Genetic rescue of remnant tropical treesby an alien pollinatorChristopher W. Dick1,2{1Department of Organismic and Evolutionary Biology, Harvard University, 26 Oxford Street, Cambridge, MA 02138-2902, USA2Biological Dynamics of Forest Fragments Project, Instituto Nacional de Pesquisas da Amazo? nia, CP 478, Manaus, AM 69011- 970,Brazil Habitat fragmentation is thought to lower the viability of tropical trees by disrupting their mutualismswith native pollinators. However, in this study, Dinizia excelsa (Fabaceae), a canopy-emergent tree, wasfound to thrive in Amazonian pastures and forest fragments even in the absence of native pollinators.Canopy observations indicated that African honeybees (Apis mellifera scutellata) were the predominant?oral visitors in fragmented habitats and replaced native insects in isolated pasture trees. Trees in habitatfragments produced, on average, over three times as many seeds as trees in continuous forest, and micro-satellite assays of seed arrays showed that genetic diversity was maintained across habitats. A paternityanalysis further revealed gene ?ow over as much as 3.2 km of pasture, the most distant pollinationprecisely recorded for any plant species. Usually considered only as dangerous exotics, African honeybeeshave become important pollinators in degraded tropical forests, and may alter the genetic structure ofremnant populations through frequent long-distance gene ?ow.Keywords: Apismellifera; Amazonia; habitat fragmentation ; microsatellite;Dinizia excelsa; paternity inference
1. INTRODUCTIONLowland tropical rainforests are renowned for theirspecies richness and ecological complexity. Woody plants,the main structural elements of tropical forests, occur inassemblages of 280 species in a single hectare (Phillips etal. 1994; De Oliveira & Mori 1999) and upwards of 1000species in 25 ha of pristine forest (Bermingham & Dick2001). Concomitant with high alpha diversity, tropicaltrees occur in low population densities, typically less thanone individual per hectare (Hubbell & Foster 1983).Despite low population densities, self-fertilization is rarein natural settings. In most tropical trees, outcrossing isenforced either through self-incompatibility (Bawa et al.1985a), dioecy or high genetic loads (Alvarez-Buylla et al.1996). Animals, rather than wind, are the presumedpollen vectors for over 99% of lowland rainforest species(Bawa et al. 1985b). Animals also disperse the fruits andseeds of most tropical plants.Human encroachment of tropical forests throughlogging and agriculture threatens to fragment the world?sremaining rainforests (Laurance 1998). Forest fragment-ation has two immediate e?ects on the breeding structureof rainforest plants. First, it increases the distancesbetween potential mates. Second, it can alter the speciescomposition, relative abundance and foraging patterns ofpollinators. The disruption of pollinators can result inreduced fecundity of the host plants due to pollen limit-ation (e.g. Levin 1995; Ghazoul et al. 1998) and geneticerosion through drift (Young et al. 1996) if plants in forestpatches become reproductively isolated.Large trees left standing in tropical pastures representan extreme case of ecological deprivation. They may beisolated from forest-dwelling pollinators and seed disper-sers, or surrounded by habitat that is unsuitable for seed-
ling establishment. Pasture trees have been described asthe `living dead? ( Janzen 1986), persisting by virtue oftheir longevity but contributing little to forest regenera-tion and therefore meriting minimal conservation atten-tion. Nevertheless, several genetic studies have recordedpollen ?ow to pasture trees separated by hundreds ofmetres (Chase et al. 1996; Aldrich & Hamrick 1998),indicating that spatially isolated trees may not necessarilybe reproductively isolated.I examined the viability of remnant populations ofDinizia excelsa (Fabaceae), a prominent Amazonian tree,in light of the `living dead? hypothesis. D. excelsa is amodel species for studies of pollinator disruption becauseits generalist bee-pollination system is shared by ca. 30%of lowland rainforest species (Bawa et al. 1985b). Datawere obtained on the pollination and reproductive perfor-mance of trees in pasture, forest fragments and pristinehabitats. Microsatellite markers were then used to quan-tify outcrossing and gene ?ow and to map precisely thebreeding structure of trees in a fragmented population.
2. METHODS(a) Study species and siteD. excelsa is one of the largest Amazonian trees, attaining60 m in height and 2m in girth (Ducke 1922); it is one of theregion?s most important timber species (Barbosa 1990). Thespecies is endemic to Brazil, and occurs at natural densities ofaround one adult tree per 6 ha (Dick 2001). In the process ofclearing forest to create pasture, D. excelsa are often left standingbecause of their value for timber and shade. The pale yellow-green ?owers are small (calyx, 1^1.5 mm), hermaphroditic andborne on racemes. Small insects are attracted to their nectarand fragrance. Dormant seeds (ca. 1 cm) are borne in indehis-cent pods and dispersed primarily by wind and gravity. Thetrees in this study were located in the reserve system of theBiological Dynamics of Forest Fragments Project (Lovejoy &Bierregaard 1990), located ca. 90 km north of Manaus, Brazil
Proc. R. Soc. Lond. B (2001) 268, 2391^2396 2391 ? 2001 The Royal SocietyReceived 2 March 2001 Accepted 13 June 2001
doi 10.1098/rspb.2001.1781
{Present address: Smithsonian Tropical Research Institute, Unit 0948,APOAA 34002, USA (dickc@naos.si.edu).
(?gure 1). Trees were mapped in three habitat types: 15^20-year-old pasture of Colosso ranch, forest fragments in Colosso, andadjacent continuous forest west (Cabo Frio) and east (Km 41) ofColosso. Field studies were performed in 1995, 1996 and 1999.(b) Canopy studiesRope-climbing techniques (Laman 1994; Smith & Padgett1996) were used to reach the canopy, at an average height of
25 m (?gure 2). Flower-visiting insects were observed andcollected on seven large densely ?owering trees to assess changesin pollination. In addition, all mapped trees were observed fromthe ground with ?10 binoculars during the course of phenolo-gical observations. The climbed trees were located in thefollowing habitats: an isolated 10 ha reserve in Colosso ranch(n ? 2), isolated 100 ha reserves in Porto Alegre and Dimona(n ? 2), Colosso pasture (n ? 1) and an undisturbed forest at siteKm 41 (n ? 2). The intensively studied tree `Col.06? was sepa-rated from its nearest neighbour by over 600m of open pasture(?gure 4). Continuous-forest trees (`41.01? and `41.13?) were situ-ated 700 m apart.Over 52 h were spent (on 17 visit days) in canopy observationduring the mornings (06.00^12.00), when insect visitation wasat its peak. During the afternoon and evening periods, 10 h werespent in observation (on 4 days). Observations were dividedbetween trees in pristine forest (27 h) and trees in disturbedhabitats (26.5 h). The putative pollinators were captured withnets, mounted and identi?ed by specialists (Dick 2001).(c) Reproductive performanceFecundity and regeneration were used as indicators of repro-ductive performance. Fecundity was calculated as the totalnumber of pods and seeds produced per tree. The pods werecounted from the ground with binoculars in Cabo Frio andKm 41 and in the pastures, gallery forests and 10 ha reserve atColosso. Juvenile trees were excluded from the analyses. Colossopasture and fragment populations were analysed separately, and
2392 C.W. Dick Genetic rescue of remnant tropical trees
Proc. R. Soc. Lond. B (2001)
 
 
Figure 1. Reserves of the Biological Dynamics of Forest Fragments Project (28300 S, 608W). Dinizia excelsa ( 40 cm diameter atbreast height) were mapped in pasture (shaded areas) and forest fragments (black areas) in Dimona, Porto Alegre and Colossoranches, and in undisturbed forest at Cabo Frio, Km 41 (unshaded) and Reserva Ducke (not shown, located ca. 70 km southeastof the Biological Dynamics of Forest Fragments Project reserves).
Figure 2. The author climbing pasture tree `Col.06? forcanopy observations (see ? 2b and ?gure 4).
Table 1. Microsatellite loci used in this study.(The repeat motif is of the cloned sequence. The allelenumbers shown are from 240 seed and tree genotypes fromColosso ranch.)locus repeat sequence number of allelesDE27 (AAG)8 4DE37 (AC)20 12DE44 (GT)13 7DE48 (GA)27 34DE54 (CT)39 23
then combined for comparison with undisturbed forest. Seedswere counted from 2177 fallen pods from trees in Colosso (n ? 25trees), Km 41 (n ? 45) and Cabo Frio (n ? 23) to obtain esti-mates of mean seed set per pod. Seedling regeneration wassurveyed and mapped in secondary vegetation throughoutColosso ranch.(d) Genetic analysesLeaf tissue was collected from all of the mapped adult treesin Colosso (n ? 36). DNA was extracted from adult leaves andfrom 11 seed families (25^55 seeds per family) collected belowmaternal trees located in the di?erent habitat types. Genotypeswere obtained for ?ve microsatellite loci using primers devel-oped for D. excelsa (Dick & Hamilton 1999). A total of 80 allelesin the seed and adult populations provided an average exclusionprobability of greater than 99% (table 1). The outcrossing rate(t) for each maternal tree was calculated as the proportion of itsseed array that contained non-maternal alleles. The minimumnumber of sires for each seed family was calculated by dividingthe number of paternal alleles for the most variable locus bytwo, since the species is diploid. Seed paternity was assigned onthe basis of multilocus segregation probabilities using theprogram CERVUS (Marshall et al. 1998).
3. RESULTS(a) Pollinator shiftsOnly native insects were observed on the densely ?ow-ering trees in continuous forest. These included stinglessbees (tribe Meliponini), predatory wasps and small beetles(2^5 mm in length) (Dick 2001). Stingless bees (14 species,12 genera) were judged to be the primary outcross pollina-tors of these trees, as they were the only insects observedforaging among branches and ?ying away from the trees.More than 10 species of small beetles (from eight families)were observed in the ?oral cups of D. excelsa. These beetleswere not observed moving beyond individual racemes, butmay be a vector for self-pollination.Although the species composition and the abundanceof native insects were similar in continuous forest, 10 haand 100 ha forest fragments, African honeybees occurredonly in disturbed habitats, where they greatly outnum-bered all native insects (Dick 2001). Honeybees werevirtually the only pollinating insects on the pasture treeCol.06, which was devoid of beetles and visited rarely by
native bees. African honeybees outnumbered the nativebees on the pasture tree by thousands. No behaviouralinteractions were observed between African honeybeesand native insects on any of the trees. The co-occurrenceof native and exotic bees on ?owers in the forest isolatessuggests that African bees did not displace native insectsfrom pasture trees. Rather, the long distance of openpasture probably inhibits foraging by native bees, whichtend to nest in standing forest (Roubik 1989; Aizen &Feinsinger 1994).(b) Reproductive performanceThe trees in pasture and forest fragments (n ? 26)produced more than three times as many pods per tree asdid trees from the adjacent continuous-forest populations(n ? 20 and n ? 45; Welsch test, p5 0.01; ?gure 3). Nosigni?cant di?erence in fecundity was observed betweentrees in pasture (n ? 10) and trees in forest fragments(n ? 16) (p4 0.05). Senescent pasture trees increased thevariance of the Colosso population. The mean seed set ofca. three seeds per pod across habitats indicates that podsare an accurate proxy for seed set. The mean seed set atColosso was ca. 10 000 seeds per tree, with the mostisolated pasture tree, Col.06, at the mean.Although D. excelsa regeneration was rarely observed inundisturbed forest, seedlings were abundant in aban-doned pastures near remnant trees, with stem diameters
Genetic rescue of remnant tropical trees C.W. Dick 2393
Proc. R. Soc. Lond. B (2001)
 
Figure 3. Mean pod set of remnant trees in Colosso ranch andadjacent forest (Cabo Frio and Km 41).
3634
North
10
1316
1 km
31 8 7
forest
6
3
pasture
54
118213022232829 26
Figure 4. Map of Colosso ranch showing the mating patternsquanti?ed in table 3. Pasture and secondary vegetation isshown in white; the shaded areas represent primary forest.The 10 ha fragment harbours trees 18^26 and 30. Diniziaexcelsa seedlings were mapped near trees 03, 06, 07, 08, 11, 13,27 and 29.
ranging from 1mm to over 6.5 cm and with heights of upto 8m. Over 30 seedlings were mapped in 2 year oldsecondary vegetation between the 1ha and 10 ha frag-ments in Colosso. Over 40 saplings were mapped on theforest edges near Col.13, Col.18 and Col.16. Dozens ofseedlings were found near Col.03, Col.06, Col.07 andCol.08. These observations indicate that demographicgrowth is possible in the disturbed habitats.(c) Genetic resultsOutcrossing rates were high (ca. 95%) and did not di?erbetween trees in forest fragments and trees in continuousforest (p 4 0.05, n ? 7). However, pasture trees experi-enced 10% higher rates of self-fertilization (t-test, p 50.01,n ? 6; table 2). Increased sel?ng has been observed forother canopy species in pasture (Aldrich & Hamrick 1998)and selectively logged forest (Murawski et al. 1994).Gene ?ow within the fragmented population was exten-sive. Paternity inference of seed genotypes from sevenmaternal trees in Colosso yieldedpaternity assignments withhigh exclusion probabilities for 77 out of 240 seeds (table 3),indicating that about two-thirds of the seeds were sired byunmapped trees in the surrounding forest. Out of the 77inferred pollinations, 22 resulted from self-fertilization. Outof the 55 cross-fertilized seeds, 26 resulted from repro-duction between trees separated by 1km or more of pasture(range, 128^3200m) (table 3 and ?gure 4). The mean polli-nation distance for the Colosso pasture and gallery foresttrees, 1288m (n ? 45 seeds), was much greater than themean distance to nearest neighbour (DNN) of 235m(p5 0.001, n ? 18 trees). The mean pollination distance inthe 10 ha fragment, 417m (n ?10 seeds), also greatlyexceeded the mean DNN of 50.5m (n ? 12 trees). Phenolo-gical observations con?rmed that nearest neighbours ?ow-ered in synchrony and, thus, were potentialmates.The high genetic diversity of seeds was maintainedacross habitats, and even isolated trees sampled most of
2394 C.W. Dick Genetic rescue of remnant tropical trees
Proc. R. Soc. Lond. B (2001)
Table 2. Genetic diversity of seed arrays.( t, multilocus outcrossing rate (see ? 2d). The numbers below each locus indicate the number of unique paternal alleles in theseed array, which are summed across loci under `total alleles?. The bottom row describes the allelic diversity of Colosso adultgenotypes. Asterisks indicate samples of fewer than 10 seeds. Seed arrays are from the 1995 ?owering, except for Col.06`93? andCol.07`93? from 1993. Canopy observations were made on the trees indicated in italic.)
maternal number locus total minimumnumber oftree habitat of seeds t DE27 DE37 DE44 DE48 DE54 alleles siresCol.06 `95? pasture 35 0.84 1 5 3 7 11 27 6Col.06 `93? pasture 20 0.80 1 3 4 3 5 16 3Col.07`95? pasture 25 0.80 1 4 1 5 5 16 3Col.07`93? pasture 25 0.87 1 4 3 9 8 25 5Col.08 pasture 25 0.96 1 2 3 6 4* 16 3Col.29 pasture 25 0.80 0 4 3 7 1* 15 4Col.13 gallery 25 0.92 3 4 3 12 11 33 6Col.18 10 ha 25 1.00 2 4 1 9 8 24 5Col.26 10 ha 25 0.96 2 5 4 14 12 37 7Duk.10 forest 28 0.91 0 3 3 8 11 25 6Duk.32 forest 25 0.96 1 4 3 12 8 28 6Km 41.13 forest 25 0.90 1 4 3 8 9 25 5CF.26 forest 25 1.00 1 4 0 7 5 17 4Colosso adults (n ? 36) 3 6 6 21 18 54 n/a
Table 3. Paternity assignments (see also ?gure 4).(Maternal tree, single non-excluded male (`sire?), number ofidentical parentage events (`seeds?), distance between parentsand the exclusion probability given the maternal and seedgenotypes. In the case of more than one event, the range ofexclusion probabilities is shown. Exclusion probabilitiesgreater than 0.995 are rounded to 1.0. Canopy observationswere performed on the mother trees indicated in italic.)
maternaltree sire seeds distancebetweenparents (m) exclusionprobabilityCol.06 Col.07 1 1053 0.98Col.06 Col.08 1 1181 1.0Col.06 Col.04 1 1600 0.99Col.06 Col.26 2 1500 0.98^1.0Col.06 Col.29 1 1600 0.98Col.07 Col.08 7 128 0.99^1.0Col.07 Col.31 6 400 0.97^0.99Col.07 Col.06 1 1053 0.99Col.07 Col.21 1 2800 1.0Col.07 Col.03 5 3200 0.92^1.0Col.08 Col.07 4 128 0.98^0.99Col.08 Col.22 1 2900 1.0Col.08 Col.01 3 3000 0.98^0.99Col.10 Col.03 2 400 1.0Col.18 Col.29 2 500 1.0Col.18 Col.22 2 250 1.0Col.18 Col.04 1 300 1.0Col.18 Col.30 1 270 1.0Col.18 Col.01 1 400 0.99Col.26 Col.34 1 1100 0.95Col.26 Col.28 1 300 1.0Col.26 Col.36 1 1100 1.0Col.29 Col.03 1 1000 1.0Col.29 Col.21 1 500 0.88Col.29 Col.31 4 2200 0.97^1.0Col.29 Col.36 2 1200 0.92^0.97Col.29 Col.23 1 450 0.97
the local adult alleles in their seed arrays (table 2). Forexample, the `1995? seeds from pasture tree Col.06(n ? 35) had at least six sires. This tree was separatedfrom its two nearest neighbours by 600m and 1300m. Incontrast, forest tree Km 41.13, with at least ?ve sires forits 25 sampled seeds, was surrounded by 18 potentialmates within a 600m radius. The seed array from Col.13(n ? 25) contained 11 out of 18 alleles found in the adultpopulation for locus DE48.
4. DISCUSSIONHoneybees have a long and global association withhumans. However, their appearance in the neotropics isrelatively recent. African honeybees (A. mellifera scutellata)escaped from research apiaries in southern Brazil in 1956and spread rapidly, hybridizing with European honeybeeswhile retaining behavioural and morphological traits ofthe African race. There were no Apis colonies in theAmazon basin prior to the invasion of African honeybeesin the 1970s (Roubik 1989). Now there are an estimated50^100 million African honeybee colonies throughout theneotropics (Winston 1992). Globally, feral honeybees visitroughly one-third of the plant species in local ?oras, buttheir e?ect on the genetic structure of native plants islargely unknown (Butz-Huryn 1997). Honeybees areexpected to follow an optimal foraging strategy ; in rare?ights between densely ?owering trees they are expectedto visit only nearest neighbours (Butz-Huryn 1997; T.Seeley, personal communication).The genetic data presented here suggest that inter-treeforaging is not necessarily rare or limited to nearest ?ow-ering neighbours. For example, pasture tree Col.06 wasvisited almost exclusively by honeybees and set ca. 8000outcrossed seeds. The nearest ?owering neighbours,located 600 m (Col.13) and 1000m (Col.10) away, didnot fertilize any of the seeds assayed (n ? 55), althoughCol.06 did receive pollen from a stand of densely ?ow-ering trees in the 10 ha fragment (Col.26) (?gure 4).Similar patterns of pollen ?ow have been reported amongpasture trees in Costa Rica (Chase et al. 1996), althoughthe insect pollinators in that study were not known. Thegene ?ow between Col.07 and Col.03 represents the mostdistant pollination precisely measured in any plantspecies (Chase et al. 1996).African honeybees may increase the neighbourhoodarea of fragmented D. excelsa above natural levels, sincethey have a vast foraging area (212 km2) compared withnative stingless bees (12.5 km2) (Roubik 1989). Moreover,the large number of unassigned fathers in the Colossoprogeny indicates that pollinator movements were notcon?ned to the disturbed habitats. This result has prac-tical conservation implications, since one guideline forreforestation is to avoid using seeds from isolated treesbecause of their presumed low genetic quality.African honeybees can be considered to be at least apartial cause of the increased fecundity of trees in thedegraded habitat. Fecundity is a?ected by a host ofphysical and biotic factors, such as competition for lightand soil nutrients. Since the emergent canopy of D. excelsais fully exposed to light in all habitats, the most importantphysical factor should be nutrient release. However, therewas no signi?cant di?erence in fecundity between popula-
tions in the 10 ha forest fragment and those in openpasture. African honeybees, which were generallyabundant in disturbed sites, have been shown to increaseseed set above natural levels in other tropical plant species(Aizen & Feinsinger 1994; Roubik 2001). Moreover,increased seed production may contribute to the demo-graphic growth of D. excelsa in degraded habitats, since itdoes regenerate in pasture.In conjunction with their new pollinators, isolated treesperform two conservation roles. First, they act asstepping-stones for gene ?ow between fragmented popu-lations, and thereby help to preserve genetic diversitywithin the species. This is consistent with genetic studiesof isolated tropical trees with unknown pollinators (Chaseet al. 1996; Nason & Hamrick 1997; White et al. 1999).Second, because of their high output of geneticallydiverse seeds, isolated trees serve as foci for forestregeneration in abandoned pasture. In this dynamicagricultural landscape, an alien pollinator has helped tobring D. excelsa back from the ranks of the living dead.I thank Peter Ashton, Elizabeth Kellogg, Richard Lewontin,Peter Stevens and Stephen Palumbi for advising on this PhDresearch; Matt Hamilton, DanWeinreich and other members ofthe Lewontin, Palumbi and Fleischer laboratories for technicalassistance; Douglas Yu, Chris Jiggins and Eldredge Berminghamand David Roubik for helpful comments on the manuscript.Claude Gascon and Heraldo Vasconcelos for facilitating myresearch in Brazil. Anto? nio Ribeiro Mello, Romeu Moura Car-doso, Sebastia? o Salvino de Souza and Alae?rcio Marajo? dos Reisfor their dedicated assistance in the ?eld.This work was funded bya National Science Foundation pre-doctoral fellowship, theArnold Arboretum, National Geographic Society, Sigma Xi, theBiological Dynamics of Forest Fragments Project, the HarvardCommittee on Latin American Studies and the Department ofOrganismic and Evolutionary Biology at Harvard.
REFERENCESAizen, M. A. & Feinsinger, P. 1994 Forest fragmentation, polli-nation, and plant reproduction in a Chaco dry forest,Argentina. Ecology 75, 330^351.Aldrich, P. & Hamrick, J. L. 1998 Reproductive dominance ofpasture trees in a fragmented tropical forest mosaic. Science281, 103^105.Alvarez-Buylla, E. R., Garc|?a-Barrios, R., Lara-Moreno, C. &Mart|?nez-Ramos, M. 1996 Demographic and genetic modelsin conservation biology: applications and perspectives fortropical rain forest tree species. A. Rev. Ecol. Syst. 27, 387^421.Barbosa, R. I. 1990 Ana?lise do setor madeireiro do estado deRoraima. Acta Amazonica 20, 193^209.Bawa, K. S., Perry, D. R. & Beach, J. H. 1985a Reproductivebiology of tropical lowland rainforest trees. I. Sexual systemsand incompatibility mechanisms. Am. J. Bot. 72, 331^345.Bawa, K. S., Bullock, S. H., Perry, D. R., Colville, R. E. &Grayum, M. H. 1985b Reproductive biology of tropicallowland rainforest trees. II. Pollination mechanisms. Am. J.Bot. 72, 346^356.Bermingham, E. & Dick, C. 2001 The Inga: Newcomer ormuseum antiquity? Science 293, 2214^2216.Butz-Huryn, V. M. 1997 Ecological impacts of introducedhoneybees. Q. Rev. Biol. 72, 275^297.Chase, M. R., Moller, C., Kesseli, R. & Bawa, K. S. 1996Distant gene ?ow in tropical trees. Nature 383, 398^399.De Oliveira, A. A. & Mori, S. 1999 A central Amazonian terra?rme forest. I. High tree species richness on poor soils. Biodiv.Conserv. 8, 1219^1244.
Genetic rescue of remnant tropical trees C.W. Dick 2395
Proc. R. Soc. Lond. B (2001)
Dick, C. 2001 Habitat change, African honeybees and fecundityin the Amazonian tree Dinizia excelsa (Fabaceae). In Lessonsfrom Amazonia: the ecology and conservation of a fragmented trop icalforest (ed. R. O. Bierregaard, C. Gascon, T. E. Lovejoy & R.Mesquita), pp. 146^157. New Haven, CT: Yale UniversityPress.Dick, C. & Hamilton, M. B. 1999 Microsatellites from theAmazonian tree Dinizia excelsa (Fabaceae). Mol. Ecol. 8, 1765^1766.Ducke, A. 1922 Plantes nouvelles ou peu connues de la re?gionamazonienne. Archivos do Jardim Botanico do Rio de Janeiro 3,2^269.Ghazoul, J., Liston, K. A. & Boyle, T. J. B. 1998 Disturbance-induced density-dependent seed set in Shorea siamensis(Dipterocarpaceae), a tropical forest tree. J. Ecol. 86, 462^473.Hubbell, S. P. & Foster, R. B. 1983 Diversity of canopy trees in aneotropical forest and implications for conservation. InTropical rain forest: ecology and management (ed. S. L. Sutton,T. C.Whitmore & A. C. Chadwick), pp. 25^41. Oxford, UK:Blackwell Scienti?c Publications.Janzen, D. H. 1986 The future of tropical ecology. A. Rev. Ecol.Syst. 17, 305^324.Laman,T. 1994 Safety recommendations for climbing rain foresttrees with s`ingle rope technique?. Biotropica 27, 406^409.Laurance, W. F. 1998 A crisis in the making: responses ofAmazonian forests to land use and climate change.Trends Ecol.Evol. 13, 411^415.Levin, D. 1995 Plant outliers: an ecogenetic perspective. Am.Nat. 145, 109^118.Lovejoy, T. & Bierregaard, R. 1990 Central Amazonian forestsand the minimum critical size of ecosystems project. In Four
neotrop ical forests (ed. A. Gentry), pp. 60^75. New Haven, CT:Yale University Press.Marshall, T. C., Slate, J., Kruuk, L. E. B. & Pemberton, J. M.1998 Statistical con?dence for likelihood-based paternityinference in natural populations. Mol. Ecol.7, 639^655.Murawski, D. A., Gunatilleke, I. A. U. & Bawa, K. 1994 Thee?ects of selective logging on inbreeding in Shorea megistophylla(Dipterocarpaceae) from Sri Lanka. Conserv. Biol. 8, 997^1002.Nason, J. D. & Hamrick, J. L. 1997 Reproductive and geneticconsequences of forest fragmentation: two case studies ofneotropical canopy trees. J. Heredity 88, 264^276.Phillips, O. L., Hall, P., Gentry, A. H., Sawyer, S. A. &Vasquez, R. 1994 Dynamics and species richness of tropicalrain forests. Proc. Natl Acad. Sci. USA 91, 2805^2809.Roubik, D. W. 1989 Ecology and natural history of tropical bees.Cambridge University Press.Roubik, D. W. 2001 Feral African bees augment neotropicalco?ee yield. In International Workshop on the Conservation andSustainable Use of Pollinators (ed. P. G. Kevan, V. Imperatriz-Fonseca). Cardi?, UK: International Bee ResearchAssociation. (In the press.)Smith, B. & Padgett, A. 1996 On rope: North American vertical ropetechniques. Huntsville, AL: National Speleological Society.White, G. M., Boshier, D. H. & Powell, W. 1999 Geneticvariation within a fragmented population of Swietenia humilisZucc. Mol. Ecol. 8, 1899^1909.Winston, M. L. 1992 The biology and management ofAfricanized honey bees. A. Rev. Entomol. 37, 173^193.Young, A., Boyle, T. & Brown, T. 1996 The population geneticconsequences of habitat fragmentation for plants. Trends Ecol.Evol. 11, 413^418.
2396 C.W. Dick Genetic rescue of remnant tropical trees
Proc. R. Soc. Lond. B (2001)