Abstract Evidence concerning mechanisms hypothe-

sized to explain species coexistence in hyper-diverse

communities is reviewed for tropical forest plants. Three
hypotheses receive strong support. Niche differences are
evident from non-random spatial distributions along mi-
cro-topographic gradients and from a survivorship-
growth tradeoff during regeneration. Host-specific pests
reduce recruitment near reproductive adults (the Janzen-
Connell effect), and, negative density dependence occurs
over larger spatial scales among the more abundant spe-
cies and may regulate their populations. A fourth hy-
pothesis, that suppressed understory plants rarely come
into competition with one another, has not been consid-
ered before and has profound implications for species
coexistence. These hypotheses are mutually compatible.
Infrequent competition among suppressed understory
plants, niche differences, and Janzen-Connell effects
may facilitate the coexistence of the many rare plant spe-
cies found in tropical forests while negative density de-
pendence regulates the few most successful and abun-
dant species.
Keywords Alpha diversity Density dependence
Disturbance Host-specific pests Janzen-Connell
hypothesis
Introduction
A central question in community ecology concerns the
control of alpha diversity, or the number of species able
to coexist at small spatial scales. Plant alpha diversity
reaches astonishing levels in equatorial forests. For ex-
ample, a single hectare of Amazonian forest can support
more than 280 tree species with diameter at breast height
(d.b.h.) 10 cm (Valencia et al. 1994; Oliveira and Mori
1999). Tree diversity is equally remarkable at slightly
larger spatial scales. A 0.52-km
2
plot in Borneo and a
0.25-km
2
plot in Ecuador support 1,175 and 1,104 tree
species with diameter at breast height (d.b.h.) 1 cm, re-
spectively (LaFrankie 1996; R Condit, personal commu-
nication). In contrast, the 4.210
6
km
2
of temperate for-
ests that cover Europe, North America and Asia support
just 1,166 tree species with maximum height >7 m
(Latham and Ricklefs 1993). For trees, alpha diversity in
the tropics can rival hemispheric diversity in the North
Temperate Zone. Other life forms augment alpha diversi-
ty in tropical forests, particularly in rain forests where
trees comprise just 25% of the plant species (Gentry and
Dodson 1987). Plant alpha diversity is greater in equato-
rial rain forests than in any other vegetation type (Gentry
1988).
Evolutionary biogeography contributes to this diversi-
ty. Angiosperms dominate tropical forests (Prance 1977).
Warm, wet climates, which are now restricted to equato-
rial latitudes, predate the first angiosperms and charac-
terized immense landmasses throughout the angiosperm
radiation (Terborgh 1973; Morley 2000). Hyper-diverse
angiosperm floras have repeatedly developed through
long distance dispersal and in situ evolution within
510 million years after a warm, wet climate replaced a
cool, dry climate (Morley 2000). Today, the most diverse
tree floras are found in Borneo and in a broad band
across the central and western Amazon basin (Ashton
1993; Ter Steege et al. 2000). The persistent, extensive,
and benign physical environment combine to insure that
the pool of tropical forest plant species is large.
Here I ask how hundreds of these species can coexist
within a single hectare. Competition limits diversity
when superior competitors increase in abundance until
other species are excluded. More than 100 mechanisms
have been proposed that delay or prevent competitive ex-
clusion. Palmer (1994) organized this huge mass of
competing hypotheses around six conditions required to
realize competitive exclusion. I use the same six condi-
tions to organize this review (Table 1). Each condition is
violated by one or more of the principal hypotheses ad-
S.J. Wright (

)
Smithsonian Tropical Research Institute,
Apartado 2072, Balboa, Ancn, Republic of Panama
e-mail: [email protected]
Tel.: +507-212-8132, Fax: +507-212-8148
Oecologia (2002) 130:114
DOI 10.1007/s004420100809
S. Joseph Wright
Plant diversity in tropical forests:
a review of mechanisms of species coexistence
Received: 10 August 2000 / Accepted: 14 August 2001 / Published online: 12 October 2001
Springer-Verlag 2001
vanced to explain the coexistence of tropical forest
plants (Table 1). The long generation times of most trop-
ical forest plants have precluded experiments to evaluate
these hypotheses relative to alpha diversity. This has left
two approaches to study plant species coexistence in
tropical forests.
The first assumes Palmers corollary holds (see final
line in Table 1) and evaluates correlations between alpha
diversity and proxy variables, which are believed to re-
flect the degree to which a condition of the competitive
exclusion principle is violated. Three examples follow.
Evidence for the (1) niche differentiation and nutrient re-
source ratio hypotheses, (2) regeneration niche and gap
dynamic hypotheses, and (3) Janzen-Connell and com-
pensatory mortality hypotheses comes from correlations
between alpha diversity and (1) soil attributes believed
to reflect spatial heterogeneity (Ashton 1993; Tilman and
Pacala 1993), (2) tree mortality rates believed to reflect
temporal environmental variation (Phillips et al. 1994);
and (3) several conditions believed to reflect pest pres-
sure (Givnish 1999). This valuable approach has recently
been reviewed elsewhere and is not considered here
(Wright 1999).
The second approach to the study of plant species co-
existence in tropical forests concerns the detection and
strength of hypothesized mechanisms. If a mechanism
cannot be demonstrated to occur, then it can be discount-
ed. If a mechanism can be detected but its signal is weak,
then its influence on plant diversity is doubtful. A large
literature has demonstrated that the mechanisms underly-
ing the niche differentiation, regeneration niche, gap dy-
namic, and Janzen-Connell hypotheses operate in tropi-
cal forests (reviewed by Denslow 1987; Hammond and
Brown 1998; Sollins 1998). Hubbell and associates have,
however, marshaled evidence that each of these mecha-
nisms is too weak to maintain observed levels of plant
alpha diversity (Hubbell 1979, 1980; Hubbell and Foster
1986, 1992; Hubbell et al. 1990, 1999; Welden et al.
1991; Condit et al. 1992, 1994). This review re-evaluates
the evidence and this conclusion for conditions 14 and
the associated hypotheses from Table 1. Conditions 5
and 6 from Table 1 are omitted because mechanisms pos-
tulated for the mass effect and nutrient resource ratio hy-
potheses are unexplored for tropical forests.
Several critical attributes characterize plants and for-
est environments. As sessile autotrophs, plants must dis-
perse pollen and seeds, survive environmental variation
and pest depredations in situ, and tolerate competition
with a few near neighbors for water, mineral nutrients
and light. In forests, herbivore and pathogen activity
may be intense, low resource levels often limit plants,
and resource levels and in particular understory light lev-
els can vary tremendously in both time and space. These
attributes of plants and forest environments underlie the
hypotheses considered here.
Are rare species favored?
Negative density dependence
Negative density dependence occurs when nearby con-
specifics impair performance. Allelopathy, intraspecific
competition, and pest facilitation may all contribute. Neg-
ative density dependence constrains locally abundant spe-
cies, which opens space for otherwise less successful spe-
cies and facilitates species coexistence. Three methods
have been used to detect density dependence among trop-
ical forest plants. The Janzen-Connell hypothesis has mo-
tivated comparisons of juvenile performance near and far
from reproductive conspecifics. A second single-species
approach has contrasted performance among plots char-
acterized by the density of a focal species; and a multi-
species approach has contrasted performance among spe-
cies characterized by different density. I review each
method as applied to tropical forest plants. Throughout
performance refers to survival, growth, or recruitment.
2
Table 1 The principal hypotheses advanced to explain species co-
existence among tropical forest plants. Each hypothesis violates
one (or more) of six conditions required to realize competitive ex-
clusion. The hypotheses are associated with the principal condi-
tion violated. Citations are to studies that applied the hypotheses
to tropical forest plants. Adapted from Palmer (1994)
The competitive exclusion principle:
Given a suite of species, interspecific competition will result in the exclusion of all but one species
Conditions of the competitive exclusion principle Hypotheses that violate the condition
(1) Rare species are not favored demographically Janzen-Connell hypothesis (Janzen 1970; Connell 1971)
Compensatory mortality (Connell et al. 1984)
(2) Species have the opportunity to compete Recruitment limitation (Hubbell et al. 1999); Low understory densities (this review)
(3) (a) The environment is temporally constant Regeneration Niche or Gap dynamics (Grubb 1977; Denslow 1987)
(b) The environment has no spatial variation Many authors (reviewed by Sollins 1998; Svenning 2001)
(4) Time has been sufficient to allow exclusion Dynamic equilibrium (Huston 1994) and Intermediate disturbance (Connell 1978)
Chance population fluctuations (Hubbell 1979)
(5) Growth is limited by one resource Nutrient resource ratios (Ashton 1993; Tilman and Pacala 1993)
(6) There is no immigration Mass effects (Stevens 1992)
Corollary:
The greater the degree to which these conditions are broken, the greater the number of species that can coexist
The Janzen-Connell hypothesis
Janzen (1970) and Connell (1971) hypothesized that
host-specific pests reduce recruitment near conspecific
adults (or where conspecific seed density is greatest)
thereby freeing space for other plant species. Hammond
and Brown (1998) reviewed 46 studies that compared
seed or seedling performance near and far from conspe-
cific adults. Performance was lower near conspecific
adults for 15 of 19 populations whose principal herbi-
vore was an insect, but for just 2 of 27 populations
whose principal herbivore was a vertebrate. Vertebrates
forage over large areas, are polyphagous, and do not
reduce seed and seedling performance disproportionately
near conspecific adults. Insects and the few micro-
bial pathogens studied to date do, however, act in a
manner consistent with the Janzen-Connell hypothesis
(Augspurger 1984; Gilbert et al. 1994; Gilbert and
DeSteven 1996; Packer and Clay 2000). Do insects and
pathogens reduce recruitment near conspecific adults for
most species? Does this, in turn, influence alpha diversi-
ty? Community-level analyses are required to answer
these questions.
There have been four community-level evaluations of
the distance or density dependence postulated by Janzen
and Connell. Condit et al (1992) contrasted distances to
reproductive adults for conspecific versus heterospecific
recruits on Barro Colorado Island (BCI), Panama. Re-
cruits were concentrated near conspecific adults for 27
species, were indifferent to the identity of the nearest
adult for 38 species, and were repelled from the vicinity
of conspecific adults for 15 species. The repelled distri-
bution extended for just one crown width for 14 of the fi-
nal 15 species. Condit et al. (1992) concluded recruit-
ment is consistent with the Janzen-Connell hypothesis
for very few species and then only over very short dis-
tances. This conclusion must be qualified. Limited seed
dispersal and appropriate microhabitat (as demonstrated
by the presence of a conspecific adult) favor recruits
near conspecific trees. Host-specific pests could reduce
recruits substantially without offsetting these advantages
(Hamill and Wright 1986). Condit et al. (1992) recog-
nized this limitation of their analysis. They did not, in
fact, test the hypothesis that juvenile performance de-
clines near conspecific adults.
The three remaining community-level analyses pro-
vide evidence for distance or density dependence.
Hubbell et al. (1990) contrasted saplings located under
conspecific versus heterospecific adults for 11 abundant
species on BCI. Nearby conspecific adults reduced
growth for 9 species for small (significant for 5 species)
and intermediate sized saplings (significant for 3 spe-
cies) and also reduced survival significantly for the 2
most abundant species. Connell et al. (1984) contrasted
plants whose nearest larger neighbor was conspecific
versus heterospecific for two sites in northern Australia.
Species were pooled to increase sample size and 26 ana-
lyses were performed (two sites, first and second nearest
neighbors, multiple juvenile size classes; I disregard
near and far nearest neighbors.). Near conspecific
neighbors reduced growth in 17 analyses (5 significant-
ly) and survival in 24 analyses (14 significantly). Harms
et al. (2000) evaluated density dependence during the
seed-to-seedling transition for 53 species from BCI.
They used 200 census stations to fit recruit (R) and seed
(S) density to the following function: R=cS
b
, where c
and b are fitted constants. Negative density dependence
(b<1) was evident for every species, and the median b-
value was just 0.23. These three studies provide strong
community-level evidence for the distance and density
dependent recruitment postulated by Janzen and Connell.
It remains to explore the consequences for alpha
diversity. The theoretical development of the Janzen-
Connell hypothesis has been contentious. Janzen (1970)
reasoned that recruitment would fail completely near
conspecific adults freeing space for other species.
Hubbell (1980) showed that when recruitment failed
completely within 1, 2, 3, 4, 5, and 6 crown diameters of
a conspecific adult just 3, 7, 12,19, 27, and 37 species
could coexist at equilibrium, respectively. Hubbell con-
cluded that this effect was quantitatively unimportant be-
cause reduced performance occurs over relatively short
distances and most tropical forests support hundreds to
thousands of plant species. Becker et al. (1985) ques-
tioned Hubbells choice of equilibrium conditions and
showed that a minimum spacing rule could greatly pro-
long species persistence times in simulated communities.
Armstrong (1989) developed an analytical model to
demonstrate that a minimum spacing rule always stabi-
lizes the coexistence of competing species. This theory
has limited relevance to real forests where there is no
minimum spacing rule. Recruitment does not fall to zero
near conspecific adults. Rather, the large numbers of
seeds falling near conspecific adults overwhelm distance
and density dependent differences in per capita recruit-
ment so that recruit density is greatest where seeds were
initially most numerous (Hubbell 1980; Condit et al.
1992; Harms et al. 2000). A renewed theoretical effort is
needed to relate alpha diversity to partial reduction of re-
cruitment near conspecifics.
There has, in the meantime, been one empirical at-
tempt to relate alpha diversity to partial reduction of re-
cruitment near conspecifics (Harms et al. 2000). The
BCI seed rain is heavily dominated by a few species
while first-year seedling recruits are much more diverse.
The plant directly overhead dominated the seed rain at
the small spatial scale of single census stations (~0.5 m
2
)
while a few species that produced copious numbers of
small seeds dominated the seed rain at the larger spatial
scale of 200 census stations (50 ha). Harms et al. (2000)
partitioned the increase in diversity between seed rain
and seedling recruits into components due to density de-
pendent recruitment and species-specific differences in
seed-to-seedling transition probability (generally greater
for species with larger seeds). Density dependence made
a substantial contribution to the increase in diversity
even though recruit density increased with conspecific
seed density for 41 of the 53 focal species (0<b<1). The
3
partial reduction of recruitment near conspecifics in-
creased alpha diversity.
To summarize, the depredations of host-specific in-
sects and possibly pathogens are greatest near reproduc-
tive host trees. As a consequence, juvenile performance
is often impaired near conspecific adults and where con-
specific seed density is greatest. This effect has almost
certainly been underestimated because performance has
been evaluated for short time intervals for plants of dif-
ferent sizes with no attempt to integrate the cumulative
effect over time. Negative effects on growth and survival
are particularly likely to reinforce one another over lon-
ger time intervals, and the eventual reduction in recruit-
ment near conspecific adults may be much greater than
is now appreciated. As demonstrated by Harms et al.
(2000), even a partial reduction of recruitment near con-
specifics can increase alpha diversity. The Janzen-
Connell mechanism is likely to play an important role in
the maintenance of plant diversity in tropical forests.
Analyses among plots characterized by density
of a single species
The second method used to detect density dependence
contrasts performance among plots characterized by dif-
ferent densities of single focal species. Density and per-
formance are implicitly assumed to fluctuate asynchro-
nously among plots and to vary over a similar range for
each plot. Temporal and spatial variation may falsify
both assumptions. Interannual variation in climate syn-
chronizes seed production, tree growth, and tree mortali-
ty over large areas and may override a spatial signature
of density dependence (Clark and Clark 1994; Condit et
al. 1995; Wright et al. 1999). Spatially heterogeneous re-
sources may constrain the potential range of density and
performance in different plots and even introduce appar-
ent positive density dependence, with elevated perfor-
mance and population density in plots with appropriate
resources. For these reasons, plot-based analyses may
rarely detect negative density dependence.
Plot-based analyses have, in fact, repeatedly detected
negative density dependence for abundant species [An
appendix critiques reports of negative density depen-
dence among rarer species (Wills et al. 1997; Wills and
Condit 1999)]. Hubbell and Foster (1986) used 50 1-ha
plots to fit the densities of juveniles (J) and adults (A) of
their 48 most abundant species to the following model:
J=a
0
+a
1
A+a
2
A
2
, where a
0
, a
1
and a
2
are fitted constants.
Strong negative density dependence was evident for the
most abundant species (a
1
<0), weak negative density de-
pendence was evident for another 20 species (a
1
>0;
a
2
<0), and there was no evidence for density dependence
for 27 species. Negative density dependence was also
detected in five of six studies of single abundant species
(Martinez-Ramos et al. 1988; Alvarez-Buylla 1994;
Condit et al. 1994; Gilbert et al. 1994; Silva Matos et al.
1999). Schupp (1992) reported positive density depen-
dence for an abundant small tree on BCI; however, both
seedling recruitment and sapling performance were neg-
atively density dependent in subsequent studies of the
same population (Condit et al. 1994; Harms et al. 2000).
Plot-based analyses frequently detect negative density
dependence among the more abundant plant species in
tropical forests.
Hubbell and associates reasoned that most tropical
trees are too rare to be regulated by density dependence,
which therefore is unlikely to promote species coexis-
tence (Hubbell 1979; Hubbell and Foster 1986; Hubbell
et al. 1990). Density-dependent population regulation is
difficult to demonstrate for long-lived organisms like
tropical trees. The one approach attempted has been to
incorporate observed density-dependent vital rates into
numerical simulations of population fluctuations. Nega-
tive density dependence regulated population density at
observed levels for Trichilia tuberculata, the most abun-
dant tree on BCI (Hubbell et al. 1990). Populations were,
in contrast, growing rapidly at all densities for Euterpe
edulis, Astrocaryum mexicanum and Cecropia obtusifo-
lia (Alvarez-Buylla 1994; Alvarez-Buylla et al. 1996; Si-
lva Matos et al. 1999). Additional studies are needed to
determine whether negative density dependence regu-
lates population size and facilitates coexistence among
tropical forest plants.
Analyses among species:
the community compensatory trend
Connell et al. (1984) coined the phrase community com-
pensatory trend (CCT) to describe an inverse interspecif-
ic relationship between performance and population den-
sity. A CCT will facilitate species coexistence by pro-
tecting rare species from extinction and preventing com-
mon species from further increases at the expense of rare
species. A CCT is consistent with equilibrium species
composition if one aspect of performance improves with
population density and offsets the CCT (C.O. Webb, per-
sonal communication). A CCT otherwise implies that
species composition fluctuates through time.
There have been four attempts to detect a CCT.
Growth and survival were independent of conspecific
density in northern Australia (Connell et al. 1984; I dis-
count their regressions between per capita recruitment
and adult density, which took the form y/x versus x). Sap-
ling survival increased with population size, which is the
reverse of a CCT, at Pasoh Forest, Malaysia (He et al.
1997). CCTs were observed for recruitment on BCI and
for sapling survival at Gunung Palung, Borneo (Welden
et al 1991; Webb and Peart 1999). Density dependence
may be the mechanism responsible for the two observed
CCTs; however, other possibilities should be considered.
A CCT might be observed if rare species were associ-
ated with a rare microhabitat characterized by high re-
cruitment, rapid growth, or high survival. Gap-dependent
pioneers are rare in most tropical forests, and gaps favor
recruitment, growth, and survival (Hubbell and Foster
1986; Denslow 1987; Brown and Whitmore 1992). The
4
recruitment CCT observed for BCI spanned a severe El
Nio drought, which increased tree mortality and num-
bers of tree fall gaps (Welden et al. 1991; Condit et al.
1996). An alternative explanation for this CCT might be
an increase in habitat available to rare pioneers. Other
analyses have indicated that rare species perform poorly
on BCI (Hubbell and Foster 1986). The survival CCT
observed for Gunung Palung contrasts with the opposite
pattern observed for a second Dipterocarp forest (He et
al. 1997; Webb and Peart 1999). This may reflect differ-
ent methods. He et al. (1997) evaluated all saplings in a
50-ha plot, the vast majority was in deep shade, and rare
pioneers probably survived poorly. In contrast, Webb and
Peart (1999) evaluated saplings from similar numbers of
plots in treefall gaps and the shaded understory. An alter-
native explanation for their CCT might be that pioneers
tended to be rare and pioneer saplings were dispropor-
tionately represented in tree fall gaps where high light
levels favored survival. Future efforts to detect a CCT
should consider microhabitat associations and other pos-
sible alternative mechanisms.
The opportunity for competition
Potential competitors must encounter one another to
compete. Both recruitment limitation and low population
densities reduce such encounters.
Recruitment limitation
Recruitment limitation occurs when species fail to reach
otherwise suitable regeneration sites. Recruitment limita-
tion may facilitate species coexistence when (1) each
species is a superior competitor somewhere in the envi-
ronment, (2) each species fails to reach some otherwise
suitable regeneration sites, and (3) inferior competitors
win those sites by default (Hurtt and Pacala 1995).
Hubbell et al. (1999) addressed the second condition by
identifying 1.310
6
seeds collected in 520 weekly cen-
suses of 200 0.5-m
2
seed traps located on BCI. Just 3 of
314 species reached all 200 traps, and a mean of just 31
species arrived at each seed trap in 10 years. Hubbell et
al. (1999) concluded that dispersal limitation and, hence,
recruitment limitation was nearly universal.
This conclusion is premature. Dispersal limitation de-
creases as (1) each seed is dispersed more effectively, (2)
each tree becomes more fecund, and/or (3) the number of
reproductive trees increases (Nathan and Muller-Landau
2000). The third possibility is particularly problematical
for species coexistence. If a superior competitor became
sufficiently abundant, it could escape dispersal and re-
cruitment limitation. This may describe Trichilia tuber-
culata, which comprises 13% of all canopy trees on BCI
(Muller-Landau et al., in press). The modestly sized seed
(140 mg dry mass) is embedded in a rich aril and is dis-
persed by vertebrates capable of long distance move-
ments. In years of heavy seed set, T. tuberculata seeds
arrived at more than 90% of the 200 seed traps and seed-
lings recruited into more than 66% of 600 1-m
2
plots
(Fig. 1). The seedling of T tuberculata is among the most
shade tolerant on BCI, and the recruitment pulse ob-
served in 1997 is persistent with seedlings present in
more than 66% of 1-m
2
plots from 1997 through 1999
(Fig. 1). The small 1-m
2
seedling plots contrast with the
21-m
2
canopy area of the smallest (20 cm d.b.h.) repro-
ductive adults of T. tuberculata (OBrien et al. 1995).
One or more T. tuberculata seedlings are expected in 14
of 21 1-m
2
subplots at the spatial scale relevant to regen-
eration. This species has escaped recruitment limitation.
More generally, the combination of high adult density,
shade-tolerant seedlings, and well-dispersed seeds will
minimize recruitment limitation (Muller-Landau et al. in
press). To summarize, recruitment limitation has the po-
tential to limit interactions among the many rare tropical
trees; however, some other mechanism must check the
abundance of species like T. tuberculata that are charac-
terized by high adult density, shade-tolerant seedlings,
and well-dispersed seeds.
Low population density among understory plants
Plants compete when zones of resource depletion of near
neighbors overlap. Hence, low plant densities can pre-
5
Fig. 1 Evidence that Trichilia tuberculata escapes dispersal limi-
tation on Barro Colorado Island (BCI), Panama. The lower panel
presents seed rain density (number of seeds m
2
month
1
) estimat-
ed from 200 randomly located, 0.5-m
2
seed traps (for methods see
Wright et al. 1999). The upper panel presents percentages of seed
traps that received at least one seed (filled bars) and percentages
of 600 nearby 1-m
2
plots where at least one seedling recruited
(hatched bars) or where at least one seedling was present regard-
less of its age (open bars). Seeds reached most traps in most years
and particularly in the 1997 El Nio year when seed production
was very heavy (Wright et al. 1999). Seedlings are shade tolerant,
and the surge of seedling recruits following the 1997 El Nio per-
sisted until 1999 and beyond (personal observation)
clude direct competition particularly for light, water, and
nutrients. Canopy trees and lianas have contiguous or
overlapping crowns (and probably also root systems).
The situation may be very different among understory
plants.
Two types of interaction limit the summed density of
all understory species. The first is the competitive asym-
metry between forest strata. Canopy trees and lianas
dominate forest light and rooting environments. The can-
opy intercepts or reflects as much as 99.5% of photosyn-
thetically active radiation, and low light levels limit all
understory plants (Canham et al. 1990; Chazdon and
Pearcy 1991). Trees and lianas also dominate the rooting
environment. Dried fine root (2 mm diameter) biomass
averages 372 g m
2
for the upper 30 cm of soil on BCI
(Yavitt and Wright, in press). In contrast, the entire root
systems of 48 excavated understory shrubs averaged just
34 g m
2
(excavation methods follow Wright et al.
1992). Canopy plants account for more than 90% of the
roots in the soil volume occupied by understory shrubs
on BCI. Not surprisingly, understory plants often grow
more rapidly after experimental trenches sever compet-
ing roots both in the shaded understory and in tree fall
gaps (Coomes and Grubb 1998, 2000; Lewis and Tanner,
2000). Root and shade competition from canopy plants
suppresses understory plants.
Herbivores further reduce understory plant density.
Seedling density was 230% greater where most mamma-
lian herbivores were absent compared to a second forest
with an intact mammal fauna (Dirzo and Miranda 1991).
Exclusion experiments provide additional insight. Insect
plus vertebrate exclusion increased the growth rate of an
understory shrub by an order of magnitude in Panama
(Sagers and Coley 1995). Vertebrate exclusion alone had
a greater positive impact on seedling survival than did
the difference in microenvironments between tree fall
gaps and the shaded understory in both Panama and Aus-
tralia (Fig. 2). Vertebrate exclusion also increased seed-
ling recruitment in Panama and Peru and understory
plant density in Panama and tropical Australia (Terborgh
and Wright 1994; W.P. Carson and J.H. Connell, person-
al communication). Herbivores and competition with
canopy plants both suppress understory plant density.
Do low population densities preclude competition
among understory plants? The experimental removal of
part of the understory could provide an answer. If com-
petition were important, the remaining understory plants
should experience competitive release. This experiment
has been performed twice. Seedling germination and sur-
vival were unaffected by the removal of all plants small-
er than 5 cm d.b.h. from the understory of a Costa Rican
forest (Marquis et al. 1986). In the second experiment,
all non-dipterocarps were removed from half of ten tree
fall gaps and saplings of three focal dipterocarp species
were monitored in the removal and control halves of
each gap for 40 months (Brown and Whitmore 1992).
Survival and height growth were similar in both halves
of the nine smaller gaps (canopy openness 17.5%) and
were consistently reduced in the removal half of a single
large gap (30% canopy openness). Competitive release
was absent after understory plants were removed.
The implications of the possible lack of competition
among understory plants for light, water and nutrients
are profound. The understory is home to juveniles for
canopy trees and to all life stages for smaller trees,
shrubs and terrestrial herbs. The notion of competitively
superior and competitively inferior species becomes
moot for the smaller life forms if low density prevents
competition with similar sized plants. Juveniles of cano-
py trees and lianas might also be shielded from direct
competition with all but the canopy tree directly over-
head. Every species able to tolerate the understory envi-
ronment could potentially coexist.
Additional experiments will be required to explore
the role of competition in the understory. Understory
plants might be manipulated along environmental gradi-
ents that alter the level of suppression of understory
plants. Competition among understory plants would be
predicted to increase in importance as the level of sup-
pression by herbivores and canopy plants declined. If
this prediction were realized, forests with more favorable
understory environments (more light, fewer herbivores),
greater understory plant density, and a greater potential
for competitive exclusion among understory plants
would be predicted to have lower alpha diversity.
Spatial and temporal variation in the environment
Plants potentially compete for pollinators, seed dispersal
agents, light, water, and mineral nutrients. Plant species
6
Fig. 2 Evidence that vertebrate herbivores have a greater impact
on seedling survival than do microhabitat differences between the
shaded understory and recent tree fall gaps for northern Australia
and Panama. Experimental treatments indicated along the horizon-
tal axis include unfenced controls open to vertebrates and fenced
exclosures that excluded vertebrates. Seedling survival was great-
er in tree fall gaps (open bars) than in the shaded understory (solid
bars), inside vertebrate exclosure than in unfenced controls, and
for fenced seedlings in the deeply shaded understory than for un-
fenced seedlings in recent tree fall gaps. Mean survival ( 1 SE) is
presented for six species and two sites for Queensland, Australia
and for three species for BCI, Panama. Drawn from data presented
by Howe (1990), Osunkoya et al. (1992), and Molofsky and Fisher
(1993)
may coexist by utilizing these resources in different
ways. This possibility has been evaluated separately for
biotic and abiotic resources.
Biotic resources
Plants serviced by different animal species avoid compe-
tition for pollination and seed dispersal. Figs, for exam-
ple, never compete for pollinators because each fig spe-
cies is pollinated by a different wasp species. Such spe-
cies-specific mutualisms could greatly facilitate plant
species coexistence. Species-specific mutualisms are,
however, rare. Flower morphology and observations of
flower visitors suggest that most plant species attract one
or more large groups of taxonomically related pollina-
tors, but rarely just one pollinator species (Feinsinger
1983 Bawa et al. 1985; Bawa 1990; Roubik 1992; Kress
and Beach 1994; Momose et al. 1998). The species of in-
sects that visit conspecific flowers can also vary widely
among years (Horvitz and Schemske 1990). Most seed
dispersal agents are equally non-specific. Fruit morphol-
ogy and observations of seed dispersal suggest a small
number of syndromes each involving many plant and an-
imal species (Howe and Smallwood 1982). Most plant
species share seed dispersal agents and potential pollina-
tors with many other plant species.
Temporal segregation of flower and seed production
could still minimize competition for shared pollinators
and seed dispersal agents and thereby facilitate species
coexistence (Snow 1965; Stiles 1977; Ashton et al.
1988; Poulin et al. 1999). The potential for temporal
segregation is, however, limited. Most species bear
flowers (mature fruit) for weeks to months, and more
than 500 species may flower (fruit) in every calendar
month in species rich tropical forests (Croat 1978).
Flowers, fruits, and reproductive phenologies are also
evolutionarily conservative (Lord et al. 1995; Wright
and Calderon 1995). Closely related species are likely to
have similar flowers and fruits, to share pollinators and
seed dispersal agents, and to flower and mature fruit at
similar times. Generalized pollinators and seed dispersal
agents service most tropical forest plants, and temporal
segregation of flower and seed production is rare. I con-
clude that the potential for pollinators and seed dispersal
agents to promote plant species coexistence is severely
limited.
Abiotic resources: spatial heterogeneity
The mobility of pollinators and seed dispersal agents
precludes spatial heterogeneity at scales relevant to alpha
diversity. Many abiotic resources are, in contrast, fixed
in space. Micro-topography, for example, causes drain-
age, moisture, and possibly nutrients to vary from ridges
to slopes and nearby streams often over just tens of me-
ters. Many tropical forest plants are non-randomly dis-
tributed with respect to such micro-topographic gradients
(Clark et al. 1998; 1999; Svenning 1999; earlier studies
reviewed by Sollins 1998).
A detailed case study hints at the implications for al-
pha diversity. Drainage influences the distributions of two
congeners with contrasting physiology in Guyana (Ter
Steege 1994). Mora excelsa occurs near streams, its seeds
float and survive well in water, and its seedlings transpire
rapidly. Mora gongrijpii occurs a few meters above
stream level, its seeds sink and die once submerged, and
its seedlings transpire slowly and are able to extract water
from relatively dry soil. Each Mora species occupies the
end of the drainage gradient where its seed and seedling
physiology provides a regeneration advantage. Together
the two Mora species occupy the entire drainage gradient.
These last two points are crucial and recur elsewhere. For
example, just two species were associated with intermedi-
ate slopes in a survey of the distributions of 110 tree spe-
cies along micro-topographic gradients in Costa Rica
(Clark et al. 1999). Species restricted to the middle of mi-
cro-topographic gradients are rare.
This places an upper limit on the potential contribu-
tion of micro-topography to alpha diversity. If equal
numbers of species were restricted to each end of a mi-
cro-topographic gradient and no species were widely dis-
tributed, then micro-topography would make a two-fold
contribution to alpha diversity. The actual contribution
will be lower because many species are widely distribut-
ed across micro-topographic gradients (Sollins 1998).
Spatial variation alone is unlikely to explain the high
levels of plant alpha diversity observed in tropical for-
ests unless important sources of spatial heterogeneity
have gone undetected.
Abiotic resources: spatiotemporal variation
Theory suggests that indefinite numbers of sessile spe-
cies may coexist when resources vary through space at
any instant in time and also through time at any point in
space (Comins and Noble 1985; Tilman 1994). Such spa-
tiotemporal variation characterizes understory environ-
ments, particularly light levels and possibly rooting envi-
ronments. Tree death and replacement is the greatest
source of variation, and expected resource distributions
are identical for all locations given a sufficiently long
time period (many cycles of tree death and replacement).
This spatiotemporal variation underlies the regenera-
tion niche and gap dynamic hypotheses (Grubb 1977;
Ricklefs 1977; Denslow 1980; Orians 1982). The death of
a branch, a tree or several neighboring trees opens a gap.
This alters light levels, rooting environments, litter depth
and/or surface soils. Each new gap includes a range of
microhabitats from its center to the adjacent understory,
and larger gaps include more extreme conditions in their
more open centers. Early in regeneration species segre-
gate spatially in and around gaps according to their re-
quirements for establishment and onward growth. None-
theless, species with very different regeneration require-
ments are often near neighbors as adults for two reasons.
7
First, environmental gradients change over short distanc-
es within gaps. Second, a patchwork develops through
time as new gaps are superimposed on older gaps. Estab-
lished plants that regenerated in one microhabitat in an
earlier gap are often located near a very different micro-
habitat when a new gap forms. Species with different re-
generation requirements coexist through spatial segrega-
tion during regeneration, yet end up as near neighbors as
adults. As a consequence, alpha diversity is high.
Recent criticisms of the gap dynamic hypothesis focus
on spatial segregation during regeneration. The role of
gap dynamics is unexplored for the overwhelming major-
ity of tropical tree species, which are shade tolerant as ju-
veniles (Brown and Jennings 1998). More than 80% of
the shade-tolerant saplings present before gap formation
survive after gap formation (Uhl et al. 1988; Fraver et al.
1998). The size advantage of survivors often offsets the
high growth rates achieved by light-demanding pioneers
that germinate in gaps (Brown and Whitmore 1992;
Grubb 1996). As a consequence, both alpha diversity and
species composition are largely independent of gap size
and the history of gap forming events (Lieberman et al.
1995; Hubbell et al. 1999). The spatial heterogeneity in-
troduced by gaps does not create predictable regeneration
niches in space for most tropical plant species.
Most trees also regenerate too slowly to germinate
and reach the canopy in a single gap cycle. Temperate
tree saplings, for example, show repeated cycles of large
and small annual growth rings (Wu et al. 1999). These
saplings survived and grew slowly beneath a closed can-
opy, survived a gap-forming event and grew more rapid-
ly in the gap, and then repeated this cycle many times.
Clearly, spatial heterogeneity within any single gap can
only make a limited contribution to the successful regen-
eration of such species. A longer temporal scale that in-
tegrates survival and growth in deep shade and in gaps is
required to understand the regeneration of slow-growing,
shade-tolerant trees.
This temporal scale is addressed by studies that dem-
onstrate a tradeoff between potential growth rates in gaps
and survivorship in deep shade. For example, Shorea
johorensis and Hopea nervosa grew to mean heights of
10 m and 4 m in large gaps and had survivorship of 35%
and 60% in the shaded understory, respectively, over
77 months at Danum Valley, Sabah (Whitmore and
Brown 1996). Among three relatively light-demanding
pioneers on BCI, Trema micrantha grew most rapidly
but only survived in the largest gaps, Cecropia insignis
was intermediate on both counts, and Miconia argentea
grew most slowly and survived in the smallest gaps
(Brokaw 1987). Similar tradeoffs have been documented
in multi-species comparisons for BCI trees. Growth rates
measured with ample resources in the growing house
were inversely related to survivorship measured for
wildings in the shaded understory for 18 species
(Kitajima 1994). Finally, sapling growth rates measured
in gaps were inversely related to sapling survival mea-
sured beneath closed canopy forest (Fig. 3, Hubbell and
Foster 1992).
The tradeoff in Fig. 3 defines combinations of traits
that permit species to coexist. An inferior species whose
growth and survivorship fell below the relationship
would be excluded, while a super species whose growth
and survivorship fell above the relationship would ex-
clude other species. Super species are presumably impos-
sible because the tradeoff reflects allocation of a fixed to-
tal amount of resource to functions that maximize growth
(leaves, stem extension, roots) or survivorship (storage,
defenses against herbivores) (Kitajima 1994). The critical
role for herbivore defenses is illustrated by the reduction
in seedling survivorship caused by vertebrate herbivores,
which is greater than the reduction observed between
gaps and shaded understory (Fig. 2). The tradeoff in
Fig. 3 also suggests a mechanism for coexistence given
strict spatiotemporal variation. Conservative species
(slow growth, high survival) regenerate where resource
availability is low over time, and aggressive species (rap-
id growth, low survival) regenerate where resource avail-
ability is higher over time. This tradeoff facilitated spe-
cies coexistence in the SORTIE model of temperate forest
dynamics (Pacala et al. 1996).
Has there been sufficient time
for competitive exclusion to occur?
The temporal dynamics of competitive exclusion has be-
come intertwined with the possible coexistence of eco-
8
Fig. 3 The tradeoff between survivorship and growth for trees
from BCI, Panama. The annual survival rate (vertical axis) is for
saplings in the shaded understory. The annual growth rate (hori-
zontal axis) is for saplings in tree fall gaps. Sapling d.b.h. was ini-
tially between 1 cm and 4 cm. Many slowly growing species have
median growth rates of zero because d.b.h. was measured to the
nearest 5 mm and slow growth was undetectable. Drawn from data
in Welden et al. (1991) following Hubbell and Foster (1992)
logically equivalent species in the tropical forest litera-
ture. Tropical forests may harbor large numbers of eco-
logically equivalent plant species (Aubrville 1938;
Federov 1966; Van Steenis 1969; Hubbell and Foster
1986). Ecological equivalence may arise through com-
mon descent (Federov 1966; Van Steenis 1969) or
through convergent evolution for a generalized ability to
tolerate diffuse competition (Hubbell and Foster 1986).
In either case, ecologically equivalent species with non-
overlapping distributions would enhance diversity at
large spatial scales (Federov 1966; Van Steenis 1969).
Ecologically equivalent species will not, however, en-
hance local alpha diversity unless their prolonged coex-
istence is possible.
Slow population dynamics may permit ecologically
equivalent species to coexist. Huston (1994) considered
relative rates of population growth and of disturbances
that reduce population size. Each species persists where
the rate of disturbance permits population recovery but
prevents competitive exclusion between disturbances.
The proposed balance between population dynamics and
disturbance has yet to be evaluated for tropical forest and
seems unlikely for three reasons. First, rapid population
dynamics characterize many tropical forest plants with
population density fluctuations frequently exceeding 5%
per annum (Alvarez-Buylla et al. 1996; Condit et al.
1996, 1999; Silva Matos et al. 1999). Second, few distur-
bances reset population sizes (reviewed by Whitmore and
Burslem 1998). Anthropogenic disturbance can be dis-
counted because tropical paleofloras that predate man
were as diverse as modern floras (Burnham and Graham
1999; Morley 2000). Extreme windstorms can be dis-
counted because the advanced regeneration survives and
species composition is quickly re-established (Burslem et
al., 2000). The final reason to discount the Huston hy-
pothesis is that tree alpha diversity is often greatest where
tree population dynamics are most rapid and the two
types of disturbance that do reset population sizes (fire
and climate change) are least likely. In the western Ama-
zon for example, tree alpha diversity is greatest (Gentry
1988), tree population dynamics are most rapid (Phillips
et al. 1994), and fire and climatic drying are least likely
where annual rainfall now exceeds 4,000 mm. These in-
consistent data contrast with an impressive array of con-
sistent data marshaled by Huston (1994). A direct test of
the dynamic equilibrium hypothesis is needed.
Chance population dynamics may permit ecologically
equivalent species to coexist even in the absence of dis-
turbances. Hubbell (1979, 1997) developed the commu-
nity drift model to demonstrate that identical species can
coexist indefinitely when births and deaths occur by
chance. The community drift model is considered to be
the single plausible explanation for coexistence without
niche differentiation (Chesson and Huntly 1997).
Recent criticism of the community drift model has fo-
cused on the assumption that all species are identical. In
the community drift model, each individual has the same
probability of death, each survivor has the same proba-
bility of successful reproduction, and individuals are ar-
bitrarily divided among species. Zhang and Lin (1997)
and Yu et al. (1998) relaxed the assumptions of identical
probabilities of reproduction and survival, respectively.
Expected times to extinction declined by two orders of
magnitude when species-specific probabilities differed
by just one or two percent in both cases. Since probabili-
ties of reproduction and survival differ by at least 2%
among real species, Zhang and Lin (1997) and Yu et al.
(1998) concluded that the community drift model was
unable to explain prolonged coexistence.
In real forests, a mixture of chance and species-spe-
cific events influence recruitment and survival. Falling
debris, lightning strikes and wind throws introduce
chance. For example, falling debris killed 33% of the li-
ana saplings that died in 2 years on BCI and buried 19%
of artificial seedlings in 1 year at the La Selva Biological
Station (Aide 1987; Clark and Clark 1989). Demograph-
ic events only occurred with species-specific probabili-
ties in the models of Zhang and Lin (1997) and Yu et al.
(1998). How would coexistence times change if instead
some portion of the births and deaths occurred by chance
and the remainder occurred according to species-specific
probabilities?
I used the community drift model to answer this ques-
tion with demographic events partitioned among chance
and species-specific causes. Hubbell (1979) simulated a
community of K individuals initially divided into equally
abundant species and selected D individuals (without re-
placement) to die and D individuals from among the sur-
vivors (with replacement) to reproduce completing one
model iteration. To introduce species-specific mortality, I
first determined the species of a randomly selected indi-
vidual, then compared its species-specific probability of
dying to a random uniform number, and repeated the
process until D deaths were recorded (without replace-
ment). An analogous procedure (but with replacement)
was used to introduce species-specific reproduction. The
proportion of demographic events attributed to chance
versus species-specific causes was then varied, and each
set of conditions was simulated 100 times to estimate the
mean number of model iterations until half of the species
initially present were lost.
Coexistence times decayed rapidly as the proportion
of demographic events attributed to species-specific
causes increased from zero to 10% or 20% of all events
(Fig. 4). The diversity reducing power of a few species-
specific demographic events was unexpected. In retro-
spect, demographic events controlled by species-specific
probabilities recur, and, even when the proportion of
such events is small, their directional effect accumulates
over model iterations and disadvantaged species are rap-
idly lost. Figure 4 was created using probabilities of
death observed for BCI trees (taken from Condit et al.
1995). Additional simulations varied the variance of spe-
cies-specific probabilities of birth and death and demon-
strated a strict equivalence between this variance and the
proportion of demographic events occurring by chance
(not shown). Chance births and deaths can do very little
to enhance transitory species coexistence unless chance
9
accounts for a very large proportion (80% or more) of all
demographic events or interspecific differences are arti-
ficially small. Real differences among species quickly
overwhelm chance demographics and diversity declines.
Synthesis and future directions
There is ample evidence among tropical forest plants for
(1) niche differences associated with micro-topography,
(2) niche differences associated with a trade-off between
survivorship and growth, (3) the reduction of recruitment
near fruiting conspecifics by host-specific pests (the
Janzen-Connell effect), and (4) negative density depen-
dence among the more abundant species at larger spatial
scales. These mechanisms of species coexistence are mu-
tually compatible. Niche segregation and Janzen-Connell
effects could facilitate the coexistence of the many rare
species found in tropical forests while negative density
dependence checked populations of the most successful
and abundant species.
Janzen-Connell effects may be much more important
than is now generally realized. The full extent of the re-
duction of recruitment near fruiting conspecifics has
been systematically underestimated by short-term studies
focused on the very earliest stages of regeneration. With
one exception, distance dependence has been evaluated
for initial seedling establishment, first-year seedling per-
formance, or the onward growth of small seedlings of
unknown age (reviewed by Hammond and Brown 1998).
The one exceptional study documented reductions in
performance for saplings as large as 8 cm d.b.h. when lo-
cated near larger conspecific adults (Hubbell et al. 1990).
Distance dependent reductions in performance must ac-
cumulate as juveniles grow. Future studies that integrate
distance dependence from seed to first reproduction
are required to appreciate the real strength of Janzen-
Connell effects.
There may also be an interesting synergism between
recruitment limitation and the life history niches de-
scribed in Abiotic resources: spatio-temporal variation
above. Many tree species pass through several cycles of
gap formation and closure before reaching the canopy.
Their saplings experience a wide range of light and root-
ing environments as gaps form and close. No single sap-
ling is the best possible competitor in all environments.
Established saplings enjoy a size advantage, however,
and may resist displacement by a superior competitor
particularly when the environmental conditions that fa-
vor the superior competitor change relatively rapidly
(i.e., as gaps close). Inferior competitors may routinely
hold regeneration sites, making recruitment limitation an
ongoing process (Brown et al. 1999). The implications
for species coexistence are unexplored.
We are, finally, left with intriguing possibilities. Field
measurements only demonstrate that niche differences,
Janzen-Connell effects, and negative density dependence
occur. Implications for species coexistence and plant di-
versity remain conjectural. Large size, low population
density, and long generation times may well preclude ex-
perimental evaluation of mechanisms of plant species
coexistence in tropical forests. A promising alternative
uses field measurements to parameterize multi-species
models, which are then used to determine which field at-
tributes facilitate species coexistence (e.g., Pacala et al.
1996). The Center for Tropical Forest Science coordi-
nates a network of 16 forest plots from 15 to 52 ha in ar-
ea where all trees and shrubs larger than 1 cm d.b.h. are
enumerated at 5-year intervals (http://www.ctfs.si.edu).
These enumerations will soon make it possible to param-
eterize species rich models to evaluate how plant diversi-
ty is affected by negative density dependence, Janzen-
Connell effects, and niche differences associated with
the trade-off between survivorship and growth.
Acknowledgements I thank Allen Herre, Will Edwards, and two
anonymous reviewers for comments. The manuscript was rewrit-
ten during a sabbatical at the Rainforest CRC, James Cook Uni-
versity, Cairns, Australia. The Environmental Sciences Program of
the Smithsonian Institution funded Fig. 1.
Appendix 1
Critique of Wills et al. (1997) and
Wills and Condit (1999)
Wills and associates used a Monte Carlo simulation to
evaluate density dependence for two 50-ha plots where
all trees and shrubs larger than 1 cm d.b.h. were cens-
used twice. They found evidence for negative density de-
pendent recruitment for most species. In contrast, Condit
10
Fig. 4 The transitory maintenance of community diversity by
chance (vertical axis) decays rapidly as the proportion of deaths
determined by chance declines and the proportion determined by
species-specific causes increases (horizontal axis). Hubbell (1979)
simulated the indicated conditions. In contrast, Yu et al. (1998)
simulated conditions similar to those at the opposite extreme along
the horizontal axis. I used annual probabilities of mortality ob-
served for different tree species on BCI (Condit et al. 1995) for
species-specific deaths and equal probabilities for every individual
regardless of species for chance deaths and for all births. Follow-
ing Hubbell (1979), the simulated community included 1,600 trees
initially divided among 40 equally abundant species with 160 tree
deaths per model iteration
et al. (1992) concluded that recruits were spatially aggre-
gated near conspecific adults for one of the same 50-ha
plots (The Janzen-Connell hypothesis above). To re-
solve this discrepancy, I describe the Monte Carlo simu-
lations then offer an alternative explanation for the re-
sult.
The Monte Carlo simulation held tree locations con-
stant and swapped d.b.h. and status (recruited, survived,
or died) for two censuses for large numbers of randomly
selected pairs of conspecifics. Let N equal the number of
conspecifics observed in a particular subplot in the two
censuses. Then N=R+S+D, where R recruits first ap-
peared in the second census, S survivors were present in
both censuses, and D individuals died between censuses.
The Monte Carlo simulations maintained N but allowed
R, S, and D to vary. The final response variables were
correlation coefficients. Each 50-ha plot was divided into
contiguous subplots and correlation coefficients were
calculated among subplots for pairs of metrics derived
from observed or simulated R, S, D, and d.b.h.. The
choice of correlation coefficients is critical. Correlation
coefficients vary from 1 to 1 and squared correlation
coefficients (or coefficients of determination) vary with
the proportion of variation in one variable explained by
variation in the second variable (Sokal and Rohlf 1995,
p 564).
Consider the expected values of the simulated corre-
lation coefficients. Wills et al. (1997) simulated correla-
tion coefficients between per capita recruitment (R/A)
and conspecific basal area (BA) for 84 species from BCI.
Adults (A) were larger than a species-specific d.b.h. in
the first census. Subplots without an adult were excluded
to insure that per capita recruitment was defined. The ex-
pected value of simulated correlation coefficients is zero
because the expected value for each simulated R/A is
R/A, where sums are over all subplots. Wills and Con-
dit (1999) simulated correlation coefficients between R
and S+D (or BA) for 100 species from BCI and 100 spe-
cies from Pasoh, Malaysia. The expected simulated slope
of the relationship between R and S+D equals
R(S+D) because expected simulated values of R
equal N(RN) and expected simulated values of S+D
equal N[(S+D)N], where sums are again over all
subplots. Expected simulated correlation coefficients are
positive because expected simulated slopes are positive.
Similar logic applies to the relationship between R and
BA. Evidence for negative density dependence occurred
when observed correlation coefficients were smaller than
2.5% of the simulated correlation coefficients.
Now consider how biological phenomena affect ob-
served correlation coefficients. Any phenomenon that
concentrated recruits where S+D or BA was low will re-
duce the slope of relationships between R/A and BA or
between R and S+D (or BA). If the slope becomes nega-
tive, the correlation coefficient also becomes negative.
In addition, any phenomenon that tended to aggregate
recruitment independently of S+D or BA would also re-
duce observed correlation coefficients. Negative density
dependence would concentrate recruits where S+D and
BA were low. Unfortunately, recruitment in treefall
gaps or low canopy sites, where S+D and BA are al-
ways small, would also have this effect. Recruits were
significantly aggregated in low canopy sites for 70 of
155 BCI tree species (Welden et al. 1991). A number of
additional phenomena cause spatial aggregation of re-
cruitment. These phenomena include spatial heteroge-
neity in resource availability, localized seed dispersal,
and aggregated mortality due to wind, lightning, or pest
outbreaks (Hamill and Wright 1986). Given that Condit
et al. (1992) demonstrated that recruits were concentrat-
ed near conspecific adults for many BCI species, it is
probable that the analyses of Wills and associates de-
tected nearly ubiquitous spatial aggregation of recruit-
ment rather than negative density dependent recruit-
ment.
References
Aide TM (1987) Limbfalls: a major cause of sapling mortality for
tropical forest plants. Biotropica 19:284285
Alvarez-Buylla ER (1994) Density dependence and patch dynam-
ics in tropical rain forests: matrix models and applications to a
tree species. Am Nat 143:155191
Alvarez-Buylla ER, Garca-Barrios R, Lara-Moreno C, Martnez-
Ramos M (1996) Demographic and genetic models in conser-
vation biology: applications and perspectives for tropical rain
forest tree species. Annu Rev Ecol Syst 27:387421
Armstrong RA (1989) Competition, seed predation, and species
coexistence. J Theor Biol 141:191195
Ashton PS (1993) Species richness in plant communities. In Fiedler
PL, Jain SK (eds) Conservation biology. Chapman and Hall,
New York, pp 422
Ashton PS, Givnish TJ, Appanah S (1988) Staggered flowering in
the Dipterocarpaceae: new insights into floral induction and
the evolution of mast fruiting in the aseasonal tropics. Am Nat
132:4466
Aubrville A (1938) La foret de la Cote dIvoire. Bull Com Afr
Occie Franc 15:205261
Augspurger CK (1984) Seedling survival of tropical tree species:
interactions of dispersal distance, light gaps, and pathogens.
Ecology 65:17051712
Bawa KS (1990) Plant-pollinator interactions in tropical rain for-
ests. Annu Rev Ecol Syst 21:399422
Bawa KS, Bullock SH, Perry DR, Coville RE, Grayum MH
(1985) Reproductive biology of tropical lowland rain forest
trees. II. Pollination systems. Am J Bot 72:346356
Becker P, Lee LW, Rothman ED, Hamilton WD (1985) Seed pre-
dation and the coexistence of tree species: Hubbells models
revisited. Oikos 44:382390
Brokaw N (1987) Gap-phase regeneration of three pioneer tree
species in a tropical forest. J Ecol 75:919
Brown ND, Jennings S (1998) Gap-size niche differentiation by
tropical rainforest trees: a testable hypothesis or a broken-
down bandwagon. In: Newbery DM, Prins HHT, Brown ND
(eds) Dynamics of tropical communities. Blackwell, Oxford,
pp 7994
Brown ND, Whitmore TC (1992) Do dipterocarp seedlings really
partition tropical rain forest gaps? Philos Trans R Soc Lond B
335:369378
Brown ND, Press M, Bebber D (1999) Growth and survivorship of
dipterocarp seedlings: differences in shade persistence create a
special case of dispersal limitation. Philos Trans R Soc Lond
B 354:18471855
Burnham RJ, Graham A (1999) The history of neotropical vegeta-
tion: new developments and status. Ann Missouri Bot Gard
86:546589
11
Burslem DFRP, Whitmore TC, Brown GC (2000) Short-term ef-
fects of cyclone impact and long-term recovery of tropical rain
forest on Kolombangara, Solomon Islands. J Ecol 88:1063
1078
Canham CD, Denslow JS, Platt WJ, Runkle JR, Spies TA, White
PS (1990) Light regimes beneath closed canopies and tree-fall
gaps in temperate and tropical forests. Can J For Res 20:620
631
Chazdon RL, Pearcy RW (1991) The importance of sunflecks for
forest understory plants. BioScience 41:760766
Chesson P, Huntly N (1997) The roles of harsh and fluctuating
conditions in the dynamics of ecological communities. Am
Nat 150:519553
Clark DA, Clark DB (1994) Climate-induced annual variation in
canopy tree growth in a Costa Rican tropical rainforest. J Ecol
82:865872
Clark DB, Clark DA (1989) The role of physical damage in the
seedling mortality regime of a neotropical rain forest. Oikos
55:225230
Clark DB, Clark DA, Read JM (1998) Edaphic variation and the
mesoscale distribution of tree species in a neotropical rain for-
est. J Ecol 86:101112
Clark DB, Palmer MW, Clark DA (1999) Edaphic factors and the
landscape-scale distributions of tropical rain forest trees. Ecol-
ogy 80:26622675
Comins HN, Noble IR (1985) Dispersal, variability, and transient
niches: species coexistence in a uniformly variable environ-
ment. Am Nat 126:706723
Condit R, Hubbell SP, Foster RB (1992) Recruitment near conspe-
cific adults and the maintenance of tree and shrub diversity in
a neotropical forest. Am Nat 140:261286
Condit R, Hubbell SP, Foster RB (1994) Density dependence in
two understory tree species in a neotropical forest. Ecology
75:671680
Condit R, Hubbell SP, Foster RB (1995) Mortality rates of 205
neotropical tree and shrub species and the impact of a severe
drought. Ecol Monogr 65:419439
Condit R, Hubbell SP, Foster RB (1996) Changes in tree species
abundance in a neotropical forest: impact of climate change.
J Trop Ecol 12:231256
Condit R, Ashton PS, Manokaran N, LaFrankie JV, Hubbell SP,
Foster RB (1999) Dynamics of the forest communities at
Pasoh and Barro Colorado: comparing two 50-ha plots. Philos
Trans R Soc Lond B 354:17391748.
Coomes DA, Grubb PJ (1998) Responses of juvenile trees to
above- and belowground competition in nutrient-starved Ama-
zonian rainforest. Ecology 79:768782
Coomes DA, Grubb PJ (2000) Impacts of root competition in for-
ests and woodlands: a theoretical framework and review of ex-
periments. Ecol Monogr 70:171207
Connell JH (1971) On the role of natural enemies in preventing
competitive exclusion in some marine animals and in rain for-
est trees. In: den Boer PJ, Gradwell GR (eds) Dynamics of
populations. Centre for Agricultural Publishing and Documen-
tation, Wageningen, The Netherlands, pp 298313
Connell JH (1978) Diversity in tropical rain forests and coral
reefs. Science 199:13021309
Connell JH, Tracey JG, Webb LJ (1984) Compensatory recruit-
ment, growth, and mortality as factors maintaining rain forest
tree diversity. Ecol Monogr 54:141164
Croat TB (1978) Flora of Barro Colorado Island. Stanford Univer-
sity Press, Stanford, Calif. USA
Denslow JS (1980) Gap partitioning among tropical rainforest
trees. Biotropica [Suppl] 12:4755
Denslow JS (1987) Tropical rainforest gaps and tree species diver-
sity. Annu Rev Ecol Syst 18:431451
Dirzo R, Miranda A (1991) Altered patterns of herbivory and di-
versity in the forest understory: a case study of the possible
consequences of contemporary defaunation. In: Price PW,
Lewinsohn PW, Fernandes GW, Benson WW (eds) Plant-ani-
mal interactions: evolutionary ecology in tropical and temper-
ate regions. Wiley, New York, pp 273287
Federov AA (1966) The structure of the tropical rain forest and
speciation in the humid tropics. J Ecol 54:111
Feinsinger P (1983) Coevolution and pollination. In: Futuyma DJ,
Slatkin M (eds) Coevolution. Sinauer, Sunderland, Mass. pp
282310
Fraver S, Brokaw NVL, Smith AP (1998) Delimiting the gap
phase in the growth cycle of a Panamanian forest. J Trop Ecol
14:673681
Gentry AH (1988) Changes in plant community diversity and
floristic composition on environmental and geographical gra-
dients. Ann Mo Bot Gard 75:134
Gentry AH, Dodson C (1987) Contribution of nontrees to species
richness of a tropical rain forest. Biotropica 19:149156
Gilbert GS, DeSteven D (1996) A canker disease of seedlings and
saplings of Tetragastris panamensis (Burseraceae) caused by
Botryosphaeria dothidea in a lowland tropical forest. Plant Dis
80:684687
Gilbert GS, Hubbell SP, Foster RB (1994) Density and distance-
to-adult effects of a canker disease of trees in a moist tropical
forest. Oecologia 98:100108
Givnish TJ (1999) On the causes of gradients in tropical tree di-
versity. J Ecol 87:193210
Grubb PJ (1977) The maintenance of species-richness in plant
communities: the importance of the regeneration niche. Biol
Rev 52:107145
Grubb PJ (1996) Rainforest dynamics: the need for new para-
digms. In: Edwards DS (ed) Tropical rainforest research cur-
rent issues. Kluwer, Dordrecht, pp 215233
Hamill DN, Wright SJ (1986) Testing the dispersion of juveniles
relative to adults: a new analytic method. Ecology 67:952957
Hammond DS, Brown VK (1998) Disturbance, phenology and life-
history characteristics: factors influencing distance/density-
dependent attack on tropical seeds and seedlings. In: Newbery
DM, Prins HHT, Brown ND (eds) Dynamics of tropical com-
munities. Blackwell, Oxford, pp 5178
Harms KE, Wright SJ, Caldern O, Hernndez A, Herre EA
(2000) Pervasive density-dependent recruitment enhances
seedling diversity in a tropical forest. Nature 404:493495
He F, Legendre P, LaFrankie JV (1997) Distribution patterns of
tree species in a Malaysian tropical rain forest. J Veg Sci
8:105114
Horvitz CC, Schemske DW (1990) Spatio-temporal variation in
insect mutualists of a neotropical herb. Ecology 71:10851097
Howe HF (1990) Survival and growth of juvenile Virola surina-
mensis in Panama: effects of herbivory and canopy closure.
J Trop Ecol 6:259280
Howe HF, Smallwood J (1982) Ecology of seed dispersal. Annu
Rev Ecol Syst 13:201228
Hubbell SP (1979) Tree dispersion, abundance, and diversity in a
tropical dry forest. Science 203:12991309
Hubbell SP (1980) Seed predation and the coexistence of tree spe-
cies in tropical forests. Oikos 35:214229
Hubbell SP (1997) A unified theory of biogeography and relative
species abundance and its application to tropical rain forests
and coral reefs. Coral Reefs 16:S9-S21
Hubbell SP, Foster RB (1986) Biology, chance, and history and
the structure of tropical rain forest tree communities. In:
Diamond J, Case TJ (eds) Community ecology. Harper and
Row, New York, pp 314329
Hubbell SP, Foster RB (1992) Short-term dynamics of a neotropi-
cal forest: why ecological research matters to tropical conser-
vation and management. Oikos 63:4861
Hubbell SP, Condit R, Foster RB (1990) Presence and absence of
density dependence in a neotropical tree community. Philos
Trans R Soc Lond B 330:269281
Hubbell SP, Foster RB, OBrien ST, Harms KE, Condit R,
Weschler B, Wright SJ, Loo de Lao S (1999) Light gap distur-
bances, recruitment limitation, and tree diversity in a neotropi-
cal forest. Science 283:554557
Hurtt GC, Pacala SW (1995) The consequences of recruitment
limitation: reconciling chance, history and competitive differ-
ences between plants. J Theor Biol 176:112
12
Huston MA (1994) Biological diversity. Cambridge University
Press, Cambridge
Janzen DH (1970) Herbivores and the number of tree species in
tropical forests. Am Nat 104:501528
Kitajima K (1994) Relative importance of photosynthetic traits
and allocation patterns as correlates of seedling shade toler-
ance of 13 tropical trees. Oecologia 98:419428
Kress WJ, Beach JH (1994) Flowering plant reproductive systems.
In: McDade LA, Bawa KS, Hespenheide HA, Hartshorn GS
(eds) La Selva, ecology and natural history of a tropical rain
forest. University of Chicago Press, Chicago, Ill. pp 161182
LaFrankie J (1996) Initial findings from Lambir: trees, soils and
community dynamics. Cent Trop For Sci 1995:5
Latham RE, Ricklefs, RE (1993) Continental comparisons of tem-
perate-zone tree species diversity. In: Ricklefs RE, Schluter D
(eds) Species diversity in ecological communities. University
of Chicago Press, Chicago, pp 294314
Lewis SL, Tanner EVJ (2000) Effects of aboveground and below-
ground competition on growth and survival of rain forest tree
seedlings. Ecology 81:25252538
Lieberman M, Lieberman D, Peralta R, Hartshorn GS (1995) Can-
opy closure and the distribution of tropical forest trees at La
Selva, Costa Rica. J Trop Ecol 11:161177
Lord J, Westoby M, Leishman M (1995) Seed size and phylogeny
in six temperate floras: constraints, niche conservatism, and
adaptation. Am Nat 146:349364
Marquis RJ, Young HJ, Braker HE (1986) The influence of un-
derstory vegetation cover on germination and seeding estab-
lishment in a tropical lowland wet forest. Biotropica 18:273
278
Martnez-Ramos M, Sarukhn JK, Piero DD (1988) The demog-
raphy of tropical trees in the context of forest gap dynamics:
the case of Astrocaryum mexicanum at Los Tuxtlas tropical
rain forest. In: Davy DJ, Hutchings MJ, Watkinson AR (eds)
Plant population ecology. Blackwell, Oxford, pp 293313
Molofsky J, Fisher BL (1993) Habitat and predation effects on
seedling survival and growth in shade-tolerant tropical trees.
Ecology 74:261265
Momose K, Yumoto T, Nagamitsu T, Kato M, Nagamasu H, Sakai
S, Harrison RD, Itioka T, Hamid AA, Inoue T (1998) Pollina-
tion biology in a lowland dipterocarp forest in Sarawak, Ma-
laysia. I. Characteristics of the plant-pollinator community in a
lowland dipterocarp forest. Am J Bot 85:14771501
Morley RJ (2000) Origin and evolution of tropical rainforests.
Wiley, New York
Muller-Landau M, Wright SJ, Calderon O, Hubbell SP, Foster RB
(in press) Assessing recruitment limitation: concepts, methods,
and case studies from a tropical forest. In: Levey DJ, Silva
WR, Galetti M (eds) Seed dispersal and frugivory: ecology,
evolution and conservation. CAB International, Oxford
Nathan R, Muller-Landau HC (2000) Spatial patterns of seed dis-
persal, their determinants and consequences for recruitment.
Trends Ecol Evol 15:278285
OBrien ST, Hubbell SP, Spiro P, Condit R, Foster RB (1995) Di-
ameter, height, crown, and age relationships in eight Neotropi-
cal tree species. Ecology 76:19261939
Oliveira AA de, Mori SA (1999) A central Amazonian terra firme
forest. I. High tree species richness on poor soils. Biodivers
Conserv 8:12191244
Orians GH (1982) The influence of tree-falls in tropical forests in
tree species richness. Trop Ecol 23:255279
Osunkoya OO, Ash JE, Hopkins MS, Graham AW (1992) Factors
affecting survival of tree seedlings in North Queensland rain-
forests. Oecologia 91:569578
Pacala SW, Canham CD, Saponara J, Silander JA, Kobe RK,
Ribbens E (1996) Forest models defined by field measure-
ments: estimation, error analysis and dynamics. Ecol Monogr
66:143
Packer A, Clay K (2000) Soil pathogens and spatial patterns of
seedling mortality in a temperate tree. Nature 404:278281
Palmer MW (1994) Variation in species richness: towards a unifi-
cation of hypotheses. Folia Geobot Phytotaxon 29:511530
Phillips OL, Hall P, Gentry AH, Sawyer SA, Vquez R (1994) Dy-
namics and species richness of tropical rain forests. Proc Natl
Acad Sci USA 91:29052809
Poulin B, Wright SJ, Lefebvre G, Caldern O (1999) Interspecific
synchrony and asynchrony in the fruiting phenologies of con-
generic bird-dispersed plants in Panama. J Trop Ecol 15:213
227
Prance GT (1977) Floristic inventory of the tropics: where do we
stand? Ann Mo Bot Gard 64:659684
Ricklefs RE (1977) Environmental heterogeneity and plant species
diversity: a hypothesis. Am Nat 111:376381
Roubik DW (1992) Loose niches in tropical communities: why are
there so few bees and so many trees? In: Hunter MD, Ohgushi
T, Price PW (eds) Effects of resource distribution on animal-
plant interactions. Academic Press, San Diego, pp 327354
Schupp EW (1992) The Janzen-Connell model for tropical tree di-
versity: population implications and the importance of spatial
scale. Am Nat 140:526530
Sagers CL, Coley PD (1995) Benefits and costs of defense in a
neotropical shrub. Ecology 76:18351843
Silva Matos DM, Freckleton RP, Watkinson AR (1999) The role
of density dependence in the population dynamics of a tropical
palm. Ecology 80:26352650
Snow DW (1965) A possible selective factor in the evolution of
fruiting seasons in tropical forest. Oikos 15:274281
Sokal RR, Rohlf FJ (1995) Biometry, 3rd edn. Freeman, New
York
Sollins P (1998) Factors influencing species composition in tropi-
cal lowland rain forest: does soil matter? Ecology 79:2330
Stevens GC (1992) The elevational gradient in altitudinal range:
an extension of Rapoports latitudinal rule to altitude. Am Nat
140:893911
Stiles FG (1977) Coadapted competitors: the flowering seasons of
hummingbird-pollinated plants in a tropical forest. Science
198:11771178
Svenning J-C (1999) Microhabitat specialization in a species-rich
palm community in Amazonian Ecuador. J Ecol 87:5565
Svenning J-C (2001) On the role of microenvironmental heteroge-
neity in the ecology and diversification of neotropical rain-for-
est palms (Arecaceae). Bot Rev 67:153
Terborgh J (1973) On the notion of favorableness in plant ecology.
Am Nat 107:481501
Terborgh J, Wright SJ (1994) Effects of mammalian herbivores on
plant recruitment in two neotropical forests. Ecology 75:1829
1833
Ter Steege H (1994) Flooding and drought tolerance of two Mora
species segregated along a soil hydrological gradient in the
tropical rain forest of Guyana. Oecologia 100:356367
Ter Steege H, Sabatier D, Castellanos H, Van Andel T,
Duivenvoorden J, Oliveira AA de, EK R, Lilwah R, Maas P,
Mori S (2000) An analysis of the floristic composition and di-
versity of Amazonian forests including those of the Guiana
Shield. J Trop Ecol 16:801828
Tilman D (1994) Competition and biodiversity in spatially struc-
tured habitats. Ecology 75:216
Tilman D, Pacala S (1993) The maintenance of species richness in
plant communities. In: Ricklefs RE, Schluter D (eds) Species
diversity in ecological communities. University of Chicago
Press, Chicago, pp 1325
Uhl C, Clark K, Dezzeo N, Maquirino P (1988) Vegetation dy-
namics in Amazonian treefall gaps. Ecology 69:751763
Valencia R, Balslev H, Paz y Mio G (1994) High tree alpha-
diversity in Amazonian Ecuador. Biodivers Conserv 3:2128
Van Steenis CGGJ (1969) Plant speciation in Malesia, with special
reference to the theory of non-adaptive saltatory evolution.
Biol J Linn Soc 1:97133
Webb CO, Peart DR (1999) Seedling density dependence pro-
motes coexistence of Bornean rain forest trees. Ecology 80:
20062017
Welden CW, Hewett SW, Hubbell SP, Foster RB (1991) Sapling
survival, growth, and recruitment: relationship to canopy
height in a neotropical forest. Ecology 72:3550
13
Whitmore TC, Brown ND (1996) Dipterocarp seedling growth in
rain forest canopy gaps during six and a half years. Philos
Trans R Soc Lond B 351:11951203
Whitmore TC, Burslem DFRP (1998) Major disturbances in tropi-
cal rain forests. In: Newbery DM, Prins HHT, Brown ND (eds)
Dynamics of tropical communities. Blackwell, Oxford, pp
549565
Wills C, Condit R (1999) Similar non-random processes maintain
diversity in two tropical rainforests. Proc R Soc Lond B 266:
14451452
Wills C, Condit R, Foster RB, Hubbell SP (1997) Strong density-
and diversity-related effects help to maintain tree species di-
versity in a neotropical forest. Proc Natl Acad Sci USA 94:
12521257
Wright SJ (1999) Plant diversity in tropical forests. In: Pugnaire
FI, Valladares F (eds) Handbook of functional plant ecology.
Dekker, New York, pp 449472
Wright SJ, Caldern O (1995) Phylogenetic constraints on tropical
flowering phenologies. J Ecol 83:937948
Wright SJ, Machado JL, Mulkey S, Smith AP (1992) Drought ac-
climation among tropical forest shrubs (Psychotria, Rubiac-
eae). Oecologia 89:457463
Wright SJ, Carrasco C, Caldern O, Paton S (1999) The El Nio
Southern Oscillation, variable fruit production and famine in a
tropical forest. Ecology 80:16321647
Wu X, McCormick JF, Busing RT (1999) Growth pattern of Picea
rubens prior to canopy recruitment. Plant Ecol 140:245253
Yavitt JB, Wright SJ (in press) Seasonal drought and irrigation ef-
fects on fine root dynamics in a tropical moist forest on Barro
Colorado Island, Panama. Biotropica
Yu DW, Terborgh JW, Potts MD (1998) Can high tree species rich-
ness be explained by Hubbells null model? Ecol Lett 1:193
199
Zhang D, Lin K (1997) The effects of competitive asymmetry on
the rate of competitive displacement: how robust is Hubbells
community drift model? J Theor Biol 188:361367
14