Ecology, 97(12), 2016, pp.

33263336
2016 by the Ecological Society of America

How mammalian predation contributes to tropical tree

community structure
C. E. TIMOTHY PAINE,1,5 HARALD BECK,2 AND JOHN TERBORGH3,4
1
Biological and Environmental Sciences, University of Stirling, Stirling, FK9 4LA United Kingdom
Department of Biological Sciences, Towson University, 8000 York Road, Towson, Maryland 21252 USA
3
Duke University Center for Tropical Conservation, Nicholas School of the Environment and Earth Sciences, PO Box 90318,
Durham, North Carolina 27708 USA
4
Florida Museum of Natural History, University of Florida, Gainesville, Florida 32611 USA
2

Abstract. The recruitment of seedlings from seeds is the key demographic transition for
rain forest trees. Though tropical forest mammals are known to consume many seeds, their
effects on tree community structure remain little known. To evaluate their effects, we monitored
8,000 seeds of 24 tree species using exclosure cages that were selectively permeable to three size
classes of mammals for up to 4.4 years. Small and medium-bodied mammals removed many
more seeds than did large mammals, and they alone generated beta diversity and negative
density dependence, whereas all mammals reduced diversity and shaped local species
composition. Thus, small and medium-bodied mammals more strongly contributed to
community structure and promoted species coexistence than did large mammals. Given that
seedling recruitment is seed limited for most species, alterations to the composition of the
community of mammalian seed predators is expected to have long-term consequences for tree
community structure in tropical forests.
Key words: agouti; beta diversity; defaunation; negative density dependence; peccary ; Peru; seed predation;
seed size; species composition.

INTRODUCTION
The recruitment of seedlings from seeds is the key
demographic transition for trees in tropical forests.
Mortality rates are not only greater during this than any
other ontogenetic stage, but they are also the most predictable and species specific, often caused by host-specific
natural enemies (Terborgh 2012). Mortality patterns
become increasingly stochastic over ontogeny (Green
et al. 2014). Thus, to understand the determinants of
tropical tree community structure, one must study
seedling recruitment. Though their relative importance
continues to be debated, it is certain that terrestrial
mammals, pathogenic fungi and herbivorous insects play
strong roles in this transition, consuming and destroying
many seeds and seedlings (Notman and Villegas 2005,
Paine and Beck 2007, Alvarez-Loayza and Terborgh
2011, Bagchi et al. 2014). The objective of this study was
to determine the extent to which mammalian predation
contributes to the community structure of tropical forest
trees.
Previous examinations of this topic have followed one
of two approaches (Beck et al. 2013). The first compares
seedling recruitment in intact and defaunated forests,
from which anthropogenic hunting has extirpated largebodied vertebrates (Asquith et al. 1997, Terborgh et al.
2008, Harrison et al. 2013). Because humans hunt both
Manuscript received 31 May 2016; revised 26 August 2016; accepted 30 August 2016. Corresponding Editor: Thomas J. Valone.
5
E-mail: [email protected]

arboreal and terrestrial animals, however, such investigations can confound their potentially contrasting effects
(Kurten et al. 2015). Arboreal vertebrates are largely
frugivorous, consuming fruit pulp and dispersing seeds.
Terrestrial vertebrates, on the other hand, mostly
consume seeds and seedlings destructively, although
scatter-hoarding rodents are also important vectors of
secondary seed dispersal (Vander Wall et al. 2005, Hirsch
et al. 2012). Moreover, abiotic factors may vary among
sites, influencing recruitment patterns (Beck et al. 2013).
In a second, more direct approach, the experimental use
of selectively permeable cages (i.e., exclosures) allows
investigators to manipulate the access of terrestrial vertebrates to seeds and seedlings in relatively homogeneous
abiotic conditions (Daubenmire 1940). Exclosure technique is particularly powerful when coupled with the
addition of seeds, through which investigators can generate artificial communities of known age and species
composition (DeMattia et al. 2004, Paine and Beck 2007).
Unfortunately, many studies of this type have been of
very short duration, often less than two years (DeMattia
et al. 2004, Hautier et al. 2010, Kuprewicz 2013).
Moreover, the few long-duration studies have not
included seeds of enough species to make strong inferences about the effects of mammals on tree community
structure (Notman and Villegas 2005, Norghauer et al.
2006).
Paine and Beck (2007) provide the most thorough
analysis to date of the effects of mammalian predation on
tropical tree community structure. Their study, however,

3326

December 2016

HOW MAMMALS SHAPE TREE COMMUNITIES

suffered from a number of shortcomings, which we

remedy in the current contribution. First, Paine and Beck
(2007) analyzed diversity using species richness per individual. Though this metric is frequently assessed (Hubbell
et al. 1999), it cannot be considered a diversity index as it
does not account for the relative abundance of species
(Magurran 2013). In fact, it is maximal when evenness is
minimal. Second, their study was incomplete, as it did not
examine important aspects of community structure such
as functional traits, species composition, or beta diversity.
We expand upon Paine and Beck (2007) by studying 24
species for up to 4.4 years, compared with 14 species and
2.2 yr in Paine and Beck (2007). Finally, we take advantage
of new data on functional traits and tree demography to
assess the effects of mammalian predation on all
important aspects of tree community structure.
We sought to understand the relative effects of three
size classes of mammals on tree community structure.
Though it is well known that terrestrial rain forest vertebrates consume many tree seeds (Paine and Beck 2007,
Hautier et al. 2010, Beck et al. 2013, Kurten et al. 2015),
their relative effects in generating tree community
structure is less evident. We consider three size classes.
Small mammals, with adult body mass <1 kg, include
mice (Muridae) and spiny rats (Echimyidae). Mediumsized mammals (112 kg) are caviomorph rodents and
include acouchis, agoutis, and pacas (Myoprocta pratti,
Dasyprocta variegata, and Cuniculus paca, respectively).
Large mammals (>20 kg) are predominantly peccaries
(Pecari tajacu and Tayassu pecari, Tayassuidae), but also
include deer (Cervidae) and tapirs (Tapirus terrestris,
Tapiridae). Assessing the effects of each size class separately is critical for predicting the effects of anthropogenic activities, such as hunting-induced defaunation, on
the tree community. Large terrestrial mammals become
locally scarce in lightly hunted forests, whereas even
medium-sized mammals can be extirpated from intensively hunted forests (Endo et al. 2010). Small mammals
are not typically hunted, but their populations frequently
expand following hunting, presumably because of
reduced competition from larger mammals (Asquith
et al. 1997, Peres and Palacios 2007).
We posit five hypotheses linking predation by each size
class of mammal to tree community structure. First,
mammals will shape tree community structure only if
they generate interspecific variation in seed survival.
Only if this is the case can mammalian predation affect
the relative abundance of tree species. Second, we hypothesize that predation will reduce evenness and thus species
diversity (Paine and Beck 2007, Theimer et al. 2011, Beck
et al. 2013). Third, because mammalian feeding preferences can vary spatially, we hypothesize that mammalian
predation will affect local species composition and beta
diversity, the change in species composition over space.
Although distance-limited seed dispersal is understood to
be the primary generator of beta diversity (Chave and
Leigh 2002), environmental filtering, in the form of mammalian feeding preferences, may also play a role. Fourth,

3327

we hypothesize that mammals preferentially prey upon

larger seeds, as nutritional rewards scale with seed mass,
assuming that seeds do not vary in detectability or handling time (Paine and Beck 2007). If mammals disproportionately prey upon large-seeded species, they may also
affect the distribution of wood density across the tree
community, owing to a weakly positive association
between seed mass and wood density in tropical forests
worldwide (Wright et al. 2007). Thus, mammalian predation may affect the distribution of wood density among
species recruiting as seedlings, with potential long-term
effects on biomass and carbon sequestration (Peres et al.
2015). Finally, we hypothesized that mammalian predation generates a negative relationship between seedling
recruitment and population density. Such negative
density dependence is pervasive in seedling recruitment,
and is essential for stable species coexistence (Chesson
2000, Harms et al. 2000). Therefore, we hypothesized
that mammals may disproportionately prey upon species
that are common as adults, because they may have
stronger search images for such species.
METHODS
This study was conducted in tropical moist forest in the
vicinity of Cocha Cashu Biological Station (CCBS),
Manu National Park, Peru (12 S, 71 W, ~350 m elevation; see site description in Gentry 1990). The forested
floodplain of the Manu River is extremely diverse, with
almost 350 species of trees that attain a diameter of 10 cm
at breast height (dbh). Average annual precipitation is
2,200 mm, falling mainly between October and April.
The vastness and physical isolation of the 1.9 million
hectare Manu National Park, together with neighboring
protected areas, have facilitated the preservation of the
diverse vertebrate community of CCBS, making it one of
few sites worldwide that remains intact and accessible for
study (Endo et al. 2010). It is thus an ideal location to
detail the effects of terrestrial mammals on seedling
recruitment.
To determine the individual effects of small, medium,
and large mammals, we established exclosures that differed in their permeability to each size class. We built
exclosures in eight randomly located blocks, separated
by at least 250 m, within an area of 3 km2. In each block,
we located one 2 2 m exclosure cages of each of five
types 20 m apart along a randomly oriented transect.
There were 40 exclosures in total. NONE exclosures,
which were impermeable to all terrestrial mammals,
were 90-cm tall wire hardware cloth (mesh size 1 cm),
reinforced with iron rebar at the corners and the middle
of each side. SMALL exclosures were identical, but with
7 7 cm holes cut along the bottom edge of the walls,
making them permeable to small mammals. MEDIUM
LARGE exclosures consisted of 20 cm tall sheet-metal
barriers to small mammals (Appendix S1: Fig. S1).
Medium and large mammals could easily step over the
walls to enter the exclosure. MEDIUM combined the

3328

C. E. TIMOTHY PAINE ET AL.

sheet metal of MEDIUMLARGE with a wrapping of

barbed wire, which barred the entry of large mammals,
making them permeable only to medium-sized mammals.
Finally, ALL treatments were only marked with rebar
at the four corners, permitting the entry of all terrestrial
mammals.
Tree species were included in the study based upon
three criteria. First, their fruit had to be single-seeded
and their seeds had to be sufficiently large to be easily
cleaned, sown, and monitored. Second, fruiting adults
had to be sufficiently common and fecund to provide
enough seeds for placement in the exclosures. Third, as
seeds were placed into the exclosures in four batches,
fruit needed to be available at the beginning of one of the
four experimental periods: November 1999January
2000, June 2001, AprilJune 2004, or February 2005.
These criteria yielded 24 species, representing 17 families, including 18 trees, five palms, and one liana
(Sparattanthelium tarapotanum). All species reach the
canopy as adults and are primarily dispersed by
mammals. Seed mass was measured for at least 30 seeds
per species. Sapwood samples were obtained using an
increment borer from up to three adults per species.
Wood density was assessed with the water displacement
method. Seed mass and wood density values were each
missing for a single species, and were obtained from the
Kew Seed Information Database and Chave et al. (2009),
respectively (database available online).6 Adult abundance (individuals 10 cm dbh), was determined in 38
permanent plots totaling 25 ha in the floodplain and
uplands of the Manu River watershed (Manu Plant
Network; J. Terborgh, unpublished data). Seed mass
varied over two orders of magnitude, from 50 to 5400 mg
(median 1800 mg), whereas wood density varied from
0.22 to 0.76 g/cm3 (median 0.54 g/cm3). Adult stem
density of the focal species ranged from ~0.01 to 90
adults/ha (median 0.67 adults/ha), encompassing almost
the entire range of densities observed among adult trees
in this region. The distinguishing characteristics of
species are presented in Table 1. We added seeds to
exclosures and monitored their fates as in Paine and
Beck (2007), except that seeds were placed in conspecific
groups of six in experimental periods 1 and 2, and conspecific groups of 10 in experimental periods 3 and 4. We
used seed removal as a proxy for seed mortality, given
the uncertainty in the precise fate of missing seeds
(Vander Wall et al. 2005). See Paine and Beck (2007) for
further experimental details.
Data analysis
All analyses were performed on the sum of surviving
seeds and germinating seedlings. We evaluated the effects
of mammalian predation on seed and seedling survival
using a parametric survival regression, in which survival
was predicted from the interacting effects of species and
6

http://data.kew.org/sid

Ecology, Vol. 97, No. 12

treatment. As mortality risk is likely to decrease over time

for seeds and seedlings, residuals were assumed to follow
a Weibull distribution.
The effects of mammalian predation on species
evenness, diversity, plot-mean seed mass, plot-mean
wood density, and plot-mean adult stem density were
assessed using linear mixed-effect models. Evenness and
diversity were expressed as Pielous J and the effective
number of species (eH ; Magurran 2013), respectively.
Plot-mean seed mass, wood density, and adult stem
density were calculated at each census time using species-mean trait values, weighted by the abundance of the
species remaining in each exclosure. All five response variables were predicted on the basis of the interacting effects
of treatment and observation day. To account for spatial
variation in mammalian effects, blocks were included as
random effect. Experimental periods entered the model
with random slopes and intercepts, because species composition varied among them. All five response variables
were log-transformed prior to analysis to control
heteroscedasticity.
We assessed the degree to which predation by each size
class of vertebrates shaped species composition in two
ways. First, to assess the effects of mammalian predation
on local species composition, we calculated the BrayCurtis dissimilarity in species composition caused by predation by each size class of mammal within each of the
eight geographical blocks at each time of observation.
Separately, we assessed the degree to which mammalian
predation generated beta diversity by calculating the
Bray-Curtis dissimilarity in species composition within
each exclosure type among all pairwise combinations of
blocks. Because Bray-Curtis dissimilarity cannot exceed
one (Magurran 2013) and because all exclosures began
with identical species compositions, we modeled these
compositional dissimilarities using asymptotic mixedeffect models that were forced through the origin
(Pinheiro and Bates 2000). Asymptotes and rate constants were allowed to vary among mammal size classes
as fixed effects. We included experimental period as a
random effect in the beta-diversity model, and both block
and experimental period in the local species composition
model.
For all aspects of tree community structure, our
interest regarded the effects of each size class of vertebrates, rather than of the treatments themselves. We used
a priori orthogonal contrasts among treatments to test
the separate effects of each mammal size class on each
aspect of tree community structure. The impact of each
mammalian size class was determined by contrasting the
pair of exclosure treatments that differed only in their
permeability to that size class. Accordingly, we contrasted NONE vs. SMALL exclosures to estimate the
effect of small mammals, NONE vs. MEDIUM for
medium mammals, and MEDIUM vs. MEDIUM
LARGE for large mammals. For species composition,
dissimilarities between treatments were analyzed directly,
obviating the need for orthogonal contrasts.

December 2016

HOW MAMMALS SHAPE TREE COMMUNITIES

3329

TABLE 1. Names and key attributes of the 24 woody plant species studied at Cocha Cashu Biological Station, Manu National Park,
Peru.

Family
Annonaceae
Arecaceae

Clusiaceae
Combretaceae
Ebenaceae

Hernandiaceae
Icacinaceae

Lecythidaceae
Malvaceae
Meliaceae
Moraceae
Myristicaceae

Nyctaginaceae
Olacaceae
Rubiaceae
Salicaceae
Verbenaceae

Species
Duguetia quitarensis
Benth.
Astrocaryum murumuru
Mart.
Attalea butyracea
(Mutis ex L.f.)
Wess.Boer
Iriartea deltoidea
Ruiz & Pav.
Mauritia flexuosa L.f.
Socratea exorrhiza
(Mart.) H.Wendl.
Calophyllum
brasiliense Cambess.
Buchenavia grandis
Ducke
Diospyros artanthifolia
Mart. ex Miq.
Diospyros subrotata
Heirn
Sparattanthelium
tarapotanum Meisn.
Calatola microcarpa
Gentry ex Duno &
Janovec
Calatola costaricensis
Standl.
Bertholletia excelsa
Bonpl.
Matisia cordata Bonpl.
Swietenia macrophylla
King
Clarisia racemosa Ruiz
& Pav.
Otoba parvifolia
(Markgr.) Gentry
Virola calophylla
(Spruce) Warb.
Neea sp. nov. Foster
5005
Heisteria nitida Engl.
Genipa americana L.
Casearia sp. nov.
Huillca-Aedo 3561
Vitex cymosa Bertero
ex Spreng.

Experimental
period

Number
of
censuses

Seed
mass
(mg)

Wood
density
(g/cm3)

Adult
density
(per ha)

Germination
rate (%)

Final
survival
(%)

12

410

0.612

2.36

2.5

0.5

24

6,000

0.508

37.44

6.3

3.3

11

54,700

0.326

27.19

7.1

2.5

13

3,860

0.267

89.89

31.3

10.0

1
3

18
13

13,840
3,440

0.557
0.226

3.55
0.12

0.0
34.8

0.0
4.3

23

2,520

0.579

0.35

29.4

1.8

13

1,930

0.755

0.12

15.8

0.5

610

0.535

0.08

30.0

7.5

13

790

0.498

0.95

22.8

9.0

550

0.432

0.03

25.3

17.5

11

6,000

0.472

1.10

37.1

15.0

11

16,000

0.545

4.02

52.1

45.4

11

7,500

0.624

0.39

7.5

3.8

1
1

24
26

5,290
442

0.373
0.522

4.61
0.04

14.6
50.4

0.0
6.3

1,780

0.585

3.23

59.0

12.5

1,800

0.426

27.39

13.5

9.0

26

1,310

0.329

0.16

7.5

0.0

26

102

0.664

0.12

15.8

1.3

3
4
4

13
7
7

220
50
610

0.602
0.643
0.658

1.10
0.35
0.03

15.3
39.8
49.0

0.8
1.3
9.3

490

0.570

0.32

0.3

0.3

Notes: Adult density is based on observations on 38 permanent plots, totaling 25 ha, distributed across the floodplain and uplands
of the Manu River watershed. Germination rate indicates the percentage of seeds that yielded observed seedlings, whereas final
survival indicates the percentage of seeds that yielded seedlings at the end of the experiment.

For all response variables, the effects of mammalian

predation were compared at two points in time: 1.4 and
4.4 years, which were the durations of the shortest- and
longest-duration experimental periods, respectively
(Table 1). At each of these time points, each response variable was predicted using 1,000 parametric bootstrap replicates. The effects of each size class of mammals on the

response variables was assessed as the base-10 logarithm

of the ratio of the response variable in exclosures permeable to the given mammal size class to its value in exclosures from which the mammal size class was excluded.
Log10-ratios of 1 or 1 indicate that a mammal size class
caused a 10-fold increase or decrease in the response variable, respectively. Mammalian effects were deemed

3330

C. E. TIMOTHY PAINE ET AL.

significant if the 95% confidence intervals of the bootstrap

replicates did not include zero. Analyses were performed
in R 3.2.3 (R Core Team 2015). Survival and species composition models were fit using the survival and nlme
libraries, respectively (Pinheiro and Bates 2000), whereas
all other analyses were implemented using the lme4
library (Bates et al. 2014). All raw data and R code used in
this study are available online (see Data Availability).
RESULTS
Over the eight-year duration of the study, 8,000 seeds
of 24 species were placed into the exclosures, yielding a
total of 1,917 seedlings, 515 of which survived to the end
of the experimental period. Germination rate varied
among species from 0% to 59%. Three species recruited
no seedlings (Table 1). One of these, Mauritia flexuosa, is
a swamp specialist, the seeds of which were rapidly consumed by terrestrial termites. Virola calophylla and
Matisia cordata germinated weakly (7.5% and 14.6%,
respectively), but all their seedlings perished, potentially
as a result of host-specific natural enemies (AlvarezLoayza and Terborgh 2011). In contrast, 45% of the seeds
of Calatola costaricensis, a large-seeded tree that is rare
as an adult, survived as seedlings to the end of the study
(Appendix S1: Fig. S2).
Our first hypothesis, that mammals generate interspecific variation in seed survival, was strongly supported by
the data. Small, medium, and large mammals reduced
median survival time by up to 10.3, 15.1, and 2.0 months,
respectively (Fig. 1). The strongest effects were generated
by small and medium-sized mammals, which significantly
reduced the survival of 17 and 14 species, respectively.
Large mammals, on the other hand, significantly reduced
the survival of only four species, all of which were also
significantly impacted by small- or medium-bodied
mammals. Survival of a few species was modestly
increased by exposure to mammals; why this occurred is
unclear.
Given their differential effects upon survival, mammalian predation also reduced evenness and species
diversity, supporting our second hypothesis. All mammalian size classes reduced species evenness (Pielous J),
with effects that strengthened over the duration of the
experiment (Fig. 2A). All three size classes of mammals also
significantly reduced the effective number of species (eH ),
with effects that also strengthened over time (Fig. 2B). For
both evenness and diversity, the effects of large mammals
were weaker than those of small and medium-bodied
mammals (Fig. 2). Our third hypothesis found strong
support, as predation by all three mammal size classes
caused rapid and significant changes in local species composition, which lasted through the end of the experiment
(Fig. 3A). Beta diversity showed a strikingly different
pattern. Small and medium-sized mammals rapidly generated significant beta diversity, which lasted throughout
the experiment, whereas large mammals did not significantly contribute to beta diversity at any time (Fig. 3B).

Ecology, Vol. 97, No. 12

There was strong support for the first part of our fourth
hypothesis, that mammals preferentially preyed upon
large-seeded species. Predation by all three size classes of
mammals lead to significant decreases in the plot-mean
seed mass over time. Small mammals had the strongest
effect, reducing plot-mean seed mass more than 10-fold
(log10 ratio, 1.01), whereas the effects of medium-sized
and large mammals were weaker (Fig. 4A). The evident
preference of mammals for larger-seeded species generated
only weak effects on plot-mean wood density, however,
despite the significant negative relationship between seed
mass and wood density (Appendix S1: Fig. S3). After
1.4 years, wood density was significantly increased by
small and large mammal predation, and significantly
decreased by medium-sized mammals. Only the effects of
large mammals persisted through the end of the experiment, but they only increased wood density by 4% (log10
ratio 0.039; Fig. 4B). Small and medium-sized mammals,
on the other hand, strongly and significantly reduced
plot-mean adult density by disproportionately removing
seeds of species that are common as adults, thus generating
negative density dependence (Fig. 4C). Both size classes
reduced plot-mean adult density by at least 100-fold by the
end of the experiment (log10 ratios, 2.43 and 2.13 for
small and medium mammals, respectively). Large
mammals reduced plot-mean adult density initially,
though this effect disappeared by the end of the study.
DISCUSSION
By following the fates of seeds of 24 tree species for up
to 4.4 years in a well-replicated experiment, we were able
to assess aspects of community structure, such as beta
diversity, that were beyond the scope of previous studies.
Overall, mammalian predation on seeds and seedlings
had strong and predictable effects on tree community
structure. Small, medium, and large-bodied species all
reduced species evenness and diversity (Fig. 2). This
finding directly contradicts that of Paine and Beck (2007),
who claimed that predation by small mammals increased
diversity, measured as species richness per stem. Their
error was that species richness per stem is not a diversity
index, as it does not account for relative species abundance (Magurran 2013). In Paine and Beck (2007) and
the current study, it would have been impossible for
mammals to increase species diversity. They could not
have increased species richness, as experimentally placed
seeds were clearly distinguishable from naturally dispersed ones. Nor could they have increased evenness, as
it was maximized at the beginning of each experimental
period by the placement of equal numbers of seeds of
each species in each exclosure.
Small and medium-bodied mammals more strongly
affected tree community structure than did large
mammals, in accordance with previous studies (Asquith
et al. 1997, DeMattia et al. 2004, Norghauer et al. 2006,
Paine and Beck 2007, Hautier et al. 2010). Not only did
they remove more seeds than did large-bodied mammals,

December 2016

HOW MAMMALS SHAPE TREE COMMUNITIES

3331

Neea sp. nov. Foster 5005

Calophyllum brasiliense
Astrocaryum murumuru
Bertholletia excelsa
Iriartea deltoidea
Otoba parvifolia
Socratea exorrhiza
Diospyros subrotata
Clarisia racemosa
Virola calophylla
Attalea butyracea
Diospyros artanthifolia
Duguetia quitarensis
Calatola costaricensis
Buchenavia grandis
Matisia cordata
Genipa americana
Swietenia macrophylla
Vitex cymosa
Mauritia flexuosa
Heisteria nitida
Casearia sp. nov. Huillca-Aedo 3561

Small
Medium
Large

Calatola microcarpa
Sparattanthelium tarapotanum

2.0

1.0

0.0

0.5

1.0

Change in median survival time (yr)

FIG. 1. The effects of mammalian predation on the median survival time of seeds and seedlings of each species at Cocha Cashu
Biological Station, Manu National Park, Peru. Solid points indicate significant effects of mammals (P 0.05), whereas open points
indicate nonsignificant effects. Lines indicate 95% confidence intervals around the estimated effect derived from a parametric
survival regression. Small and medium-sized mammals reduced the survival of most species, whereas large mammals had far weaker
effects. Species are sorted by magnitude of the effect of small mammals on survival.

they also generated beta diversity and negative density

dependence through their actions (Figs. 1, 3B, and 4C).
The effects of large mammals were altogether weaker,
although they alone favored the recruitment of species
with dense wood by disproportionately preying upon
species with low wood density (Fig. 4B). These findings
broaden and generalize those of the few previous studies
that have attempted to link the actions of mammalian
seed predators to tree community structure (DeMattia
et al. 2004, Paine and Beck 2007, Theimer et al. 2011,
Kurten et al. 2015).
Notably, ours is the first study, to our knowledge, to
demonstrate that mammalian predation can contribute
to beta diversity, the change in species composition over

space. Beta diversity is generally assumed to arise from

distance-limited seed dispersal (Chave and Leigh 2002),
though biogeographical history also makes an important
contribution (Dexter et al. 2012). We suggest that spatial
variation in canopy tree composition, together with the
relatively small home ranges of small and medium-bodied
mammals, could lead to spatial variation in search images
for preferred food items. This, in turn, would lead to
spatial variation in species-specific rates of seedling
recruitment. Such a process could amplify and contribute
to the patterns of beta diversity observed in tropical tree
communities.
Given the central role of negative density dependence
for the maintenance of diversity and its pervasive nature

3332

C. E. TIMOTHY PAINE ET AL.

0.5

***
***

1.0

***
***

1.5
2.0
2.5

A 1.0
***
***

0.1

***

0.2

***
***

Small
Medium
Large

0.3
0.4
0

BrayCurtis dissimilarity in
community composition

log(effective species ratio)

B
0.0

mammalian seed predators outnumber medium- and

large-bodied ones by an order of magnitude (419, 14, and
12 individuals/km2, respectively; Janson and Emmons
1990). The population-level biomass of large mammalian
seed predators at this site, however, far exceeds that of
medium or small ones (590, 10, and 12 kg/km2, respectively; Janson and Emmons 1990, Endo et al. 2010).
Moreover, the large body size and rooting behavior of
Tayassu pecari (white-lipped peccary), the dominant large
terrestrial mammal at CCBS and across the Neotropics,
cause them to have very strong per capita effects (Beck
2005, Beck et al. 2013). The observation that small and
medium-bodied mammals had consistently stronger
effects on tree community structure indicates that
ubiquity, imparted by very large population sizes,

1
2
3
4
5
Time since exposure (yr)

FIG. 2. Predation by mammals led to changes in (A) Pielous

evenness (J) and (B) Shannons diversity index, expressed as the
effective number of species (eH ). The effects of mammals are
represented as the log10 ratio of the response variable in exclosures
to which the mammals had access vs. those from which they were
excluded. Predicted effects and confidence intervals are derived
from mixed-effect models. *P < 0.05, ***P < 0.001.

(Chesson 2000, Harms et al. 2000), understanding its generative mechanisms is of great interest. Our findings add
to the body of evidence that small-bodied mammals can
generate negative density dependence (Paine and Beck
2007), thus contributing to stabilizing niche differences
and thus species coexistence. Arthropods and pathogenic
fungi can also do so (Notman and Villegas 2005, AlvarezLoayza and Terborgh 2011, Bagchi et al. 2014). There
remains a need for studies that evaluate the relative
importance of the primary biotic sources of mortality on
tropical tree seedlings: fungi, insects and mammals, so
that we may better understand the mechanisms that generate tropical tree community structure.
Abundance vs. biomass
The relative effects of small, medium, and large-bodied
mammals on tree community structure is not easy to
predict a priori. At Cocha Cashu Biological Station, small

***

***

0.8

***
***

***
***
0.6
0.4
Small
Medium
Large

0.2
0.0

B
Log ratio of betadiversity

log10(evenness ratio)

A
0.0

Ecology, Vol. 97, No. 12

0.6

***

***

***

0.4
ns
0.2
ns
0.0

0.2
0

Years since exposure (yr)

FIG. 3. Predation by all mammal size classes caused changes
in (A) tree species composition through time, whereas (B) only
small and medium-sized mammals generated significant beta
diversity. The effects of mammals on beta diversity are
represented as the log10 ratio of the Bray-Curtis dissimilarity
among exclosures to which the mammals had access, vs. that
from which they were excluded. Predicted effects and confidence
intervals are derived from nonlinear mixed-effect models. *P <
0.05, ***P < 0.001; ns, not significant.

December 2016

HOW MAMMALS SHAPE TREE COMMUNITIES

log10(seed mass ratio)

***
***
***
***

***

Small
Medium
Large

***

B
log10(wood density ratio)

*
ns
ns

***
***
*

log10(adult density ratio)

C
ns
***
***
***

***
***
3

Time since exposure (yr)

facilitates stronger trophic interactions than does great

individual biomass. In other words, a seed predators
ability to locate seeds is a better predictor of its ecological
impacts than is its jaw strength, at least for the 24 plant
species used in this study. Notably, all mammal size
classes were able to consume seeds (or seedlings) of all
studied species. Had this not been the case, for example,
if some species had been chemically defended, other outcomes would have been observed (Kuprewicz 2013).

3333

FIG. 4. Predation by mammals led to changes in plot-mean

(A) seed mass (measured in mg), (B) wood density (measured in
g/cm3), and (C) adult density (measured as no./ha). The effects
of mammals are represented as the log10 ratio of the response
variable in exclosures to which the mammals had access, vs.
those from which they were excluded. Predicted effects and
confidence intervals are derived from mixed-effect models. Note
that the y-axis scales vary among panels. *P < 0.05, ***P <
0.001; ns, not significant.

Why were large mammals observed to have such weak

effects on tree community structure? The scale of the
experimental exclosures may have played a role. Herds of
T. pecari, containing up to 200 individuals, travel approximately 10 km/d as they bulldoze through the understory (Wyatt and Silman 2004, Beck 2005). The 4-m2
exclosure plots used in this study may have been too small
to attract the attention of these wide-ranging mammals.
Notably, T. pecari is the dominant seed predator of the
palm Astrocaryum murumuru when it is found in highdensity aggregations, whereas Proechimys spp. and other
small rodents are the primary seed predators of isolated
A. murumuru individuals (Beck and Terborgh 2002). Thus,
T. pecari may have weak effects on the tree community as
a whole, but strong effects on a few species (Silman et al.
2003, Wyatt and Silman 2004). It is likely that larger-scale
exclosures would have more equitably assessed the effects
of large mammals on tree community composition (Kurten
et al. 2015), although it would have been logistically challenging to achieve sufficient replication with them.
In several regions, population sizes of T. pecari vary
erratically, even repeatedly becoming undetectably scarce
for years at a time (Vickers 1991, Reyna-Hurtado et al.
2009). Infectious disease, which could presumably spread
rapidly in this highly social species, may cause these
large-scale extirpations (summarized by Richard-Hansen
et al. 2014). T. pecari were effectively absent from CCBS
between 1978 and 1990, and disappeared again in 2012
(Silman et al. 2003). Nevertheless, they were present and
abundant throughout the eight-year duration of the
current study, meaning that population fluctuations
should not have weakened their apparent effects on tree
community structure.
Taken at face value, our results suggest that the consequences of hunting on tree community structure should
be relatively minor, because small and medium-sized
mammals, which are less-often hunted, more strongly
contributed to tree community structure than did largebodied mammals, which are preferred by hunters (Peres
and Palacios 2007, Endo et al. 2010). However, two considerations make this conclusion overly simplistic. First,
in the absence of mammalian seed predation, many seeds
are destroyed by fungal pathogens, bacteria, or
arthropods, some of which can generate negative density
dependence (Bagchi et al. 2014). This was observed in our
study, as few seeds or seedlings survived to the end of the
experiment, even in the CLOSED treatment (Appendix
S1: Fig. S2). Thus, the consequences for the tree community of the local extirpation of large mammalian seed

3334

C. E. TIMOTHY PAINE ET AL.

predators could be, at least in part, compensated for by

the actions of smaller-bodied organisms (Asquith et al.
1997). Second, hunting by humans extirpates largebodied arboreal primates as well as terrestrial mammals,
thus affecting both seed dispersal and seed predation.
Across Amazonia, heavily hunted sites retain approximately 10% of the population density of ateline primates
(primarily Ateles and Lagothrix), compared to nonhunted sites of equivalent productivity (Peres and
Palacios 2007). These genera of frugivorous primates
provide the irreplaceable ecological service of seed dispersal to roughly 23% of genera of Neotropical trees
(Peres et al. 2015). Thus, although the ecological consequences of human hunting on seed predation are likely to
be relatively modest, its impacts on seed dispersal are
substantial and detrimental (Kurten et al. 2015).
Observational studies of hunted forests show strong
directional change in tree species composition (Terborgh
et al. 2008, Harrison et al. 2013). We suggest that these
changes are more likely to be driven by reduced seed dispersal than by reduced seed predation.
Terborgh (2012) levelled three criticisms at studies
such as the current one, which he referred to as seed
presentation trials. First, they are conducted with relatively large seeds. Though seeds masses in the current
study spanned two orders of magnitude, tree seeds in
Neotropical forests vary over seven orders of magnitude
(Wright et al. 2007). This could be seen as limiting our
ability to make inferences on the effects of mammalian
predation on small-seeded species. Most small-seeded
species require high light to germinate, however, and our
results suggest that very small-seeded species are likely to
escape the notice of mammals (Fig. 4A). Thus, their
recruitment dynamics are likely to be relatively independent of mammalian seed predators.
Second, abundant seeds are displayed conspicuously. In
contrast, naturally dispersed seeds are often scattered as
they fall from the canopy and are subsequently covered by
leaf litter. In the current study, conspecific seeds were
placed in groups in each 4-m2 exclosure, a far greater
density than the one viable seed per square meter observed
in a concurrent seed-trapping at CCBS (Swamy et al.
2010). By artificially elevating the natural density of seed
rain, the exclosure cages could have become more
attractive foraging sites for seed predators than they
might otherwise have been. This increase in food abundance is likely to have influenced mammal behavior.
Although this elevated abundance was unavoidable and
necessary for efficient data collection, we partially
addressed the concern about conspicuousness by replacing
leaf litter that had fallen on the seeds after every census.
Third, the seeds used are typically undispersed, and have
been cleaned of pulp. Such pre-treatment can alter the
olfactory cues that attract both seed predators and secondary dispersers, and can thus affect the probabilities of
being eaten or buried. Secondary dispersal and burial by
rodents or dung beetles increase dispersal, reduce predation, and enhance survival to the seedling stage

Ecology, Vol. 97, No. 12

(Andresen and Levey 2004, Vander Wall et al. 2005,

Hirsch et al. 2012). Cleaning seeds may have shaped the
outcome of our experiment in two ways: our seeds may
have been less likely to be found by mammalian seed
predators than ones reeking of dung, and seeds buried by
dung beetles or scatter-hoarding mammals were counted
as dead. The former effect would lead us to underestimate
the effects of mammalian predation, because more seeds
would have been consumed, had they been covered in
dung. The latter, on the other hand, would lead us to
overestimate the effects of mammalian predation,
because seeds that were removed or buried, and subsequently germinated, were considered by us to have died.
The relative magnitude of these countervailing effects
remains unclear. Nevertheless, we stand by our decision
to clean seeds prior to their placement in the exclosures.
To have done otherwise would have been impractical.
CONCLUSIONS
Mammals contribute strongly to tropical tree community structure through the consumption of seeds and
seedlings. Multiple lines of evidence suggest that small
and medium-bodied mammals play a stronger role in the
seed-to-seedling transition than do large mammals. They
significantly reduce survival of many species, generate
beta diversity and crucially, they contribute to the negatively density dependent nature of seedling recruitment
by disproportionately preying upon tree species that are
common as adults. In contrast, large mammalian seed
predators had minimal effects on tree community
structure. Predation, especially by rodents, plays an
important role in maintaining tree diversity and shaping
tree community dynamics in tropical forests.
ACKNOWLEDGMENTS
The Peruvian Servicio Nacional de reas Naturales
Protegidas por el Estado (SERNANP) granted us permission to
conduct research in Manu National Park. We thank the many
people involved in the establishment and monitoring of this
study, including Gabriela Nuez, Cornelio Machaca, Pamela
Weisenhorn, Patricia Alvarez-Loayza, Cecilia Carrasco, Beth
Pringle, and Karim Ledesma. Comments from Kirsten
Hazelwood improved the manuscript.
LITERATURE CITED
Alvarez-Loayza, P., and J. W. Terborgh. 2011. Fates of seedling
carpets in an Amazonian floodplain forest: Intra-cohort
competition or attack by enemies? Journal of Ecology 99:
10451054.
Andresen, E., and D. J. Levey. 2004. Effects of dung and seed
size on secondary dispersal, seed predation, and seedling
establishment of rain forest trees. Oecologia 139:4554.
Asquith, N. M., S. J. Wright, and M. J. Clauss. 1997. Does
mammal community composition control recruitment in neotropical forests? Evidence from Panama. Ecology 78:
941946.
Bagchi, R., R. E. Gallery, S. Gripenberg, S. J. Gurr, L. Narayan,
C. E. Addis, R. P. Freckleton, and O. T. Lewis. 2014.

December 2016

HOW MAMMALS SHAPE TREE COMMUNITIES

Pathogens and insect herbivores drive rainforest plant diversity and composition. Nature 506:8588.
Bates, D. M., M. Maechler, B. M. Bolker, and S. Walker. 2014.
lme4: Linear mixed-effects models using Eigen and S4. R
package.
Beck, H. 2005. Seed predation and dispersal by peccaries
throughout the neotropics and its consequences: a review and
synthesis. Page 77 in P. M. Forget, J. E. Lambert, P. E.
Hulme, and S. B. Vander Wall, editors. Seed fate: predation,
dispersal and seedling establishment. CABI Publishing,
Wallingford, UK.
Beck, H., and J. W. Terborgh. 2002. Groves versus isolates:
how spatial aggregation of Astrocaryum murumuru palms
affects seed removal. Journal of Tropical Ecology 18:
275288.
Beck, H., J. W. Snodgrass, and P. Thebpanya. 2013. Long-term
exclosure of large terrestrial vertebrates: implications of
defaunation for seedling demographics in the Amazon rainforest. Biological Conservation 163:115121.
Chave, J. J., D. A. Coomes, S. Jansen, S. L. Lewis, N. G.
Swenson, and A. E. Zanne. 2009. Towards a worldwide wood
economics spectrum. Ecology letters 12:351366.
Chave, J., and E. G. Leigh. 2002. A spatially explicit neutral
model of beta-diversity in tropical forests. Theoretical
Population Biology 62:153168.
Chesson, P. 2000. Mechanisms of maintenance of species
diversity. Annual review of Ecology and Systematics 31:
343366.
Daubenmire, R. F. 1940. Exclosure technique in ecology.
Ecology 21:514515.
DeMattia, E. A., L. M. Curran, and B. J. Rathcke. 2004. Effects
of small rodents and large mammals on neotropical seeds.
Ecology 85:21612170.
Dexter, K. G., J. W. Terborgh, and C. W. Cunningham. 2012.
Historical effects on beta diversity and community assembly
in Amazonian trees. Proceedings of the National Academy of
Sciences USA 109:77877792.
Endo, W., C. A. Peres, E. Salas, S. Mori, J. L. Sanchez-Vega,
G. H. Shepard, V. Pacheco, and D. W. Yu. 2010. Game vertebrate densities in hunted and nonhunted forest sites in
Manu National Park, Peru. Biotropica 42:251261.
Gentry, A. H. 1990. Four neotropical forests. Yale University
Press, New Haven, Connecticut, USA.
Green, P. T., K. E. Harms, and J. H. Connell. 2014.
Nonrandom, diversifying processes are disproportionately
strong in the smallest size classes of a tropical forest.
Proceedings of the National Academy of Sciences USA 111:
1864918654.
Harms, K. E., S. J. Wright, O. Caldern, A. Hernandez, and
E. A. Herre. 2000. Pervasive density-dependent recruitment
enhances seedling diversity in a tropical forest. Nature
404:493495.
Harrison, R. D., S. Tan, J. B. Plotkin, F. Slik, M. Detto, T. Brenes,
A. Itoh, and S. J. Davies. 2013. Consequences of defaunation
for a tropical tree community. Ecology Letters 16:687694.
Hautier, Y., P. Saner, C. D. Philipson, R. Bagchi, R. C. Ong,
and A. Hector. 2010. Effects of seed predators of different
body size on seed mortality in Bornean logged forest. PLoS
ONE 5:e11651.
Hirsch, B. T., R. Kays, V. E. Pereira, P. A. Jansen, and
M. Rejmanek. 2012. Directed seed dispersal towards areas
with low conspecific tree density by a scatter-hoarding rodent.
Ecology Letters 15:14231429.
Hubbell, S. P., R. B. Foster, S. T. OBrien, K. E. Harms, R. S.
Condit, B. Wechsler, S. J. Wright, and S. L. de Lao. 1999.
Light-Gap disturbances, recruitment limitation, and tree
diversity in a neotropical forest. Science 283:554557.

3335

Janson, C. H., and L. H. Emmons. 1990. Ecological structure of the nonflying mammal community at Cocha
Cashu Biological Station, Manu National Park, Peru.
Pages 314338 in A. H. Gentry, editor. Four Neotropical
rainforests. Yale University Press, New Haven, Connecticut,
USA.
Kuprewicz, E. K. 2013. Mammal abundances and seed
traits control the seed dispersal and predation roles of
terrestrial mammals in a Costa Rican forest. Biotropica 45:
333342.
Kurten, E. L., S. J. Wright, W. P. Carson, and T. M. Palmer.
2015. Hunting alters seedling functional trait composition in
a Neotropical forest. Ecology 96:19231932.
Magurran, A. E. 2013. Measuring biological diversity. John
Wiley & Sons, Hoboken, New Jersey, USA.
Norghauer, J. M., J. R. Malcolm, B. Zimmerman, and J. M.
Felfili. 2006. An experimental test of density- and distantdependent recruitment of mahogany (Swietenia macrophylla)
in southeastern Amazonia. Oecologia 148:437446.
Notman, E., and A. C. Villegas. 2005. Patterns of seed predation by vertebrate versus invertebrate seed predators among
different plant species, seasons and spatial distributions.
Pages 5575 in P. Forget, editor. Seed fate: predation,
dispersal and seedling establishment. CABI Publishing,
Wallingford, UK.
Paine, C. E. T., and H. Beck. 2007. Seed predation by
Neotropical rain forest mammals increases diversity in seedling recruitment. Ecology 88:30763087.
Peres, C., and E. Palacios. 2007. Basin-wide effects of game harvest on vertebrate population densities in Amazonian forests:
implications for animal-mediated seed dispersal. Biotropica
39:304315.
Peres, C. A., E. Thaise, J. Schietti, S. J. M. Desmoulieres, and
T. Levi. 2015. Dispersal limitation induces long-term
biomass collapse in overhunted Amazonian forests.
Proceedings of the National Academy of Sciences USA 113:
892897.
Pinheiro, J. C., and D. M. Bates. 2000. Mixed-effects models in
S and S-PLUS. Springer Verlag, New York, New York,
USA.
R Core Team. 2015. R: a language and environment for statistical computing. R Foundation for Statistical Computing,
Vienna, Austria. https://www.r-project.org/
Reyna-Hurtado, R., E. Rojas-Flores, and G. W. Tanner. 2009.
Home range and habitat preferences of white-lipped peccaries
(Tayassu pecari) in Calakmul, Campeche, Mexico. Journal of
Mammalogy 90:11991209.
Richard-Hansen, C., N. Surugue, K. Khazraie, M. Le Noc, and
P. Grenand. 2014. Long-term fluctuations of white-lipped
peccary populations in French Guiana. Mammalia 78:
291301.
Silman, M. R., J. W. Terborgh, and R. A. Kiltie. 2003.
Population regulation of a dominant rain forest tree by a
major seed predator. Ecology 84:431438.
Swamy, V., J. W. Terborgh, K. G. Dexter, B. D. Best,
P. Alvarez, F. Cornejo, and P. Alvarez-Loayza. 2010. Are
all seeds equal? Spatially explicit comparisons of seed fall
and sapling recruitment in a tropical forest. Ecology Letters
14:195201.
Terborgh, J. W. 2012. Enemies maintain hyperdiverse tropical
forests. American Naturalist 179:303314.
Terborgh, J. W., G. Nuez-Iturri, N. C. A. Pitman, F. H. C.
Valverde, P. Alvarez-Loayza, V. Swamy, E. G. Pringle, and
C. E. T. Paine. 2008. Tree recruitment in an empty forest.
Ecology 89:17571768.
Theimer, T. C., C. A. Gehring, P. T. Green, and J. H. Connell.
2011. Terrestrial vertebrates alter seedling composition and

3336

C. E. TIMOTHY PAINE ET AL.

richness but not diversity in an Australian tropical rain forest.

Ecology 92:16371647.
Vander Wall, S. B., K. M. Kuhn, and M. J. Beck. 2005. Seed
removal, seed predation, and secondary dispersal. Ecology
86:801806.
Vickers, W. T. 1991. Hunting yields and game composition over
ten years in an Amazon Indian territory. Neotropical Wildlife
Use and Conservation 400:5381.

Ecology, Vol. 97, No. 12

Wright, I. J., et al. 2007. Relationships among ecologically

important dimensions of plant trait variation in seven neotropical forests. Annals of Botany 99:10031015.
Wyatt, J. L., and M. R. Silman. 2004. Distance-dependence in
two Amazonian palms: effects of spatial and temporal variation in seed predator communities. Oecologia 140:2635. Data
AvailabilityRaw data and R code used in this study are available from GitHub (http://dx.doi.org/10.5281/zenodo.154042).

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article at http://onlinelibrary.wiley.com/
doi/10.1002/ecy.1586/suppinfo
DATA AVAILABILITY
Raw data and R code used in this study are available from GitHub (http://dx.doi.org/10.5281/zenodo.154042).