LETTER A highly resolved food web for insect seed predators in a
species-rich tropical forest
Sofia Gripenberg,1,2,3,4*
Yves Basset,5,6,7,8 Owen T. Lewis,3
J. Christopher D. Terry,3
S. Joseph Wright,2
Indira Sim
on,2
D. Catalina Fern
andez,2
Marjorie Cede~no-Sanchez,2
Marleny Rivera,2,8 H
ector Barrios,8
John W. Brown,9
Osvaldo Calder
on,2
Anthony I. Cognato,10
Jorma Kim,11 Scott E. Miller,9
Geoffrey E. Morse,12
Sara Pinz
on-Navarro,8
Donald L. J. Quicke,13
Robert K. Robbins,9
Juha-Pekka Salminen11 and
Eero Vesterinen4,14
Abstract
The top-down and indirect effects of insects on plant communities depend on patterns of host
use, which are often poorly documented, particularly in species-rich tropical forests. At Barro Col-
orado Island, Panama, we compiled the first food web quantifying trophic interactions between
the majority of co-occurring woody plant species and their internally feeding insect seed predators.
Our study is based on more than 200 000 fruits representing 478 plant species, associated with
369 insect species. Insect host-specificity was remarkably high: only 20% of seed predator species
were associated with more than one plant species, while each tree species experienced seed preda-
tion from a median of two insect species. Phylogeny, but not plant traits, explained patterns of
seed predator attack. These data suggest that seed predators are unlikely to mediate indirect inter-
actions such as apparent competition between plant species, but are consistent with their proposed
contribution to maintaining plant diversity via the Janzen–Connell mechanism.
Keywords
Apparent competition, Barro Colorado Island, host specialisation, interaction network, Janzen–
Connell hypothesis, Panama, plant traits, quantitative food web, seed predation.
Ecology Letters (2019)
INTRODUCTION
Natural enemies of plants such as herbivores and pathogens
are important determinants of the fitness of individual plants,
the dynamics of plant populations, and the diversity and com-
position of entire plant communities (Crawley 1989; Burdon
et al. 2006; Bradley et al. 2008; Bagchi et al. 2014). Plant-en-
emy interactions are particularly frequent and intense in tropi-
cal forests (Schemske et al. 2009), where the diversity of both
plants and enemies peaks (Willig et al. 2003). However, pat-
terns of enemy attack across plant species, and patterns of
host use by individual enemy species, are often particularly
poorly known in tropical forests. As a result, these patterns
are often inferred from other information. For example, clo-
sely related plant species are expected to be attacked by simi-
lar sets of enemies, leading to phylogenetic clustering of
susceptibility and patterns of enemy attack (Gilbert & Webb
2007; Gilbert et al. 2015).
Plant species vary greatly in their susceptibility to natural
enemies (Coley 1983; Fritz & Simms 1992; Marquis et al.
2001). Understanding the causes and consequences of this
variation is a central aim for ecologists studying plant-enemy
interactions (e.g. Carmona et al. 2011; Loranger et al. 2012;
Turcotte et al. 2014). A variety of traits and ecological cir-
cumstances have been hypothesised to drive interspecific vari-
ation in plant susceptibility to enemies. For example, plant
species that invest heavily in defences are likely to be less
prone to enemy attack and to have fewer and more specialised
consumer species (Walters 2011), as are plant species with
unpredictable distributions in space or time (Feeny 1976).
Most studies linking interspecific variation in enemy attack
to plant traits have focused on leaf herbivory (e.g. Coley
1983; Schuldt et al. 2012; C
ardenas et al. 2014). However,
leaves represent only one component of plant biomass. Traits
influencing plant susceptibility to enemy attack will differ
among plant tissues and organs (Sang et al. 1984; Brown
et al. 2003), and may change during ontogeny (Boege & Mar-
quis 2006) for example as a result of developmental changes
in plant tolerance to enemy attack (Boege et al. 2007). Thus,
new insights into the role of enemies in structuring plant com-
munities could emerge from studying variation in the suscepti-
bility of plants to enemies attacking non-foliar plant tissues
1School of Biological Sciences, University of Reading, Reading, UK
2Smithsonian Tropical Research Institute, Balboa, Republic of Panama
3Department of Zoology, University of Oxford, Oxford, UK
4Biodiversity Unit, University of Turku, Turku, Finland
5ForestGEO, Smithsonian Tropical Research Institute, Balboa, Republic of
Panama
6Faculty of Science, University of South Bohemia, Ceske Budejovice, Czech
Republic
7Biology Centre of the Czech Academy of Sciences, Institute of Entomology,
Ceske Budejovice, Czech Republic
8Maestria de Entomologia, Universidad de Panam
a, Panama, Republic of
Panama
9National Museum of Natural History, Smithsonian Institution, Washington,
DC, USA
10Department of Entomology, Michigan State University, East Lansing, MI,
USA
11Department of Chemistry, University of Turku, Turku, Finland
12Biology Department, University of San Diego, San Diego, CA, USA
13Integrative Ecology Laboratory, Department of Biology, Faculty of Science,
Chulalongkorn University, Bangkok, Thailand
14Department of Ecology, Swedish University of Agricultural Sciences,
Uppsala, Sweden
*Correspondence: E-mail: s.gripenberg@reading.ac.uk
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use,
distribution and reproduction in any medium, provided the original work is properly cited.
Ecology Letters, (2019) doi: 10.1111/ele.13359
and different stages of the plant life cycle. Of particular inter-
est are the seed and early seedling stages, which represent
important demographic bottlenecks (e.g. Fenner & Thompson
2005; Leck et al. 2008; Green et al. 2014). Despite calls for
more focus on the seed stage (e.g. Janzen 1969; Lewis &
Gripenberg 2008), few studies have investigated plant suscepti-
bility to seed enemies (but see e.g. Zalamea et al. 2018).
Information on the identity and specificity of plant enemies
is also important for predicting their community-level effects.
Such effects have received particular attention in tropical for-
ests, where plant enemies are thought to contribute to main-
taining the high diversity of woody plant species. According
to the Janzen–Connell hypothesis (Janzen 1970; Connell
1971), plant enemies can facilitate plant species coexistence
and promote diversity, provided that these enemies are rela-
tively host-specific (Sedio & Ostling 2013). Existing data sug-
gest that, while specificity varies markedly among different
guilds and taxa, relatively few enemies are specialised on a
single plant species (e.g. Novotny et al. 2002, 2010). Where
enemies have host ranges spanning multiple species, they
might instead contribute to eroding plant diversity (by exclud-
ing the species that are least tolerant to enemy attack), or to
structuring the plant community through processes such as
‘apparent competition’ and ‘apparent mutualism’ (Holt 1977;
Lewis & Gripenberg 2008). If enemies are specialised on sev-
eral closely related plant species (e.g. Gilbert et al. 2012), we
expect the potential for enemy-mediated indirect interactions
to depend on the degree of phylogenetic relatedness among
plants. While enemy-mediated indirect interactions have been
documented in a number of systems, they have so far received
little attention in the context of tropical forest plant commu-
nities (but see Garzon-Lopez et al. 2015; Downey et al. 2018).
Obtaining detailed, multi-species information on plant-en-
emy interactions and their consequences can be challenging,
particularly in species-rich tropical forests. For example,
although leaves that have experienced herbivory are often
ubiquitous, herbivores are rarely observed in the process of
feeding (Elton 1975). Moreover, even if host associations of
plant enemies can be established, it can be difficult to quantify
the fitness consequences for long-lived plants. In this context,
the internally feeding insect seed predators attacking develop-
ing and mature seeds offer a convenient and relevant study
system. By killing seeds, either after they have been dispersed
or while they are still attached to the mother plant, these
insects have direct and measurable fitness consequences for
their hosts (e.g. Crawley 2000; Kolb et al. 2007). In a classic
paper, Janzen (1980) assessed patterns of host use by 110 spe-
cies of beetles associated with seeds of 100 plant species in
Guanacaste province, Costa Rica. His study showed that the
specificity of beetles was high, consistent with a contribution
to maintaining plant species diversity via the Janzen–Connell
hypothesis.
In this study, we document a food web including the great
majority of woody plants and their endophagous seed preda-
tors at our study site, Barro Colorado Island in Panama. Our
principal aims are (1) to describe community-level patterns of
insect seed predator attack; (2) to assess the role of phylogeny
and of a set of relevant plant traits (Table 1) in influencing
how prone plant species are to seed predator attack; (3) to
measure the degree of seed predator host-specificity and the
extent to which it varies among taxa; and (4) to test whether
there is potential for enemy-mediated interactions between
plant species via shared seed predators.
METHODS
Study site
Our study targeted tree, shrub and liana species on Barro Col-
orado Island (BCI), a 15.6 km2 island, located in Lake Gatun
in central Panama. Its flora is exceptionally well known (e.g.
Croat 1978; Leigh 1999), and long-term studies of a 50-ha for-
est dynamics plot (Condit 1998; Hubbell et al. 1999, 2005)
have yielded unprecedented data on spatial and temporal pat-
terns of seed and fruit production (e.g. Harms et al. 2000;
Wright et al. 2005; Wright & Calder
on 2006). Seed predation
by insects has been studied in detail on a small number of
plant species (e.g. Wright 1983; Jones & Comita 2010; Visser
et al. 2011), but there has been limited investigation of com-
munity-level patterns of insect seed predation (but see Pinz
on-
Navarro et al. 2010).
Documenting plant-seed predator relationships
Host relationships for endophagous insect seed predators can
be established and quantified conveniently by rearing insects
from collected seeds or fruits. We aimed to collect at least 200
seeds/fruits of each woody plant species fruiting during our
study. To achieve this, we adopted several approaches. First,
sections of the ~ 40 km BCI trail network were walked each
week and freshly fallen, intact-looking seeds (individual dias-
pores) and fruits (seed-bearing structures with one or more
seeds) were collected from the forest floor. Seeds/fruits on
branches that could be reached by hand or using a telescoping
pole pruner were also occasionally collected to increase sam-
ples sizes. Second, each month we generated a target list that
included species known to produce seeds/fruits at that time of
the year (based on data from S. J. Wright’s long-term seed
monitoring project; Zimmerman et al. 2007) and for which
the opportunistic trail walks had not yet yielded the target of
200 seeds/fruits. A team of botanical field technicians was
tasked with collecting seeds from species on this list, generat-
ing samples of species not encountered during the trail walks
and boosting sample sizes for rarer species. Third, we also
obtained seeds and fruits from a network of seed fall traps in
the 50-ha forest dynamics plot (for details, see below). Seed
collections took place between July 2010 and November 2013.
Although our target was the plant community of BCI, a small
subset (2.4%) of seed/fruit samples were collected on sur-
rounding mainland peninsulas within approximately 2 km of
BCI with a similar forest composition.
Seeds and fruits were sorted according to species and degree
of maturity (mature/immature), counted, and stored in plastic
pots lined with absorbent paper and covered with fine nylon
mesh. All samples (seeds/fruits of a given species collected on
a given day at a particular site) were identified to species. The
pots were stored on shelves in a shade house in conditions
resembling ambient forest understory conditions for three
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
2 S. Gripenberg et al. Letter
months. Given the biology of endophagous seed predators, we
consider the possibility of transmission of insects between
individual seeds/fruits in the rearing pots extremely unlikely.
During the rearing period, each pot was checked approxi-
mately twice a week and emerging adults and larvae were
removed, processed and identified by morphological or molec-
ular methods (Supporting Information, Appendix S1). At the
end of the 3-month rearing period each seed was carefully
examined and dissected to check for evidence of insect attack
such as exit holes, feeding tunnels, presence of frass or dead
or living insects. Seeds with apparent insect damage but where
other possibilities (such as fungal attack causing the interior
of the seed to deteriorate) could not be ruled out were not
scored as predated by insects.
We derived three variables reflecting susceptibility of plant
species to seed predator attack: the incidence of seed predators
(whether seed predators were reared from the plant species),
seed predator richness (number of seed predator species reared
from the plant species) and seed predation rate (proportion of
dissected seeds showing signs of seed predator attack). To
account for dependence of the incidence of seed predators and
seed predator richness on sample size, we restricted certain
analyses (identified below) to a subset of the data comprising
plant species with ≥ 200 seeds (hereafter referred to as well-
sampled plant species). Above this sample size, a statistically
significant relationship between the likelihood of seed preda-
tor detection and sample size could not be detected (Fig. S1).
To check the robustness of our results to identification errors,
we repeated several analyses excluding interactions docu-
mented only once (details below).
Role of plant phylogeny and traits on patterns of seed predator
host use
To test for phylogenetic signal in patterns of seed predator
attack across the plant community we used the D statistic
(Fritz & Purvis 2010) for incidence and Blomberg’s K (Blom-
berg et al. 2003) for seed predator richness and seed predation
rates. In all cases, we used a phylogeny provided by David
Erickson (Smithsonian Institution) that was constructed fol-
lowing methods in Kress et al. (2009). This phylogeny
included 362 of our sampled plant species. Where sampled
plant species were missing from the phylogeny but congeners
were present (n = 59 species), they were added at the root of
Table 1 Traits hypothesised to influence plant species’ susceptibility to attack by internally-feeding seed predators. In the context of our study, the prone-
ness of plant species to seed predator attack was assessed as incidence of seed predators, seed predator richness (number of seed predator species observed
on each plant species) and seed predation rates (assessed by seed dissection). For details on how individual variables were estimated, see Appendix S2
Variable Predicted relationship
Variables reflecting resource availability at various spatial scales
Local seed
abundance
Species that are locally abundant at the seed stage are more prone to seed predator attack than species that are rare since they are
more likely to be colonised by and to sustain viable seed predator populations (e.g. Pacala & Crawley 1992; Hanski 2001)
Maximum tree
height
Species with large growth forms (e.g. canopy trees) are more prone to seed predator attack than species with smaller growth forms
(e.g. shrubs and understory trees) since local seed crop sizes are likely to be bigger and more apparent to enemies (Janzen 1968)
Confamilial species
on BCI
Species with many confamilial species in the local plant community are more prone to seed predator attack than phylogenetically
isolated species (Janzen 1968), since the abundance of resources available to seed predators specialised at the family level will be
higher
Congeneric species
on BCI
Species with many congeneric species in the local plant community are more prone to seed predator attack than phylogenetically
isolated species (Janzen 1968) since the abundance of resources available to seed predators specialised at the genus level will be
higher
Local abundance of
adult trees
Species that are locally abundant as adults are more prone to seed predator attack than species that are locally rare since they are
more likely to be colonised by and to sustain viable seed predator populations (e.g. Pacala & Crawley 1992; Hanski 2001)
Variables reflecting seed size and investment in seed defences
Seed mass Species with large seeds are more prone to seed predator attack than species with small seeds since their seeds provide larger
quantities of resources to developing seed predators (cf. Fenner et al. 2002)
Endocarp
investment
Species that invest little in protective tissues surrounding the seeds are more prone to seed predator attack than species that invest
large amounts of resources in seed protection (cf. Kuprewicz & Garc
ıa-Robledo 2010)
Polyphenol
concentration
Species with low investment in polyphenol production are more prone to seed predator attack than species with high polyphenol
concentrations in their seeds (Janzen 1971; McArt et al. 2013)
Variables reflecting temporal patterns in fruit fall
Interannual
variation in seed
crop sizes
Species with temporally predictable fruiting patterns are more prone to seed predator attack than species with large interannual
variation in fruit crop sizes (Janzen 1971, 1976) since they provide a more stable resource base for specialist seed predators
Fruiting season The proneness to seed predation varies between species fruiting in the wet versus the dry season if the abundance of seed predators
varies with patterns of rainfall (cf. Wolda 1988)
Overlap in fruit
production by
other species
Species fruiting at times of the year when few other species fruit are more prone to seed predator attack than species that fruit
when many co-fruiting species fruit (Kelly 1994)
Other
Growth form Lianas are more prone to seed predator attack than trees, since they invest less in defence chemicals (Asner & Martin 2012; but see
Gripenberg et al. 2018)
Relative growth rate
(RGR)
Fast-growing species tend to invest less in defense and are therefore more prone to seed predator attack than slow-growing species
(Coley et al. 1985)
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
Letter Host use by insect seed predators 3
the genus using the add.species.to.genus function of the R
package phytools (Revell 2012). Analyses were implemented
using the phylo.d function of the R package caper (Orme
et al. 2018) and the multiPhylosignal function in the R pack-
age picante (Kembel et al. 2010). For details, see
Appendix S1.
To assess whether particular traits or trait values hypothe-
sised to make plants particularly susceptible to enemies
(Table 1) influenced seed predator attack we compiled data on
relevant plant species traits from a range of sources
(Appendix S2). This data set was at least 50% complete for
all of our well-sampled species (Appendix S1). We used ran-
dom forest models (Cutler et al. 2007) to examine whether
these traits were informative for predicting seed predator inci-
dence, richness, and seed predation rates. Random forest
algorithms combine many classification or regression tree
models to produce more robust predictions. The technique
can identify valuable predictors from heterogeneous data
sources with missing values and does not require the prior
specification of relationships between traits and response vari-
ables. Analyses were conducted for each of our three response
variables. Only well-sampled plant species (n = 214) were
included in the main analyses. Since we tested a mixture of
categorical and continuous predictors, to minimise variable
selection bias we used conditional inference trees via the cfor-
est function in the R package party (Strobl et al. 2007).
Model hyperparameters are given in Appendix S1. To assess
the overall ability to predict seed predator incidence, we used
Cohen’s j Accuracy
Baseline Accuracy1
Baseline Accuracy

 
to determine the extent to
which the model can improve upon an expected level of classi-
fication accuracy (Landis & Koch 1977). For seed predator
richness and seed predation rates we based our assessment on
pseudo-R2 values. We compared the relative importance of
each predictor trait using variable importance plots, which
measure the mean decrease in accuracy of the model when the
focal trait is randomised. We also conducted parallel analyses
in which all plant species were included along with sample size
as an additional predictor variable.
Constructing quantitative plant-seed predator food webs
To provide a visual and quantitative description of patterns
of host use by seed predators, a set of quantitative food webs
(one overall web and two taxon-specific webs focusing on
Coleoptera and Lepidoptera, respectively) were constructed
and analysed using the R package bipartite (Dormann et al.
2008). To construct a fully quantified food web (e.g. M€uller
et al. 1999; Lewis et al. 2002), information on host densities
and the frequency of trophic interactions between hosts and
enemies is needed.
Estimating host densities
Data on species-specific seed abundances per unit area were
obtained from S. J. Wright’s long-term study on seed and fruit
production in the 50-ha plot (see Appendix S1). In brief, we
used information on the number of seeds and fruits falling into
a network of 200 seed traps (each 0.5 m2) during 1987–2015,
along with information on the average number of seeds per
fruit. Since many seed predator species in our data set are likely
to be pre-dispersal seed predators (S. Gripenberg, pers. obs.),
we quantified host seed abundances based on both immature
and mature fruits and seeds. Of 486 plant species observed in
the seed traps, 357 (74.5%) were collected for insect rearing. Of
the 129 species we failed to collect, 75 were not detected in the
seed traps during the period of our study and the remainder
were collected in very small numbers (<25 seeds). Conversely,
seeds were collected for insect rearing from 122 plant species
that did not appear in the seed traps; these species were
excluded from the food webs but used in other analyses.
Estimating the number of seeds killed per seed predator species
For each predator 9 prey association in the data set, we esti-
mated the typical number of insect individuals emerging per
attacked seed by dividing the number of insect individuals
emerging by the number of seeds scored as predated. In most
cases, only one insect could develop successfully in a seed.
Since assessing the number of Hymenoptera individuals per
infested seed proved difficult, our food webs do not include
Hymenoptera (which comprised only 2% of seed predator
individuals). We encountered multiple cases where one or sev-
eral seed predator individuals fed inside fruits containing mul-
tiple small seeds (e.g. Ficus spp., Apeiba spp.), killing an
unknown proportion of seeds. Since our methods did not
allow us to estimate the proportion of seeds killed by these
seed predators, these species (n = 16, 4.5% of the species that
could potentially have been included in the food web) were
excluded. A further 22 (6.2%) plant species were excluded
from the food webs since the metadata needed to assess inter-
action frequencies (e.g. mean number of seeds per fruit) were
not available. The resulting food web data set included infor-
mation for 322 predator-prey associations.
Seed predator specialisation
The degree of seed predator specialisation is reflected visually
in the quantitative food web plots, and was assessed quanti-
tatively using Bl€uthgen et al.’s species-level specialisation
index (d’). This metric can be interpreted as the ‘deviation of
the actual interaction frequencies from a null model which
assumes that all partners are used in proportion to their
availability’ (Bl€uthgen et al. 2006). To facilitate comparisons
between taxa, we calculated d’ values for species within each
order (Coleoptera, Lepidoptera) in the quantitative food web
as well as for all seed predator taxa combined. We also cal-
culated a specialisation index for the entire network (H2’;
Bl€uthgen et al. 2006). Using the larger data set that included
species not found in the seed traps in the 50-ha forest
dynamics plot we also produced histograms showing the fre-
quency distribution of species against different diet richness
values (the number of plant species recorded in the diet of
each seed predator species; Futuyma & Gould 1979). As for
d’, we examined the distribution of diet richness for all seed
predators combined, as well as separately for each insect
order (including Hymenoptera). Since limited sampling effort
could inflate apparent specialisation, we limited diet richness
analyses to insect species with a sample size of ≥ 10
individuals.
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
4 S. Gripenberg et al. Letter
Potential for enemy-mediated indirect interactions
Using the food web data set, we assessed the potential for
seed predator-mediated indirect interactions following M€uller
et al. (1999):
dij ¼
X
k
ajkP
l ail
ajkP
m amk
 
where dij is the probability that that a seed predator attacking
species i developed on species j, a is the link frequency, the
summation m is over all host species from 1 to H (where H is
the total number of host species), k is a seed predator species,
and the summation l is over all seed predator species from 1
to P (the total number of seed predator species). Pairwise dij
values were calculated for the quantitative food web data set
using the PAC function in bipartite. To assess whether the
potential for apparent competition was highest for closely-re-
lated plant species, we used Spearman’s rank correlation to
assess whether there was a relationship between PAC values
and pairwise phylogenetic distances. Species pairs for which
PAC = 0 were excluded from this analysis.
RESULTS
Community-level patterns of seed predator attack
We collected 9325 samples totalling 207 201 seeds and fruits
representing 478 plant species, 281 genera and 78 families
(Table S1). We reared 21 604 adult insects and 3843 larvae
from these samples. 13 293 (61.5%) insect individuals repre-
senting 369 (morpho)species were considered highly likely to
be seed predators based on principles outlined in
Appendix S1. Of these, the majority (221 species; 11,063
individuals) were Coleoptera (notably Curculionidae and
Bruchinae), while the remainder were Lepidoptera (111 species
representing ten families; 1965 individuals) and Hymenoptera
(37 species of Eurytomidae; 254 individuals). A total of 471
interactions between plant species and seed predator species
was documented. Table S2 provides a list of these interactions
along with taxonomic information and Barcode Index Num-
bers (Ratnasingham & Hebert 2013) of seed predator species.
Data on sequenced insect specimens (including images) are
available on BOLD (www.boldsystems.org) accessible through
https://doi.org/10.5883/DS-BCISP.
Seed predators were reared from 199 of the 478 sampled
plant species (41.6%). Since the likelihood of rearing seed
predators increased with sample size (Figure S1), the true pro-
portion of plant species attacked by seed predators is likely to
be higher. Considering only the 223 well-sampled plant spe-
cies, 142 species (63.6%) were attacked by seed predators. The
average number of seed predator species per attacked plant
species was 2.4 (range 1–8; median 2) for the full data set,
and slightly higher (2.6; range 1–8; median 2) for well-sampled
species (Fig. 1). Across the network, estimated interaction
coverage was 0.74 (Chao1 estimator; Appendix S3).
Seed predators were associated with plant species across a
wide range of the plant phylogeny (Fig. 2). For well-sampled
plant species, a weak phylogenetic signal in incidence of seed
predators was detected (D = 0.779, PD<1 = 0.006; see Table S3
for results from order-specific analyses which are qualitatively
similar). Although close to being statistically significant, the
phylogenetic signals in seed predator richness (K = 0.004,
P = 0.051) and seed predation rate (K = 0.006, P = 0.082)
were negligible.
Plant traits as predictors of community-level patterns of seed
predator attack
Using the data set which included only well-sampled plant spe-
cies, the variables examined in the random forest models only
explained a small proportion of the variation in observed pat-
terns of seed predator attack across the plant community. In
terms of seed predator incidence, Cohen’s j was 0.049, indicating
that the predictor variables added very little information (Landis
& Koch 1977). Likewise, the pseudo-R2 values of the models
assessing the effect of seed predator richness (R2 = 0.098) and
seed predation rate (R2 = 0.047) were very small.
Across the three response variables, the most important pre-
dictor variables were seed mass and maximum tree height
(Appendix S4). Taller trees and larger seeds were more likely
to be predated (Fig. 3), although the overall explanatory
power of these relationships was weak. The relationship
between seed predator incidence and all studied plant traits
are shown in Figure S3. The results from the analyses on the
full data set (n = 478 species) were qualitatively similar
(Appendix S4).
Host specificity of seed predators
The quantitative plant-seed predator web included 254 seed
predator species feeding on 141 plant species (Fig. 4). Of
35 814 potential links only 322 were realised, yielding a
0
20
40
60
80
1 2 3 4 5 6 7 8
Observed number of seed
N
um
be
r o
f p
la
nt
 s
pe
ci
es
Seed Unit
Sample Size
< 200
= 200
predator species
Figure 1 Frequency histogram showing the number of plant species
associated with different numbers (n = 1–8) of seed predator species.
Poorly sampled plant species (< 200 seeds/fruits collected for insect
rearing) are shown in light grey. The number of species with no seed
predator species were 95 and 212 for well-sampled and poorly-sampled
plant species, respectively.
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
Letter Host use by insect seed predators 5
connectance value of 0.009. The H2’ value of the network was
0.970, that is, close to the maximum possible value of 1. H2’
for the Coleoptera and Lepidoptera webs was 0.975 and
0.974, respectively. Specialisation values at the level of individ-
ual species, d’, ranged from 0.118 to 1 (median 0.830;
Fig. 5a). Host specialisation (d’) did not differ significantly
between Coleoptera and Lepidoptera (Wilcoxon rank sum
test, W = 691, P = 0.284; analysis including insect species with
≥ 10 individuals).
The diet richness of seed predators (examined using the full
insect rearing data set) showed that seed predator species fed on
an average of 1.28 plant species (range 1–7). The diet richness
was highly skewed towards monophagous species: of the 135
best-sampled seed predator species (≥ 10 individuals reared), 85
(63%) were reared from a single plant species (Fig. 5b). The per-
centage of extreme specialists (species associated with a single
host plant species) was 66.3, 50.0 and 70.0%) for Coleoptera,
Lepidoptera and Hymenoptera respectively.
Not predated
Predated
Figure 2 Presence (black circle) or absence (open circle) of seed predators and incidence of seed predator orders (Coleoptera, Lepidoptera, Hymenoptera)
plotted against a plant phylogeny that includes only plant species with a minimum sample size of 200 seeds/fruits. Plant species names can be seen in the
larger phylogeny shown in Figure S2. The figure was drawn using the R package ggtree (Yu et al. 2017).
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
6 S. Gripenberg et al. Letter
In 37.5% of cases where a seed predator species was reared
from multiple plant species the host species were congeneric
and in 61.1% they were confamilial. Measures of specialisa-
tion were robust to the inclusion of poorly-sampled interac-
tions and seed predator species (Table S4).
Potential for seed predator mediated indirect interactions between
plant species
The potential for seed predator-mediated apparent competi-
tion between plant species was very low. In the data for the
141 plant species that were part of the quantitative food web,
only 68 (0.34%) of the 19,740 potential pairwise enemy-medi-
ated interactions between plant species showed PAC values
> 0.1 (corresponding to a situation where 10% of the enemies
attacking the focal plant species are predicted to have devel-
oped as larvae on the alternative host species). Considering
the species pairs for which PAC > 0, there was a significant
negative relationship between PAC and pairwise phylogenetic
distances (Spearman’s r = 
0.249, P < 0.001; Fig. S6).
DISCUSSION
Documenting and understanding the ecological implications
of the myriad of plant-enemy interactions in hyper-diverse
forest communities is among the most challenging tasks facing
tropical forest ecologists. Most such studies focus on particu-
lar plant or enemy taxa (e.g. Endara et al. 2017), or on sub-
sets of plant species (e.g. Novotny et al. 2002). Our study
includes approximately three quarters of the 652 woody plant
species recorded at the study site (Croat 1978), and 90.5% of
the species recorded as producing fruit there during the study.
To the best of our knowledge, this is the largest and most
complete data set yet available on plant-enemy interactions in
any tropical community, and it focuses on a guild of plant
enemies which, through their direct effects on seed mortality,
are expected to have pronounced effects on plant populations
and communities.
We identified and quantified host associations for 369 seed
predator species. While there was some evidence of phyloge-
netic clustering in patterns of attack, seed predators were
widely distributed across the plant community (Fig. 1), with
few plant clades and only 36.3% of well-sampled plant species
escaping this enemy guild. Clades where the majority of plant
species were attacked were found within the Malpighiaceae
and Fabaceae, whereas shrubs in the genus Psychotria (Rubi-
aceae) and the Melastomataceae genera Mouriri, Conostegia,
Leandra and Miconia typically lacked seed predators.
One key aim of our study was to link patterns of seed
predator attack to plant traits. Previous studies have found
that interspecific variation in enemy attack can be explained
by interspecific variation in particular traits or suites of traits
(e.g. Carmona et al. 2011; Turcotte et al. 2014), including
studies of leaf herbivory in tropical forests (e.g. Coley 1983,
Schuldt et al. 2012, C
ardenas et al. 2014). However, our pre-
dicted associations between plant traits and seed predator
attack were only weakly supported. The variables explaining
the greatest proportion of observed variation were maximum
tree height and seed mass. Traits associated with abundant
resources were therefore associated with enemy attack, consis-
tent with our predictions. Nevertheless, the plant traits investi-
gated were poor predictors of community-level patterns of
seed predator attack, suggesting that seed predator host use is
likely to be either stochastic or driven primarily by factors not
investigated in this study. For example, seed predators may
respond to chemical compounds other than polyphenols (e.g.
Birch et al. 1986), or their patterns of host use might be influ-
enced by predation risk (Lill et al. 2002).
Our most striking result is the remarkable degree of host
specificity of seed predators. Although seed samples were
Figure 3 Spine plots showing relationships between the incidence of seed predation and (a) tree height and (b) seed mass. Equivalent plots for other traits
are given in Figure S3 and partial dependence plots showing the predictions of random forest models as trait variables are changed are shown in
Appendix S4. Note that panel a) includes only free-standing species, since height (which is used as a proxy for seed crop size; Table 1) is not a meaningful
trait for lianas. Internally-feeding seed predators were never observed on species with seed masses smaller than 10
3 g (indicated by a star in panel b).
Below this seed size, individual seed predators were consuming multiple seeds within a fruit.
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
Letter Host use by insect seed predators 7
obtained from 478 plant species, no seed predator species was
recorded from more than seven host species, and 63% of the
best-sampled seed predator species (those with ≥ 10 individu-
als reared) were restricted to a single plant species. Given the
local rarity and patchy distribution of most plant species in
tropical forests (Condit et al. 2000), these specialised insects
must have evolved highly efficient strategies for host location.
Host specificity was consistently high across insect orders.
This contrasts with results from studies on fruit-associated
insects in Papua New Guinea (Ctvrtecka et al. 2014; Sam
et al. 2017), where Curculionidae showed higher levels of spe-
cialisation than Lepidoptera. A possible explanation for the
discrepancies in results between the two sites might be our
more explicit focus on seed-eating insects (i.e. plant antago-
nists). Had we included Lepidopteran families with members
that are primarily pulp-eaters or detritivores rather than seed-
eaters (e.g. Tineidae, Blastobasidae), we would likely have
detected broader host ranges in Lepidoptera (S. Gripenberg;
pers. obs.).
The high degree of host specificity is consistent with a role
for seed predators in contributing to the maintenance of tropi-
cal forest plant diversity. The Janzen–Connell hypothesis (Jan-
zen 1970; Connell 1971) proposes that enemy-mediated
negative density-dependence favours plants that are locally
rare, thereby promoting diversity at the community level. The
Janzen–Connell mechanism requires that enemies are reason-
ably host specific (Sedio & Ostling 2013). Like the seed-eating
beetles of Guanacaste studied by Janzen (1980), the seed
predators on BCI fulfil this criterion. Studies exploring spatial
variation in seed predation rates on individual plant species
are needed to evaluate further the role of seed predators in
suppressing the reproductive output of their hosts in areas
where the host is common (Gripenberg 2018). Our whole-
community focus meant that we were unable to quantify spa-
tial variation in seed predation in relation to seed or plant
abundance. Nevertheless, predation rates varied greatly
among seed samples for individual species (S. Gripenberg,
unpublished data), suggesting high spatial variation in rates of
insect attack. Further studies are needed to investigate the
extent to which this variation is caused by spatial variation in
local seed densities.
Studies on tropical insect folivores have revealed lower
levels of specialisation than previously assumed (e.g.
Novotny et al. 2002). This opens up the possibility that
insect enemies link the dynamics of tree populations indi-
rectly, via apparent competition. However, few studies have
tested for such indirect interactions, probably because pre-
dicting which species are most likely to interact via shared
enemies requires accurate quantification of networks of
trophic interactions across whole communities, data which
are particularly challenging to collect in species-rich tropical
forests. On BCI we found little potential for apparent com-
petition mediated by shared insect seed predators, one conse-
quence of the high degree of seed predator host specificity
documented. Given the unique features of insect seed preda-
tors such as their sessile feeding habit and their dependency
on a resource that is typically only available for a small part
of the year, the results from our study must not be uncriti-
cally extrapolated to other types of plant-eating insects such
as free-feeding leaf herbivores. For example, seed predators
are likely to be more host-specific than many other herbivo-
rous insect guilds (Novotny et al. 2010), leading to different
effects on their host plant communities. Similarly, our results
from BCI should also not be extrapolated uncritically to
other sites: As noted above, Ctvrtecka et al. (2014) and Sam
et al. (2017) found that seed- and fruit-feeding insects in the
lowland rainforests of Papua New Guinea were less spe-
cialised than those on BCI, so we cannot rule out a role for
seed predators as agents of apparent competition in other
tropical forests.
In summary, we have compiled and quantified the most
complete network of trophic interactions yet documented for
rainforest plants and their consumers. Focusing on a poorly-
studied group of plant enemies we found that species varied
widely in their susceptibility to insect seed predator attack,
partly explained by variables reflecting local resource
A
ea
tu
s 
sp
. c
ur
26
3S
G
C
ur
c 
sp
. c
ur
18
0S
G
A
ea
tu
s 
co
st
ul
at
us
A
ea
tu
s 
sp
. c
ur
24
2S
G
P
yr
a 
sp
. l
ep
81
S
G
A
ea
tu
s 
sp
. c
ur
26
2S
G
C
ur
c 
sp
. c
ur
18
4S
G
C
ry
pt
or
hy
nc
hu
s 
sp
. c
ur
54
S
G
C
ly
do
no
pt
er
on
 p
om
po
ni
us
A
ea
tu
s 
sp
. c
ur
10
8S
G
C
ur
c 
sp
. c
ur
17
7S
G
S
em
no
rr
hy
nc
hu
s 
sp
. c
ur
44
S
G
C
ur
c 
sp
. c
ur
18
8S
G
S
em
no
rr
hy
nc
hu
s 
fu
lv
op
ic
tu
s
C
on
ot
ra
ch
el
us
 v
er
tic
al
is
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
18
2S
G
S
es
i s
p.
 le
p1
01
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
13
8S
G
C
os
m
 s
p.
 le
p4
25
S
G
M
yr
m
ex
 s
p.
 c
ur
48
S
G
M
yr
m
ex
 s
p.
 1
28
S
G
G
el
e 
sp
. l
ep
26
4S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
18
1S
G
C
he
lo
to
ny
x 
sp
. 1
S
G
G
el
e 
sp
. l
ep
19
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
14
1S
G
A
nc
ho
nu
s 
sp
. c
ur
62
S
G
C
ur
c 
sp
. c
ur
18
7S
G
R
hy
ss
om
at
us
 s
p.
 c
ur
72
S
G
C
ur
c 
sp
. c
ur
12
9S
G
A
m
bl
yc
er
us
 p
er
fe
ct
us
A
m
bl
yc
er
us
 s
p.
 b
ru
17
S
G
C
ur
c 
sp
. c
ur
17
4S
G
Le
ch
rio
ps
 s
p.
 c
ur
97
S
G
D
ep
r E
la
c 
sp
. l
ep
8S
G
A
po
ge
sh
na
 s
te
ni
al
is
M
eg
ac
er
us
 lu
nu
la
tu
s
S
pi
lo
m
el
in
ae
 s
p.
 8
0Y
B
O
ec
o 
sp
. l
ep
4S
G
H
ei
lip
us
 s
p.
 c
ur
15
S
G
C
ur
c 
sp
. c
ur
17
5S
G
P
yr
a 
sp
. l
ep
30
5S
G
C
ur
c 
sp
. c
ur
17
9S
G
C
ur
c 
sp
. c
ur
20
4S
G
H
ei
lip
us
 d
ra
co
R
ic
ul
or
am
ph
a 
an
cy
lo
id
es
Pa
gi
oc
er
us
 fr
on
ta
lis
H
is
tu
ra
 p
an
am
an
a
P
yr
a 
sp
. l
ep
34
6S
G
P
yr
a 
sp
. l
ep
24
8S
G
N
eo
po
ta
m
ia
 s
tre
bl
op
a
H
ei
lip
us
 s
p.
 c
ur
24
3S
G
H
ei
lip
us
 s
p.
 c
ur
50
S
G
C
ur
c 
sp
. c
ur
17
6S
G
G
el
e 
sp
. l
ep
42
7S
G
O
ec
o 
sp
. l
ep
29
3S
G
O
ec
o 
sp
. l
ep
28
9S
G
C
os
m
 s
p.
 le
p2
74
S
G
A
try
pa
riu
s 
sp
. 1
S
G
N
ea
r G
ym
na
nd
ro
so
m
a 
sp
. l
ep
35
6S
G
A
m
bl
yc
er
us
 s
p.
 b
ru
21
S
G
P
yr
a 
sp
. l
ep
44
9S
G
C
on
ot
ra
ch
el
us
 tu
rb
at
us
E
ub
ul
us
 s
p.
 c
ur
90
S
G
C
ur
c 
sp
. c
ur
19
0S
G
C
ur
c 
sp
. c
ur
10
3S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
16
3S
G
M
yr
m
ex
 s
p.
 c
ur
94
S
G
C
ur
c 
sp
. c
ur
20
6S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
15
2S
G
C
ry
pt
or
hy
nc
hu
s 
sp
. c
ur
26
S
G
Le
pt
ur
ge
s 
sp
. 1
S
G
C
er
am
by
ci
da
e 
sp
. c
ol
ce
r1
S
G
C
er
am
by
ci
da
e 
sp
. c
ol
ce
r3
S
G
Le
ch
rio
ps
 p
ar
ot
ic
a
C
ol
ob
ot
he
a 
sp
. 1
S
G
P
la
ty
no
ta
 s
p.
 le
p1
95
S
G
E
ul
ec
hr
io
ps
 s
p.
 c
ur
53
S
G
H
el
i s
p.
 le
p2
53
S
G
C
ur
c 
sp
. c
ur
19
8S
G
C
ur
c 
sp
. c
ur
19
3S
G
C
on
ot
ra
ch
el
us
 p
um
ili
o
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
92
S
G
C
ur
c 
sp
. c
ur
20
3S
G
C
ur
c 
sp
. c
ur
20
1S
G
C
ur
c 
sp
. c
ur
21
3S
G
S
ch
ac
on
tia
 A
A
P
19
91
P
yr
a 
sp
. l
ep
29
1S
G
P
yr
a 
sp
. l
ep
30
7S
G
G
eo
by
rz
a 
no
di
fe
ra
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
30
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
22
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
16
4S
G
A
nc
ho
nu
s 
sp
. c
ur
32
S
G
C
ur
c 
sp
. c
ur
18
3S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
12
4S
G
G
el
e 
sp
. l
ep
10
6S
G
C
ar
m
en
ta
 s
p.
 le
p5
39
S
G
C
ar
m
en
ta
 s
p.
 le
p1
S
G
P
yr
a 
sp
. l
ep
31
0S
G
C
ur
c 
sp
. c
ur
17
3S
G
O
ec
o 
sp
. l
ep
13
S
G
O
ec
o 
sp
. l
ep
46
4S
G
D
ep
r E
la
c 
sp
. l
ep
18
6S
G
B
ot
hr
yo
pt
er
on
 d
ar
lin
gt
on
i
A
nt
ho
no
m
in
ae
 s
p.
 2
47
S
G
A
nt
ho
no
m
us
 s
p.
 c
ur
13
3S
G
B
ot
hr
yo
pt
er
on
 s
p.
 c
ur
85
S
G
G
el
e 
sp
. l
ep
32
6S
G
A
po
ge
sh
na
 s
te
ni
al
is
D
H
J0
2
R
hy
ss
om
at
us
 s
p.
 c
ur
17
4S
G
P
yr
a 
sp
. l
ep
9S
G
E
ub
ul
us
 fu
lv
os
qu
am
is
C
ry
pt
as
pa
sm
a 
pe
rs
ea
na
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
10
4S
G
C
ar
m
en
ta
 s
p.
 le
p1
11
S
G
B
or
is
au
le
te
s 
sp
. c
ur
1S
G
P
yr
a 
sp
. l
ep
11
7S
G
B
ar
is
 s
p.
 c
ur
7S
G
P
se
ud
ob
ar
is
 s
p.
 c
ur
19
S
G
P
se
ud
ob
ar
is
 s
p.
 c
ur
14
5S
G
Pa
ch
ym
er
us
 s
p.
 b
ru
67
S
G
P
ac
hy
m
er
us
 b
ac
tri
s
M
et
am
as
iu
s 
sp
. c
ur
13
4S
G
P
la
ty
no
ta
 s
p.
 le
p9
9S
G
A
nt
ho
no
m
us
 s
p.
 c
ur
45
S
G
G
el
e 
sp
. l
ep
30
8S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
13
2S
G
B
ru
c 
sp
. b
ru
45
S
G
B
ru
c 
sp
. b
ru
44
S
G
A
m
bl
yc
er
us
 s
p.
 b
ru
59
S
G
A
m
bl
yc
er
us
 s
p.
 b
ru
22
S
G
A
m
bl
yc
er
us
 c
ha
m
pi
on
i
A
ea
tu
s 
ve
st
itu
s
La
m
pr
os
pi
lu
s 
co
llu
ci
a
H
yp
oc
os
m
ia
 b
im
ac
ul
al
is
C
ar
m
en
ta
 fo
ra
se
m
in
is
B
us
ck
ie
lla
 le
cy
th
id
is
A
ea
tu
s 
sp
. c
ur
14
4S
G
P
yc
no
th
ea
nt
is
 s
p.
 c
ur
47
S
G
G
er
ae
us
 s
p.
 c
ur
51
S
G
C
ur
c 
sp
. c
ur
19
2S
G
C
ur
c 
sp
. c
ur
17
2S
G
C
os
m
 s
p.
 le
p3
4S
G
G
ra
ph
ol
ita
 y
ur
ub
in
a
E
rio
so
ci
a 
gu
tti
fe
ra
G
el
e 
sp
. l
ep
41
S
G
C
ur
c 
sp
. c
ur
21
0S
G
C
ur
c 
sp
. c
ur
17
8S
G
A
tra
ct
om
er
us
 c
al
ig
at
us
Le
ch
rio
ps
 s
p.
 c
ur
40
S
G
E
ub
ul
us
 s
p.
 c
ur
24
S
G
C
ur
c 
sp
. c
ur
19
6S
G
A
nt
ho
no
m
us
 s
p.
 c
ur
14
S
G
R
ic
ul
a 
sp
. l
ep
21
8S
G
P
yr
a 
sp
. l
ep
23
1S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
17
5S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
10
S
G
G
el
e 
sp
. l
ep
3S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
8S
G
C
er
at
op
us
 s
p.
 c
ur
2S
G
G
el
e 
sp
. l
ep
30
2S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
68
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
24
9S
G
A
na
ca
m
ps
is
 p
hy
to
m
ie
lla
C
yr
io
ny
x 
sp
. c
ur
91
S
G
C
yr
io
ny
x 
sp
. c
ur
69
S
G
P
hy
m
at
op
ho
su
s 
sp
. c
ur
19
7S
G
P
hy
m
at
op
ho
su
s 
sp
. c
ur
10
7S
G
C
ur
c 
sp
. c
ur
19
9S
G
R
ic
ul
a 
sp
. 5
45
S
G
R
ic
ul
a 
la
ci
st
em
a
C
ur
c 
sp
. c
ur
18
6S
G
H
el
i s
p.
 le
p5
2S
G
H
el
i s
p.
 le
p1
02
S
G
C
oc
hy
lis
 s
p.
 le
p9
7S
G
S
pe
ci
om
er
us
 g
ig
an
te
us
B
ru
c 
sp
. b
ru
66
S
G
B
ru
c 
sp
. b
ru
54
S
G
P
se
ud
om
op
si
s 
sp
. c
ur
34
S
G
P
se
ud
om
op
si
s 
sp
. c
ur
25
4S
G
A
nt
ho
no
m
us
 s
p.
 c
ur
15
0S
G
P
se
ud
om
op
si
s 
sp
. c
ur
16
8S
G
P
hy
m
at
op
ho
su
s 
sp
. c
ur
37
S
G
A
nt
ho
no
m
us
 s
p.
 c
ur
13
6S
G
S
en
ni
us
 s
p.
 b
ru
8S
G
S
en
ni
us
 s
p.
 b
ru
58
S
G
A
m
bl
yc
er
us
 ta
ch
ig
al
ia
e
P
yr
a 
sp
. l
ep
31
2S
G
O
ec
o 
sp
. l
ep
31
1S
G
O
xy
te
no
pt
er
us
 s
p.
 c
ur
55
S
G
O
ec
o 
sp
. l
ep
25
0S
G
M
im
os
es
te
s 
en
te
ro
lo
bi
i
G
el
e 
sp
. l
ep
29
5S
G
M
ae
m
ac
te
s 
sp
. c
ur
75
S
G
G
el
e 
sp
. l
ep
27
5S
G
E
ut
ox
us
  s
p.
 c
ur
21
S
G
C
ur
c 
sp
. c
ur
20
5S
G
C
ur
c 
sp
. c
ur
18
2S
G
C
ur
c 
sp
. c
ur
18
1S
G
C
os
m
 s
p.
 le
p5
42
S
G
C
os
m
 s
p.
 le
p4
99
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
18
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
17
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
17
6S
G
C
on
ot
ra
ch
el
us
 p
os
tic
at
us
C
er
co
no
ta
 a
no
ne
lla
C
ur
c 
sp
. c
ur
20
S
G
C
ar
m
en
ta
 s
p.
 le
p1
31
O
ec
o 
sp
. l
ep
19
2S
G
B
ar
is
 s
p.
 c
ur
46
S
G
A
nt
ho
no
m
us
 s
p.
 c
ur
24
5S
G
A
nt
ho
no
m
us
 s
p.
 c
ur
15
8S
G
C
ur
c 
sp
. c
ur
19
4S
G
A
nt
ho
no
m
in
ae
 s
p.
 2
48
S
G
C
os
m
 s
p.
 le
p3
13
S
G
A
m
bl
yc
er
us
 v
eg
ai
P
yr
al
id
ae
 s
p.
 4
96
S
G
P
yr
a 
sp
. l
ep
47
8S
G
P
se
ud
ob
ar
is
 s
p.
 c
ur
20
2S
G
P
se
ud
ob
ar
is
 s
p.
 c
ur
18
8S
G
O
ec
o 
sp
. l
ep
46
6S
G
O
ec
o 
sp
. l
ep
44
1S
G
O
ec
o 
sp
. l
ep
41
9S
G
N
ea
lc
id
io
n 
sp
. 1
S
G
H
el
i s
p.
 le
p4
38
S
G
G
ra
ph
ol
ita
 s
p.
 le
p5
5S
G
G
er
ae
us
 s
p.
 c
ur
64
S
G
G
el
e 
sp
. l
ep
39
S
G
G
el
e 
sp
. l
ep
28
1S
G
E
ub
ul
us
 s
p.
 c
ur
15
1S
G
C
yd
ia
 rh
od
as
pi
sD
H
J0
2
C
ur
c 
sp
. c
ur
21
1S
G
C
ur
c 
sp
. c
ur
20
9S
G
C
ur
c 
sp
. c
ur
20
8S
G
C
ur
c 
sp
. c
ur
20
2S
G
C
ur
c 
sp
. c
ur
18
9S
G
C
ur
c 
sp
. c
ur
18
5S
G
C
ur
c 
sp
. c
ur
13
1S
G
C
os
m
 s
p.
 le
p4
06
S
G
C
os
m
 s
p.
 le
p2
71
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
58
S
G
C
on
ot
ra
ch
el
us
 s
p.
 c
ur
29
S
G
C
er
am
by
ci
da
e 
sp
. c
ol
ce
r7
S
G
C
ar
ye
de
s 
br
as
ili
en
si
s
B
ot
hr
yo
pt
er
on
 s
p.
 2
3S
G
B
ar
is
 s
p.
 c
ur
12
1S
G
B
ar
is
 s
p.
 c
ur
10
0S
G
A
pi
on
 s
p.
 c
ur
18
3S
G
A
pi
on
 s
p.
 c
ur
18
0S
G
A
nc
ho
nu
s 
sp
. c
ur
53
S
G
A
m
bl
yc
er
us
 s
p.
 b
ru
28
S
G
A
m
bl
yc
er
us
 s
p.
 b
ru
27
S
G
A
m
bl
yc
er
us
 p
te
ro
ca
rp
ae
A
m
ph
ilo
ph
iu
m
 p
an
ic
ul
at
um
Ta
be
bu
ia
 ro
se
a
A
m
ph
ilo
ph
iu
m
 c
ru
ci
ge
ru
m
C
al
lic
hl
am
ys
 la
tif
ol
ia
B
ig
no
ni
a 
ae
qu
in
oc
tia
lis
G
ua
re
a 
gr
an
di
fo
lia
G
ua
re
a 
gu
id
on
ia
H
ur
a 
cr
ep
ita
ns
ys
op
hy
llu
m
 a
rg
en
te
um
Q
ua
ra
rib
ea
 a
st
er
ol
ep
is
C
om
br
et
um
 fr
ut
ic
os
um
C
om
br
et
um
 la
xu
m
P
ro
tiu
m
 te
nu
ifo
liu
m
M
ar
ip
a 
pa
na
m
en
si
s
O
co
te
a 
w
hi
te
i
N
ec
ta
nd
ra
 li
ne
at
a
N
ec
ta
nd
ra
 c
is
si
flo
ra
C
in
na
m
om
um
 tr
ip
lin
er
ve
O
co
te
a 
ob
lo
ng
a
B
ei
ls
ch
m
ie
di
a 
to
va
re
ns
is
O
co
te
a 
pu
be
ru
la
S
ou
ro
ub
ea
 s
ym
pe
ta
la
In
ga
 m
uc
un
a
In
ga
 p
un
ct
at
a
Tr
at
tin
ni
ck
ia
 a
sp
er
a
P
la
ty
po
di
um
 e
le
ga
ns
M
ac
ha
er
iu
m
 m
ill
ef
lo
ru
m
Po
ut
er
ia
 re
tic
ul
at
a
Lo
nc
ho
ca
rp
us
 h
ep
ta
ph
yl
lu
s
P
ic
ra
m
ni
a 
la
tif
ol
ia
C
ei
ba
 p
en
ta
nd
ra
S
te
rc
ul
ia
 a
pe
ta
la
P
se
ud
ob
om
ba
x 
se
pt
en
at
um
H
ei
st
er
ia
 c
on
ci
nn
a
O
ur
at
ea
 lu
ce
ns
C
oc
co
lo
ba
 m
an
zi
ne
lle
ns
is
C
oc
co
lo
ba
 e
xc
el
sa
C
ap
pa
rid
as
tru
m
 fr
on
do
sa
C
al
op
hy
llu
m
 lo
ng
ifo
liu
m
Fa
ra
m
ea
 o
cc
id
en
ta
lis
C
ou
ss
ar
ea
 c
ur
vi
ge
m
m
ia
C
up
an
ia
 s
ee
m
an
ni
i
C
up
an
ia
 la
tif
ol
ia
C
ne
st
id
iu
m
 ru
fe
sc
en
s
C
on
na
ru
s 
tu
rc
za
ni
no
w
ii
Pa
ul
lin
ia
 p
in
na
ta
C
on
na
ru
s 
pa
na
m
en
si
s
M
as
ca
gn
ia
 d
iv
ar
ic
at
a
H
ira
ea
 re
cl
in
at
a
A
de
lp
hi
a 
hi
ra
ea
In
ga
 la
ur
in
a
P
rio
ria
 c
op
ai
fe
ra
D
ol
io
ca
rp
us
 o
liv
ac
eu
s
D
ol
io
ca
rp
us
 m
ul
tif
lo
ru
s
D
ol
io
ca
rp
us
 m
aj
or
B
ac
tri
s 
m
aj
or
A
st
ro
ca
ry
um
 s
ta
nd
le
ya
nu
m
Zu
el
an
ia
 g
ui
do
ni
a
La
et
ia
 p
ro
ce
ra
C
as
ea
ria
 g
ui
an
en
si
s
C
or
di
a 
bi
co
lo
r
G
us
ta
vi
a 
su
pe
rb
a
H
an
dr
oa
nt
hu
s 
gu
ay
ac
an
La
cm
el
le
a 
pa
na
m
en
si
s
O
en
oc
ar
pu
s 
m
ap
or
a
G
ar
ci
ni
a 
m
ad
ru
no
G
ar
ci
ni
a 
re
co
nd
ita
S
ym
ph
on
ia
 g
lo
bu
lif
er
a
E
ug
en
ia
 n
es
io
tic
a
E
ug
en
ia
 o
er
st
ed
ia
na
V
iro
la
 n
ob
ili
s
V
iro
la
 s
eb
ife
ra
Tr
ic
hi
lia
 tu
be
rc
ul
at
a
tra
ga
st
ris
 p
an
am
en
si
s
S
po
nd
ia
s 
ra
dl
ko
fe
ri
A
na
ca
rd
iu
m
 e
xc
el
su
m
Po
m
ba
lia
 p
ru
ni
fo
lia
D
es
m
op
si
s 
pa
na
m
en
si
s
tra
pt
er
ys
 g
ou
do
tia
na
La
ci
st
em
a 
ag
gr
eg
at
um
G
ua
pi
ra
 s
ta
nd
le
ya
na
A
tta
le
a 
ro
st
ra
ta
G
ua
tte
ria
 lu
ce
ns
Te
tra
pt
er
ys
 d
is
co
lo
r
Ta
ch
ig
al
i p
an
am
en
si
s
C
av
an
ill
es
ia
 p
la
ta
ni
fo
lia
U
no
no
ps
is
 p
itt
ie
ri
E
nt
er
ol
ob
iu
m
 s
ch
om
bu
rg
ki
i
E
nt
er
ol
ob
iu
m
 c
yc
lo
ca
rp
um
In
ga
 s
ap
in
do
id
es
In
ga
 c
oc
le
en
si
s
H
ei
st
er
ia
 a
cu
m
in
at
a
S
oc
ra
te
a 
ex
or
rh
iz
a
Fr
id
er
ic
ia
 fl
or
id
a
H
irt
el
la
 tr
ia
nd
ra
Po
ut
er
ia
 s
tip
ita
ta
Li
ca
ni
a 
pl
at
yp
us
A
nn
on
a 
sp
ra
gu
ei
A
nn
on
a 
ha
ye
si
i
S
try
ch
no
s 
pa
na
m
en
si
s
D
av
ill
a 
ni
tid
a
M
yr
ci
a 
sp
le
nd
en
s
H
et
er
op
te
ry
s 
la
ur
ifo
lia
C
or
di
a 
al
lio
do
ra
S
er
ja
ni
a 
de
ca
pl
eu
ria
A
nt
ho
do
n 
pa
na
m
en
se
P
la
ty
m
is
ci
um
 p
in
na
tu
m
Fr
id
er
ic
ia
 c
an
di
ca
ns
Pa
ul
lin
ia
 ru
go
sa
ic
ho
sp
er
m
um
 g
al
eo
tti
i
Li
nd
ac
ke
ria
 la
ur
in
a
na
em
on
ta
na
 a
rb
or
ea
N
ee
a 
am
pl
ifo
lia
H
as
se
lti
a 
flo
rib
un
da
S
im
ar
ou
ba
 a
m
ar
a
M
ou
rir
i m
yr
til
lo
id
es
In
ga
 th
ib
au
di
an
a
Po
ur
ou
m
a 
bi
co
lo
r
In
ga
 u
m
be
lli
fe
ra
Pe
tre
a 
vo
lu
bi
lis
Te
rm
in
al
ia
 o
bl
on
ga
C
el
tis
 ig
ua
na
ea
G
ue
tta
rd
a 
fo
lia
ce
a
S
po
nd
ia
s 
m
om
bi
n
Ja
ca
ra
nd
a 
co
pa
ia
Te
rm
in
al
ia
 a
m
az
on
ia
Pa
ul
lin
ia
 fi
br
ig
er
a
A
st
ro
ni
um
 g
ra
ve
ol
en
s
Tr
ip
la
ris
 c
um
in
gi
an
a
S
lo
an
ea
 te
rn
ifl
or
a
A
nd
ira
 in
er
m
is
D
io
cl
ea
 w
ils
on
ii
H
ira
ea
 s
m
ila
ci
na
tra
ce
ra
 p
or
to
be
lle
ns
is
D
es
m
on
cu
s 
or
th
ac
an
th
os
M
ac
ha
er
iu
m
 a
rb
or
eu
m
M
ac
ha
er
iu
m
 s
ee
m
an
ni
i
Po
ul
se
ni
a 
ar
m
at
a
D
ry
pe
te
s 
st
an
dl
ey
i
M
en
do
nc
ia
 g
ra
ci
lis
P
te
ro
ca
rp
us
 h
ay
es
ii
Figure 4 Bipartite network depicting the interactions between seeds and internally feeding insect seed predators in the 50-ha forest dynamics plot on Barro
Colorado Island. The lower bars show individual plant species and the upper bars show individual seed predator species. The widths of the bars reflect the
abundance of seeds (lower bars) and the number of seeds killed by each seed predator species (upper bars), respectively. Species are organised according to
the size of the network compartment they belong to. Order-specific webs (Coleoptera and Lepidoptera) and a food web from which interactions
documented only once (‘singletons’) have been excluded are shown in Fig. S4 and Fig. S5, respectively. An interactive version of the food web is available
at http://bl.ocks.org/jcdterry/29fd8e581a27f0e861a71915ccaec938.
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
8 S. Gripenberg et al. Letter
abundance. The exceptionally high level of host specificity of
seed predators of BCI suggests that these insects might con-
tribute to maintaining the diversity of tree and liana species at
our study site, while they are less likely to play an important
role as mediators of indirect interactions among plant species.
ACKNOWLEDGEMENTS
We thank Xoaqu
ın Moreira and two anonymous reviewers
for providing valuable feedback on a previous version of our
manuscript. We thank Sebastian Bernal, Rufino Gonz
alez
and Omar Hern
andez for collecting many seed and fruit
samples and for helping with seed identifications, Betzi Perez
for assisting with seed collection and insect rearing,
and Annette Aiello and the staff of the Smithsonian Tropical
Research Institute field station on Barro Colorado Island for
help with research logistics. The Biodiversity Institute of
Ontario, University of Guelph, provided DNA barcodes.
Tiia K€arkk€ainen assisted with the laboratory work related to
DNA barcoding of larvae. David Adamski, Mark Metz, and
Alma Solis helped with identifications of Lepidoptera. Sup-
port from the following sources is gratefully acknowledged:
0
10
20
30
40
50
0.00 0.25 0.50 0.75 1.00
Specialisation, d'
N
um
be
r o
f i
ns
ec
t s
pe
ci
es
(i) Coleoptera
0
5
10
15
20
0.00 0.25 0.50 0.75 1.00
Specialisation, d'
N
um
be
r o
f i
ns
ec
t s
pe
ci
es
(ii) Lepidoptera
Insect
Sample size
< 10
= 10
0
50
100
150
1 2 3 4 5 6
Diet richness
N
um
be
r o
f i
ns
ec
t s
pe
ci
es
(i) Coleoptera
0
20
40
60
80
1 2 3 4 5 6
Diet richness
N
um
be
r o
f i
ns
ec
t s
pe
ci
es
(ii) Lepidoptera
0
10
20
30
1 2 3 4 5 6
Diet richness
N
um
be
r o
f i
ns
ec
t s
pe
ci
es
(iii) Hymenoptera
Insect
sample size
Singleton
<10
= 10
(b)
(a)
Figure 5 The distribution of diet specialisation among seed predator species displayed as (a) frequency histograms of d’ values of individual species in the
food web focusing on the 50-ha plot (displayed separately for Coleoptera and Lepidoptera) and (b) diet richness of seed predators using the full material
(including insect and plant species not included in the food web). In panel (b), the histograms show the number of seed predator species in the data set
reared from 1, 2, (. . .) and 7 plant species. Since the documented diet richness depends on sampling effort, the proportions of insect species in each diet
richness category being singletons, species represented by 2–9 individuals, and species with ≥ 10 individuals are shown.
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
Letter Host use by insect seed predators 9
Academy of Finland (Postdoctoral research project to SG),
The Smithsonian barcoding initiative FY014 and FY015
(YB), The Royal Society (SG’s University Research Fellow-
ship), The John Fell OUP Research Fund (OL, SG), Natural
Environment Research Council award NE/J007463/1 (OL),
Rachadaphiseksomphot Fund, Graduate School, Chula-
longkorn University (DQ), SENACYT (YB and HB) and
GACR 16-20825S (YB). The datasets on non-chemical plant
traits and the barcode phylogeny were funded by the Frank
Levinson Family Foundation and the Smithsonian Tropical
Research Institute. The Carl Cedercreutz Foundation funded
analyses on seed polyphenols. Data on the abundance of
adult trees were obtained from the BCI forest dynamics
research project. The BCI forest dynamics research project
was founded by S.P. Hubbell and R.B. Foster and is now
managed by S. Davies and R. Perez under the Center for
Tropical Forest Science and the Smithsonian Tropical
Research in Panama. Numerous organizations have provided
funding, principally the U.S. National Science Foundation,
and hundreds of field workers have contributed.
AUTHORSHIP
SG and OL conceived the study. Protocols for data collection
were designed by SG, OL, YB and SJW with input from SP-N.
Collection of seeds and fruits and insect rearing, sorting and
morphotyping was carried out by IS, CF and MC-S under the
supervision of SG, SJW and YB. OC identified seed and fruit
samples. MR morphotyped a subset of the insect specimens.
Insect specimens were identified by HB, JB, AC, GM, DQ and
RR. DNA barcoding of larvae was conducted by EV. SM
coordinated Lepidoptera identifications and helped curate the
DNA barcode library (managed by YB and SG). J-PS and JK
conducted analyses on seed polyphenols. CT conducted statisti-
cal analyses and prepared all figures. SG wrote the first draft
of the manuscript. CT, OL, YB and SJW contributed to early
revisions and all authors contributed to later revisions.
DATA AVAILABILITY STATEMENT
Data associated with this manuscript are deposited in the
Dryad Digital Repository : https://doi.org/10.5061/dryad.
230j5ch. Code for analyses is available on GitHub (https://
github.com/jcdterry/BCI_Seed_Predator). Data for sequenced
insect specimens (including images) are available on BOLD
(www.boldsystems.org), accessible by DOI (https://doi.org/10.
5883/DS-BCISP).
REFERENCES
Asner, G.P. & Martin, R.E. (2012). Contrasting leaf chemical traits in
tropical lianas and trees: implications for future forest composition.
Ecol. Lett., 15, 1001–1007.
Bagchi, R., Gallery, R.E., Gripenberg, S., Gurr, S.J., Narayan, L., Addis,
C.E. et al. (2014). Pathogens and insect herbivores drive rainforest
plant diversity and composition. Nature, 506, 85–88.
Birch, A.N.E., Fellows, L.E., Evans, S.V. & Doherty, K. (1986). Para-
aminophenylalanine in vigna: possible taxonomic and ecological
significance as a seed defence against bruchids. Phytochemistry, 25,
2745–2749.
Blomberg, S.P., Garland, T. & Ives, A.R. (2003). Testing for phylogenetic
signal in comparative data: behavioral traits are more labile. Evolution,
57, 717–745.
Bl€uthgen, N., Menzel, F. & Bl€uthgen, N. (2006). Measuring specialization
in species interaction networks. BMC Ecol., 6, 9.
Boege, K. & Marquis, R.J. (2006). Plant quality and predation risk
mediated by plant ontogeny: consequences for herbivores and plants.
Oikos, 115, 559–572.
Boege, K., Dirzo, R., Siemens, D. & Brown, P. (2007). Ontogenetic
switches from plant resistance to tolerance: minimizing costs with age?
Ecol. Lett., 10, 177–187.
Bradley, D.J., Gilbert, G.S. & Martiny, J.B.H. (2008). Pathogens promote
plant diversity through a compensatory response. Ecol. Lett., 11, 461–469.
Brown, P.D., Tokuhisa, J.G., Reichelt, M. & Gershenzon, J. (2003).
Variation of glucosinolate accumulation among different organs and
developmental stages of Arabidopsis thaliana. Phytochemistry, 62, 471–481.
Burdon, J.J., Thrall, P.H. & Ericson, A.L. (2006). The current and future
dynamics of disease in plant communities. Annu. Rev. Phytopath., 44,
19–39.
C
ardenas, R.E., Valencia, R., Kraft, N.J., Argoti, A. & Dangles, O.
(2014). Plant traits predict inter- and intraspecific variation in
susceptibility to herbivory in a hyperdiverse Neotropical rain forest tree
community. J. Ecol., 102, 939–952.
Carmona, D., Lajeunesse, M.J. & Johnson, M.T. (2011). Plant traits that
predict resistance to herbivores. Funct. Ecol., 25, 358–367.
Coley, P.D. (1983). Herbivory and defensive characteristics of tree species
in a lowland tropical forest. Ecol. Monogr., 53, 209–233.
Coley, P.D., Bryant, J.P. & Chapin, F.S. (1985). Resource availability and
plant antiherbivore defense. Science, 230, 895–899.
Condit, R. (1998). Tropical Forest Census Plots: Methods and Results
from Barro Colorado Island, Panama and a Comparison with Other
Plots. Springer-Verlag, Berlin.
Condit, R., Ashton, P.S., Baker, P., Bunyavejchewin, S., Gunatilleke, S.,
Gunatilleke, N. et al. (2000). Spatial patterns in the distribution of
tropical tree species. Science, 288, 1414–1418.
Connell, J.H. (1971). On the role of natural enemies in preventing
competitive exclusion in some marine animals and in rain forest trees.
In: Dynamics of Populations (eds. DenBoer, P.J. & Gradwell, G.R.).
PUDOC, Wageningen, pp. 298–312.
Crawley, M.J. (1989). Insect herbivores and plant-population dynamics.
Annu. Rev. Entomol., 34, 531–564.
Crawley, M.J. (2000). Seed predators and plant population dynamics. In
Seeds, 3rd Edition: The Ecology of Regeneration in Plant Communities
(ed Gallagher, R.S.). CAB International, Wallingford, Oxfordshire, pp.
167–182.
Croat, T.B. (1978). Flora of Barro Colorado Island. Stanford University
Press, Stanford, CA.
Ctvrtecka, R., Sam, K., Brus, E., Weiblen, G.D. & Novotny, V. (2014).
Frugivorous weevils are too rare to cause Janzen-Connell effects in
New Guinea lowland rain forest. J. Trop. Ecol., 30, 521–535.
Cutler, D.R., Edwards, T.C., Beard, K.H., Cutler, A., Hess, K.T.,
Gibson, J. et al. (2007). Random forests for classification in ecology.
Ecology, 88, 2783–2792.
Dormann, C.F., Gruber, B. & Fr€und, J. (2008). Introducing the bipartite
package: analysing ecological networks.Rnews, 8.
Downey, H., Lewis, O.T., Bonsall, M.B., Fernandez, D.C. & Gripenberg,
S. (2018). Insect herbivory on seedlings of rainforest trees: effects of
density and distance of conspecific and heterospecific neighbors. Ecol.
Evol., 8, 12702–12711.
Elton, C.S. (1975). Conservation and the low population density of
invertebrates inside neotropical rain forest. Biol. Conserv., 7, 3–15.
Endara, M.-J., Coley, P.D., Ghabash, G., Nicholls, J.A., Dexter, K.G.,
Donoso, D.A. et al. (2017). Coevolutionary arms race versus host
defense chase in a tropical herbivore–plant system. Proc. Natl Acad.
Sci. U. S. A., 114, E7499–E7505.
Feeny, P.P. (1976). Plant apparency and chemical defense. Rec. Adv.
Phytochem., 10, 1–40.
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
10 S. Gripenberg et al. Letter
Fenner, M. & Thompson, K. (2005). The Ecology of Seeds. Cambridge
University Press, Cambridge, UK.
Fenner, M., Cresswell, J., Hurley, R. & Baldwin, T. (2002). Relationship
between capitulum size and pre-dispersal seed predation by insect
larvae in common Asteraceae. Oecologia, 130, 72–77.
Fritz, S.A. & Purvis, A. (2010). Selectivity in mammalian extinction risk
and threat types: a new measure of phylogenetic signal strength in
binary traits. Conserv. Biol., 24, 1042–1051.
Fritz, R.S. & Simms, E.L. (1992). Plant Resistance to Herbivores and
Pathogens: Ecology, Evolution, and Genetics. University of Chicago
Press, Chicago, IL.
Futuyma, D.J. & Gould, F. (1979). Associations of plants and insects in
deciduous forest. Ecol. Monogr., 49, 33–50.
Garzon-Lopez, C.X., Ballesteros-Mejia, L., Ordo~nez, A., Bohlman, S.A.,
Olff, H. & Jansen, P.A. (2015). Indirect interactions among tropical
tree species through shared rodent seed predators: a novel mechanism
of tree species coexistence. Ecol. Lett., 18, 752–760.
Gilbert, G.S. & Webb, C.O. (2007). Phylogenetic signal in plant
pathogen–host range. Proc. Natl Acad. Sci. U. S. A., 104, 4979–4983.
Gilbert, G.S., Magarey, R., Suiter, K. & Webb, C.O. (2012).
Evolutionary tools for phytosanitary risk analysis: phylogenetic signal
as a predictor of host range of plant pests and pathogens. Evol. Appl.,
5, 869–878.
Gilbert, G.S., Briggs, H.M. & Magarey, R. (2015). The impact of plant
enemies shows a phylogenetic signal. PLoS ONE, 10, e0123758.
Green, P.T., Harms, K.E. & Connell, J.H. (2014). Nonrandom,
diversifying processes are disproportionately strong in the smallest size
classes of a tropical forest. Proc. Natl Acad. Sci. U. S. A., 111, 18649–
18654.
Gripenberg, S. (2018). Do pre-dispersal insect seed predators contribute
to maintaining tropical forest plant diversity? Biotropica, 50, 839–845.
Gripenberg, S., Rota, J., Kim, J., Wright, S.J., Garwood, N.C., Fricke,
E.C. et al. (2018). Seed polyphenols in a diverse tropical plant
community. J. Ecol., 106, 87–100.
Hanski, I. (2001). Spatially realistic theory of metapopulation ecology.
Naturwissenschaften, 88, 372–381.
Harms, K.E., Wright, S.J., Calderon, O., Hernandez, A. & Herre, E.A.
(2000). Pervasive density-dependent recruitment enhances seedling
diversity in a tropical forest. Nature, 404, 493–495.
Holt, R.D. (1977). Predation, apparent competition, and structure of prey
communities. Theor. Popul. Biol., 12, 197–229.
Hubbell, S.P., Foster, R.B., O’Brien, S.T., Harms, K., Condit, R.,
Wechsler, B. et al. (1999). Light-gap disturbances, recruitment
limitation, and tree diversity in a neotropical forest. Science, 283, 554–
557.
Hubbell, S.P., Condit, R. & Foster, R.B. (2005). Barro Colorado Forest
Plot Census Data. http://ctfs.si.edu/webatlas/datasets/bci.
Janzen, D.H. (1968). Host plants as islands in evolutionary and
contemporary time. Am. Nat., 102, 592–595.
Janzen, D.H. (1969). Seed-eaters versus seed size, number, toxicity and
dispersal. Evolution, 23, 1–27.
Janzen, D.H. (1970). Herbivores and number of tree species in tropical
forests. Am. Nat., 104, 501–528.
Janzen, D.H. (1971). Seed predation by animals. Annu. Rev. Ecol. Syst.,
2, 465–492.
Janzen, D.H. (1976). Why bamboos wait so long to flower. Ann. Rev.
Ecol. Syst., 7, 347–391.
Janzen, D.H. (1980). Specificity of seed-attacking beetles in a Costa Rican
deciduous forest. J. Ecol., 68, 929–952.
Jones, F.A. & Comita, L.S. (2010). Density-dependent pre-dispersal seed
predation and fruit set in a tropical tree. Oikos, 119, 1841–1847.
Kelly, D. (1994). The evolutionary ecology of mast seeding. Trends Ecol.
Evol., 9, 465–470.
Kembel, S.W., Cowan, P.D., Helmus, M.R., Cornwell, W.K., Morlon,
H., Ackerly, D.D. et al. (2010). Picante: R tools for integrating
phylogenies and ecology. Bioinformatics, 26, 1463–1464.
Kolb, A., Ehrl
en, J. & Eriksson, O. (2007). Ecological and evolutionary
consequences of spatial and temporal variation in pre-dispersal seed
predation. Perspect. Plant Ecol. Evol. Syst., 9, 79–100.
Kress, W.J., Erickson, D.L., Jones, F.A., Swenson, N.G., Perez, R.,
Sanjur, O. et al. (2009). Plant DNA barcodes and a community
phylogeny of a tropical forest dynamics plot in Panama. Proc. Natl
Acad. Sci. U. S. A., 106, 18621–18626.
Kuprewicz, E.K. & Garc
ıa-Robledo, C. (2010). Mammal and insect
predation of chemically and structurally defended Mucuna holtonii
(Fabaceae) seeds in a Costa Rican rain forest. J. Trop. Ecol., 26, 263–
269.
Landis, J.R. & Koch, G.G. (1977). The measurement of observer
agreement for categorical data. Biometrics, 36, 207–216.
Leck, M.A., Parker, V.T. & Simpson, R.L. (2008). Seedling Ecology and
Evolution. Cambridge University Press, Cambridge, UK.
Leigh, E.G.J. (1999). Tropical Forest Ecology. A View from Barro
Colorado Island. Oxford University Press Inc., New York.
Lewis, O.T. & Gripenberg, S. (2008). Insect seed predators and
environmental change. J. Appl. Ecol., 45, 1593–1599.
Lewis, O.T., Memmott, J., Lasalle, J., Lyal, C.H.C., Whitefoord, C. &
Godfray, H.C.J. (2002). Structure of a diverse tropical forest insect-
parasitoid community. J. Anim. Ecol., 71, 855–873.
Lill, J.T., Marquis, R.J. & Ricklefs, R.E. (2002). Host plants influence
parasitism of forest caterpillars. Nature, 417, 170–173.
Loranger, J., Meyer, S.T., Shipley, B., Kattge, J., Loranger, H., Roscher,
C. et al. (2012). Predicting invertebrate herbivory from plant traits:
evidence from 51 grassland species in experimental monocultures.
Ecology, 93, 2674–2682.
Marquis, R.J., Diniz, I.R. & Morais, H.C. (2001). Patterns and correlates
of interspecific variation in foliar insect herbivory and pathogen attack
in Brazilian cerrado. J. Trop. Ecol., 17, 127–148.
McArt, S.H., Halitschke, R., Salminen, J.-P. & Thaler, J.S. (2013). Leaf
herbivory increases plant fitness via induced resistance to seed
predators. Ecology, 94, 966–975.
M€uller, C.B., Adriaanse, I.C.T., Belshaw, R. & Godfray, H.C.J. (1999).
The structure of an aphid-parasitoid community. J. Anim. Ecol., 68,
346–370.
Novotny, V., Basset, Y., Miller, S.E., Weiblen, G.D., Bremer, B., Cizek,
L. et al. (2002). Low host specificity of herbivorous insects in a tropical
forest. Nature, 416, 841–844.
Novotny, V., Miller, S.E., Baje, L., Balagawi, S., Basset, Y., Cizek, L.
et al. (2010). Guild-specific patterns of species richness and host
specialization in plant–herbivore food webs from a tropical forest.
J. Anim. Ecol., 79, 1193–1203.
Orme, D., Freckleton, R., Thomas, G., Petzoldt, T., Fritz, S., Isaac, N.
et al. (2018). caper: comparative analyses of phylogenetics and
evolution in R. R package version 1.0.1. https://CRAN.R-project.org/
package=caper.
Pacala, S.W. & Crawley, M.J. (1992). Herbivores and plant diversity. Am.
Nat., 140, 243–260.
Pinz
on-Navarro, S., Barrios, H., M
urria, C., Lyal, C.H.C. & Vogler, A.P.
(2010). DNA-based taxonomy of larval stages reveals huge unknown
species diversity in neotropical seed weevils (genus Conotrachelus):
relevance to evolutionary ecology. Mol. Phyl. Evol., 56, 281–293.
Ratnasingham, S. & Hebert, P.D.N. (2013). A DNA-based registry for all
animal species: The Barcode Index Number (BIN) system. PLoS ONE,
8, e66213.
Revell, L.J. (2012). phytools: an R package for phylogenetic comparative
biology (and other things). Methods Ecol. Evol., 3, 217–223.
Sam, K., Ctvrtecka, R., Miller, S.E., Rosati, M.E., Molem, K., Damas,
K. et al. (2017). Low host specificity and abundance of frugivorous
lepidoptera in the lowland rain forests of Papua New Guinea. PLoS
ONE, 12, e0171843.
Sang, J.P., Minchinton, I.R., Johnstone, P.K. & Truscott, R.J.W. (1984).
Glucosinolate profiles in the seed, root and leaf tissue of cabbage,
mustard, rapeseed, radish and swede. Canad. J. Plant Sci., 64, 77–93.
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
Letter Host use by insect seed predators 11
Schemske, D.W., Mittelbach, G.G., Cornell, H.V., Sobel, J.M. & Roy, K.
(2009). Is there a latitudinal gradient in the importance of biotic
interactions? Annu. Rev. Ecol. Evol. Syst., 40, 245–269.
Schuldt, A., Bruelheide, H., Durka, W., Eichenberg, D., Fischer, M.,
Kr€ober, W. et al. (2012). Plant traits affecting herbivory on tree
recruits in highly diverse subtropical forests. Ecol. Lett., 15, 732–
739.
Sedio, B.E. & Ostling, A.M. (2013). How specialised must natural
enemies be to facilitate coexistence among plants? Ecol. Lett., 16,
995–1003.
Strobl, C., Boulesteix, A.-L., Zeileis, A. & Hothorn, T. (2007). Bias in
random forest variable importance measures: Illustrations, sources and
a solution. BMC Bioinformatics, 8, 25.
Turcotte, M.M., Davies, T.J., Thomsen, C.J.M. & Johnson, M.T.J.
(2014). Macroecological and macroevolutionary patterns of leaf
herbivory across vascular plants. Proc. R. Soc. Lond. B: Biol. Sci., 281,
20140555.
Visser, M.D., Muller-Landau, H.C., Wright, S.J., Rutten, G. & Jansen,
P.A. (2011). Tri-trophic interactions affect density dependence of seed
fate in a tropical forest palm. Ecol. Lett., 14, 1093–1100.
Walters, D.R. (2011). Plant Defense: Warding off Attack by Pathogens,
Herbivores, and Parasitic Plants. Wiley-Blackwell, Chichester, UK.
Willig, M.R., Kaufman, D.M. & Stevens, R.D. (2003). Latitudinal
gradients of biodiversity: pattern, process, scale, and synthesis. Annu.
Rev. Ecol. Evol. Syst., 34, 273–309.
Wolda, H. (1988). Insect seasonality: why? Ann. Rev. Ecol. Syst., 19,
1–18.
Wright, S.J. (1983). The dispersion of eggs by a bruchid beetle among
Scheelea palm seeds and the effect of distance to the parent palm.
Ecology, 64, 1016–1021.
Wright, S.J. & Calder
on, O. (2006). Seasonal, El Ni~no and longer term
changes in flower and seed production in a moist tropical forest. Ecol.
Lett., 9, 35–44.
Wright, S.J., Muller-Landau, H.C., Calder
on, O. & Hernand
ez, A. (2005).
Annual and spatial variation in seedfall and seedling recruitment in a
neotropical forest. Ecology, 86, 848–860.
Yu, G., Smith, D.K., Zhu, H., Guan, Y. & Tsan-Yuk Lam, T. (2017).
GGTREE: An R package for visualization and annotation of
phylogenetic trees with their covariates and other associated data.
Methods Ecol. Evol., 8, 28–36.
Zalamea, P.-C., Dalling, J.W., Sarmiento, C., Arnold, A.E., Delevich, C.,
Berhow, M.A. et al. (2018). Dormancy-defense syndromes and
tradeoffs between physical and chemical defenses in seeds of pioneer
species. Ecology, 99, 1988–1998.
Zimmerman, J.K., Wright, S.J., Calder
on, O., Pagan, M.A. & Paton, S.
(2007). Flowering and fruiting phenologies of seasonal and aseasonal
neotropical forests: the role of annual changes in irradiance. J. Trop.
Ecol., 23, 231–251.
SUPPORTING INFORMATION
Additional supporting information may be found online in
the Supporting Information section at the end of the article.
Editor, Ferenc Jordan
Manuscript received 1 July 2019
Manuscript accepted 7 July 2019
© 2019 The Authors. Ecology Letters published by CNRS and John Wiley & Sons Ltd.
12 S. Gripenberg et al. Letter