. Biodiversity and Con- servation 2:258?282. Sanders, H. 1968. Marine benthic diversity: a comparative study. American Naturalist 102:243-282. Sayre, R., E. Roca, G Sedaghatkish, B. Young, S. Keel, R. Roca, and S. Sheppard, editors. 2000. Nature in focus: rapid ecological assessment. Island Press, Washington, D.C., USA. Shmida, A., and M. V. Wilson. 1985. Biological determinants of species diversity. Journal of Biogeography 12:1-20. Simberloff, D. 1978. Use of rarefaction and related methods in ecology. Pages 150-165 in K. L. Dickson, J. Cairns, Jr., and R. J. Livingston, editors. Biological data in water pol- lution assessment: quantitative and statistical analyses. American Society for Testing and Materials, Philadelphia, Pennsylvania, USA. Soberon, M. J., and J. Llorente B. 1993. The use of species accumulation functions for the prediction of species rich- ness. Conservation Biology 7:480?488. Stork, N. E. 1988. Insect diversity: facts, fiction and spec- ulation. Biological Journal of the Linnean Society 35:321? 337. Sugihara, G. 1980. Minimal community structure: an expla- nation of species abundance patterns. American Naturalist 116:770-787. Talbot, M. 1975. A list of the ants (Hymenoptera: Formici- dae) of the Edwin S. George Reserve, Livingston County, Michigan. Great Lakes Entomologist 8:245?246. Terborgh, J., S. K. Robinson, T. A. Parker, C. A. Munn, and N. Pierpont. 1990. Structure and organization of an Am- azonian forest bird community. Ecological Monographs 60: 213-238. Ugland, K. I., and J. S. Gray. 1982. Lognormal distributions and the concept of community equilibrium. Oikos 39: Wi- Van Pelt, A. F. 1956. The ecology of the ants of the Welaka Reserve, Florida (Hymenoptera: Formicidae). American Midland Naturalist 56:358-387. Verhaagh, M. 1990. The Formicidae of the rain forest in Panguana, Peru: the most diverse local ant fauna ever re- corded. Pages 217-218 in G. K. Veeresh, B. Mallick, and C. A. Viraktamath, editors. Social insects and the environ- ment. Oxford and IBH Publishing, New Delhi. Verhaagh, M. 1991. Clearing a tropical rain forest?effects on the ant fauna. Pages 59?68 in W. Erdelen, N. Ishwaran, and P. Miiller, editors. Proceedings of the International and Interdisciplinary Symposium on Tropical Ecosystems. Margraf Scientific, Weikersheim, Germany. Williams, C. B. 1964. Patterns in the balance of nature and related problems in quantitative ecology. Academic Press, London, UK. Wilson, J. B. 1993. Would we recognize a Broken-Stick com- munity if we found one? Oikos 67:181-183. APPENDIX Descriptions of methods used in the inventory of the ants of La Selva Biological Station, Costa Rica, are available in ESAs Electronic Data Archive: Ecological Archives E083-011-A1. SUPPLEMENT An inventory of species occurrences of ants at La Selva Biological Station in both viewable and downloadable form is available in ESAs Electronic Data Archive: Ecological Archives E083-011-S1.