The University of Chicago
Pollen Dispersal in Low-Density Populations of Three Neotropical Tree Species
Author(s): E. A. Stacy, J. L. Hamrick, J. D. Nason, S. P. Hubbell, R. B. Foster and R. Condit
Reviewed work(s):
Source: The American Naturalist, Vol. 148, No. 2 (Aug., 1996), pp. 275-298
Published by: The University of Chicago Press for The American Society of Naturalists
Stable URL: http://www.jstor.org/stable/2463455 .
Accessed: 10/12/2012 15:39
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
 .
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
 .
The University of Chicago Press, The American Society of Naturalists, The University of Chicago are
collaborating with JSTOR to digitize, preserve and extend access to The American Naturalist.
http://www.jstor.org 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
Vol. 148, No. 2 The American Naturalist August 1996 
POLLEN DISPERSAL IN LOW-DENSITY POPULATIONS OF THREE 
NEOTROPICAL TREE SPECIES 
E. A. STACY,l * J. L. HAMRICK,1 J. D. NASON,1 S. P. HUBBELL,2'4 
R. B. FOSTER,3'4 AND R. CONDIT4 
'Department of Botany, 2502 Miller Plant Sciences Building, University of Georgia, Athens, 
Georgia 30602; 2Department of Biology, Princeton University, Princeton, New Jersey 08544; 
3Department of Botany, Field Museum of Natural History, Chicago, Illinois 60605; 4Smithsonian 
Tropical Research Institute, Unit 0948, APO AA 34002-0948 
Submitted December 6, 1994; Revised October 23, 1995; Accepted November 6, 1995 
Abstract-Studies of mating patterns of tropical trees, typically involving common species, 
have revealed that most species are outcrossed and that, in some cases, a significant fraction 
of outcross pollen moves long distances. We evaluated mating systems and effective pollen 
dispersal for three hermaphroditic, insect-pollinated Neotropical tree species, Calophyllum lon- 
gifolium, Spondias mombin, and Turpinia occidentalis, all of which occurred at low adult densi- 
ties at the study site. Mating patterns were estimated for each maternal tree within 84-ha popula- 
tions of C. Longifolium and S. mombin in 1992 and 1993 and within a 50-ha population of T. 
occidentalis in 1993. Each population was 100% outcrossed. Multilocus paternity exclusion 
analyses indicated that in C. Longifolium, a minimum of 62% of effective pollen moved at least 
210 m. For S. mombin, estimates of apparent pollen flow greater than 300 m were 5.2% and 
2.5% in 1992 and 1993, respectively. For all species, pollen dispersal was strongly affected by 
the spatial distribution of reproductive trees. Where flowering adults were evenly spaced, a 
large fraction of effective pollen moved at least a few hundred meters and well beyond the 
nearest reproductive neighbors. Conversely, where flowering trees were clumped, the majority 
of matings were among near neighbors. The minimum area required to encompass a natural 
breeding unit was estimated for each population. 
The large intertree distances characteristic of many tropical tree populations 
have long fascinated evolutionary biologists. Early workers, perplexed by the 
broad spacing often observed among conspecifics, predicted that such trees 
should be predominantly self-fertilizing or inbred (e.g., Corner 1954; Baker 1959; 
Fedorov 1966). This expectation was based on the following eneralizations. 
First, unlike the primarily wind-pollinated temperate flora, tropical woody species 
are almost exclusively animal pollinated (Frankie 1975; Janzen 1975; Opler et al. 
1980). In view of the high tree species diversity and structural complexity of the 
rain forest, it was assumed that these pollinators would be unlikely to move 
reliably among widely spaced conspecifics (Corner 1954; Baker 1959; Fedorov 
1966). It was also observed (though in unnatural settings) that tropical trees flower 
* Present address: Department of Biology, 5 Cummington Street, Boston University, Boston, Mas- 
sachusetts 02215. E-mail: estacy(cabio.bu .edu. 
Am. Nat. 1996. Vol. 148, pp. 275-298. 
C 1996 by The University ofChicago. 0003-0147/96/4802-0003$02.00. All rights reserved. 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
276 THE AMERICAN NATURALIST 
asynchronously within populations (Corner 1954; Fedorov 1966). Thus, opportu- 
nities for outcrossing among tropical trees, particularly over long distances, were 
expected to be limited. 
Based on his studies of dipterocarp forests in Southeast Asia, Ashton (1969) 
advanced the contrasting view that, although selling may be common, ample 
opportunity exists for outcrossing in tropical trees. The generalist nature of many 
tropical pollinators, as well as the highly synchronized flowering characteristic 
of dipterocarp forests, were cited among the reasons for this expectation (Ashton 
1969). Indeed, more recent observations of the reproductive phenology of tropical 
trees have revealed that flowering within atural populations is largely synchro- 
nized and governed by seasonal processes (e.g., Frankie et al. 1974). Further 
support for the view that widely spaced tropical trees are outcrossed came from 
Janzen's (1971) observation that female Euglossine bees, important pollinators in 
the Neotropics, regularly "trapline" over tens of kilometers in their daily foraging 
for nectar and pollen. Janzen (1971) cited additional reports that other Neotropical 
pollinators, including hummingbirds, sphyngid moths, and other large bees, regu- 
larly move long distances among nonspecific plants, shifting to new host species 
as they become available. 
Despite the challenge presented by these opposing views, relatively few studies 
have focused on mating patterns of tropical trees. Direct investigations oftropical 
tree breeding systems, involving surveys of floral morphology and/or hand polli- 
nation experiments, revealed that obligate outcrossing isthe chief mode of repro- 
duction in these species (Ashton 1969; Bawa 1974; Kalin Arroyo 1976; Opler and 
Bawa 1978; Chan 1981; Koptur 1984; Lack and Kevan 1984; Bawa et al. 1985b; 
Kevan and Lack 1985; Dayanandan et al. 1990). High outcrossing rates in tropical 
trees are achieved largely through a combination of widespread self- 
incompatibility in hermaphroditic species and a high frequency of dioecy (22% in 
a Costa Rican dry forest [Bawa 1974], 23% in a Costa Rican moist forest [Bawa 
et al. 1985b], and 26% in a mixed dipterocarp forest in Sarawak [Lewis 1942]). 
In the first analysis of the mating system of a tropical tree species using genetic 
markers, O'Malley and Bawa (1987) estimated a population outcrossing rate of 
95% for Pithecellobium pedicellare in a Costa Rican rain forest. Subsequent anal- 
yses have revealed high rates of outcrossing in many other hermaphroditic ropi- 
cal trees (O'Malley et al. 1988; Murawski and Hamrick 1991; Murawski et al. 
1994). Though mixed mating systems, in which a significant proportion of progeny 
arise through selling, have also been found (e.g., Cavanillesia platanifolia [Mura- 
wski et al. 1990; Murawski and Hamrick 1992a], Ceiba pentandra [Murawski and 
Hamrick 1992b], and Shorea trapezifolia [Murawski et al. 1994]), such species 
appear to be relatively uncommon. For two species of Bombacaceae with mixed 
mating systems (C. platanifolia nd C. pentandra), outcrossing rates of individual 
trees were positively correlated with the number of flowering nonspecific neigh- 
bors (Murawski et al. 1990; Murawski and Hamrick 1991, 1992b). Furthermore, 
for these and six other tropical species, deviation from random mating was greater 
in more spatially isolated trees, which presumably received pollen from fewer 
donors (Murawski and Hamrick 1991). 
Studies of patterns of pollen movement in tropical woody species are few, but 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 277 
in some cases they have revealed that a significant fraction of outcross pollen 
moves over long distances. For two butterfly-pollinated sp cies, the herb Cnidos- 
colus urens and the vine Psiguria warscewiczii, work with fluorescent dye parti- 
cles indicated pollen movement over distances of 31 m and up to 600 m, respec- 
tively (Webb and Bawa 1983; Murawski 1987). Analogous work with the 
hummingbird-pollinated shrub Malvaviscus arboreus indicated pollen flow over 
225 m between flowering plants (Webb and Bawa 1983). More recently, genetic 
markers and paternity exclusion techniques have been used to infer directly pat- 
terns of effective pollen movement in populations of tropical trees. Using this 
approach, other researchers (Hamrick and Murawski 1990) found that, in popula- 
tions of the hermaphroditic anopy trees, Platypodium elegans and Tachigali 
versicolor, 20% of successful pollen moved at least 750 m, and 25% moved at 
least 500 m, respectively. Both of these species (Leguminosae: Papilionoideae) 
are pollinated by bees that presumably forage over extensive areas (Janzen 1971). 
Similar work with a third Neotropical tree species, Cordia alliodora, which is 
likely pollinated by long-tongued bees and/or butterflies (Croat 1978), showed 
that, although the majority of effective pollen movement was restricted to about 
75 m, a small but significant proportion moved up to 280 m (Boshier et al. 1995). 
Most studies of mating patterns of tropical trees have focused on conspicuous 
species. Much less is known about tropical tree species that normally occur at 
very low densities (e.g., one adult per 10 ha), for which reproduction is presum- 
ably more challenging. For low-density populations, a key question is whether 
cross-pollination is regularly accomplished or whether selling is more common 
as a result of the limited availability of outcross pollen. 
For instance, in an indirect study of mating systems of lowland rain forest 
trees in Costa Rica, Bawa et al. (1985b) found an unusually high frequency of 
self-compatibility among low-density hermaphroditic species. It was suggested 
(Murawski and Hamrick 1991) that the underlying incompatibility mechanisms 
and constraints on selling normally found in tropical trees may be relaxed in 
species that typically occur at extremely low densities. On the other hand, if 
reproduction i  low-density populations is primarily by outcrossing, as it is for 
the more common tree species, then how far does pollen move among flowering 
adults? More specifically, are maternal trees mating primarily with nearest neigh- 
bors, or are they receiving the bulk of their pollen from donors located farther 
away? If the dispersal of outcross pollen is highly localized (i.e., restricted to 
nearest neighbors), then the potential for inbreeding in these populations hould 
be considerable (Fedorov 1966; Bawa et al. 1985b), especially if seed dispersal is 
also limited. 
In this study, we examine the mating system and patterns of effective pollen 
movement of three hermaphroditic tree species that occur at low adult densities 
in the tropical moist forest of Barro Colorado Island (BCI), Republic of Panama. 
In contrast o tropical tree species that have been analyzed previously, these 
three species are pollinated by a variety of small insects, including bees, beetles, 
flies, wasps, moths, and butterflies (Bawa et al. 1985a). Although the behavior 
of these insects as pollinators i largely unknown, their potential for long-distance 
pollen movement is presumably less than that for the larger tropical bee species 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
278 THE AMERICAN NATURALIST 
that have been studied more intensively (Frankie et al. 1976; Bawa 1977; Bawa 
et al. 1985a). 
For each study species, the following questions were asked: Are reproductive 
individuals primarily selfed or outcrossed? If they are outcrossed, what is the 
pattern of effective pollen movement among individuals within the population? 
Specifically, what proportion of pollen movement is restricted to nearest neigh- 
bors versus that over longer distances? What is the effect of the spatial distribu- 
tion of reproductive adults on the mating system and patterns of effective pollen 
movement? How do the mating system and the breeding structure of the popula- 
tion vary between reproductive events? Finally, what is the minimum area re- 
quired to encompass a natural breeding unit? 
STUDY SPECIES 
Calophyllum longifolium Willd. (Clusiaceae) is a large, polygamous canopy 
tree known from moist forests of Panama, Colombia, Surinam, Brazil, and Peru 
(Croat 1978) (here, polygamous refers to individuals possessing both unisexual 
and bisexual flowers). Its flowers, produced in vast numbers, are small, greenish- 
white, fragrant, and characterized as primitive and open (Croat 1978). In Panama, 
flowering occurs principally during the rainy season (October-November), with 
fruits maturing throughout the late dry and early rainy seasons (March-July; 
Croat 1978). The green, single-seeded fruits are dispersed by mammals (primarily 
bats; Yazquez-Yanes et al. 1975) but may also be moved by water (Ridley 1930). 
At the study site, seedlings and saplings of C. longifolium are common, but adults 
occur at a density of about one tree per 3.5 ha. 
Spondias mombin L. (Anacardiaceae), also a canopy tree, occurs naturally 
throughout tropical America nd as an introduced species in tropical Africa and 
the East Indies (Croat 1978). In Panama, the small, white, actinomorphic flowers 
of S. mombin are typically bisexual and occasionally pistillate (Croat 1978). Flow- 
ering occurs primarily from March to June, with fruits maturing from July to 
October (Croat 1978). The fleshy, tasty mesocarp of the multiseeded fruit is an 
important food source for a wide range of frugivores, ome of which may serve 
as effective seed dispersers (Smythe 1970; Heithaus et al. 1975; Yasquez-Yanes 
et al. 1975). On BCI, S. mombin is frequent in secondary forest and along lake 
edges, but it is restricted within the old forest o an adult density of approximately 
one tree per 6 ha. 
Turpinia occidentalis G. Don subsp. breviflora Croat (Staphyleaceae) is a 
smaller tree found occasionally in subcanopies of moist forests of southern Mex- 
ico through Colombia and the West Indies (Croat 1978). The flowers of T. occi- 
dentalis are tiny, fragrant, white, and bisexual (Croat 1978). In Panama, flowers 
are present mainly from April to June, with fruits maturing from July through 
September (Croat 1978). Fruits are yellow at maturity and contain several small 
orange-brown seeds that are dispersed by arboreal frugivores and secondarily by 
terrestrial mammals (Croat 1978). Although juveniles at the study site are com- 
mon (about one individual with >1-cm diameter at breast height [dbh] per 0.6 
ha), adults occur at a much lower density (one tree per 2.2 ha). 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 279 
MATERIAL AND METHODS 
Study Site and Field Methods 
We sampled populations of Calophyllum longifolium, Spondias mombin, and 
Turpina occidentalis on BCI, the largest island (16 kM2) in Lake Gatun. Lake 
Gatun was formed after the damming of the Chagres River during the construction 
of the Panama Canal from 1907 to 1914. The vegetation on BCI is semievergreen, 
moist tropical forest, including both primary and secondary stands (Knight 1975; 
Croat 1978). Our study site was the 50-ha Forest Dynamics Project (FDP) plot, 
managed by the Smithsonian Tropical Research Institute (fig. 1) (Hubbell and 
Foster 1983). The Smithsonian Tropical Research Institute maintains an extensive 
database on the demography and sizes of all woody species occurring on the FDP 
plot, updating the information every 5 yr through recensusing. Using this database 
and published records of fruiting phenology (e.g., Croat 1978; Foster 1992), we 
selected three species that occur at low adult densities on the FDP plot and that 
produce accessible seeds. For the two larger species, C. longifolium and S. mom- 
bin, we surveyed a 100-m-wide border around the plot perimeter (previously 
established by J. L. Hamrick and D. A. Murawski). This approach increased the 
total study area to 84 ha for those two species. All adults found within this zone 
were tagged, measured for diameter at breast height, scored as with or without 
fruit, and mapped relative to the 50-ha FDP plot. 
Calophyllum longifolium and S. mombin were sampled in both 1992 and 1993, 
whereas sampling of T. occidentalis was restricted to 1993. All potentially repro- 
ductive individuals within the study area were scored for the percentage of their 
crown that contained fruit. During the longer 1993 field season, the total reproduc- 
tive output of individual trees (from 0% to 100% crown) was determined by 
repeatedly observing the canopy and the ground surrounding each tree throughout 
the period of fruit maturation. 
Dispersion analysis for reproductive trees was performed for each species ac- 
cording to the nearest-neighbor method of Clark and Evans (1954). The index of 
dispersion (R) was calculated as the ratio of observed to expected mean distances 
between nearest reproductive neighbors, the latter calculated on the assumption 
that trees are distributed randomly. Values of R less than 1.0 indicate clumping, 
whereas those greater than 1.0 indicate a more even distribution. 
Using a wrist slingshot, we sampled leaf material from each adult and subadult 
within the study area for each species (i.e., either 84 ha or 50 ha). Subadults were 
included in the sampling, as estimates of minimum size (dbh; FDP data) at sexual 
maturity are not always accurate. Thus, all individuals of C. longifolium -18-cm 
dbh according to the 1990 plot census were sampled. Similarly, we sampled all 
individuals of S. mombin and T. occidentalis recorded in 1990 as >-20-cm and 
-15-cm dbh, respectively. Sampled tissue was freeze-dried on BCI for 48 h fol- 
lowing an earlier method (Hamrick and Loveless 1986) and shipped in airtight 
plastic bags to the University of Georgia within 2 wk of sampling. At the lab, all 
freeze-dried material was held at - 70'C until enzyme xtraction was performed. 
Fruit or seed was collected from the ground beneath each reproductive adult and 
either freeze-dried on BCI (for C. longifolium, about one-third embryo) or 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
280 THE AMERICAN NATURALIST 
REPUBLIC OF A _ 
S |~~FDP PLOT| 
Kilometers 
BARRO COLORADO ISLAND 
FIG. 1.-Republic of Panama with Barro Colorado Island, located approximately inthe 
center of the Panama Canal. The location of the 50-ha Forest Dynamics Project (FDP) plot 
on Barro Colorado Island is also shown. 
shipped directly to the University of Georgia for germination under standard 
greenhouse conditions (single seeds of T. occidentalis and cleaned, air-dried en- 
docarps of S. mombin). 
Enzyme extraction followed the procedure of Mitton et al. (1979). All material 
was crushed in liquid nitrogen with a pinch of clean sea sand. An extraction 
buffer was added, and the resulting solution was absorbed onto filter-paper wicks 
and stored at - 70TC before electrophoresis. 
Electrophoresis and Statistical Analysis 
For all adults and progeny sampled, we employed starch-gel ectrophoresis to
determine genotypes at all polymorphic loci showing high resolution. Five gel/ 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 281 
TABLE 1 
POLYMORPHIC Loci RESOLVED FOR THREE SPECIES OF TROPICAL TREES 
AND THE GEL/ELECTRODE BUFFER SYSTEMS USED FOR EACH 
Buffer Calophyllum Spondias Turpinia 
System longifolium mombin occidentalis 
7 ... ... AAT- 1(2) 
... ... AAT-2(2) 
MNR- 1(2) 
8 LAP-1(3) DIA-1(2) DIA-1(2) 
PGI-2(2) FE-1(2) PER-1(2) 
LAP- 1(2) PGI-2(2) 
... PGI-2(4) TPI- 1(2) 
10 6PGD-1(3) . 
11 MDH- 1(2) PGM- 1(2) MDH- 1(2) 
... ... IDH-1(2) 
6 CE-2(2) ... 
FE-2(2) . . 
PER- 1(2) 
NOTE.-Descriptions and recipes of buffer systems can be found 
in Soltis et al. (1983). The number of alleles at each locus is shown 
in parentheses. Abbreviations: AAT, Aspartate aminotransferase; 
CE, colorimetric esterase; DIA, diaphorase; FE, fluorescent ester- 
ase; IDH, isocitrate dehydrogenase; LAP, leucine aminopeptidase; 
MDH, malate dehydrogenase; MNR, menadione reductase; PER, 
peroxidase; 6PGD, 6-phosphogluconate dehydrogenase; PGI, 
phosphoglucoisomerase; PGM, phosphoglucomutase; and TPI, trio- 
sephosphate isomerase. 
electrode buffer systems were used to resolve five to nine polymorphic allozyme 
loci, depending on the species (table 1). Based on observed levels of genetic 
diversity at each locus, multilocus exclusion probabilities were calculated for 
each population (Brown 1989). An exclusion probability indicates the resolving 
power of paternity data and is defined as the mean probability of excluding a
randomly selected male as the father of a given offspring from a randomly chosen 
maternal individual (Chakraborty etal. 1988). 
The mating system, or proportion of selfing versus outcrossing, was estimated 
for each species using the multilocus mixed-mating model of Ritland and Jain 
(1981; Ritland 1990). This maximum-likelihood method provides single- and 
multilocus estimates of outcrossing rate, as well as estimates of pollen allele 
frequencies at both the population and individual levels. Standard errors for each 
value were estimated based on the construction of 100 bootstrap data sets, follow- 
ing the methods of Ritland and Jain (1981). Model assumptions are discussed by 
Ritland and Jain (1981). 
The distribution of pollen movement o each outcrossed adult was evaluated 
using fractional paternity exclusion analysis (Devlin et al. 1988). For each mater- 
nal tree, the straight-line distance to the nearest edge of the study area was 
determined and treated as the radius of a hypothetical circle encompassing the 
"known neighborhood" for that tree (fig. 2). The known neighborhood was there- 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
282 THE AMERICAN NATURALIST 
- g N~~~~~~~~~~~~~ 
E I 
0 N 
01N 
exlso anayse codceMnte5-aFPpo n urudn 0--iebre 
7 
/ 0 
zone 
S..lOom 
0 I~~~~~~~~~~~~~~~~~~~0 
fore the maximum circular area surrounding each maternal tree within which all 
potential pollen donors had been located and genotyped. More centrally located 
maternal trees had larger known neighborhoods, the largest possible neighbor- 
hood being 700 m in diameter (38.5 ha in area) for trees located near the center 
of the east-west midline of the FDP plot. A series of paternity exclusion analyses 
was performed for all progeny sampled from each maternal tree. In the first 
analysis, only the maternal tree was included in the pool of potential paternal 
parents. Subsequent analyses included additional trees in the paternal pool, one 
at a time, from the nearest neighbor outward until the edge of the known neighbor- 
hood was reached. After each analysis, the proportion of progeny whose geno- 
type could not be assigned to any tree in the paternal pool was recorded. Apparent 
pollen flow was estimated as the proportion of total pollen flow for which no 
donor within the paternal pool could be identified given the available genetic data 
(Devlin and Ellstrand 1990). Since this approach evaluates levels of apparent 
pollen movement only, it provides conservative, minimum estimates of pollen 
dispersal distances to each reproductive adult over the spatial scale of its known 
neighborhood. In addition to the above analyses, further analyses were conducted 
in which all known potential fathers occurring within a zone ("extended neighbor- 
hood") of 100-150 m beyond the known neighborhood were included. Estimates 
of incoming apparent pollen flow from beyond the extended neighborhood may 
be inflated, if unidentified reproductive adults occurred in unsurveyed areas. 
A second means of examining pollen flow was used for S. mombin. Three 
reproductive adults were each heterozygous for a rare marker allele. These indi- 
viduals were treated as pollen donors, and the distribution of their rare alleles 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 283 
Ml H I 0 Om 
iN17 9*10 N 
16I5 El 
F 5( 4M *7 14 ! 
013 
| ? *012 8lO9m 
E 
j41 ~~~~~~~~C Jw *~~~D* * 
FIG. 3.-The 84-ha plot indicating the locations and identifications of all 25 adults of 
Calophyllum longifolium (i.e., all individuals with ?30-cm dbh). Symbols are in two halves, 
the left and right halves depicting reproductive activity in 1992 and 1993, respectively. Circles 
denote reproductive individuals (solid circles for moderately to highly reproductive trees; 
open circles for minimally reproductive trees). Squares denote nonreproductive adults. As- 
terisks indicate adults alive in 1992 that died before the 1993 fruiting season. 
was examined in progeny sampled from surrounding adults. Also, at each of 
three loci, an allele was found in the sampled progeny that was absent among 
reproductive adults within the study area. Variation in the frequency and distribu- 
tion of each immigrant allele among sampled families was examined for insights 
into patterns of long-distance gene movement in S. mombin. 
Based on the observed distributions of effective pollen movement, we esti- 
mated the minimum area occupied by a natural breeding unit of each species. 
This was defined as the area within which, on average, 95% of the pollen received 
by a centrally located adult originates. 
RESULTS 
Spatial and Reproductive Patterns 
Calophyllum longifolium. -The minimum size required for reproduction in Ca- 
lophyllum longifolium is about 30-cm dbh (FDP data). Fruit production by adults 
less than 49-cm dbh, however, was limited or absent in both years. A total of 25 
adult (fig. 3) and six subadult trees were sampled from the 84-ha study area. Of 
the 25 adults present in August 1992, 23 survived to the following fruiting season. 
In 1992 and 1993, respectively, about half (14 and 12) of these adults did not 
reproduce, and a few (one and four) were minimally reproductive (i.e., with 10% 
of their canopy containing fruit), which left only 8-10 highly reproductive trees 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
284 THE AMERICAN NATURALIST 
- 1000 m - 
*13 *2 N 13 t 
E 9 lo lo left ar l dinediavi312 A -1116 3 
5 / 1 ~~~~~4 "A 
8~~~~~~~~~~~~8 
CI 
FIG. 4.-The 84-ha plot indicating the locations and identifications ofall 14 adults of 
Spondias mombin (i.e., all individuals with >-30-cm dbh). Symbols are in two halves, the 
left and right halves depicting reproductive activity in 1992 and 1993, respectively. Circles 
denote reproductive individuals ( olid circles for moderately to highly reproductive trees; 
open circles for minimally reproductive trees). Squares denote nonreproductive adults. 
(50%-100% of the crown with fruit). Total fruit production in these trees was 
consistent across the two study seasons. In sum, the density of C. longifolium 
adults with substantial reproduction was only about one tree per 10 ha. The 
spatial distribution of reproductive trees was slightly clumped (index of disper- 
sion, R = 0.89), with a mean nearest-neighbor distance between reproductive 
trees of 99.8 m. Two adults, I and J, were tightly clustered (fig. 3). Removing 
these two trees from the dispersion analysis resulted in a slightly hyperdispersed 
distribution (R = 1.06) with a mean reproductive nearest-neighbor distance of 
108.4 m. 
Up to 60 seeds were collected from each reproductive tree per season. Exclud- 
ing minimally reproductive trees, means of 32 and 56 mature single-seeded fruits 
per tree were collected in 1992 and 1993. The larger sample sizes in 1993 were 
due to a longer field season in that year, which permitted greater opportunity to
collect from trees with slightly different phenologies. 
Spondias mombin. -As for Calophyllum longifolium, the minimum size of sex- 
ually mature adults of Spondias mombin is approximately 30-cm dbh (FDP data). 
A total of 14 trees of adult size occurred within the 84-ha site, 12 of which 
reproduced uring the study period (fig. 4). Eleven adults produced fruit during 
both reproductive seasons, whereas the twelfth, a small adult, fruited possibly 
for the first ime in 1993. In sum, the observed density of reproductive S. mombin 
in the study area was approximately one tree per 7 ha. The distribution ofrepro- 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 285 
ductive S. mombin adults was clumped (fig. 4; R = 0.65), with a mean nearest- 
neighbor distance for reproductive adults of 86.5 m. The two adults that failed to 
reproduce in either year had heavy vine loads, which could have restricted flow- 
ering. In addition to the 14 adult trees, five subadults between 20- and 29.9-cm 
dbh (1990 census) were sampled for electrophoretic analysis. 
Fruiting of reproductive trees involved nearly 100% of the crown with few 
exceptions. In 1992 and 1993, about 150 mature fruits were collected from the 
ground beneath each reproductive adult. Ultimately, 576 seedlings from 10 fami- 
lies and 430 seedlings from 11 families were analyzed from the 1992 and 1993 
fruit crops, respectively. The additional family sampled in 1993 was permitted by 
the longer field season that year, which overlapped with fruit drop of all adults. 
Turpinia occidentalis. -We observed and sampled Turpinia occidentalis in 1993 
only. Although the minimum size of sexually mature trees was previously reported 
as 30-cm dbh (FDP data), we found evidence of low fruit production by individuals 
as small as 17-cm dbh, which were consequently classified as adults. A total of 30 
trees -15-cm dbh (1990 census) were sampled from the 50-ha study area, including 
23 adults (fig. 6, below). Dispersion analysis of all potentially reproductive trees 
(i.e., -17-cm dbh) revealed a slightly clumped spatial distribution (R = 0.95), with 
a mean reproductive nearest-neighbor distance of 67.1 m. Fruit production varied 
considerably among individuals, with seven failing to fruit altogether. Of the 16 re- 
maining trees, eight showed minimal fruit yield (< 10% of canopy), with most of the 
fruit crop restricted to the largest rees. Five of the six largest adults had crowns 
50%-100% full of fruit, whereas reproduction by the sixth tree was apparently re- 
stricted by vines. The five highly reproductive trees occurred in two clusters (i.e., 
no individual was spatially isolated; fig. 6, below). 
We collected from 34 to approximately 800 seeds from the seed shadows of 
each reproductive tree. Germination rates were fairly low across all families, in 
large part because of infestation of some seeds. Ultimately, 172 progeny from six 
families were electrophoresed (6-43 progeny each; mean = 28). 
Genetic Variation and Exclusion Probabilities 
The quality of allozyme xpression varied among species and tissue types. The 
number of resolvable loci was greatest in freeze-dried embryos of Calophyllum 
longifolium and fresh and freeze-dried leaves of Turpinia occidentalis, whereas 
it was somewhat reduced in all other lyophilized material and fresh Spondias 
mombin. For all three species, particularly C. longifolium and S. mombin, accu- 
rate scoring of some loci was restricted to seed/seedling material, because of 
stronger expression in these tissues. Maternal genotypes at these loci were, there- 
fore, inferred based on genotypic ratios in progeny arrays (Brown and Allard 
1970). In all, seven polymorphic loci (with two or three alleles each) were scored 
in C. longifolium, five (with two to four alleles each) in S. mombin, and nine 
(each with two alleles) in T. occidentalis (table 1). Based on observed levels of 
allozyme variation, exclusion probabilities calculated for each population and 
season were as follows: C. longifolium, 0.67 and 0.74 in 1992 and 1993, respec- 
tively; 0.53 for S. mombin in both years; and 0.78 for T. occidentalis in 1993. 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
286 THE AMERICAN NATURALIST 
Mating System 
All three populations were essentially 100% outcrossed. No notable variation 
in mating system was found within individuals between reproductive s asons or 
among fruiting adults within populations. Therefore, there seems to be little ffect 
of year or local density of reproductive adults (i.e., number of near neighbors) 
on individual outcrossing rates. Here we report multilocus and mean single-locus 
estimates of population outcrossing rates, along with standard errors for each 
based on 100 bootstraps. 
Calophyllum longifolium.-Estimates of the mating system were based on a 
set of eight reproductive adults in each year. Multilocus estimates of outcrossing 
rate (tm) were very similar over the 2 yr at 1.030 ? 0.085 (mean ? SE) and 1.031 
+ 0.035 in 1992 and 1993, respectively. Mean single-locus estimates (tQ) for the 
two seasons were lower, 0.974 ? 0.062 and 0.948 ? 0.042, which possibly indi- 
cates biparental inbreeding (Brown 1989). 
Spondias mombin.-Multilocus estimates of population outcrossing rate were 
tm = 0.989 ? 0.163 and tm = 1.304 ? 0.108 in 1992 and 1993, respectively, 
based on 10 and 11 families ampled in the 2 yr. Mean single-locus estimates 
were ts = 0.996 ? 0.166 in 1992 and ts = 1.280 ? 0.124 in 1993. The high 
variances associated with these estimates may be due to skewed allele frequencies 
at three of the five loci used. 
Additional evidence that adults of Spondias mombin were completely out- 
crossed was observed in the progeny genotype data of the three reproductive 
adults heterozygous for a rare allele. No individuals homozygous for the rare 
allele were found among the progeny sampled in either year (156, 129, and 64 
progeny each, pooled across years). 
Turpinia occidentalis.-Population-level estimates of outcrossing rate based 
on all six families with appreciable progeny arrays were tm = 1.006 ? 0.090 and 
ts = 1.071 ? 0.109. 
As unequal family sizes can negatively bias multilocus estimates of the out- 
crossing rate (Shaw et al. 1981), mating system analyses were repeated for each 
species minus the one or two smallest families in each set. The resulting estimates 
of population outcrossing rate were elevated slightly. 
Patterns of Effective Pollen Movement 
Calophyllum longifolium.-In each reproductive s ason, a substantial propor- 
tion of pollen moved over distances of at least a few hundred meters and well 
beyond the nearest reproductive neighbors. Averaging the results of the paternity 
exclusion analyses over both years for the five most centrally located maternal 
trees (414 total progeny), we found that 62% of all successful pollen moved at 
least 210 m (table 2), which is more than twice the mean distance between nearest 
reproductive neighbors. For all but two maternal trees (trees I and J) for which 
the nearest reproductive neighbor occurred within the known neighborhood, from 
40% to 91% of all pollen received came from trees located beyond the nearest 
neighbor. Trees I and J occurred just 2 m apart at breast height (fig. 3). In sharp 
contrast to the high levels of moderate-distance pollen movement observed 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 287 
TABLE 2 
MINIMUM ESTIMATES OF APPARENT INCOMING POLLEN FLOW (%APF) FROM BEYOND THE KNOWN 
AND EXTENDED NEIGHBORHOODS FOR FIVE CENTRALLY LOCATED MATERNAL ADULTS OF 
CALOPHYLLUM LONGIFOLIUM 
Radius of 
Radius of Extended 
Material Number of Known Neighborhood Neighborhood 
ID Progeny (m) %APF (m) %APF 
1 41 200 48.8 316 9.8* 
2 90 260 38.9 382 6.7 
4 43 300 48.8* 375 9.3* 
8 121 140 73.4 290 46.4* 
12 119 212 77.3 230 23.7 
Means 82.8 209.5t 62.lt ... ... 
73.3t 237.9t 57.3t ... ... 
NOTE.-The extended neighborhood for each maternal tree includes the known neighborhood plus 
an additional area up to 150 m wide that includes the next nearest known pollen donor(s). Values are 
pooled over two reproductive s asons except where noted. 
* Estimate for one reproductive s ason only. 
t Means are weighted by family size. 
t Mean values if tree 8 (with smallest known neighborhood) is excluded. 
among the more widely distributed trees, paternity exclusion results howed that 
each of these highly reproductive adults could have sired 73% (I) and 72% (J) of 
the progeny of its nearest neighbor. Thus, where reproductive trees occurred 
randomly, pollen flow ranged over at least a few hundred meters and well beyond 
the nearest reproductive neighbors. Conversely, in the single instance where 
adults were clustered, a very large fraction of matings were between nearest 
neighbors. The observed distribution ofeffective pollen flow was highly consis- 
tent over both years of the study. 
To evaluate total apparent pollen flow on a broader scale, we performed exclu- 
sion analyses that treated the 50- and 84-ha populations as pollen sinks. Analyses 
involving all maternal trees on the 50-ha FDP plot and the same adults in the 
pool of potential fathers revealed 24.1% and 35.2% apparent pollen flow coming 
from beyond the 50-ha border in 1992 and 1993, respectively. When the paternal 
pool was expanded to include all reproductive adults in the 84-ha area, the esti- 
mated level of gene flow from beyond the study plot was 5.8% and 6.2% in these 
years. Consequently, in 1992 and 1993, approximately 18.3% and 29.1% of the 
progeny from maternal trees within the 50-ha plot may have been sired by adult 
trees within the 100-m zone surrounding the FDP plot. Finally, when all families 
within the 84-ha area were analyzed and all reproductive trees within the same 
area were included in the paternal pool, total apparent pollen flow from beyond 
the 84-ha plot was 5.0% and 7.7% in 1992 and 1993, respectively. Averaged over 
the 2 yr, maternal trees located within the 50-ha FDP plot received about 6% of 
their total incoming pollen from outside the 84-ha study area, whereas mothers 
located within the 100-m border received approximately 8% (9.3% in 1993 with 
increased sampling). 
Spondias mombin.-The 84-ha Spondias mombin population was character- 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
288 THE AMERICAN NATURALIST 
TABLE 3 
DISTRIBUTION AND FREQUENCY OF LONG-DISTANCE POLLINATION EVENTS (IMMIGRATION) WITHIN AND 
AMONG Six ADULTS OF SPONDIAS MOMBIN CLUSTERED NEAR THE CENTER OF THE FDP PLOT 
NUMBER OF NUMBER WITHOUT PERCENTAGE 
PROGENY SAMPLED FATHERS IMMIGRATION 
MATERNAL 
ID 1992 1993 1992 1993 1992 1993 
4 96 60 1 1 1.04 1.67 
5 ... 74 ... 3 ... 4.05 
6 69 60 11 0 15.94 0 
7 48 35 1 3 2.08 8.57 
3 62 37 4 1 6.45 2.70 
9 73 60 1 0 1.37 0 
Means 70 54 3.6 2.3 5.17 2.45 
NOTE.-Results shown are of paternity exclusion analyses for 2 yr including all reproductive trees 
within the 84-ha plot in the pool of potential fathers. Ellipses indicate no sample. 
ized by a low but substantial level of apparent pollen flow over distances of at 
least 300-350 m. A cluster of six reproductive trees located in a seasonal swamp 
along the east-west midline of the study plot (fig. 4) permitted quantification of
apparent pollen flow over this range of distances. Treating the swamp cluster as 
a pollen sink, paternity exclusion analyses revealed that a minimum of 5.2% (of 
348 pooled 1992 progeny) and 2.5% (of 326 pooled 1993 progeny) of the apparent 
pollen flow originated from outside the 84-ha plot. Within this group, the propor- 
tion of progeny per maternal tree sired by unknown donors ranged from 0% to 
roughly 16% (of 69 progeny) in 1992 and 0% to almost 9% (of 74 progeny) in 1993 
(table 3). The seemingly haphazard distribution of apparent pollen flow events 
among mothers within this relatively small area, as well as within mothers be- 
tween years, suggests that long-distance pollen movement in this population is 
highly idiosyncratic. 
Paternity exclusion analyses were also performed for the entire population of 
maternal trees within the 84-ha plot, including all known potential fathers in the 
paternal pool. Viewing the study population as a whole, a minimum of 4.2% (24/ 
576 pooled 1992 progeny) and 2.3% (10/430 pooled 1993 progeny) of all matings 
resulted from immigrant pollen from outside the 84-ha area. Unlike Calophyllum 
longifolium, adults of S. mombin at the edge of the study area received no more 
detectable immigrant pollen than did more centrally located trees. This observa- 
tion, however, may be an artifact of the lower exclusion probability for S. 
mombin. 
At three of five polymorphic loci, we found an allele in the progeny that did 
not occur in the adults within the 84-ha plot. Each progeny carrying a foreign 
allele therefore represented a pollen flow event from a paternal tree located out- 
side the 84-ha area. Even though all alleles at polymorphic loci were included in 
the multilocus analysis, examination of the on-plot distribution a d frequency of 
each foreign allele in sampled families allowed additional insights into patterns 
of long-distance pollen flow. At PGI-2 in 1992, the foreign allele (PGI-2c) was 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 289 
present in very low frequencies in families ampled throughout the 84-ha plot 
(one to three progeny from each of five families). In 1993, the same allele was 
sampled in the progeny of only two maternal trees (two progeny of tree 7 and 
three progeny of tree B). Ironically, these two maternal trees were located 849 
m apart (fig. 4). Observation of a second immigrant allele (FE-lb) was restricted 
to 1992 and in one family only (in seven of 69 progeny), which was located in the 
cluster of trees in the swamp. A third foreign allele, LAP-lb, appeared in both 
years and was restricted to progeny from the swamp cluster. The LAP-lb allele 
was observed in 1992 in each of the four swamp cluster families ampled (nine 
progeny; from one to five per family), and it was found in 1993 in three families in 
that group (five progeny; one, one, and three per family). Based on the observed 
frequencies of LAP-lb (the most common immigrant allele), estimated rates of 
apparent gene flow from beyond the 84-ha border to the swamp cluster (>314-m 
distance) were 2.6% and 1.5% in 1992 and 1993, respectively. 
Based on the pooled results of the single-tree paternity exclusion analyses, a 
high proportion of successful pollen moved short distances between mates, usu- 
ally among members of a cluster. Pooling data across all but the two most isolated 
maternal trees, we found that 90% + 3.6% of the progeny of a given mother 
were sired by the nearest neighbor (or two nearest neighbors in the tight cluster 
of adults including trees 4, 5, and 6; fig. 4). The observed pattern of effective 
pollen movement was consistent over both years of the study. 
Analysis of the distribution ofrare marker alleles about paternal parents indi- 
cates that pollen dispersal among adults of S. mombin was typically restricted to 
near neighbors. Because each donor was heterozygous for a rare allele, the total 
proportion of progeny sired by a donor on a particular maternal tree was deter- 
mined by multiplying by two the proportion of offspring carrying that allele. This 
method should provide unbiased estimates of pollen dispersal among members 
of a cluster, though not necessarily over longer distances (i.e., the probability of 
sampling a rare allele decreases with increasing distance from the source). Each 
of the three pollen donors included in this analysis (trees 1, 4, and 6) occurred in 
an area of clumped reproductive adults (two groups of three and six trees each; 
fig. 4). The rare marker alleles were detected in progeny of adults within a radius 
of 75, 87, and 138 m from the three donor trees (fig. 5a-c). The only exception 
to this pattern was the observation in 1993 of one progeny sampled from tree 2 
(60 total progeny sampled) with a rare allele for which the two only known sources 
occurred at distances of 392 and 465 m (fig. Sb, c). As tree 2 occurred 103 m from 
the nearest edge of the study area, it is possible that the donor of the rare allele 
was an off-plot adult located closer than either of the two known donors. How- 
ever, the complete absence of the rare allele in all but one of 110 progeny sampled 
from tree 2 over the 2 yr makes this latter scenario unlikely. 
The distribution of the three rare marker alleles among sampled families was 
consistent over the 2-yr period. In all cases, the rare allele was present in each 
family within the cluster that included the donor tree. The sole exception to this 
pattern occurred in 1993, when we failed to find DIA-lb from tree 1 in any of 19 
progeny sampled from tree B located in the same cluster (fig. 4). Dispersal of 
rare alleles decreased as a function of distance between parents within a cluster 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
80 
70 a. 
*60 1992 
o - 1993 
50 
c40 
30 
20 
10 
8 10 68 82 87 249 384 688 721 792 
Distance From Donor Tree 6 (m) 
80 
70 b. 
61992 
?C - 1993 
a 50 
40O 
30 
C 20 
10 
8 10 75 90 91 241 392 694 727 798 
Distance From Donor Free 4 (iM) 
80 
-70 C. 
F 60 z 1992 
C - " 1993 
50 
40 
>,30 
20 
10 
0 
40 135 465 613 623 688 694 697 727 900 
Distance From Donor Tree 1 (m) 
FIG. 5.-Effective pollen dispersal from three individual donor trees of Spondias mombin 
as indicated by rare marker alleles over two reproductive seasons. a, Tree 6 using PGI-2e; 
b, tree 4 using DIA-lb; c, tree I using DIA-lb). Zero values are indicated by the absence 
of a vertical bar, while missing data are indicated by an asterisk. 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 291 
- 1000 m - 
| 16 N 
08 072 t 
7 2~~~~~1U | *~ ~ ~ ~~~~ ~ ~~~18 17;a 
1 06 010 
0 0 
Uo 
50 9 
*22 2 4 
03 *19 *21 
20K 23 
FIG. 6.-The 50-ha FDP plot indicating the locations, identifications, and fruiting activity 
in 1993 of all 23 adults of Turpinia occidentalis (i.e., all individuals with >17-cm dbh). Circles 
denote reproductive individuals (solid circles for moderately to highly reproductive trees; 
open circles for minimally reproductive trees). Squares denote nonreproductive adults. 
(r2 = 0.25, slope = -0.061, F = 9.88, df = 1, P = .004, pooled across pollen 
donors and years; fig. 5a-c). However, the frequencies of rare alleles sampled 
within families varied between years (fig. 5a-c). 
Turpinia occidentalis.-The five highly reproductive adults of Turpinia occi- 
dentalis occurred in two clusters, 35 and 57 m in diameter, located approximately 
235 m apart (fig. 6). Paternity exclusion analyses indicated that the majority of 
successful pollen moved among individuals within clusters but that a fraction of 
the matings occurred over longer distances. Mating between the two adults 
of the more northerly cluster (trees 1 and 2), located 35 m apart, was such that 
each could have sired up to 62.5% (20/32) and 89.5% (17/19) of the progeny of 
its nearest neighbor. Pooling progeny over both individuals, 27.5% (14/51) re- 
sulted from apparent pollen flow from beyond these two trees. In the second 
cluster, composed of three adults, even higher proportions of progeny resulted 
from near-neighbor matings. For two of these maternal trees, 100% of the progeny 
sampled (72 total) could have been sired by one or two of the other adults, which 
were located within 42 m of the maternal tree. For the third adult (tree 4), how- 
ever, pollen flow over longer distances was detected. Paternity analyses indicated 
that the nearest neighbor (tree 9, 21 m away) could have sired up to 83.7% (36/ 
43) of progeny sampled from tree 4. Including the second-nearest neighbor (tree 
3, 57 m away) in the analysis accounted for 93% of the progeny sampled. Adding 
the third-nearest neighbor (tree 11, 145 m away), which was well outside the 
cluster and slightly outside the known neighborhood for tree 4, accounted for all 
but one, or 97.7%, of the sampled progeny. Thus, pooling progeny over all three 
maternal trees, we found that less than 1% (i/115) resulted from apparent pollen 
flow from beyond the cluster of nearest neighbors and over a minimum distance 
of 130 m. 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
292 THE AMERICAN NATURALIST 
DISCUSSION 
Although patterns of pollen movement varied among these three Neotropical 
tree species, they also had many features in common. First, these species, like 
the majority of tropical trees examined to date, are insect pollinated and predomi- 
nantly outcrossed (e.g., O'Malley and Bawa 1987; O'Malley et al. 1988; Murawski 
and Hamrick 1991, 1992; Murawski et al. 1994). These trees probably experience 
high rates of intracrown pollination. The high observed rates of outcrossing, 
however, do not necessarily indicate self-incompatibility. Spondias mombin, the 
only study species examined previously for its breeding system, is reportedly 
self-incompatible in Costa Rica (Bawa 1974; Bawa and Opler 1975). If the other 
two species are self-compatible, their high outcrossing rates might be due to 
preferential bortion of selfed seeds (Dayanandan et al. 1990; Ramirez 1993). 
For instance, adults of Calophyllum longifolium experienced heavy fruit abortion 
throughout the fruit maturation period (E. Stacy, personal observation). Although 
we sampled and electrophoresed mbryos from aborted fruits of C. longifolium, 
analysis of the mating system was impossible because of insufficient enzyme 
activity. 
Another characteristic ofthe breeding system common to the three study spe- 
cies is that a substantial proportion of pollen moved at least a few hundred meters, 
and possibly well beyond that range, in each population. Actual pollen flow is 
expected to be substantially greater than the levels of apparent pollen movement 
detected. For the three species, the probability of exclusion ranged from 0.53 to 
0.78. As a result of these modest values, the single-tree paternity exclusion analy- 
ses were unable to exclude some of the potential but not actual fathers for a 
portion of the progeny analyzed. As more potential fathers were included in the 
paternal pool in each analysis, starting with the nearest neighbor and proceeding 
outward, our inability to exclude nonfathers esulted in the underestimation f the 
distances between members of a mating pair. Moreover, our inability to exclude a 
higher proportion of pollen donors within the study plot decreased the number 
of detectable pollen flow events from off-plot fathers. The ratio of total to ap- 
parent gene flow for populations with higher exclusion probabilities (>0.80) is 
2-3: 1 (Devlin and Ellstrand 1990). In contrast, with a lower exclusion probability 
(e.g., S. mombin), estimated total gene flow would be greater, possibly three to 
four times the apparent rate. As a result, the paternity exclusion analysis esti- 
mates of pollen dispersal distances reported here should be conservative underes- 
timates of their actual values. 
Estimates of gene movement into the 84-ha S. mombin population based on 
rare marker alleles are also conservative. First, reported rates of immigrant pollen 
flow over 300 m (2.6% in 1992; 1.5% in 1993) are underestimates that can be 
translated to total gene flow by dividing by the marker allele frequencies in the 
donor population. For example, if the global frequency of a marker allele is 0.20, 
estimated total gene flow in 1992 would be 13% (2.6%/0.20). Second, our analyses 
of pollen movement within the 84-ha study plot based on rare marker alleles in 
individual donors revealed only one case of pollen flow over a few hundred meters 
(392 or 465 m; fig. 5b, c). Certainly, many more cross-pollinations occurred over 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 293 
this range of distances. However, as the frequency of a rare allele in the pollen 
pool is initially low and decreases with increasing distance from the donor, the 
likelihood of detecting long-distance pollen movement by this method is low given 
the progeny sample sizes obtained. 
To our knowledge, no studies exist on the pollination behavior of the diverse 
array of small insects that occur in Neotropical rain forests. The capacity for 
long-distance pollen movement by such species has been suggested to be limited, 
however, because of their restricted foraging ranges (Frankie et al. 1976; Bawa 
1977). In this study, we identified substantial levels of pollen movement over 300 
m for two species pollinated by diverse small insects. Because of the conservative 
nature of our estimates and because the size and shape of the study plot limited 
detection of pollen flow to less than 350 m, actual pollen dispersal over distances 
greater than 350 m undoubtedly occurred but could not be quantified. 
Previous work on the mating patterns of tropical trees has also revealed sub- 
stantial evels of pollen flow over relatively long distances (e.g., Hamrick and 
Loveless 1989; Hamrick and Murawski 1990). These studies, however, focus on 
species primarily pollinated by bees (small to large). In one (Hamrick and Love- 
less 1989), it is estimated that 25% of progeny arising within an isolated cluster 
of five adults of Tachigali versicolor esulted from pollen flow from at least 500 
m outside the group. In a second study (Hamrick and Murawski 1990), patterns 
of pollen movement were assessed over a 3-yr period for Platypodium elegans 
occurring in low reproductive adult density (8-12 individuals) on the 50-ha FDP 
plot. Using fractional paternity methods, at least 20% of the pollen movement 
within this population is estimated to have occurred over distances of 750 m. As 
both pollinator type and experimental design differ among our study and these 
investigations, it is difficult to assess the effect of pollinator type on patterns 
of pollen movement directly. However, because the P. elegans study was also 
conducted on the 50-ha FDP plot and 100-m border zone, a comparison can be 
made of the level of apparent pollen flow received by maternal trees located 
within the 50-ha plot from trees located outside the 84-ha area. For P. elegans, 
minimum estimates of long-distance gene flow via pollen into the 50-ha plot are 
17.5%, 36.1%, and 40.1% for three consecutive years (Hamrick and Murawski 
1990). The same estimates for C. longifolium are 5.8% and 6.2% for two seasons; 
for S. mombin (with a lower exclusion probability), 4.3% and 2.0% for the same 
2 yr. The greater estimates of moderate- to long-distance gene flow in the bee- 
pollinated P. elegans are consistent with the view that tropical bees are capable 
of moving pollen over greater distances than are smaller insects. 
Overall patterns of pollen flow were strongly affected by the spatial distribution 
of reproductive trees. Where flowering adults occurred in clusters, a large fraction 
of matings were among nearest neighbors (although the proportion of near- 
neighbor matings is overestimated by our methods, as explained earlier). Con- 
versely, where reproductive trees were spaced more evenly, the majority of pol- 
len dispersal occurred over larger areas, often bypassing the nearest reproductive 
trees. For each species, we assumed that flowering activity was restricted to 
those individuals producing fruit, however minimal. Because of the clumped dis- 
tributions of S. mombin and T. occidentalis, violation of this assumption would 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
294 THE AMERICAN NATURALIST 
not affect he observed patterns of pollen movement for either population. If, 
however, substantial flowering occurred in adults of C. longifolium classified as 
nonreproductive, the estimated istances of pollen movement would be upwardly 
biased. Flowering of C. longifolium adults classified as nonreproductive was 
likely nominal or absent. Reproduction was assessed even for trees with minimal 
fruiting activity. In 1993, all adult trees were monitored for evidence of reproduc- 
tion beginning early in the period of fruit maturation, and flowering was confined 
to those adults classified as reproductive. 
Whether the strong association observed between the spatial distribution of 
flowering trees and the degree of near-neighbor mating can be generalized to 
tropical trees pollinated by other kinds of vectors remains to be seen, as we are 
not aware of similar studies for comparison. Our results, however, may be consis- 
tent with a previous observation (Murawski and Hamrick 1991) that trees oc- 
curring at low density appear to receive pollen from relatively few donors (i.e., 
one or a few near neighbors). In a related study based on pollinator flight patterns 
and movement of marker dyes, Linhart (1973) concludes that pollen flow in two 
hummingbird-pollinated sp cies of Heliconia was strongly influenced by the spa- 
tial dispersion of flowering individuals. Where flowering shrubs were clumped 
(H. latispatha), the majority of pollen movement was within the cluster, whereas 
where shrubs were scattered (H. acuminata), a linear relationship between dis- 
tance and pollen flow was observed (Linhart 1973). The similarity between the 
observed patterns of hummingbird- and small-insect-mediated pollen movement 
suggests that, in general, foraging behavior of tropical pollinators (and hence 
pollen dispersal patterns) is strongly affected by the spatial distribution of flow- 
ering plants. 
Examination of mating patterns on a finer scale revealed some inconsistencies 
in pollen flow. For example, apparent pollen dispersal from individual donors of 
S. mombin, as indicated by rare marker alleles, was typically restricted to mem- 
bers of a cluster. The distribution of rare alleles was, however, erratic across 
years over any given distance, as well as across families within a year. Such 
variation in mating patterns within clusters of flowering trees could be explained 
by lowered fitness of offspring resulting from biparental inbreeding (assuming 
near neighbors are related), asynchronous flowering phenologies of neighboring 
trees, or inconsistent pollinator behavior. Although genetic relatedness of neigh- 
boring adults cannot be ruled out, spatial autocorrelation studies of other tropical 
species have shown that genetic relatedness of adult-sized neighboring trees is 
extremely low (Hamrick et al. 1992). Slight differences among the flowering pe- 
riods of near neighbors i a more likely explanation, because initial fruit drop was 
not completely simultaneous among adults at the study site (E. Stacy, personal 
observation). It seems, therefore, that slightly asynchronous flowering phenolo- 
gies and variable pollinator behavior are the most likely causes of the inconsisten- 
cies observed in pollen movement among neighboring trees. 
The appearance of immigrant marker alleles from donors outside the 84-ha plot 
also was idiosyncratic, interms of both frequency and distribution among families 
of S. mombin. Immigrant alleles were detected no more frequently in progeny 
sampled near the edge of the plot than in families ampled from the plot center. 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 295 
In fact, the majority (76%) of all detected foreign alleles occurred in families in 
the swamp cluster (roughly 350 m from the nearest plot border), although at 
variable frequencies among families. As the probability of sampling rare alleles 
decreases with distance from the source (here, a minimum of 300 or 350 m), 
variation in frequency of immigrant alleles in sampled families could be explained, 
in part, by sampling error. It seems equally likely, however, that pollen movement 
by small insects between widely spaced trees, or clusters of trees, is highly er- 
ratic. The only consistent pattern with respect o immigrant alleles was observed 
for LAP-lb, the most common immigrant allele. The LAP-lb allele was found in 
1992 in all families ampled within the swamp cluster, and it was observed in 
1993 in three of the five families ampled there. This finding may indicate that 
once a cluster of flowering trees is discovered by foraging pollinators, most or 
all individuals within the group are visited. 
From our analyses, we suggest he following eneralizations for low-density 
tropical trees pollinated by diverse small insects. In populations characterized by 
a high degree of clumping (e.g., S. mombin and T. occidentalis), a large propor- 
tion (up to 90%) of the matings are among nearest neighbors, with a smaller but 
substantial fraction occurring over distances of a few to possibly several hundred 
meters. Conversely, where the spatial distribution oftrees is more regular (e.g., 
C. longifolium), a large proportion of outcrossed pollen moves well beyond the 
closest reproductive neighbors. This general pattern may be explained by the 
foraging behavior of small insects. If, unlike many of the larger tropical pollina- 
tors, small insects do not trapline regularly among widely dispersed nonspecific 
plants, movements by these insects between trees might be more haphazard. 
Where flowering adults occur in clusters, neighboring conspecifics would be 
readily discovered and visited. Where flowering trees are more widely dispersed, 
however, locating nonspecific individuals could be more difficult. As a result, 
these pollinators may inadvertently b pass nearest neighbors in their search for 
pollen and/or nectar. 
Based on the mating patterns observed for each species, we estimate the small- 
est area required for a natural breeding unit, defined as the minimum area within 
which 95% of the pollen received by a centrally located adult originates. Using 
C. longifolium as a model of an evenly dispersed population with a low density 
of reproductive adults, we suggest hat a natural breeding unit would extend a 
minimum of 60 ha. For populations characterized by clumping of reproductive 
trees (e.g., S. mombin and T. occidentalis), a natural breeding unit would occupy 
at least 40 ha. These conservative estimates hould be useful in the development 
of management s rategies for tropical forests wherever the preservation of normal 
population dynamics and mating patterns in uncommon tree species is desirable 
(e.g., in the design of forest reserves). On a dour note, however, these values 
suggest that forest fragmentation, which is widespread throughout tropical re- 
gions, likely results in significantly reduced breeding neighborhoods for many 
trees when isolated fragments are on the order of 60 ha or smaller. 
These results should apply to a broad range of Neotropical forest species. 
Because the general patterns of gene movement were consistent across species 
and years, mating patterns observed in this study should be representative for 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
296 THE AMERICAN NATURALIST 
other low-density tropical trees pollinated by an array of small insects. Our finding 
of appreciable levels of moderate-distance pollen movement in all three popula- 
tions indicates that this class of pollinators is effective in transferring pollen 
among widely dispersed tropical trees. It has been estimated that approximately 
30% of the tree species in a Costa Rican rain forest'are pollinated by diverse 
small insects, including small bees (Bawa et al. 1985a). Although such figures for 
other tropical forests are unavailable, this group of insects is likely to be involved 
in the pollination of a significant fraction of woody species occurring throughout 
the Neotropics. 
ACKNOWLEDGMENTS 
We thank K. Ritland, S. K. Jain, B. Devlin, K. Roeder, and N. C. Ellstrand for 
providing computer programs for the estimation of mating systems and patterns of 
paternity. We are grateful to J. Wright, 0. Calderon, R. Perez, and E. Sierra for 
sharing their knowledge of the reproductive biology of tropical trees. The Univer- 
sity of Georgia greenhouse staff provided care for tropical seedlings. The manu- 
script was improved through the helpful comments of R. Wyatt, R. Malmberg, 
and J. E. Tobin. C. Hahn produced many figures for the manuscript. This work 
was supported in part by the Smithsonian Tropical Research Institute's Short- 
Term Fellowship Program and National Science Foundation grants BSR-8918420 
and DEB-9307878 to J.L.H. 
LITERATURE CITED 
Ashton, P. S. 1969. Speciation among tropical forest rees: some deductions in the light of recent 
evidence. Biological Journal of the Linnean Society 1:155-196. 
Baker, H. G. 1959. Reproductive methods as factors in speciation in flowering plants. Cold Spring 
Harbor Symposia on Quantitative Biology 24:177-199. 
Bawa, K. B. 1974. Breeding systems of tree species of a lowland tropical community. Evolution 28: 
85-92. 
1977. The reproductive biology of Cupania guatemalensis Radlk. (Sapindaceae). Evolution 
31:52-63. 
Bawa, K. B., and P. A. Opler. 1975. Dioecism in tropical forest rees. Evolution 29:167-179. 
Bawa, K. S., D. R. Perry, and J. H. Beach. 1985a. Reproductive biology of tropical owland rain 
forest rees. I. Sexual systems and incompatibility mechanisms. American Journal of Botany 
72:331-345. 
Bawa, K. S., S. H. Bullock, D. R. Perry, R. E. Coville, and M. H. Grayum. 1985b. Reproductive 
biology of tropical owland rain forest rees. II. Pollination systems. American Journal of 
Botany 72:346-356. 
Boshier, D. H., M. R. Chase, and K. S. Bawa. 1995. Population genetics of (Cordia alliodora) 
(Boraginaceae), a Neotropical tree. III. Gene flow, neighborhood, and population structure. 
American Journal of Botany 82:484-490. 
Brown, A. H. D. 1989. Genetic haracterization f plant mating systems. Pages 145-162 in A. H. D. 
Brown, M. T. Clegg, A. L. Kahler, and B. S. Weir, eds. Plant population genetics, breeding, 
and genetic resources. Sinauer, Sunderland, Mass. 
Brown, A. H. D., and R. W. Allard. 1970. Estimation of the mating system in open-pollinated maize 
populations using isozyme polymorphisms. Genetics 66:133-145. 
Chakraborty, R., T. R. Meagher, and P. E. Smouse. 1988. Parentage analysis with genetic markers 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
POLLEN FLOW IN LOW-DENSITY TROPICAL TREES 297 
in natural populations. I The expected proportion of offspring with unambiguous paternity. 
Genetics 118:527-536. 
Chan, H. T. 1981. Reproductive biology of some Malaysian dipterocarps. III. Breeding systems. 
Malaysian Forester 44:28-36. 
Clark, P. J., and F. C. Evans. 1954. Distance to nearest neighbor as a measure of spatial relationships 
in populations. Ecology 35:445-453. 
Corner, E. J. H. 1954. The evolution of tropical forest. Pages 34-46 in J. S. Huxley, A. C. Hardy, 
and E. B. Ford, eds. Evolution as a process. Allen & Unwin, London. 
Croat, T. B. 1978. Flora of Barro Colorado Island. Stanford University Press, Stanford, Calif. 
Dayanandan, S., D. N. C. Attygalla, A. W. W. L. Abeygunasekera, I. A. U. N. Gunatilleke, and 
C. V. S. Gunatilleke. 1990. Phenology and floral morphology in relation to pollination of 
some Sri Lankan dipterocarps. Pages 103-133 in K. S. Bawa and M. Hadley, eds. Reproduc- 
tive ecology of tropical forest plants. UNESCO, Paris; Parthenon, Carnforth, England. 
Devlin, B., and N. C. Ellstrand. 1990. The development and application of a refined method for 
estimating gene flow from angiosperm paternity analysis. Evolution 44:248-259. 
Devlin, B., K. Roeder, and N. C. Ellstrand. 1988. Fractional paternity assignment: heoretical devel- 
opment and comparison to other methods. Theoretical and Applied Genetics 76:369-380. 
Fedorov, A. A. 1966. The structure of the tropical rain forest and speciation in the humid tropics. 
Journal of Ecology 54:1-11. 
Foster, R. B. 1992. Seasonal cycles of fruit fall on Barro Colorado Island. Pages 219-242 in E. G. 
Leigh, A. S. Rand, and D. M. Windsor, eds. Ecology of a tropical forest: seasonal rhythms 
and long-term changes. Smithsonian I stitution Press, Washington, D.C. 
Frankie, G. W. 1975. Tropical forest phenology and pollinator-plant coevolution. Pages 192-209 in 
L. E. Gilbert and P. H. Raven, eds. Coevolution of animals and plants. University of Texas 
Press, Austin. 
Frankie, G. W., H. G. Baker, and P. A. Opler. 1974. Comparative phenological studies of trees in 
tropical owland wet and dry forest sites of Costa Rica. Journal of Ecology 62:881-919. 
Frankie, G. W., P. A. Opler, and K. S. Bawa. 1976. Foraging behavior of solitary bees: implications 
for outcrossing of a Neotropical forest ree species. Journal of Ecology 64:1049-1087. 
Hamrick, J. L., and M. D. Loveless. 1986. Isozyme variation in tropical trees: procedures and 
preliminary esults. Biotropica 18:201-207. 
1989. The genetic structure of tropical tree populations: associations with reproductive biol- 
ogy. Pages 129-146 in J. H. Bock and Y. B. Linhart, eds. The evolutionary ecology of 
plants. Westview, Boulder, Colo. 
Hamrick, J. L., and D. A. Murawski. 1990. The breeding structure of tropical tree populations. Plant 
Species Biology 5:157-165. 
Hamrick, J. L., D. A. Murawski, and J. D. Nason. 1992. The influence of seed dispersal mechanisms 
on the genetic structure of tropical tree populations. Vegetatio 108:281-297. 
Heithaus, R., T. H. Fleming, and P. A. Opler. 1975. Foraging patterns and resource utilization of 
eight species of bats in a seasonal tropical forest. Ecology 56:841-854. 
Hubbell, S. P., and R. B. Foster. 1983. Diversity of canopy trees in a Neotropical forest and implica- 
tions for conservation. Pages 24-41 in S. L. Sutton, T. C. Whitmore, and A. C. Chadwick, 
eds. Tropical rain forest: ecology and management. Blackwell Scientific, Oxford. 
Janzen, D. H. 1971. Euglossine bees as long-distance pollinators of tropical plants. Science (Washing- 
ton, D.C.) 171:203-205. 
1975. Ecology of plants in the tropics. Institute of Biology's studies in biology. No. 58. 
Arnold, London. 
Kalin Arroyo, M. T. 1976. Geitonogamy inanimal-pollinated ropical angiosperms: a stimulus for the 
evolution of self-incompatibility. Taxon 25:543-548. 
Kevan, P. G., and A. J. Lack. 1985. Pollination in a cryptically dioecious plant Decaspermum par- 
viflorum (Lam.) A. J. Scott (Myrtaceae) by pollen-collecting bees in Sulawesi, Indonesia. 
Biological Journal of the Linnean Society 25:319-330. 
Knight, D. H. 1975. A phytosociological analysis of species-rich tropical forest: Barro Colorado 
Island, Panama. Ecological Monographs 45:259-284. 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions
298 THE AMERICAN NATURALIST 
Koptur, S. 1984. Outcrossing and pollinator limitation of fruit set: breeding systems of Neotropical 
Inga trees (Fabaceae: Mimosoideae). Evolution 38:1130-1143. 
Lack, A. J., and P. G. Kevan. 1984. On the reproductive biology of a canopy tree, Syzygium syzygi- 
oides (Myrtaceae), in a rain forest in Sulawesi, Indonesia. Biotropica 16:31-36. 
Lewis, D. 1942. The evolution of sex in flowering plants. Biological Review 17:46-67. 
Linhart, Y. B. 1973. Ecological and behavioral determinants ofpollen dispersal in hummingbird- 
pollinated Heliconia. American Naturalist 107:511-523. 
Mitton, J. B., Y. B. Linhart, K. B. Sturgeon, and J. L. Hamrick. 1979. Allozyme polymorphisms 
detected in mature needle tissue of ponderosa pine. Journal of Heredity 70:86-89. 
Murawski, D. A. 1987. Floral resource variation, pollinator esponse, and potential pollen flow in 
Psiguria warscewiczii. Ecology 68:1273-1282. 
Murawski, D. A., and J. L. Hamrick. 1991. The effect of the density of flowering individuals on the 
mating systems of nine tropical tree species. Heredity 67:167-174. 
1992a. The mating system of Cavanillesia platanifolia under extremes of flowering-tree d n- 
sity: a test of predictions. Biotropica 24:99-101. 
1992b. Mating system and phenology of Ceiba pentandra (Bombacaceae) in Central Panama. 
Journal of Heredity 83:401-404. 
Murawski, D. A., J. L. Hamrick, S. P. Hubbell, and R. B. Foster. 1990. Mating systems of two 
Bombacaceous trees of a Neotropical moist forest. Oecologia (Berlin) 82:501-506. 
Murawski, D. A., B. Dayanandan, and K. S. Bawa. 1994. Outcrossing rates of two endemic Shorea 
species from Sri Lankan tropical rain forests. Biotropica 26:23-29. 
O'Malley, D. M., and K. S. Bawa. 1987. Mating system of a tropical rain forest ree species. American 
Journal of Botany 74:1143-1149. 
O'Malley, D. M., D. P. Buckley, G. T. Prance, and K. S. Bawa. 1988. Genetics of Brazil nut 
(Bertholletia excelsa Humb. & Bonpl.: Lecythidaceae). II. Mating system. Theoretical and 
Applied Genetics 76:929-932. 
Opler, P. A., and K. S. Bawa. 1978. Sex ratios in tropical forest rees. Evolution 32:812-821. 
Opler, P. A., H. G. Baker, and G. W. Frankie. 1980. Plant reproductive characteristics during second- 
ary succession in Neotropical lowland forest ecosystems. Biotropica 12:40-46. 
Ramirez, N. 1993. Producci6n ycosto de frutos ysemillas entre formas de vida. Biotropica 25:46-60. 
Ridley, H. N. 1930. The dispersal of plants throughout the world. Reeve, Ashford, England. 
Ritland, K. 1990. A series of Fortran computer programs for estimating plant mating systems. Journal 
of Heredity 81:235-237. 
Ritland, K., and S. K. Jain. 1981. A model for the estimation of outcrossing rate and gene frequencies 
using n independent loci. Heredity 47:35-52. 
Shaw, D. V., A. L. Kahler, and R. W. Allard. 1981. A multilocus e timator of mating system parame- 
ters in plant populations. Proceedings of the National Academy of Sciences of the USA 78: 
1298-1302. 
Smythe, N. 1970. Relationship between fruiting season and seed dispersal methods in a Neotropical 
forest. American Naturalist 104:25-35. 
Soltis, D. E., C. H. Haufler, D. C. Darrow, and G. J. Gastony. 1983. Starch gel electrophoresis of 
ferns: a compilation of grinding buffers, gel and electrode buffers, and staining schedules. 
American Fern Journal 73:9-15. 
Webb, C. J., and K. S. Bawa. 1983. Pollen dispersal by hummingbirds and butterflies: a comparative 
study of two lowland tropical plants. Evolution 37:1258-1270. 
Yasquez-Yanes, C., A. Orozco, G. Francois, and L. Trejo. 1975. Observations on seed dispersal by 
bats in a tropical humid region in Veracruz, Mexico. Biotropica 7:73-76. 
Associate Editor: Ruth G. Shaw 
This content downloaded by the authorized user from 192.168.52.67 on Mon, 10 Dec 2012 15:39:03 PM
All use subject to JSTOR Terms and Conditions