OIKOS 112: 131
/137, 2006

A bounded null model explains juvenile tree community structure

along light availability gradients in a temperate rain forest
Christopher H. Lusk, Robin L. Chazdon and Glenn Hofmann

Lusk, C. H., Chazdon, R. L. and Hofmann, G. 2006. A bounded null model explains
juvenile tree community structure along light availability gradients in a temperate rain
forest.
/ Oikos 112: 131
/137.
Ecologists have proposed that tree species may coexist by specialising on light
environments associated with gaps of different sizes. Remarkably few communitylevel studies, however, have actually examined juvenile tree distributions along light
availability gradients. Here we describe distributions of juvenile trees in relation to
canopy openness in a temperate rainforest, and test the hypothesis that competitive
sorting causes coexisting species to overlap less in light environment occupancy than
would be expected by chance. Average overlap of species interquartile ranges on the
canopy openness gradient was tested against a bounded domain null model of
community structure which used range-size criteria to constrain random placement of
species optima. Microsite availability was strongly skewed towards low light, with 43%
of microsites occurring at B/5% canopy openness. We therefore transformed canopy
openness values to ranks, so that equal intervals on the transformed gradient
represented equal areas of microsite availability. We then calculated the interquartile
range (25
/75%) of sample ranks occupied by juveniles of each species. About half the
assemblage was non-randomly distributed in relation to canopy openness, providing
evidence of niche expression. Average overlap of species interquartile ranges did not
depart significantly from that predicted by the bounded null model, indicating that
community structure in relation to canopy openness was mainly explained by a middomain effect. As predicted by the null model, species interquartile range mid-points
were concentrated in the centre of the rank-transformed gradient, and species richness
(overlap of interquartile ranges) peaked close to the median light environment. Most
species therefore had intermediate light requirements. The apparent lack of constraints
on pairwise overlap suggest that differences in light use are not a prerequisite for tree
species coexistence. As far as we are aware, this is the first study to identify a middomain effect on a resource availability axis.
C. H. Lusk, Depto de Botanica, Univ. de Concepcion, Concepcion, Chile
(clusk@udec.cl). Present address for CL: Dept of Biological Sciences, Maquarle
Univ., NSW 2109, Australia.
/ R. L. Chazdon, Dept of Ecology and Evolutionary
Biology, Univ. of Connecticut, Storrs, CT, USA.
/ G. Hofmann, Depto de Estadstica,
Univ. de Concepcion, Concepcion, Chile.

Ecologists have proposed that tree species may coexist by

specialisation of juveniles on different light environments
associated with canopy openings of different sizes
(Ricklefs 1977, Denslow 1980). Ample scope for such
specialization could be provided by the continuous
spatial and temporal variation of light availability in
forest understories (Lieberman et al. 1995), ranging from

large canopy gaps resulting from recent multiple treefalls, to heavily shaded understoreys beneath dense
canopies, where light availability may be B/1% of that
above the canopy (Chazdon and Fetcher 1984). This
niche specialisation hypothesis implies that assembly of
forest communities is shaped by strong interspecific
competition for regeneration opportunities, leading

Accepted 7 June 2005

Copyright # OIKOS 2006
ISSN 0030-1299
OIKOS 112:1 (2006)

131

either to competitive exclusion among species with

strong functional similarities, or else to evolutionary or
ecological divergence (MacArthur and Levins 1967).
Despite the age of this hypothesis, and recent criticism
of its assumptions (Brown and Jennings 1998, Hubbell
et al. 1999, Brokaw and Busing 2000, Wright 2002), tree
species distributions in relation to understorey light
environments remain poorly documented. A number of
studies have examined small subsets of assemblages, but
we are aware of only one study of a whole community
(Lieberman et al. 1995). Furthermore, despite the
continuous nature of variation in forest light environments (Lieberman et al. 1989), most studies have focused
on comparing abundance or performance of juvenile
trees across gap/non-gap dichotomies, or in gaps of
different size classes (Brokaw 1987, Popma and Bongers
1988, Welden et al. 1991, Sipe and Bazzaz 1995,
Whitmore and Brown 1996, Busing and White 1997,
Lusk and Smith 1998). Despite technical and methodological advances, we still know little about breadth
or specificity of species distributions across light availability gradients, or the extent to which co-occurring
species overlap in their occupancy of light environments
(Lieberman et al. 1995).
Appropriate null models of community structure
(Gotelli and Graves 1996) are a potentially valuable
tool for testing hypotheses about light gradient partitioning. How would we expect species to be distributed
in relation to light availability if competition for light did
not play a major role in structuring the community? A
continuous axis of resource availability (such as light in
forests) can be viewed as a bounded domain, analogous
to geographic gradients (Colwell and Lees 2000, Colwell
et al. 2004). The extinction of light towards the lower
end of this gradient represents a hard boundary for
photosynthetic organisms. The upper boundary is more
diffuse, but will be determined by the size of the largest
canopy openings that occur within the forest mosaic.
The geometric constraints imposed by domain boundaries must therefore inevitably influence patterns of
species overlap (and possibly also range mid-points) on
light gradients, even if competition is also a major
influence on community structure. Bounded null models
of community structure are therefore appropriate to
quantify and understand species distribution patterns
on light gradients.
Here we test for partitioning of light environments
among an assemblage of 16 tree species in an old-growth
temperate rain forest. Average pairwise overlap of the
distributions of species juveniles on a canopy openness
gradient is compared with a bounded null model of
community structure. Our null model assumes that
proximity to domain boundaries depresses species density through its influence on range size (Fig. 1). Species
with optima close to domain boundaries are necessarily
limited to a narrow range, and narrow ranges (on any
132

Fig. 1. A bounded null model of community structure modified

from the range-size model of Grytnes (2003), which assumes
that proximity to domain boundaries depresses species density
through its influence on range size. The triangle shows the
feasible range for interquartile range mid-points at a given
interquartile range breadth. The small circles within the triangle
represent the mid-points and breadth of the empirical interquartile ranges of our 16 species. Our null model resamples the
empirical interquartile range breadths with replacement and
chooses the midpoints from rectangular distributions on the
corresponding feasible interval.

part of the gradient) may predispose species to local

extinction in dynamic understorey light environments.
The null model therefore predicts that both density of
species optima and species overlap should be highest in
the domain centre (intermediate light environments),
and lowest close to the domain boundaries. This
mechanism corresponds to the range-size model
proposed by Grytnes (2003) for geographic domains.

Material and methods

Sampling
Sampling was carried out in the low-altitude forests
(350
/440 m a.s.l.) of Parque Nacional Puyehue (40839?S,
72811?W) located in the western foothills of the Andean
range in south-central Chile. This area experiences a
maritime temperate climate, with an average annual
precipitation of around 3500 mm (Almeyda and Saez
1958). The old-growth rain forest at this altitude on the
western foothills of the Andes is comprised exclusively of
broad-leaved evergreen species (Table 1).
Distributions of juvenile trees were quantified in
relation to canopy openness measurements made with
a pair of LAI-2000 canopy analysers (Li-Cor, Lincoln,
Nebraska). One instrument was used to take measurements at each sampling point, while the other, placed at
the centre of a 2 ha clearing, was programmed to take
readings at 30 s intervals. Integration of data from the
two instruments enabled estimation of percentage diffuse
irradiance at each sampling point within the forest,
OIKOS 112:1 (2006)

Table 1. Species comprising old-growth stands in lowland forest at Parque Nacional Puyehue, showing number of juveniles
represented in samples, and results of Kolmogorov
/Smirnov test comparing light environments of juveniles with the distribution of
light environments of the whole forest. P-values B/0.05 indicate that juveniles of a given species are non-randomly distributed in
relation to canopy openness.
Species
Aristotelia chilensis (Mol.) Stuntz
Amomyrtus luma (Molina) Legrand
Aextoxicon punctatum R. et P.
Azara lanceolata Hook. f.
Caldcluvia paniculata (Cav.) D.Don.
Dasyphyllum diacanthoides (Less.) Cabr.
Eucryphia cordifolia Cav.
Fuchsia magellanica Lam.
Gevuina avellana Mol.
Luma apiculata (DC.) Burret
Lomatia ferruginea (Cav.) R. Br.
Laurelia philippiana Looser.
Myrceugenia planipes (H. et A.)
Nothofagus dombeyi (Mirb.) Oerst.
Rhaphithamnus spinosus (A. L. Juss)
Weinmannia trichosperma Cav.

Family
Elaeocarpaceae
Myrtaceae
Aextoxicaceae
Flacourtiaceae
Cunoniaceae
Asteraceae
Cunoniaceae
Onagraceae
Proteaceae
Myrtaceae
Proteaceae
Atherospermataceae
Myrtaceae
Fagaceae
Verbenaceae
Cunoniaceae

equivalent to % canopy openness over the quasi-hemispherical (1488) field of view perceived by the LAI-2000
sensors. Measurements with the LAI-2000 are a good
surrogate of spatial variation in mean daily photosynthetic photon flux density within a stand (Machado and
Reich 1999, Nicotra et al. 1999).
Sampling was carried out on a series of parallel
transects run through old-growth stands, all transects
beginning from access tracks. At sample points spaced at
random intervals (2 to 10 m apart) along transects,
readings were taken at 100 and 200 cm height with the
LAI-2000. Juveniles 50
/200 cm tall were recorded in a
circular quadrat of 1 m diameter, centred on the sample
point. Juveniles B/100 cm tall were referred to the light
level measured at 100 cm height. Those ]/100 cm tall
were referred to the light level measured at 200 cm
height. A total of 1608 points were sampled, yielding a
total of 1585 juveniles, with species representation
ranging from 19 to 249 individuals (Table 1). Only tree
and large shrub species normally represented by adults
within the old-growth forest matrix were considered
(Table 1). This meant excluding Embothrium coccineum ,
a small tree which, although represented by a few
juveniles inside the forest, is more characteristic of
open sites and forest margins. Lomatia hirsuta , a species
which appeared in only two samples and was not
recorded as an adult, was also discarded, as was
Pseudopanax laetevirens, which usually establishes as
an epiphyte but occasionally appears on the forest floor.

Distributions of individual species

We used a Kolmogorov
/Smirnov test to detect nonrandomness in the distributions of each species. This test
measured the goodness-of-fit of the distribution of light
environments inhabited by juveniles of each species to
OIKOS 112:1 (2006)

Symbol

Sample size

Kolmogorov
/Smirnov test

(P-value)

Ac
Al
Ap
Az
Cp
Dd
Eu
Fm
Ga
La
Lf
Lp
Mp
Nd
Rs
Wt

119
164
113
123
158
89
99
46
41
107
68
43
249
19
93
26

B/.001
0.081
0.001
0.001
0.134
0.002
0.605
B/.001
0.593
0.296
0.237
0.062
B/.001
0.052
0.008
0.384

the empirical cumulative distribution of the forest light

environments derived from our samples.

Gradient structure
Consideration of the relative availability of different light
environments is important for questions about light
gradient partitioning. In most forests, the frequency
distribution of light environments is highly skewed
towards microsites at B/ca 5% canopy openness
(Montgomery and Chazdon 2002). Regular intervals
on an arithmetic scale of diffuse light availability therefore do not represent equal areas of microsite availability
As a result, it would probably be unrealistic to look for
evidence of light gradient partitioning by testing for9/
regular spacing of species along an arithmetic scale of
canopy openness. Given the vast amount of space
potentially suitable for establishment of shade-tolerant
species, relatively minor differences in light requirements
of such taxa could be sufficient to permit coexistence of
sizeable populations of several species, and hence prevent competitive exclusion. In contrast, microsites with
/25% canopy openness are rare, implying that larger
differences in light requirements would be required for
species coexistence in such environments, and that few
light-demanding species will be able to maintain viable
populations in old-growth forests.
One way to incorporate the relative availability of
different light environments is to work with rank order
of canopy openness at microsites, instead of actual
canopy openness values. Equal intervals on a ranktransformed gradient represent equal areas of microsite
availability. Accordingly, we ordered our 3170 samples
by light environment (% canopy openness). The median
and interquartile range (25
/75%) of each species
distribution on this gradient was then calculated. The
133

50
40

Frequency (%)

interquartile range was chosen as a robust indicator of

the spread of species distributions on the transformed
gradient. This parameter distinguished between species
better than complete ranges, which spanned most of the
gradient for many taxa. Furthermore, interquartile
range breadth on the rank transformed gradient was
not correlated with sample size (P /0.79), confirming its
robustness.

30
20
10

Null model of community structure

If species coexist by partitioning light environments,
then species interquartile ranges should show lower
average pairwise overlaps than those arising from
suitable null models. We compared average pairwise
overlap of empirical species interquartiles on the ranktransformed canopy openness gradient with a bounded
null model based on the range-size mechanism
proposed by Grytnes (2003).
We generated 1000 simulations from our bounded null
model by means of a resampling approach. As the
assemblage consisted of 16 species, 16 interquartile range
breadths were randomly chosen with replacement from
our empirical data (Colwell et al. 2004). For each
interquartile range breadth, the midpoint was randomly
assigned from a rectangular distribution on the corresponding feasible interval within the triangle shown in
Fig. 1. This interval was widest for narrow interquartile
range breadths, and most constrained for broad interquartiles. The endpoints A and B of the base of the
triangle represent the bounds within which all interquartile ranges must fit. A and B were determined such
that both the leftmost and rightmost endpoints of the 16
simulated interquartile ranges coincided on average with
the corresponding endpoints of the empirical interquartile ranges. Our use of the empirical distribution of
interquartile range breadths is preferable to using a
theoretical interquartile range breadth distribution,
because the latter could lead to unrealistic scenarios in
the null model, and hence underestimate the role of the
mid-domain effect (Colwell et al. 2004).
All analyses were carried out using S-PLUS (Insightful, Seattle, USA).

0
0

10

20

30

40

50

60

Canopy openness (%)

Fig. 2. Distribution of understorey light environments in oldgrowth rainforest at Parque Nacional Puyehue, south-central
Chile.

differed significantly from the overall distribution

of microsite light availability. Marginally significant
values were obtained for two other species with small
sample sizes of B/50 individuals (Laurelia philippiana ,
Nothofagus dombeyi ), suggesting that more extensive
sampling might show significant pattern for these
species too.
Interquartile ranges of the 16 species spanned light
environments ranging from 1.7 to 26.4% canopy
openness, but were concentrated in the centre of the
rank-transformed gradient (Fig. 3). Individual interquartiles occupied between 27 and 47% of this span on
the domain. Interquartile range breadth was marginally
negatively correlated with the distance of interquartile
mid-point from the domain median (P /0.06). Although

Results
Availability of microsites was strongly skewed towards
low light, with nearly half of samples occurring at B/5%
canopy openness (Fig. 2).
About half of the assemblage was non-randomly
distributed in relation to canopy openness (Table 1).
The Kolmogorov
/Smirnov test gave P-values B/0.05
for seven out of 16 species, showing that the distributions of their juveniles in relation to canopy openness
134

Fig. 3. Distribution of species interquartile ranges on a

rank-transformed canopy openness gradient. Species symbols
(Table 1) are placed at interquartile mid-points.
OIKOS 112:1 (2006)

mid-points of species interquartile ranges were concentrated towards the centre of the canopy openness
gradient (Fig. 3), randomisation showed that this pattern
did not depart significantly from that expected under our
bounded null model assuming independence among
species (P /0.087, one-tailed test).
Average pairwise overlap of species interquartile
ranges did not depart significantly from that expected
according to the bounded null model of community
structure (Fig. 4). Species richness (as represented by
interquartile ranges) peaked at ca 7% canopy openness
(close to the median light environment: Fig. 5), and this
peak was of a similar magnitude to that predicted by the
null model.

Discussion
Niche expression and species coexistence
Species individually showed evidence of niche expression. About half the assemblage was preferentially
associated with certain light environments (Table 1). In
contrast, Lieberman et al. (1995) reported that the
juveniles of 86.5% of tree species in a tropical forest
were randomly distributed in relation to canopy openness. This divergence is at least partly attributable to
differences in sample sizes and methods. Whereas sample
sizes of most (11/16) tree species exceeded 50 individuals
in the floristically poor temperate forest that we studied
(Table 1), most (95/104) species in the diverse tropical
forest studied by Lieberman et al. (1995) were represented by B/50 stems. Moreover, although Lieberman
et al. (1995) claimed that most species at La Selva
occupied light environments indiscriminately, their
statistical analysis does not appear to distinguish
300

250

Frequency

200

150

100

50

0
0.18

0.20

0.22

0.24

0.26

0.28

0.30

0.32

0.34

0.36

Average pairwise overlap

Fig. 4. Null distribution of average pairwise overlap of species

interquartile ranges on the canopy openness gradient, from
1000 simulations. Dotted vertical line shows empirical value of
0.282 (P/0.51, two-tailed test).
OIKOS 112:1 (2006)

Fig. 5. Empirical variation in species richness in relation to

canopy openness (solid line) coincides closely with that predicted by a bounded null model (broken line shows average of
1000 simulations). Species richness is computed as the number
of interquartile ranges overlapping at each point on the ranktransformed gradient.

between species that might be significantly associated

with central regions of the canopy openness gradient
(e.g. Azara lanceolata and Dasyphyllum diacanthoides in
our study: Table 1, Fig. 3) and those genuinely showing
no pronounced pattern (e.g. Eucryphia cordifolia ). The
diversity of niche expression that we found (Fig. 4) is
consistent with studies showing rank order changes in
performance of co-occurring species along light availability gradients in other forest communities (Pacala
et al. 1994, Kobe et al. 1995, Montgomery and Chazdon
2002).
The species light requirements indicated by the
present study (Fig. 3) closely parallel those found by
Figueroa and Lusk (2001) in another temperate rainforest in the same region. Ten species are common to
both studies, and interquartile range mid-points of these
species in the present study are significantly positively
correlated with the mean light environments occupied by
juveniles in the earlier study on Isla Grande de Chiloe
(R /0.83, P/0.0025). This close correspondence suggests that our sampling methods yielded a faithful
representation of most species light requirements, and
that shade tolerance hierarchies are relatively stable
throughout the region.
Despite the diversity of niche expression on the
canopy openness gradient, we found no evidence that
differences in this niche dimension are necessary for
species coexistence. Average pairwise overlap on the
canopy openness gradient did not depart significantly
from that predicted by a null model which incorporated boundary constrains on positions of species
range mid-points (Fig. 4), indicating that the interquartile ranges of the 16 species were essentially
independently distributed in relation to each other.
135

This finding does not appear to be consistent with the

hypothesis that partitioning of light environments is
the principal mechanism of tree species coexistence
(Ricklefs 1977, Denslow 1980).
Although forest trees have traditionally been dichotomised as pioneer or shade-tolerant (Bazzaz 1984,
Swaine and Whitmore 1988, Brown and Jennings 1998),
this is not the first empirical study to produce direct or
indirect evidence that most species in humid forests are
of intermediate shade tolerance. Figueroa and Lusk
(2001) compared the light environments occupied by
juveniles of 26 tree and shrub species in a temperate
rainforest on Isla Grande de Chiloe, Chile. Although
they presented no data on species range overlaps, they
found that the mean light environments of 13 out of 26
species occurred at between 4 and 12% canopy openness.
Similarly, the data of Lieberman et al. (1995) also
suggest a mid-gradient concentration of species average
light environments, although their analysis does not
develop this point. Although Wright et al. (2003) did not
directly measure juvenile light environments, their study
of tree population size structures on Barro Colorado
Island, Panama, suggested that most of the 73 most
abundant canopy species have intermediate light requirements.
Relating the distribution of juvenile trees to present
canopy openness has been criticised as unrealistic in
view of the dynamic nature of forest light environments
(Kobe 1999). A likely consequence for our study
is overestimation of species real amplitude on the
canopy openness gradient, as carbohydrate reserves
probably enable juvenile trees to persist for some
time in light environments in which their net carbon
gain is negative. However, there are two reasons
for believing that this phenomenon did not influence
our conclusions. Firstly, juvenile tree distributions
provided clear evidence of niche expression in
about half the assemblage (Table 1), and of niche
differences among species (Fig. 3). Secondly, any
bias of this sort that might be present in our field data
was accounted for in our null model, which resampled
our empirical distribution of interquartile range
breadths.
Niche-based theories of species coexistence are
complemented, rather than completely supplanted, by
explanations that emphasise the role of chance factors
and the sessile habit in impeding competitive exclusion
among plant species with similar requirements and
similar competitive abilities (Brown and Jennings 1998,
Hubbell et al. 1999, Brokaw and Busing 2000). The
vagaries of seed dispersal, germination and seedling
establishment reduce the potential pool of contenders
for a given microsite, and Hubbell et al. (1999)
estimated that an ultimately successful juvenile tree
interacts directly with an average of only six to seven
neighbours during recruitment. Light requirements will
136

clearly play a role in determining the pool of species

that potentially comprise this neighbourhood (Fig. 3),
but random shuffling by recruitment limitation is likely
to be a major determinant of the identity of individual
neighbours. Recruitment limitation might therefore
enable coexistence of large numbers of functionally
similar species, provided these do not differ greatly in
competitive ability. The development of more sophisticated models of species coexistence, explicitly integrating the roles of both niche and chance factors in a
biologically realistic way, is a major future challenge
for plant ecologists.

The mid-domain effect

While possible partitioning of other resources remains
relatively unexplored, our results suggest that boundary
constraints on range location could largely explain
variation in species richness along the canopy openness
gradient (Fig. 1). The empirical distribution of species
richness along the gradient closely matched a null model
which, by resampling the empirical distribution of
interquartile range breadths, implicitly used range
breadth criteria to constrain random placement of
species on a bounded domain (Grytnes 2003). Species
interquartile mid-points were concentrated in the centre
of the rank-transformed canopy openness gradient
(Fig. 3), and species overlap increased steeply towards
the gradient centre, with interquartile ranges of 14 out of
16 species overlapping close to the domain median
(Fig. 5).
Although Colwell and Lees (2000) mention the
likely occurrence of mid-domain effects on nongeographic domains, as far as we are aware, this
pattern has not been explicitly identified on a resource
gradient before. However, it seems appropriate to
treat light gradients as bounded domains, and
the evidence discussed above suggests similar effects in
other forests (Lieberman et al. 1995, Figueroa and
Lusk 2001). Gradients of soil water availability to
plants can also probably be regarded as bounded
domains
/ although some studies focus on a relatively
narrow mesic range of habitats where boundary constraints are probably not relevant (Silvertown et al.
1999). However, these gradients, being more-or-less
stable features of landscapes, and usually occurring at
much broader scales than forest light mosaics, are
closer in nature to the classic geographic domains to
which bounded null models were originally applied
(Colwell and Hurtt 1994, Lyons and Willig 1997, Lees
et al. 1999).
Acknowledgements
/ We thank FONDECYT for support
through grant 1980084, Mylthon Jimenez and Patricia
Inostroza for help with fieldwork, and Rob Colwell for his
comments on a draft manuscript. This paper forms part of the
research and outreach activities of the Millennium Center for
OIKOS 112:1 (2006)

Advanced Studies in Ecology and Research on Biodiversity

supported by Grant No. P02-051-F ICM.

References
Almeyda, E. and Saez, F. 1958. Recopilacion de datos
climaticos de Chile y mapas sinopticos respectivos.
/ Ministerio de Agricultura, Chile.
Bazzaz, F. A. 1984. Dynamics of wet tropical forest and their
species strategies.
/ In: Medina, E., Mooney, H. A. and
Vazquez-Yanes, C. (eds), Physiological ecology of plants of
the wet tropics. Junk, pp. 233
/243.
Brokaw, N. V. L. 1987. Gap-phase regeneration of three pioneer
tree species in a tropical forest.
/ J. Ecol. 75: 9
/19.
Brokaw, N. and Busing, R. 2000. Niche versus chance and tree
diversity in forest gaps.
/ Trends Ecol. Evol. 15: 183
/188.
Brown, N. D. and Jennings, S. 1998. Gap-size niche differentiation by tropical rainforest trees: a testable hypothesis or a
broken-down bandwagon?
/ In: Newbery., D. M., Prins, H.
H. T. and Brown, N. (eds), Dynamics of tropical communities. Blackwell Science Press, pp. 79
/94.
Busing, R. T. and White, P. S. 1997. Species diversity and smallscale disturbance in an old-growth temperate forest: a
consideration of gap partitioning concepts.
/ Oikos 78:
562
/568.
Chazdon, R. L. and Fetcher, N. 1984. Photosynthetic light
environments in a lowland tropical rain-forest in Costa
Rica.
/ J. Ecol. 72: 553
/564.
Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients
in species richness and a spurious Rapoport effect.
/ Am.
Nat. 144: 570
/595.
Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect:
geometric constraints on the geography of species richness.

/ Trends Ecol. Evol. 15: 70
/76.
Colwell, R. K., Rahbek, C. and Gotelli, N. 2004. The middomain effect and species richness patterns: what have we
learned so far?
/ Am. Nat. 163: E1
/E23.
Denslow, J. S. 1980. Gap partitioning among tropical rain forest
trees.
/ Biotropica (Suppl.) 12: 47
/55.
Figueroa, J. A. and Lusk, C. H. 2001. Germination requirements and seedling shade tolerance are not correlated in a
Chilean temperate rain forest.
/ New. Phytol. 152: 483
/489.
Gotelli, N. J. and Graves, G. R. 1996. Null models in ecology.
/ Smithsonian Inst. Press.
Grytnes, J. A. 2003. Ecological interpretations of the middomain effect.
/ Ecol. Lett. 6: 883
/888.
Hubbell, S. P., Foster, R. B., OBrien, S. T. et al. 1999. Light-gap
disturbances, recruitment limitation, and tree diversity in a
Neotropical forest.
/ Science 283: 554
/557.
Kobe, R. K. 1999. Light gradient partitioning among tropical
tree species through differential seedling mortality and
growth.
/ Ecology 80: 187
/201.
Kobe, R. K., Pacala, S. W., Silander, J. A. et al. 1995. Juvenile
tree survivorship as a component of shade tolerance.
/ Ecol.
Appl. 5: 517
/532.
Lees, D. C., Kremen, C. and Andriamampianina, L. 1999. A
null model for species richness gradients: bounded range

overlap of butterflies and other rainforest endemics in

Madagascar.
/ Biol. J. Linn. Soc 67: 529
/584.
Lieberman, M., Lieberman, D. and Peralta, R. 1989. Forests are
not just Swiss cheese: canopy stereogeometry of non-gaps in
tropical forests.
/ Ecology 70: 550
/552.
Lieberman, M., Lieberman, D., Peralta, R. et al. 1995.
Canopy closure and the distribution of tropical forest tree
species at La Selva, Costa Rica.
/ J. Trop. Ecol. 11: 161
/
178.
Lusk, C. H. and Smith, B. 1998. Life history differences and tree
species coexistence in an old-growth New Zealand rainforest.
/ Ecology 79: 795
/806.
Lyons, S. K. and Willig, M. R. 1997. Latitudinal patterns of
range size: methodological concerns and empirical evaluations for New World bats and marsupials.
/ Oikos 79: 568
/
580.
MacArthur, R. H. and Levins, R. 1967. The limiting similarity,
convergence and divergence of coexisting species.
/ Am.
Nat. 101: 377
/85.
Machado, J. L. and Reich, P. B. 1999. Evaluation of several
measures of canopy openness as predictors of photosynthetic photon flux density in a forest understorey.
/ Can. J.
For. Res. 29: 1438
/1444.
Montgomery, R. A. and Chazdon, R. L. 2002. Light gradient
partitioning by tropical tree seedlings in the absence of
canopy gaps.
/ Oecologia 131: 165
/174.
Nicotra, A. B., Chazdon, R. L. and Iriarte, S. V. B. 1999. Spatial
heterogeneity of light and woody seedling regeneration in
tropical wet forests.
/ Ecology 80: 1908
/1926.
Pacala, S. W., Canham, C. D., Silander, J. A. Jr. et al. 1994.
Sapling growth as a function of resources in a north
temperate forest.
/ Can. J. For. Res. 24: 2172
/2183.
Popma, J. and Bongers, F. 1988. The effect of canopy gaps on
growth and morphology of seedling of rain forest species.

/ Oecologia 75: 625
/632.
Ricklefs, R. E. 1977. Environmental heterogeneity and plant
species diversity: an hypothesis.
/ Am. Nat. 111: 376
/381.
Silvertown, J., Dodd, M. E., David, J. et al. 1999. Hydrologically defined niches reveal a basis for species richness in
plant communities.
/ Nature 400: 61
/63.
Sipe, T. W. and Bazzaz, F. A. 1995. Gap partitioning among
maples (Acer ) in central New England: survival and growth.

/ Ecology 76: 1587
/1602.
Swaine, M. D. and Whitmore, T. C. 1988. On the definition of
ecological species groups in tropical rain forest.
/ Vegetatio
75: 81
/86.
Welden, C. W., Hewett, S. W., Hubbell, S. P. et al. 1991. Sapling
survival, growth and recruitment: relationships to canopy
height in a neotropical rainforest.
/ Ecology 72: 35
/50.
Whitmore, T. C. and Brown, N. D. 1996. Dipterocarp seedling
growth in rain forest canopy gaps during six and a half
years.
/ Philos. Trans. R. Soc. B 351: 1195
/1204.
Wright, S. J. 2002. Plant diversity in tropical forests: a review
of mechanisms of species coexistence.
/ Oecologia 130: 1
/
14.
Wright, S. J., Muller-Landau, H. C., Condit, R. et al. 2003.
Gap-dependent recruitment, realized vital rates, and size
distributions of tropical trees.
/ Ecology 84: 3174
/3185.

Subject Editor: Jane Memmott

OIKOS 112:1 (2006)

137