Abstract Predation and resource availability influence
community structure in many aquatic ecosystems. Preda-
tors (odonates) and resources (leaf litter) were manipu-
lated to determine their independent effects on macroor-
ganism species richness, abundance, and composition in
water-filled tree holes of Barro Colorado Island, Pana-
ma. Interactive effects of these factors were also investi-
gated in artificial tree holes. Large odonates reduced spe-
cies richness in natural tree holes, but did not significant-
ly reduce macroorganism abundance. The presence of
larvae of the mosquito Culex urichii and the ceratopogo-
nid midge Bezzia snowi were negatively associated with
the presence of large odonate larvae. In natural tree
holes, leaf litter addition and removal respectively in-
creased and decreased richness by c. 1 species relative to
controls, and macroorganism abundance was greater in
litter addition holes than in litter removal holes. Indepen-
dent effects of predation showed similar patterns in arti-
ficial holes, but there was no predator?resource interac-
tion, partly due to the short duration of the experiment.
Predators grew faster when litter was abundant, and indi-
rectly reduced litter degradation rates when resources
were scarce in artificial holes. Both resource availability
and predation influence species richness in water-filled
tree holes, but act at different time scales; richness fol-
lows productivity (litter quantity) over a period of
weeks, whereas effects of predation may span several
months.
Keywords Abundance ? Macroinvertebrates ? 
Phytotelmata ? Species richness ? Tropics
Introduction
Top predators can affect prey diversity or abundance in
large aquatic settings such as lakes and streams (e.g.,
Brooks and Dodson 1965; Zaret 1980; Kerfoot and Sih
1987; Blois-Heulin et al. 1990; Diehl 1992; McPeek and
Peckarsky 1998), and can indirectly influence popula-
tions at basal trophic levels via trophic cascades or ?top-
down? control (e.g., Power 1990; Carpenter and Kitchell
1993; Pace et al. 1999). Although predator effects and
nutrient effects are not mutually exclusive, lake and
stream communities may also be subject to ?bottom-up?
control, whereby changes in inorganic nutrient (e.g., ni-
trogen and phosphorus) quantities alter primary produc-
tivity (e.g., Proulx et al. 1996; Spencer and Ellis 1998).
Changes in productivity are then reflected by variation in
abundance or diversity within consumer populations 
or assemblages (e.g., Hershey et al. 1988; Hart and 
Robinson 1990; Perrin and Richardson 1997). The quali-
ty and quantity of leaf-litter resources may similarly reg-
ulate detritivore populations and detritus-based assem-
blages in large aquatic systems (e.g., Gee 1988; Richard-
son 1991; Dobson 1994).
Top-down and bottom-up control of community struc-
ture are less studied in small aquatic habitats, such as
phytotelmata. However, predators influence the diversity
and abundance of some taxa (often other predators) liv-
ing in the water held by pitcher plants (Addicott 1974;
Cochran-Stafira and von Ende 1998), bromeliads (e.g.,
Lounibos et al. 1987), Heliconia spp. bracts (Naeem
1988), bamboo stumps (Sota and Mogi 1996), and tree
holes (e.g., Bradshaw and Holzapfel 1983; Fincke et al.
1997; Nannini and Juliano 1998). Predation tends to re-
duce species richness and abundance in these systems
(but see Louton et al. 1996), and these effects are partly
attributed to small habitat size (Addicott 1974). Unlike
predators in the intertidal zone (Paine 1966) and other
relatively open systems, predators in phytotelmata can
efficiently patrol the entire habitat. Thus, few temporal
or spatial prey refugia exist, and the potential for strong
top-down effects is great (but see Lounibos et al. 1997).
S.P. Yanoviak
Department of Zoology, University of Oklahoma, Norman, 
OK 73019, USA
Present address:
S.P. Yanoviak, Evergreen State College, Lab I, Olympia, 
WA 98505, USA
e-mail: yanoviak@racsa.co.cr
Tel.: +1-360-8666788
Oecologia (2001) 126:125?133
DOI 10.1007/s004420000493
Stephen P. Yanoviak
Predation, resource availability, and community structure 
in Neotropical water-filled tree holes
Received: 11 January 2000 / Accepted: 14 July 2000 / Published online: 29 August 2000
? Springer-Verlag 2000
Fewer studies have addressed the role of resource
availability in structuring phytotelm communities (e.g.,
Naeem 1990; Sota 1996). Plant detritus, especially leaf
litter, forms the nutrient base for food webs in many types
of phytotelmata, including most water-filled tree holes
(see Snow 1949; Kitching 1971 for descriptions of this
habitat). Macroorganisms in tree holes consume decaying
litter directly (e.g., Paradise and Dunson 1997), or indi-
rectly by grazing and filtering decomposer microbes from
litter surfaces and the water column (e.g., Fish and Car-
penter 1982; Walker and Merritt 1991). Litter quality and
quantity influence growth rates and life history character-
istics of some tree-hole macroinvertebrates (e.g., Carpen-
ter 1982a; Hard et al. 1989; Lounibos et al. 1993; L?on-
ard and Juliano 1995; Walker et al. 1997), and resource
availability affects community parameters in experimen-
tal tree hole analogs (Jenkins et al. 1992; Srivastava and
Lawton 1998; Yanoviak 1999a, 1999b).
Water-filled tree holes are common in lowland moist
forests of Panama. Although c. 60 different species of
macroorganisms are associated with this habitat, most in-
dividual holes contain fewer than ten species on any giv-
en wet season day (Yanoviak, in press). Detritivores and
omnivores comprise the majority of the fauna, but most
holes also contain one or more of seven top predator spe-
cies, with larvae of giant damselflies (Odonata: Pseudo-
stigmatidae) being the most common (Fincke 1999). In
addition to supporting manageable and specialized mac-
roorganism assemblages, tree holes are generally small,
easy to sample, and can be replicated with simple materi-
als (e.g., plastic containers). Thus, this is a particularly
tractable system for community-level experiments.
I designed this study to test the hypothesis that mac-
roorganism community structure (i.e., species richness,
composition, and abundance) in water-filled tree holes of
Panama is influenced by the presence or absence of top
predators (odonates) and the availability of food resourc-
es (leaf litter). Based on qualitative surveys in Panama,
and previous work on this system (Fincke et al. 1997), I
predicted that macroorganism species richness and abun-
dance would be lower in holes containing odonate larvae
than in holes without odonates.
Because the number of consumer species and individ-
uals in a habitat is often linked to resource availability
(reviewed by Srivastava and Lawton 1998), I predicted
that addition of leaf litter to tree holes would increase
macroorganism species richness and abundance, and that
removal of litter would have the opposite effect. I also
predicted that: (1) litter effects would be short-lived, and
community parameters would return to initial levels after
termination of manipulations; (2) litter addition and re-
moval would affect species distributions; and (3) species
persistence times would be greater in litter addition holes
than in litter removal holes.
Finally, Washburn et al. (1991) and Fincke et al.
(1997) showed that parasite and predator effects on mos-
quito abundance differ when nutrient levels are relatively
high or low in artificial tree holes. Based on these stud-
ies, I predicted that resource availability and predation
would have interactive effects on community structure
(i.e., the influence of odonates is respectively masked or
enhanced when leaf litter abundance is high or low).
Materials and methods
This study was conducted in the seasonally moist lowland forest
of Barro Colorado Island (BCI), Panama (see Leigh et al. 1996 for
a site description). The fauna and abiotic characteristics of BCI
tree holes are described elsewhere (Yanoviak 1999c, in press). To
standardize potential effects of hole location and size on commu-
nity structure, all holes used in this study were located in the un-
derstory (maximum height 1.5 m) and were ?3 l in volume. I mea-
sured pH, height above the ground, total volume, and water sur-
face area in each hole at the start of experiments (Tables 1, 2). Hy-
drion narrow- and broad-range colorimetric strips were used for
all pH measurements. 
I censused the macrofauna of each hole by removing its con-
tents with a turkey baster or siphon into a white pan. Macroorgan-
isms were identified and counted, and subsamples were collected
when species could not be determined in the field. After each 
census, the organisms, litter, and water were returned to the hole
and the pan was rinsed. Additional methodological details for nat-
ural and artificial tree hole sampling are summarized elsewhere 
(Yanoviak 1999c).
Predation
I used a predator exclusion experiment to determine the effects of
odonates on species richness and abundance of other macroorgan-
126
Table 1 Means (SD) of physicochemical properties, species rich-
ness, and abundance measured at the start of the predator exclu-
sion experiment (December 1995). Area is estimated water surface
area. Exclusion holes were covered with chicken wire to prevent
odonate colonization. Control holes were not covered. Volume
ranges in brackets (n=17 for each mean, df=32 for all t-tests). Vol-
ume, area, and abundance data were transformed prior to analysis
Control Exclusion tobs P
Height (cm) 55 (42) 64 (44) 1.18 0.25
Volume (ml) 1075 (922) 583 (433) 0.60 0.55
[20?3000] [17?1600]
Area (cm2) 111 (83) 62 (51) 1.66 0.11
pH 6.06 (0.95) 6.32 (0.63) 0.94 0.36
Richness 4.8 (2.4) 4.9 (2.0) 0.15 0.89
Abundance 48.7 (45.6) 29.7 (26.1) 0.56 0.58
Table 2 Means (SD) of physical characteristics, species richness,
and abundance measured at the start of the nutrient addition/re-
moval experiment (week 0). Volume ranges in brackets. Volume,
area, and abundance data were transformed prior to analysis
Litter Litter Control F2,33 P
added removed
Height (cm) 62 (42) 51 (45) 41 (35) 0.77 0.47
Volume (ml) 801 (533) 833 (865) 662 (472) 0.21 0.82
[200?2000] [200?3000] [250?2000]
Area (cm2) 138 (122) 110 (84) 134 (110) 0.11 0.89
Richness 4.6 (2.1) 5.2 (2.8) 5.6 (1.8) 0.70 0.51
Abundance 20.8 (19.0) 23.1 (18.4) 45.5 (47.0) 2.19 0.13
isms in tree holes. In December 1995, I censused the macrofauna
of 40 natural tree holes in the forest understory. I removed odo-
nates and other top predators from 20 of the holes, and placed a
canopy of chicken wire (2- to 3-cm mesh) over the hole openings
to prevent recolonization by odonates. Earlier experiments using
artificial tree holes showed that chicken wire or netting deters ovi-
position by odonates, but permits colonization by other tree hole
inhabitants (Fincke et al. 1997; Fincke 1998). I distributed varia-
tion in hole size between covered and uncovered holes as much as
possible. Hereafter, I refer to uncovered holes as the control treat-
ment and covered holes as the odonate exclusion treatment. All 
40 holes were left untouched until the start of the following wet
season.
In mid-May 1996, I surveyed all experimental holes for the
presence of odonates and other top predators, and removed any
odonate larvae found in exclusion holes (5 of 20 holes). Odonates
present in these holes most likely originated from eggs deposited
prior to installation of the chicken wire in December 1995; pseu-
dostigmatid eggs are concealed in the woody margins of tree holes
and require up to 196 days to hatch (Fincke 1998, 1999). Six of
the holes (3 control and 3 exclusion) were dry or otherwise dis-
turbed, and were removed from the experiment. I censused the re-
maining 34 holes every 14 days from 10 June until 29 July, 1996
(n=4 censuses per hole), and I measured pH in each hole during
censuses 1, 2, and 4. Any leaf litter trapped by the chicken wire
covering an exclusion hole was collected and added to the hole af-
ter each census.
Resource availability
In mid-August 1997, I divided 36 natural tree holes equally among
three litter treatments: a control group, a litter removal group, and
a litter addition group. I censused all holes 2 days before the start
of experimental manipulations (hereafter, this initial census is
identified as week 0). I assigned holes to treatment groups based
in part on general shape, tree species, and initial macroorganism
species richness to distribute potential sources of variance among
groups as much as possible. Six control holes from the predator
exclusion experiment were used again in this experiment, but none
of the exclusion holes were reused.
I removed all macrodetritus (>1 cm2) from the 12 litter remov-
al holes on week 0 and at least once each week thereafter until
week 6. Care was taken to keep disturbance to a minimum during
litter removal. Litter addition holes received 0.014 g leaf lit-
ter cm?2 water surface area on week 0, and then 0.007 g cm?2 on
weeks 2 and 4. I based the litter addition quantities on the average
oven dry weight of litter collected over a 30-day period in artifi-
cial tree holes (650-ml cups, 71-cm2 opening) tied to trees 1.0 m
above ground level (mean=0.007 g cm?2 surface area, SD=0.33).
Litter added to holes consisted of approximately equal proportions
of leaves or leaflets of Ceiba pentandra, Dipteryx panamensis,
Ficus yoponensis, and Platypodium elegans collected from tree
crowns or recent branch falls (as described in Yanoviak 1999b).
All litter was air-dried for ?7 days and then oven-dried (70?C,
?48 h) before weighing.
I recorded the number of macroorganism species present and
their abundance in each hole on weeks 2?6. I censused all holes
again on week 10 (i.e., after ?5 weeks without litter addition or re-
moval) to determine if the macrofauna had recovered from the ma-
nipulations. One of the control holes dried out during the experi-
ment and was not included in analyses.
Predator?resource interaction
On 1 October 1997, I tied 30 artificial tree holes (650-ml black
plastic cups) to trees separated by 5?10 m in the forest understory.
I evenly divided the cups into three resource availability treat-
ments based on the quantity of initial litter inputs: high=4.00 g
(?0.001), medium=0.40 g (?0.001), and low=0.04 g (?0.0005).
The composition, origin, and treatment of litter was the same as
described above. I filled the cups with filtered (0.25 mm mesh)
rain water and left them undisturbed for 3 weeks to allow commu-
nities to develop.
At the end of the third week (hereafter, day 0), I censused the
contents of all artificial tree holes and added one mid-size
Megaloprepus coerulatus larva (Odonata: Pseudostigmatidae;
10?18 mm body length excluding caudal lamellae) to half of the
cups in the experiment. The initial body length of odonates did not
differ among treatments (F2,12=1.32, P=0.30). I censused the mac-
rofauna of all cups on days 4, 8, 16, and 32. Odonates were not
counted in censuses. I checked the cups every 3 days for the pres-
ence of M. coerulatus larvae and removed any macrodetritus
(>0.25 cm2) foreign to the experiment. I collected all experimental
detritus remaining in the cups on day 32. Some litter fragments
were lost from two cups that were disturbed between days 16 and
32, so these data were excluded from the analysis of remaining lit-
ter mass.
I measured the body length of the M. coerulatus larvae on
days 8, 16 and 32. One larva was missing from the high litter
treatment on day 8, and one was missing from the medium litter
treatment on day 16. I replaced the missing individuals with new
mid-size larvae, but excluded these replicates from the analysis of
predator growth. Toxorhynchites theobaldi, another top predator in
BCI tree holes, was occasionally found in non-predator holes. I in-
cluded this species in census data, but consistently removed indi-
viduals when they occurred to reduce confounding of odonate ef-
fects. I analyzed species richness and abundance data from
day 4?32 censuses with a 2?3 factorial repeated-measures AN-
OVA using presence or absence of M. coerulatus and litter quanti-
ty as main effects.
Data were tested for normality and homogeneity of variance
with normal probability plots and F-tests prior to analysis. Ryan-
Einot-Gabriel-Welsch multiple range tests (SAS 1989) were used
to compare means when ANOVA results were significant. All G-
tests had df=1 unless otherwise noted, and G values were adjusted
with Williams' correction (Sokal and Rohlf 1981). Data used in G-
tests were pooled across time within a hole to maintain statistical
independence. Bonferroni-adjusted ? values were used to correct
for multiplicity and are noted in the text. Proportional data were
arcsine square-root transformed before analysis, and the log(x+1)
transformation was used to improve normality for all abundance
data and for other data where noted. Means shown in the results
were calculated from untransformed data.
Results
Tree hole physicochemical properties, macroorganism
species richness, and macroorganism abundance did not
differ between controls and treatments at the start of the
predator exclusion (Table 1) and resource availability
(Table 2) experiments. Average pH was similar between
control and predator exclusion holes during the 1996 cen-
suses (repeated-measures ANOVA: F1,32=0.98, P=0.33;
time?treatment P=0.095).
Predation
I predicted that macroorganism species richness and
abundance would be lower in tree holes containing odo-
nates than in holes without these top predators. The tim-
ing of the experiment (early wet season) was problematic
because odonates had not colonized 5 of the 17 control
holes by the second census date. Moreover, many of the
holes that were colonized contained only small odonate
larvae (<14 mm in body length), which have relatively
127
slow feeding and growth rates (Fincke 1992a; Fincke et
al. 1997). These problems were resolved by the third
census date, when all control holes contained at least one
large (>19 mm) odonate larva or at least two mid-size
(14?18 mm) larvae.
Analyses of data from the third and fourth censuses
showed that the abundance of macroorganisms, and the
abundance of mosquitoes alone (for comparison with Fin-
cke et al. 1997), were unaffected by the presence of odo-
nates (repeated-measures ANOVAs: F1,32 <1.70, P>0.20,
?=0.025; time?treatment P>0.11 for both tests; Fig. 1A).
Although they had no effect on macroorganism abun-
dance, large odonate larvae reduced species richness rela-
tive to exclusion holes by an average of 1.12 taxa overall
(F1,32=5.65, P=0.024; time?treatment P=0.13; Fig. 1B).
Of 13 potential prey taxa (i.e., relatively common
species with active larvae living in the water), the pres-
ence of the mosquito Culex urichii and the ceratopogo-
nid midge Bezzia snowi were negatively associated with
the presence of large odonate larvae in control holes
(G>4.72, P<0.05 in both tests). The presence of Aedes
terrens, the commonest mosquito species in natural tree
holes on BCI (Fincke et al. 1997; Yanoviak, in press),
was not associated with the presence or absence of odo-
nates (G=0.23, P>0.50).
Resource availability
I predicted that addition and removal of leaf litter would
respectively increase and decrease macroorganism spe-
cies richness and abundance relative to control holes.
Differences in macroorganism abundance among all
three treatments were non-significant during weeks 2?6
(repeated-measures ANOVA: F2,32=2.82, P=0.074;
Fig. 2A), but litter addition holes contained more macro-
organisms than litter removal holes (F1,22 =7.12,
P=0.014, ?=0.025). Addition of leaf litter to tree holes
increased the average number of species present (pooled
across time) by 1.04 over controls, whereas litter remov-
al reduced richness by 1.08 species (Fig. 2B). These dif-
ferences were significant during weeks 2?6 (F2,32=5.00,
P=0.013), and there were no time?treatment interactions
for abundance or species richness (P>0.23 for all tests).
Because soluble nutrients are rapidly leached from
leaf litter in tree holes (e.g., Carpenter 1982b), I predict-
ed that experimental effects would disappear shortly af-
ter termination of litter manipulations. By week 10
(5 weeks after manipulations were terminated), average
macroorganism species richness and average abundance
had converged near initial (week 0) values and no longer
differed among treatments (richness: F2,33=0.08, P=0.92;
abundance: F2,33=0.13, P=0.88; Fig. 2).
The cumulative number of species found in each
treatment of this experiment was similar (litter addi-
tion=26, litter removal=30, control=27), and all holes
contained one or more top predators. The odonates
Megaloprepus coerulatus, Mecistogaster spp., and Aes-
hnidae occupied 60%, 68.6%, and 20% of the holes, re-
spectively. Dendrobates auratus tadpoles and larvae of
the mosquito Toxorhynchites theobaldi occupied 5.7%
and 28.6% of the holes, respectively. Tadpoles of D. au-
128
Fig. 1 Mean (?SE) A abundance and B species richness of non-
odonate macroorganisms in control holes (odonates present) and
odonate exclusion holes; n=17 for each mean
Fig. 2 Mean (?SE) A macroorganism abundance and B species
richness in natural tree holes to which leaf litter was added on
weeks 0, 2, and 4 (Add), removed at least weekly until week 6
(Rem), or unmanipulated (Ctrl); n=11 for control means, n=12 for
all others
Fig. 3 Mean (?SE) A macroorganism abundance and B species
richness in artificial tree holes with () and without () the odo-
nate Megaloprepus coerulatus; n=15 for each mean
Fig. 4 Mean (?SE) A macroorganism abundance and B species
richness in artificial tree holes containing different initial quanti-
ties of litter (High 4.0 g, Medium 0.4 g, Low 0.04 g); n=10 for
each mean
129
ratus were found only in litter removal holes, but all oth-
er top predators occurred in all treatments and were
evenly distributed among treatments (G<2.3, df=2,
P>0.10 in all tests; aeshnids and T. theobaldi were
pooled to meet statistical assumptions).
I predicted that differences in leaf litter availability
would be reflected by differences in distribution of spe-
cies among treatments. Specifically, I expected that rela-
tively short-lived species (those completing larval devel-
opment in <6 weeks) would respond to short-term
changes in resource availability. Six taxa found in this
experiment occurred often enough for quantitative analy-
sis (i.e., in >5 holes of each treatment) and qualified as
relatively short-lived: the mosquitoes Aedes terrens,
Anopheles eiseni, Culex corrigani, and Haemagogus
(H.) spp.; the midge Bezzia snowi; and the annelid Dero
sp. Where they occurred, these taxa colonized a larger
proportion of litter addition holes (68%) than control
holes (41%) only after the start of manipulations (t=3.62,
df=10, P=0.005). All of these species also disappeared
from ?50% of the same litter addition holes after manip-
ulations were terminated (i.e., between weeks 6 and 10).
The opposite pattern did not occur in the litter removal
treatment; the proportion of litter removal holes from
which these six species became locally extinct after
week 0 did not differ from controls (t=0.82, df=7,
P=0.44; test assumed unequal variance).
Finally, I predicted that the greater energy supply pro-
vided by litter addition would lead to greater species per-
sistence (e.g., Srivastava and Lawton 1998) relative to lit-
ter removal. The mean persistence time (number of con-
secutive censuses that a species was present in a hole) did
not differ between litter removal holes and litter addition
holes (Kruskal-Wallis ?2=0.17, P=0.68). However, the
average (?SE) number of species present for ?2 consecu-
tive censuses was marginally greater in litter addition
holes (6.75?0.75) than in litter removal holes (4.33?0.47;
Kruskal-Wallis ?2=4.84, P=0.027, ?=0.025).
Predator?resource interaction
I predicted that predation and initial resource (leaf litter)
quantity would have interactive effects on macroorgan-
ism species richness and abundance within the physical-
ly standardized confines of artificial tree holes. As ex-
pected, there were no predator?resource interactions for
richness or abundance on day 0, when odonates were
first added to the experiment (F2,24<2.67, P>0.09 in both
tests). Although richness and abundance on day 0 were
similar between holes that received odonates and those
that did not (F1,24<1.76, P>0.20; Fig. 3), these communi-
ty parameters were significantly influenced by initial nu-
trient quantity. Macroorganism abundance and richness
were greatest in the high litter quantity treatment on
day 0 (F2,24>4.57, P<0.021), but did not differ between
low and medium litter quantity treatments (Fig. 4). 
As the experiment progressed, predation and resource
availability did not interact to shape macroorganism
community structure in the holes; interactions were
non?significant for both abundance (repeated-measures
ANOVA: F2,24=0.01, P=0.99) and species richness(F2,24=0.78, P=0.47) during the days 4?32 censuses(time?predator?litter P>0.23 in both tests). Total macro-
organism abundance and the abundance of mosquitoes
alone did not differ between predator and non?predator
treatments (F1,24<0.24, P>0.63, ?=0.025, time?predator
P>0.35 in both tests; Fig. 3A). However, overall mean
species richness was greater by 0.59 taxa in cups lacking
odonates (F1,24=9.91, P=0.004; Fig. 3B). This effect 
was primarily caused by data from days 16 and 32
(time?predator P=0.008).
Of the 22 taxa found in this experiment, only larvae
of the predatory mosquito Toxorhynchites theobaldi were
more common (i.e., occurred in more cups) where odo-
nates were absent (G=4.76, P<0.05). The presence of T.
theobaldi did not confound the absence of odonates;
richness and abundance results were unchanged when
the three cups containing T. theobaldi on every census
date were removed from the analysis. In low and medi-
um litter quantity treatments combined, larvae of scirtid
beetles and the mosquito Culex urichii were collectively
more likely to decline in abundance when predators were
present, and either increase or remain relatively constant
in abundance when predators were absent (Fisher's exact
two-tailed P=0.005, n=20 cups containing C. urichii, sci-
rtids, or both). Changes in abundance of these taxa were
not associated with the presence or absence of odonates
in the high litter treatment (Fisher's P=0.52, n=10).
Initial litter quantity had a significant effect on mac-
roorganism abundance over the course of the experi-
ment. Mean abundance was consistently greater in the
high litter quantity treatment (F2,24=10.06, P=0.0007,
time?treatment P=0.11; Fig. 4A). Species composition
also differed among litter treatments and over time. The
dytiscid beetle Copelatus sp. occurred only in the medi-
um litter treatment. The syrphid fly Copestylum rafaela-
num, the tipulid fly Sigmatomera spp., the psychodid fly
Telmatoscopus spp., and the mosquitoes Culex mollis
and Trichoprosopon digitatum occurred only in the high
litter treatment. Copestylum rafaelanum and C. mollis
were found only on day 0 and day 4 censuses, whereas
Sigmatomera spp. occurred only on day 16 and day 32
censuses. Despite the unequal distributions of these taxa
among treatments and census dates, resource availability
had no effect on mean species richness (F2,24=0.26,
P=0.78, time?treatment=0.49; Fig. 4B) in this experi-
ment.
Litter resources influenced predator growth in the ex-
perimental tree holes. The average length of M. coerula-
tus larvae increased with initial litter quantity over time
(F2,10=9.03, P=0.006; time?treatment P=0.03) and mean
lengths differed among all treatments on day 32 (Fig. 5).
Predation and litter quantity had interactive effects on
the degradation rate of leaf material (measured as the
proportion of litter remaining at the end of the experi-
ment; F2,22=4.03, P=0.032). A larger fraction of litter re-
mained when predators were present, but only when ini-
tial litter quantities were at low and medium levels
(Fig. 6). The proportion of litter remaining in cups with
predators was correlated with the log of initial litter mass
(Pearson r=0.618, n=14, P=0.019). The same was true in
cups without predators (r=0.898, n=14, P=0.0001). 
Discussion
Despite considerable work on the ecology of phytotelm
communities, this study is the first to show that predation
and resource availability influence the number of species
present in natural water-filled tree holes. Specifically, my
results show that tree holes contained fewer macroorgan-
ism species in the presence of large odonate larvae, and
respectively contained more or fewer species when litter
was added or removed. Because tree holes on BCI typical-
ly contain fewer than ten macroorganism species on any
wet season day, the loss or addition of a single species re-
flects a large proportional change in assemblage structure.
Predation
Effects of odonates on community structure were similar
in both natural and artificial holes: species richness was
greater in the absence of odonates, but abundance did not
differ. Results of this and related studies (Yanoviak
1999a) suggest that selective predation by odonates re-
duces macroorganism species richness by promoting lo-
130
Fig. 5 Mean (?SE) body length (mm) of larval M. coerulatus add-
ed (on day 0) to artificial tree holes containing different initial
quantities of litter (High 4.0 g, Medium 0.4 g, Low 0.04 g); n=4
for high and medium treatment means, n=5 for low treatment
means
Fig. 6 Mean (+SE) percent litter mass remaining in artificial tree
holes at the end of the predator?resource interaction experiment.
High, Medium and Low refer to different initial quantities of litter
in holes to which the predator M. coerulatus was (filled bars) or
was not (open bars) added; n=4 for the high-no predator and low-
predator means, n=5 for all other means
cal prey extinctions in tree holes. In laboratory trials,
most medium and large odonate larvae preferentially fed
on larger mosquitoes when several sizes of the same
prey species were available (Yanoviak 1999a). Most
odonates also preferred Culex spp. larvae over Aedes
terrens when similar-sized individuals of each were pre-
sented simultaneously.
In this study, the species that were negatively associ-
ated with large odonates in natural tree holes (Bezzia
snowi and Culex urichii) were among the largest prey
available (final instar body length >5 mm). Toxorhynch-
ites theobaldi, the only species that occurred significant-
ly less often when odonates were present in the artificial
tree hole experiment, is one of the largest invertebrates
in this system (Bradshaw and Holzapfel 1983). In addi-
tion, T. theobaldi is a potential competitor for mosquito
prey with odonates, and its greater frequency in cups
without odonates probably reflects the absence of intra-
guild predation and priority effects (Fincke 1999). 
Lounibos et al. (1987) found a similar negative associa-
tion between Toxorhynchites and odonates in Venezuelan
phytotelmata. The lack of association between the pres-
ence of A. terrens and odonates suggests that this species
is not a preferred prey item in natural tree holes.
Lack of prey refugia is another mechanism potentially
contributing to lower species richness in the presence of
odonates. Although odonates appear to selectively feed
on large prey, they will also take smaller organisms on
occasion (Yanoviak 1999a), and probably pursue a broad
range of prey types when food is limiting. Like other pre-
dacious aquatic insects (e.g., Formanowicz 1982), tree-
hole odonates switch between sit-and-wait and active for-
aging tactics depending on their degree of starvation (au-
thor, personal observations). Some tree-hole mosquitoes
alter their behavior in the presence of a predator (Grill
and Juliano 1996), but most species must frequently
move between the water surface and submerged detritus
to obtain oxygen and food, respectively. Thus, they are
especially vulnerable to attack by sit-and-wait predators.
An alternative explanation for reduced richness in the
presence of odonates is that some prey species avoid col-
onizing holes containing predators (i.e., by detecting
chemical or other cues). This is an unlikely mechanism
because odonates are the most common top predators in
tree holes (Fincke 1999), and most potential prey species
are endemic to this habitat (Yanoviak, in press). Any
species that depends on tree holes as a reproductive re-
source throughout the wet season would become extinct
(on a regional scale) in an attempt to avoid odonates.
Resource availability
Manipulation of leaf litter quantity had significant effects
on macroorganism species richness in natural tree holes.
There are at least three possible mechanisms for this re-
sult. First, the greater energy supply in litter addition
holes compared with litter removal holes may have in-
creased the abundance and persistence of highly competi-
tive or rare species (e.g., Srivastava and Lawton 1998).
Second, the presence or absence of litter may have influ-
enced predator efficiency. Third, some species may have
been differentially attracted to chemical or other cues
(Bentley and Day 1989) specifically associated with the
early stages of leaf decay in litter addition holes.
These mechanisms are not mutually exclusive, and
are sensitive to hole-specific conditions (e.g., nutrient in-
puts from stemflow; Carpenter 1982a) and characteris-
tics of prey assemblages (e.g., size-structure) that were
not measured in this study. However, lower macroorgan-
ism abundance in litter removal holes than in litter addi-
tion holes suggests that increased predator efficiency or
nutrient limitation caused local extinctions. Leaf litter is
functionally important as a source of both energy and
structural heterogeneity in many aquatic habitats (e.g.,
Richardson 1992). In spatially limited systems such as
tree holes, litter potentially restricts the movements of
predators and can serve as refugia for prey. Thus, the
quantity of litter present in a hole presumably affected
predator-prey encounter rates.
The lack of difference in species persistence times
shows that turnover was similar among treatments, but the
larger number of persistent species in litter addition holes
suggests that the added energy enabled a larger number of
potentially competing species to coexist. Six taxa colo-
nized holes proportionally more often after litter addition,
and then disappeared from most of the same holes after
litter manipulations were terminated. This effect is unlike-
ly to be the result of differential colonization because
these six species were not unique to litter addition holes.
Intensive laboratory and field studies are needed to more
thoroughly address each of these possible mechanisms.
In the artificial tree hole experiment, the most dramat-
ic effects of initial litter quantity on macroorganism
abundance were found during the first few censuses (up
to day 16, 5 weeks after the experiment was set up;
Fig. 4A). This pattern suggests that the major effects of
litter resources on community structure occur only dur-
ing the early stages of leaf degradation, when most solu-
ble compounds are leached into the water (e.g., Carpen-
ter 1982b; Gillon et al. 1994). Results of the litter manip-
ulation experiment in natural holes (Fig. 2; week 10)
also support this conclusion.
Predator?resource interaction
The lack of interaction between predator and resource ef-
fects on community structure in the artificial tree hole
experiment, and the lack of a predator effect on mosquito
abundance, differ from the findings of Fincke et al.
(1997). Differences between results of these two studies
are partly attributed to the type of data gathered in each,
and to the relatively short duration of the experiment in
this study. Fincke and coworkers showed that odonates
can reduce the abundance of certain size classes or stag-
es (e.g., pupae) of mosquitoes without significantly re-
ducing total abundance. Such an effect may have oc-
131
curred in this study, but macroorganism abundance was
not quantified by size or life stage.
Some taxa (C. urichii and scirtid beetles) indicated a
trend toward development of a predator?resource inter-
action, and odonates grew faster in the high litter quanti-
ty treatment (cf. Fincke et al. 1997). However, total mac-
roorganism abundance had barely stabilized by the last
two censuses (Fig. 4A). Fincke and coworkers censused
artificial tree holes over a much longer period of time
(up to 10 weeks after mosquito abundance became rela-
tively stable) and included fallen fruit as a nutrient
source. Thus, the comparatively short experimental time
span and the lack of prolonged nutrient availability pre-
cluded an interaction in this study. A replicate experi-
ment run for a longer time period, and with repeated lit-
ter inputs (as in the resource availability experiment dis-
cussed above), will provide more conclusive evidence
regarding predator?resource interactions in this system.
Given that the leaf material used in each artificial tree
hole was of similar composition, the difference in litter
degradation rates among treatments was unexpected. In-
dividual leaf fragments were progressively smaller from
the high quantity treatment to the low quantity treatment.
Because the decomposition rate of a leaf is associated
with its physical characteristics, such as permeability
and thickness (e.g., Gillon et al. 1994), the correlation of
percent litter remaining with initial litter mass is partly
attributed to the greater relative surface area of leaf frag-
ments in the low and medium litter treatments. An alter-
native explanation is that water in cups containing high
litter quantities became saturated with soluble com-
pounds, thereby slowing the rate of leaching. This is un-
likely because the experiment was conducted in the wet
season, and fresh water in the form of rain and stemflow
frequently entered all cups.
My results showed a significant interaction between
predation and litter quantity effects on litter degradation
rate (Fig. 6). Mosquito larvae and other organisms facili-
tate the degradation of leaf litter in tree holes (e.g., 
Carpenter 1982b; Fish and Carpenter 1982; Yanoviak
1999b,1999d), and the significant interaction indicates
that predation by odonates can reduce facilitation effects
when resources are limiting (i.e., a trophic cascade).
Odonates did not have an indirect effect on litter degra-
dation when resources were abundant because they were
swamped with prey. Faster odonate growth in the high
litter quantity treatment indicates that a greater absolute
amount of prey biomass was consumed. However, the
biomass consumed represents a relatively small fraction
of the total prey abundance in this treatment. The greater
frequency with which C. urichii and scirtid beetle larvae
declined in abundance in low and medium litter cups
containing odonates further explains the interaction.
Both taxa are potential prey for odonates, and both func-
tion as grazing omnivores or detritivores in tree holes
(Yanoviak, in press). A trophic cascade was not detected
in a separate tree hole microcosm experiment (Yanoviak
1999d), probably because that study employed a less vo-
racious predator and only a single grazer species.
Conclusion
My results show that predation and resource availability
influence community structure in water-filled tree holes,
and that these factors operate at different time scales.
Leaf litter inputs affect community parameters over a pe-
riod of weeks, whereas predation may act over many
months (i.e., as long as large odonates are present). On
BCI, fresh litter tends to be most abundant in the early
wet season (e.g., Foster 1996), when hole occupancy by
odonates is relatively low (Fincke 1992b). Thus, exami-
nation of predation and nutrient effects on community
structure in a seasonal context would be an appropriate
extension of this work.
Acknowledgements Comments from O.M. Fincke, V.H. Hutchison,
M.E. Kaspari, L.P. Lounibos, D.S. Srivastava, and an anonymous
reviewer improved the manuscript. C.W. Tawes and K.M. Voltura
provided logistical support during my time abroad. Portions of this
work were funded by the Smithsonian Institution, the Institute for
International Education/Fulbright Foundation, Sigma-Xi Grants-
In-Aid through the National Academy of Science, the Explorers
Club, and the University of Oklahoma.
References
Addicott JF (1974) Predation and prey community structure: an
experimental study of the effect of mosquito larvae on the pro-
tozoan communities of pitcher plants. Ecology 55:475?492
Bentley MD, Day JF (1989) Chemical ecology and behavioral as-
pects of mosquito oviposition. Annu Rev Entomol 34:401?421
Blois-Heulin C, Crowley PH, Arrington M, Johnson DM (1990)
Direct and indirect effects of predators on the dominant inver-
tebrates of two freshwater littoral communities. Oecologia 84:
295?306
Bradshaw WE, Holzapfel CM (1983) Predator-mediated, non-
equilibrium coexistence of tree-hole mosquitoes in southeast-
ern North America. Oecologia 57:239?256
Brooks JL, Dodson SI (1965) Predation, body size, and composi-
tion of plankton. Science 150:28?35
Carpenter SR (1982a) Stemflow chemistry: effects on population
dynamics of detritivorous mosquitoes in tree-hole ecosystems.
Oecologia 53:1?6
Carpenter SR (1982b) Comparisons of equations for decay of leaf
litter in tree-hole ecosystems. Oikos 39:17?22
Carpenter SR, Kitchell JF (eds) (1993) The trophic cascade in
lakes. Cambridge University Press, Cambridge
Cochran-Stafira DL, Ende CN von (1998) Integrating bacteria into
food webs: studies with Sarracenia purpurea inquilines. Ecol-
ogy 79:880?898
Diehl S (1992) Fish predation and benthic community structure:
the role of omnivory and habitat complexity. Ecology
73:1646?1661
Dobson M (1994) Microhabitat as a determinant of diversity:
stream invertebrates colonizing leaf packs. Freshwat Biol 32:
565?572
Fincke OM (1992a) Interspecific competition for tree holes: con-
sequences for mating systems and coexistence in neotropical
damselflies. Am Nat 139:80?101
Fincke OM (1992b) Behavioural ecology of the giant damselflies
of Barro Colorado Island, Panama. In: Quintero D, Aiello A
(eds) Insects of Panama and Mesoamerica. Oxford University
Press, Oxford, pp 102?113
Fincke OM (1998) The population ecology of Megaloprepus coe-
rulatus and its effect on species assemblages in water-filled
tree holes. In: Dempster JP, McLean IFG (eds) Insect popula-
tions in theory and in practice. Kluwer, Dordrecht, pp 391?416
132
Fincke OM (1999) Organization of predator assemblages in Neo-
tropical tree holes: effects of abiotic factors and priority. Ecol
Entomol 24:13?23
Fincke OM, Yanoviak SP, Hanschu RD (1997) Predation by odo-
nates depresses mosquito abundance in water-filled tree holes
in Panama. Oecologia 112:244?253
Fish D, Carpenter SR (1982) Leaf litter and larval mosquito dy-
namics in tree-hole ecosystems. Ecology 63:283?288
Formanowicz DR Jr (1982) Foraging tactics of larvae of Dytiscus
verticalis (Coleoptera: Dytiscidae): the assessment of prey
density. J Anim Ecol 51:757?767
Foster RB (1996) The seasonal rhythm of fruitfall on Barro Colo-
rado Island. In: Leigh EG Jr, Rand AS, Windsor DM (eds)
(1996) The ecology of a tropical forest, 2nd edn. Smithsonian
Institution, Washington, DC, pp 151?172
Gee JHR (1988) Population dynamics and morphometrics of
Gammarus pulex L.: evidence of seasonal food limitation in a
freshwater detritivore. Freshwater Biol 19:333?343
Gillon D, Joffre R, Ibrahima A (1994) Initial litter properties and
decay rate: a microcosm experiment of Mediterranean species.
Can J Bot 72:946?954
Grill CP, Juliano SA (1996) Predicting species interactions based
on behaviour: predation and competition in container-dwelling
mosquitoes. J Anim Ecol 65:63?76
Hard JJ, Bradshaw WE, Malarkey DJ (1989) Resource- and densi-
ty-dependent development in tree-hole mosquitoes. Oikos
54:137?144
Hart DD, Robinson CT (1990) Resource limitation in a stream
community: phosphorus enrichment effects on periphyton and
grazers. Ecology 71:149?1502
Hershey AE, Hiltner AL, Hullar MAJ, Miller MC, Vestal JR, Lock
MA, Rundle S, Peterson BJ (1988) Nutrient influence on a
stream grazer: Orthocladius microcommunities respond to nu-
trient input. Ecology 69:1383?1392
Jenkins B, Kitching RL, Pimm SL (1992) Productivity, distur-
bance and food web structure at a local spatial scale in experi-
mental container habitats. Oikos 65:249?255
Kerfoot WC, Sih A (eds) (1987) Predation: direct and indirect im-
pacts on aquatic communities. University Press of New England,
Hanover
Kitching RL (1971) An ecological study of water-filled tree-holes
and their position in the woodland ecosystem. J Anim Ecol
40:281?302
Leigh EG Jr, Rand AS, Windsor DM (eds) (1996) The ecology of
a tropical forest, 2nd edn. Smithsonian Institution, Washing-
ton, DC
L?onard PM, Juliano SA (1995) Effect of leaf litter and density on
fitness and population performance of the hole mosquito
Aedes triseriatus. Ecol Entomol 20:125?136
Lounibos LP, Frank JH, Machado-Allison CE, Ocanto P, Navarro
JC (1987) Survival, development and predatory effects of mos-
quito larvae in Venezuelan phytotelmata. J Trop Ecol 3:221?242
Lounibos LP, Nishimura N, Escher RL (1993) Fitness of a treehole
mosquito: influences of food type and predation. Oikos
66:114?118
Lounibos LP, Escher RL, Nishimura N, Juliano SA (1997) Long-
term dynamics of a predator used for biological control and
decoupling from mosquito prey in a subtropical treehole eco-
system. Oecologia 111:189?200
Louton J, Gelhaus J, Bouchard R (1996) The aquatic macrofauna
of water-filled bamboo (Poaceae: Bambusoideae: Guadua) in-
ternodes in a Peruvian lowland tropical forest. Biotropica
28:228?242
McPeek MA, Peckarsky BL (1998) Life histories and the strengths
of species interactions: combining mortality, growth, and fe-
cundity effects. Ecology 79:867?879
Naeem S (1988) Predator-prey interactions and community struc-
ture: chironomids, mosquitoes and copepods in Heliconia im-
bricata (Musaceae). Oecologia 77:202?209
Naeem S (1990) Resource heterogeneity and community structure:
a case study in Heliconia imbricata phytotelmata. Oecologia
84:29?38
Nannini MA, Juliano SA (1998) Effects of the facultative predator
Anopheles barberi on population performance of its prey
Aedes triseriatus (Diptera: Culicidae). Ann Entomol Soc Am
91:33?42
Pace ML, Cole JJ, Carpenter SR, Kitchell JF (1999) Trophic cas-
cades revealed in diverse ecosystems. Trends Ecol Evol 14:
483?488
Paine RT (1966) Food web complexity and species diversity. Am
Nat 100:65?75
Paradise CJ, Dunson WA (1997) Insect species interactions and re-
source effects in treeholes: are helodid beetles bottom-up facil-
itators of midge populations? Oecologia 109:303?312
Perrin CJ, Richardson JS (1997) N and P limitation of benthos
abundance in the Nechako River, British Columbia. Can J Fish
Aquat Sci 54:2574?2583
Power ME (1990) Effects of fish in river food webs. Science
250:811?814
Proulx M, Pick FR, Mazumder A, Hamilton PB, Lean DRS
(1996) Effect of nutrients and planktivorous fish on the phy-
toplankton of shallow and deep aquatic systems. Ecology
77:1556?1572
Richardson JS (1991) Seasonal food limitation of detritivores in a
montane stream: an experimental test. Ecology 72:873?887
Richardson JS (1992) Food, microhabitat, or both? Macroinverte-
brate use of leaf accumulations in a montane stream. Freshwa-
ter Biol 27:169?176
SAS (1989) SAS/STAT user's guide, vol 2, GLM-VARCOMP.
SAS Institute, Cary
Snow WE (1949) The arthropoda of wet tree holes. PhD disserta-
tion, University of Illinois, Urbana
Sokal RR, Rohlf FJ (1981) Biometry. Freeman, New York
Sota T (1996) Effects of capacity on resource input and the aquat-
ic metazoan community structure in phytotelmata. Res Popul
Ecol 38:65?73
Sota T, Mogi M (1996) Species richness and altitudinal variation
in the aquatic metazoan community in bamboo phytotelmata
from North Sulawesi. Res Popul Ecol 38:275?281
Spencer CN, Ellis BK (1998) Role of nutrients and zooplankton
in regulation of phytoplankton in Flathead Lake (Montana,
U.S.A.), a large oligotrophic lake. Freshwater Biol 39:755?
763
Srivastava DS, Lawton JH (1998) Why more productive sites have
more species: experimental test of theory using tree-hole com-
munities. Am Nat 152:510?529
Walker ED, Merritt RW (1991) Behavior of larval Aedes triseria-
tus (Diptera: Culicidae). J Med Entomol 28:581?589
Walker ED, Kaufman MG, Ayres MP, Riedel MH, Merritt RW
(1997) Effects of variation in quality of leaf detritus on growth
of the eastern tree-hole mosquito, Aedes triseriatus (Diptera:
Culicidae). Can J Zool 75:707?718
Washburn JO, Mercer DR, Anderson JR (1991) Regulatory role of
parasites: impact on host population shifts with resource avail-
ability. Science 253:185?188
Yanoviak SP (1999a) Community ecology of water-filled tree
holes in Panama. PhD dissertation, University of Oklahoma,
Norman
Yanoviak SP (1999b) Effects of leaf litter species on macroinver-
tebrate community properties and mosquito yield in Neotropi-
cal tree hole microcosms. Oecologia 120:147?155
Yanoviak SP (1999c) Community structure in water-filled tree
holes of Panama: effects of hole height and size. Selbyana 20:
106?115
Yanoviak SP (1999d) Effects of Mecistogaster spp. (Odonata:
Pseudostigmatidae) and Culex mollis (Diptera: Culicidae) on
litter decomposition in neotropical treehole microcosms. Fla
Entomol 82:462?468
Yanoviak SP (in press) The macrofauna of water-filled tree holes
on Barro Colorado Island, Panama. Biotropica 33
Zaret TM (1980) Predation and freshwater communities. Yale
University Press, New Haven
133