Do non-myrmocophilic epiphytes influence community structure 
of arboreal ants?
Sabine Stuntz1, Christian Linder2, Karl Eduard Linsenmair3, Ulrich Simon4, Gerhard Zotz5,6*
1 Lehrstuhl f?r Botanik II der Universit?t W?rzburg, Julius-von-Sachs-Platz 3, 97082 W?rzburg, Germany
2 Biokybernetik und Theoretische Biologie, Universit?t Bielefeld, 33615 Bielefeld, Germany
3 Lehrstuhl f?r Zoologie III der Universit?t W?rzburg, Am Hubland, 97074 W?rzburg, Germany 
4 Technische Universit?t M?nchen, Forstwissenschaftliche Fakult?t, Am Hochanger 13, 85354 Freising, Germany
5 Smithsonian Tropical Research Institute, Apdo 2072, Balboa, Panama
6 Current address: Botanisches Institut der Universit?t Basel, Basel, Switzerland
Received January 28, 2002 ? Accepted May 3, 2002
Abstract
In a one-year-survey in Panama we examined the influence of a tree crown?s epiphyte assemblage
on its ant fauna. Ants were collected with various types of insect traps in 25 crowns of Annona
glabra trees. The study trees were assigned to three different categories according to their epiphyte
load, and to an epiphyte-free control group. We collected 22,335 specimens of 91 morphospecies,
32 genera and six subfamilies. By far the most abundant species was Solenopsis zeteki, a minute
Myrmicinae, which was found in each of the 25 study trees. Many other species were also rather
common and widely distributed throughout the study area. Only six species were singletons. Mea-
sures of ?- and ?-diversity, species abundance and species composition were not affected by the epi-
phyte load of a tree. We also made direct in situ observations of ants on 34 additional Annona
glabra trees with and without epiphytes. Workers were attracted with tuna and sugar baits, and in-
terspecific interactions and nesting sites were recorded. In total, 40 species of ants were found, all of
which had also been collected in the traps. Almost half of the colonies (48%) used dead wood as
nesting substrate, while 29% nested in epiphytes. Consistent with the results of the trap survey, the
epiphyte load of the study trees had no influence on ant species richness and composition, but a sig-
nificant correlation between ant abundance and epiphyte load was detectable. In both data sets, the
lack of associations between ant species indicated that the ant assemblages were not structured in a
mosaic-like fashion. We conclude that epiphytes do hardly influence the composition of ant assem-
blages in the studied tree crowns, probably because arboreal ants are highly opportunistic with re-
spect to their host plants.
In einer einj?hrigen Studie wurde der Einfluss von Epiphytengemeinschaften in Baumkronen auf die
dort lebende Ameisenfauna untersucht. Die Ameisenfauna wurde mit zwei unterschiedlichen
Methoden erfasst. Zum einen wurden mit verschiedenen Arthropodenfallen Annona glabra B?ume
beprobt, die gem?? ihres Epiphytenbewuchses in drei verschiedene Kategorien und eine unbewach-
sene Kontrollgruppe eingeteilt wurden. Insgesamt wurden 22.335 Individuen aus 91 Arten gefan-
gen, die 32 Gattungen und sechs Unterfamilien zuzuordnen waren. Die h?ufigste Art, Solenopsis
zeteki, eine winzige Myrmicinae, wurde auf allen 25 beprobten B?umen gefunden. Viele andere
Ameisenarten waren ebenfalls recht weit und gleichm??ig ?ber das Untersuchungsgebiet verbreitet.
*Corresponding author: Gerhard Zotz, Botanisches Institut der Universit?t Basel, Sch?nbeinstrasse 6, CH - 4056, Basel,
Switzerland, Phone: +41-61 267 35 11, Fax: +41 61 267 35 04, E-mail: gerhard.zotz@unibas.ch
1439-1791/03/04/04-363 $ 15.00/0
Basic Appl. Ecol. 4, 363?374 (2003)
? Urban & Fischer Verlag
http://www.urbanfischer.de/journals/baecol
Basic and Applied Ecology
Nur sechs Arten wurden in Einzelexemplaren gefunden. Weder Abundanz der Arten noch Arten-
zusammensetzung wurden durch die Art des Epiphytenbewuchses beeinflusst. Zweitens wurden
auch in-situ-Beobachtungen an einer zus?tzlichen Gruppe von 34 B?umen derselben Baumart
durchgef?hrt. Diese B?ume waren entweder mit Epiphyten bewachsen oder epiphytenfrei. Arbeite-
rinnen wurden mit Thunfisch- und Zuckerk?dern angelockt, und zwischenartliche Interaktionen
sowie die Neststandorte registriert. In dieser Teiluntersuchung wurden 40 Arten bestimmt, die aus-
nahmslos auch in den Fallen gefunden worden waren. Etwa die H?lfte der Kolonien (48%) befand
sich in Totholz und 29% in Epiphyten. Auch hier zeigte sich kein Einfluss des Epiphytenbewuchses
auf die Zahl der Arten oder ihre Zusammensetzung, die Korrelation zwischen Ameisenabundanz
und Epiphytenbewuchs war aber signifikant. In beiden Teilstudien gab es keine signifikanten Arten-
Assoziationen, die Ameisengemeinschaften waren also nicht mosaikartig strukturiert. Aus den
Daten kann die Schlussfolgerung abgeleitet werden, dass Epiphyten die Zusammensetzung der
Ameisengemeinschaften in den untersuchten Baumkronen kaum beeinflussen, weil Ameisen
wahrscheinlich sehr opportunistisch auf unterschiedliche pflanzliche Ressourcen reagieren.
Key words: Annona glabra ? ant mosaic theory ? canopy ? insect traps ? Panama
is an important prerequisite for many canopy-nesting
ants (Longino & Nadkarni 1990). Many ant species
also nest inside non-myrmecophilic epiphytes (Schim-
per 1888; Bl?thgen et al. 2000). Richards (1996) even
claims that epiphytes provide the chief nesting sites
for arboreal ants in tropical rainforests. Epiphytes
sometimes foster a rich arthropod fauna (Cotgreave
et al. 1993; Richardson 1999; Stuntz et al. 2002b), a
potential resource for predatory ants. Moreover, epi-
phytes have a mitigating influence on the microclimat-
ic conditions in their immediate surroundings (Stuntz
et al. 2002a). Thus, we hypothesized that epiphytes
positively influence ant abundance and diversity. Al-
ternatively, ants may also be indifferent to epiphytes:
being highly opportunistic in many respects (e.g.,
Stork 1987; Bl?thgen et al. 2000). The following re-
port, which is part of a series of papers on the influ-
ence of vascular epiphytes on canopy arthropods
(Stuntz et al. 2002a,b, 2003), addresses these diver-
gent notions by comparing the ant faunas of tree
crowns bearing different sets of epiphyte assemblages,
and trees free of epiphytes. 
Our study also contributes to the ongoing discus-
sion on the occurrence and importance of ant mosaics
in the tropics. Recently, the existence of well-orga-
nized ant mosaics was refuted for high-diversity rain-
forests (Floren & Linsenmair 2000). Most of our
knowledge of ant mosaics comes from orchards, man-
groves or other areas with little diverse faunas (Room
1971; Leston 1973a; Leston 1973b; Majer 1976; Fox
& Fox 1982; Majer 1982; Cole 1983; Adams 1994;
Fowler et al. 1998). Thus, we also asked the question
whether in our model system of small trees at the for-
est edge, the ant fauna is mosaic-like structured or
rather heterogeneous and unpredictable like the one
investigated by Floren & Linsenmair (2000). 
364 Stuntz et al.
Basic Appl. Ecol. 4, 4 (2003)
Introduction
Due to their extraordinary abundance, ants are a most
remarkable component of the tropical arboreal arthro-
pod fauna (Erwin 1983; H?lldobler & Wilson 1990),
and are a useful indicator taxon, for example in order
to assess overall biodiversity (Longino & Colwell
1997), altitudinal gradients (Br?hl et al. 1999), forest
edge and fragmentation effects (Majer & Delabie
1999; Dejean & Gibernau 2000), or faunal differences
between understory and canopy (Longino & Nad-
karni 1990; Yanoviak & Kaspari 2000). Ants may be
of major importance for the structure of arboreal
arthropod communities, because they exert a constant,
high predation pressure (Tobin 1995; Floren & Lin-
senmair 1997): they have been labeled the ?most im-
portant invertebrate predators? in the tropics (H?ll-
dobler & Wilson 1990; Linsenmair 1990).
For more than a century, ant-plant interactions
have been a favorite topic in tropical ecology (e.g.,
Schimper 1888; Janzen 1974; Fiala et al. 1994; De-
jean et al. 1995). Epiphytes are frequent partners of
such mutualisms (Davidson & Epstein 1989; Benzing
1990), providing living space for ant colonies (doma-
tia) or nutrition from extrafloral nectaries, or both. In
return, the plants benefit from nutrients retrieved
from the ants? waste, or enjoy protection from herbi-
vores (reviewed in H?lldobler & Wilson 1990). In this
paper, however, we investigate whether non-myrme-
cophilic epiphytes also influence ant diversity and
abundance in the studied tropical forest canopy.
Apart from increasing the structural heterogeneity of
the canopy habitat and providing shelter from climat-
ic extremes, some epiphytes could promote ant occur-
rence by impounding large amounts of leaf litter
(Davidson & Epstein 1989; Richardson 1999), which
Materials and methods
The study was conducted in the tropical moist forest
of the Barro Colorado Nature Monument (BCNM,
9?10?N, 79?51?W) in Panama. The area receives ap-
proximately 2600 mm of annual precipitation with a
pronounced dry season from late December to April.
Detailed descriptions of climate and vegetation can be
found in Leigh & Windsor (1982).
Study trees and epiphytes
The chosen host tree, Annona glabra L., grows abun-
dantly along the shore of Lake Gat?n. Despite its rather
small stature (mean height of the study trees 4.9 m ?
0.9 SD, n = 25), the climatic conditions in these tree
crowns are similar to the upper forest canopy due to its
exposure to sun and wind along the shore (Zotz et al.
1999). It is often dominated by a single epiphyte species
(Zotz et al. 1999), which allowed us to define distinct
tree categories with rather uniform epiphyte assem-
blages: 1) trees free of epiphytes as control group, 2)
trees with the orchid Dimerandra emarginata, 3) trees
with the large tank bromeliad Vriesea sanguinolenta,
and 4) trees dominated by the medium-sized bromeliad
Tillandsia fasciculata. Hereafter, the study species are
addressed by their generic names. In order to account
for spatial heterogeneity across different locations, we
chose sites where trees of all categories grew in close
vicinity. Tillandsia-trees were found only at four of the
seven study sites (distributed all over BCNM), and were
sampled only when arthropod abundance was expected
to be high. Traps were thus closed during the second
half of the rainy season, i.e. from July to November
1998. Consequently, two different data sets were used
for comparisons among categories. When the entire
sampling period of thirteen months was included, we
compared only the categories 1?3, when all four cate-
gories were taken into account, we analyzed data from
eight months with active traps in all trees.
Trapping and processing the ants
Arthropods were collected with three different types of
traps: flight interception traps, branch traps and yellow
color traps, which remained in the tree crowns for an
entire year and were emptied every two weeks. They
are illustrated and described in Stuntz et al. (1999).
Branch traps were most effective for capturing ants,
yielding more than twice as many specimens than flight
interceptors and yellow color traps (Stuntz et al. 1999).
The captured arthropods were transferred to 70%
ethanol until further treatment. Ants were separated
from the rest of the catch, mounted and identified to
genus level with the key of Bolton (1994), then as-
signed to morphospecies. The reference collection was
sent to specialists (Philip S. Ward for the Pseu-
domyrmecinae and John T. Longino for all other sub-
families) for species identification. Vouchers are de-
posited at the Smithsonian Tropical Research Institute
in Panama, and at the Technische Universit?t
M?nchen (Freising, Germany).
In situ observation of the ant fauna
We also conducted a bait study in 34 additional An-
nona trees in the same study area. Fourteen trees were
free of epiphytes, two were dominated by Vriesea, six
by Tillandsia, and twelve supported a mixed assem-
blage of epiphyte species (including, e.g., Tillandsia
subulifera Mez., and Polypodium crassifolium L.)
Total ant abundance per tree (irrespective of species
identity) was visually estimated during a 10-minute-in-
terval prior to the placement of baits. Five abundance
classes were used: class 1 (1?2 ants observed), class 2
(appr. 10 ants observed), class 3 (appr. 100 ants), class 4
(over 100 ants) and class 5 (many hundreds of ants, busy
trails on every branch). Then, the ant community com-
position was studied on each tree at daytime
(09:00?17:00 h) by placing tuna and sugar baits on all
major branches and stems. During the next two hours,
species arriving at the baits as well as all active ants with-
in the tree crown were noted. Ant behavior, such as ag-
gressive interactions with other species, bait monopoliza-
tion, or worker recruitment was recorded. By visually
following ants departing from baits, the location and
number of nesting sites was determined. One to ten spec-
imens per species were collected and processed as de-
scribed above. Similarly, nocturnal activity was also
studied on a subset of 18 of the study trees (between
21:00?24:00 h). Species that had monopolized more than
50% of the baits in a particular tree at the end of the ob-
servation time were considered dominant on this tree.
Epiphyte biomass and other host tree traits
We estimated epiphyte biomass by measuring either
the maximum leaf length of each bromeliad or the
length of the youngest stem of each orchid stand, re-
spectively. Biomass and leaf area correlates with those
parameters (compare Stuntz et al. 2003). In general,
trees with Dimerandra had both lower epiphyte
biomass and epiphyte leaf area than trees with Vriesea
or Tillandsia. Total epiphyte biomass ranged from 90 g
dry mass in a tree with Dimerandra to 3,853 g dry
mass in a tree with abundant Tillandsia, and epiphyte
leaf area ranged from 0.21 m2 in the same Dimerandra
tree to 27.9 m2 in a tree with Vriesea. The median leaf
area of a tree was 30 m2 (host tree foliage only), com-
pared to 8 m2 for epiphytes per tree.
Do epiphytes shape arboreal ant communities? 365
Basic Appl. Ecol. 4, 4 (2003)
For the in situ observations, host tree traits (tree
height, crown volume, circumference, stem number)
were estimated. As the majority of ants nested in dead
wood cavities, the amount of dead wood in a tree
crown (mainly dead branches still attached to the tree)
was visually estimated and scored at a scale from one
to ten (0 = no dead wood; 10 = large amounts of dead
wood).
Statistics
Statistical analysis was done with STATISTICA (Stat-
Soft Inc., Oklahoma, USA). The numbers of species
and individuals of the four tree categories were not
normally distributed and thus compared with Kruskal-
Wallis-ANOVA (KW-ANOVA). Changes over time
were analyzed with repeated-measures ANOVA (RM-
ANOVA). A Spearman rank coefficient was used to
test for correlations between epiphyte load and ant
abundance. Species richness, i.e. the absolute number
of species found in one sampling unit, was used as a
measure of ?-diversity, and the S?rensen index as a
measure of ?-diversity (Magurran 1988). The
S?rensen values in Table 3 were normally distributed
and thus allowed the use of parametric one-way
ANOVA among categories. To test for differences in
the species compositions of the faunas among epiphyte
species, we ran multidimensional scaling analyses
(MDS) based on a dissimilarity matrix of 1-S?rensen
values (Southwood 1978). This multivariate ordina-
tion method was favored over factor analysis because
no normal distribution of the data is required. Finally,
association analyses were computed following the pro-
tocol of Ludwig & Reynolds (1988), which uses a se-
ries of pairwise comparisons and is based on ?2-statis-
tics.
Results
I. Trap survey
Composition of the fauna
In total, we collected 22,335 specimens from 91
species in 32 genera and six subfamilies (Table 1).
Many species were widespread throughout the study
area (Fig. 1): 26 species (29%) were found in more
than half of all study trees, eight species occurred in
over 90% of the trees, and three of those, Solenopsis
zeteki, Pheidole cf. flavens and Camponotus (Myrmo-
brachys) sp. 4 (cf. auricomas) occurred on every single
study tree. Solenopsis zeteki was by far the most abun-
dant species (4,632 specimens) and contributed one
fifth of all individuals. Myrmicinae were the most di-
verse and numerous subfamily (40 species and 15,222
individuals) (Table 1). The most species-rich genera
were Camponotus (Formicinae), Pheidole (Myrmici-
nae) and Pseudomyrmex (Pseudomyrmecinae) each
with ten species.
Comparison of tree/epiphyte categories
?-diversity. During a trapping period of eight months,
we collected a median number of 26 (range: 15?38)
species and 510 (88?1,039) individuals per tree (n =
25). There were no significant differences in the num-
ber of ant species and individuals among categories
(Table 2; KW-ANOVA, numbers of species: p = 0.62;
numbers of individuals: p = 0.39). To account for sea-
sonal fluctuations, we tested for differences between
tree groups over time: data were recorded separately in
two-week-intervals and compared with RM-ANOVA.
Seasonal variation significantly influenced both the
numbers of individuals and species (p < 0.001). Con-
firming the results of the KW-ANOVA, the tree/epi-
phyte category did not significantly influence either
the number of individuals (p = 0.30) nor species (p =
0.29). There was no interaction between time and tree
category (p > 0.55). 
At the subfamily level, differences in abundance
across tree categories were found in a single taxon:
Dolichoderinae were most numerous in Tillandsia
trees (KW-ANOVA; n = 25; p = 0.049). There, we col-
lected a median number of 139 Dolichoderinae (range:
42?278, n = 4), compared to 37 (13?166, n = 7) in
control trees, 34 (15?91, n = 7) in Dimerandra trees
and only 10 (0?115; n = 7) in Vriesea trees. In three of
the four trees with Tillandsia, hundreds of ants of the
genus Azteca inhabited (and fiercely defended) most
bromeliads, each probably being an outpost of a poly-
domous colony (Stuntz & Linder, personal observa-
tion). The abundance of the other subfamilies was in-
366 Stuntz et al.
Basic Appl. Ecol. 4, 4 (2003)
Fig. 1. Rank-abundance plot of ant species. Abundance is defined as the
proportion of study trees (n = 25), in which a certain species was collected
during a trapping period of one year. For lack of space, the species ranked
along the x-axis have been given numbers (see Table 1). 
dependent of the tree category, both in terms of species
numbers and individuals (KW-ANOVA, p > 0.1).
The category assignment was merely based on
species identity of the prevalent epiphyte in a tree, irre-
spective of the quantity of epiphytes in its crown.
Therefore we also tested for correlations between epi-
phyte biomass and numbers of species and individuals.
In neither case did we find a significant relationship
Do epiphytes shape arboreal ant communities? 367
Basic Appl. Ecol. 4, 4 (2003)
Table 1. Species list. Given are totals of specimens (n) trapped in 25 study trees during the study period. Species marked with an asterisk were also collected
during direct observations. The totals of individuals within one subfamily are given in italics, the number of species within a subfamily in parentheses behind
the family names. The species codes (?code?) are referred to in Fig. 1. Morpho-species names (genus + sp.1, sp.2 etc) relate to our voucher collection or to the
collection of J. T. Longino (JTL-001).
Species or morpho-species name n code Species or morpho-species name n code
Myrmicinae (40) 15,222 Formicinae (15) 1,952
Solenopsis zeteki* 4,678 1 Paratrechina sp. 2* 602 27
Wasmannia rochai* 2,293 15 Camponotus (Myrmobrachys) sp. 4 (cf. auricomus) 344 3
Pheidole cf. flavens* 1,893 2 Camponotus atriceps* 301 6
Solenopsis sp. 1* 1,631 10 Paratrechina sp. 1* 220 7
Monomorium floricola* 928 21 Paratrechina sp. 3 181 19
Solenopsis sp. 4 918 13 Camponotus sexguttatus* 154 12
Pheidole punctatisssima* 857 17 Camponotus novogranadensis 52 25
Cyphomyrmex rimosus complex* 567 5 Paratrechina sp. 4 32 32
Crematogaster carinata and brasiliensis* 1) 472 14 Camponotus (Myrmeurynota) sp. 7 20 37
Pheidole radoszkowskii pugnax 417 28 (cf. linnaei)*
Pheidole cocciphaga* 344 16 Camponotus mucronatus* 14 35
Pheidole sp. 7 29 75 Camponotus senex 10 51
Pheidole radoszkowski luteola 29 44 Brachymyrmex sp. 1* 8 53
Leptothorax echinatinodis 19 38 Camponotus planatus 8 52
Xenomyrmex JTL-001* 17 62 Camponotus (Tanaemyrmex) sp. 1* 5 57
Pheidole sp. 6 17 34 Camponotus sericeiventris 1 82
Pheidole sp. 0 14 55
Crematogaster brevispinosus crucis 11 45 Dolichoderinae (9) 3,619
Pyramica cf. epinotalis 10 76 Azteca cf. velox 1,410 4
Pheidole pubiventris 10 63 Azteca cf. trigona* 1,336 8
Pheidole decem 8 77 Dolichoderus bispinosus* 429 23
Cephalotes grandinosus 7 54 Dolichoderus diversus 407 18
Xenomyrmex panamanus* 6 66 Dolichoderus debilis* 15 41
Atta cephalotes 6 64 Azteca forelii* 10 46
Strumigenys borgmeieri 6 48 Tapinoma melanocephalum* 7 40
Cephalotes umbraculatus 4 68 Dolichoderus lutosus 3 70
Cephalotes atratus 4 67 Dolichoderus laminatus 2 72
Cephalotes minutus 4 58
Acromyrmex octospinosus* 3 69 Pseudomyrmecinae 543
Strumigenys emmae 3 61 Pseudomyrmex elongatus* 241 9
Crematogaster crinosa* 3 60 Pseudomyrmex gracilis* 121 24
Cardiocondyla wroughtonii 3 59 Pseudomyrmex simplex* 85 20
Strumigenys elongata 2 80 Pseudomyrmex ita* 39 31
Rogeria foreli 2 74 Pseudomyrmex filiformis 34 43
Cephalotes setulifer 2 71 Pseudomyrmex oculatus 12 56
Wasmannia auropunctata 1 92 Pseudomyrmex tenuissimus 7 54
Solenopsis sp. 6 1 91 Pseudomyrmex euryblemma 2 79
Solenopsis sp. 5 1 90 Pseudomyrmex boopis 1 88
Megalomyrmex silvestrii 1 84 Pseudomyrmex browni 1 89
Leptothorax antoniensis 1 83
Ecitoninae (10) 572
Ponerinae (7) 427 Labidus praedator 358 30
Odontomachus bauri* 225 11 Labidus coecus 149 36
Odontomachus ruginodis* 116 22 Neivamyrmex sp. 1 36 29
Pachycondyla harpax 31 49 Neivamyrmex sp. 2 11 42
Hypoponera opaciceps 30 33 Eciton hamatum 10 50
Pachycondyla villosa 18 39 Neivamyrmex pilosus 3 78
Ectatomma ruidum 6 65 Eciton burchelli 2 73
Anochetus inermis group 1 81 Neivamyrmex sp. 3 1 86
Neivamyrmex sp. 4 1 85
Nomamyrmex esenbeckii 1 87
1) These two species were lumped and assigned to one morphospecies
(Spearman rank correlation, numbers of species: p =
0.81; numbers of individuals: p = 0.18). This was also
true when analyzing the subfamilies separately (Spear-
man rank correlation, r between ?0.3: Myrmicinae
and +0.3: Dolichoderinae, p > 0.1).
?-diversity. Ant assemblages in the four tree/epiphyte
categories did not differ in species composition (Fig. 2).
The symbols were not grouped corresponding to the
four tree categories, but were rather evenly distributed.
The S?rensen indices of the ant communities in the
four categories were high and ranged from 0.69 (be-
tween Tillandsia trees and control trees) to 0.83 be-
tween trees with Vriesea and control trees.
Species assemblages of individual trees did not dif-
fer significantly either (Table 3). The S?rensen indices
between pairs of epiphyte-laden trees among each
other, of control trees among each other and of epi-
phyte-laden trees paired with control trees did not dif-
fer significantly (ANOVA, p = 0.75). Similar results
were obtained when including only those species that
were present on a minimum of three study trees (to re-
duce chance effects by the occurrence of rare species;
one-way ANOVA, p = 0.87), or when excluding the
most abundant species, reasoning that their ?general-
ist? appearance might blur subtle differences in the
composition of less abundant species (one-way
ANOVA, p = 0.10; Table 3). 
The omnipresent Solenopsis zeteki was first-ranked
in all four tree categories (Table 4). Of the 19 species
with a minimum of five specimens (median) per tree
(during eight months of trapping), three occurred
throughout all categories and another three were
abundant in three of the four categories. Nine of these
19 species were ranked among the most abundant
species in one category alone.
Association calculations for 903 species pairs re-
vealed no significant association between species, nei-
ther positively nor negatively (p > 0.05). This suggests
that a structured ant mosaic did not exist in our study
system.
2. In situ observation of the ant fauna
Faunal composition
We collected specimens of 40 species of ants in 20 gen-
era and five subfamilies (Table 1). During two hours of
observation, a median of 5 species (range 1?12, n =
34) was recorded per study tree. Most frequently ob-
served was Odontomachus ruginodis, which nested on
47% of all study trees, followed by Solenopsis zeteki,
recorded on 44% of the trees. Fifteen species (38%)
were seen on only one tree. All species found during
the observations also appeared in the traps.
Dominance and species associations
Twelve species attained dominant status on at least
one tree (i.e., occupied over 50% of all baits in a given
tree at the end of the observation period; Table 5). As-
sociation calculations were conducted between ant
species of the 34 Annona trees. To minimize chance ef-
fects, only species occurring on a minimum of five
trees were included. Three species pairs showed a posi-
tive association (Camponotus atriceps/Crematogaster
cf. carinata (association coefficient C = 0.40); Cam-
ponotus atriceps/Paratrechina sp. 1 (C = 0.68); Cam-
ponotus sexguttatus/Paratrechina sp. 1 (C = 0.68)).
368 Stuntz et al.
Basic Appl. Ecol. 4, 4 (2003)
Fig. 2. Multidimensional scaling analysis of the ant assemblages of the four
tree/epiphyte cagories. The ordination is based on a dissimilarity matrix (1-
S?rensen) of the faunas of 25 study trees; each symbol represents one tree.
Ants were collected with 125 traps during eight months.
Fig. 3. Nesting substrates used by arboreal ants in Annona glabra. In total,
we documented the nesting sites of 122 colonies in 34 study trees.
None of these associations was exclusive nor obligate.
There were no negative associations between domi-
nant species, and the latter did not maintain mutually
exclusive territories.
The influence of epiphytes and other habitat traits
The presence of epiphytes in the study trees had no in-
fluence on ant species diversity (KW-ANOVA, p = 0.99).
Epiphyte biomass was positively correlated with esti-
mated ant abundance (Spearman rank correlation,
p = 0.045), but not with species richness (Spearman
rank correlation, p = 0.49). Species composition was
also independent of epiphyte load in the Annona trees,
as revealed by MDS (not shown). The amount of dead
wood in a tree crown showed a significant correlation
with ant species richness (Spearman rank correlation,
Do epiphytes shape arboreal ant communities? 369
Basic Appl. Ecol. 4, 4 (2003)
Table 2. Numbers of ant individuals and species in the 25 study trees. Given are median values, minima and maxima of ants collected in n trees during a peri-
od of eight months.
Control trees Trees with Dimerandra Trees with Tillandsia Trees with Vriesea
Individuals per tree
Median 272 376 679 510
Min 88 297 436 269
Max 1,039 955 1,014 913
Species per tree
Median 26 27 22 29
Min 17 15 19 21
Max 38 35 30 35
n 7 7 4 7
Table 3. Average S?rensen values (means ? SD) and statistics of 300 pair wise comparisons among study trees. ?Only abundant species? includes only species
that were present on at least three trees, and ?only rarer species? excludes species that were present on more than twenty study trees.
Comparison Epiphyte-laden trees Epiphyte-laden trees Control trees ANOVA
among each other with control trees among each other p-level 
All species 0.56 ? 0.08 0.55 ? 0.08 0.54 ? 0.08 0.75
Only abundant species 0.59 ? 0.09 0.58 ? 0.09 0.57 ? 0.09 0.87
Only rarer species 0.44 ? 0.11 0.42 ? 0.11 0.39 ? 0.12 0.10
n pair wise comparisons 153 126 21
Table 4. Rank order and abundance (n = median) of the most abundant species in the four tree categories. Included are species of which at least five individu-
als (median) per tree were trapped during eight months. Species present in all four categories are displayed in bold script, and species occurring in three of the
four categories are underlined.
Control trees n Trees with Dimerandra n Trees with Tillandsia n Trees with Vriesea n
Solenopsis zeteki 96 Solenopsis zeteki 64 Solenopsis zeteki 221 Solenopsis zeteki 55
Camponotus sp.4*) 18 Monomorium floricola 53 Azteca cf. velox 139 Pheidole cf. flavens 52
Azteca cf. velox 13 Solenopsis sp.4 47 Azteca cf. trigona 37 Solenopsis sp.1 22
Dolichoderus diversus 12 Pheidole punctatisssima 36 Pheidole cf. flavens 19 Azteca cf. trigona 9
Camponotus atriceps 11 Pheidole radosz. pugnax 29 Paratrechina sp.1 11 Odontomachus bauri 8
Pheidole cf. flavens 11 Pheidole cf. flavens 14 Camponotus sp.4* 7 Pheidole punctatisssima 8
Azteca cf. trigona 10 Dolichoderus diversus 13 Cyphomyrmex rim. compl. 6 Camponotus atriceps 7
Crematogaster carinata 
and brasiliensis**) 10 Dolichoderus bispinosus 11 Solenopsis sp.1 6 Cyphomyrmex rim. compl. 5
Pseudomyrmex elongatus 6 Pseudomyrmex gracilis 10 Pseudomyrmex elongatus 6
Camponotus sp.4* 9 Camponotus atriceps 5
Solenopsis sp.1 7
Azteca cf. trigona 6
*) Camponotus (Myrmobrachys) sp.4 (cf. auricomus)
**) These two species were lumped and assigned to one morphospecies
r = 0.4; p = 0.008). Other host tree parameters like tree
height or crown volume were not correlated with ant
diversity (p > 0.3).
Nesting sites
The studied tree crowns provided a variety of sub-
strates that were used as nesting sites (Fig. 3). The ma-
jority of colonies nested in dead wood (48%), another
large fraction of ant nests were found inside Vriesea
and Tillandsia epiphytes (29%). One dominant
species, Azteca cf. trigona built large carton nests. A
few small species, e.g. Solenopsis zeteki, nested in
cracks of the fissured Annona bark, others used aban-
doned, carton-covered termite trails. Many Odon-
tomachus-colonies nested in crotches close to the
water surface. Ants of the genus Pseudomyrmex usual-
ly built their nests in thin tips of dead branches in the
crown periphery. The epiphyte-nesting ant species
were rather unspecific in respect to their hosts: nine of
the twelve species nesting in epiphytes were also pre-
sent in control trees without epiphytes, mostly nesting
in dead wood.
Discussion
Annona glabra as model system for tropical canopies?
Annona glabra seems to be a feasible model system for
the forest canopy, at least for ants. Both overall diver-
sity (91 species) and average species richness per tree
(26 species) were within the range of the species num-
bers reported in previous studies of tropical canopies
(e.g. Floren & Linsenmair 1994: 30?40 species per
tree in Malaysia; Majer 1994: 91 species in total from
a Brazilean cocoa plantation; Majer et al. 2000: 14
species per tree in Brazilean rainforest). In the canopy
of Luehea seemanii Triana & Planch, a tall tree (Croat
1978), Montgomery (1985) found 22?35 species of
ants per tree on Barro Colorado Island. More evidence
that the ant fauna in Annona is comparable to the one
in the forest canopy comes from Yanoviak & Kaspari
(2000). They found 32 ant species on baits in the
crowns of four emergent tree species in BCNM, 27 of
which could be identified to species level: 63% of
these species were also common in our samples.
Dominants, submissives and mosaics
Methodological considerations. Many ant communi-
ties have clear hierarchies, featuring few dominant
species and several subordinate species (e.g., Leston
1973a; H?lldobler & Wilson 1990). The dominance
of an ant species cannot necessarily be deduced from
its massive occurrence in insect traps, but rather in-
volves a characteristic behavior towards co-occurring
species. Dominants as consistently aggressive to work-
ers of all other species, whereas those of the subordi-
nate species avoid dominants (H?lldobler and Wilson
1990). Frequently, ant species are considered domi-
nant if they are capable of monopolizing bait, i.e. of
defending food resources successfully against other
species (Yanoviak & Kaspari 2000). Trap yields may
still reasonably reflect dominance, because dominants
have large colonies and quickly recruit many workers
to food resources (e.g., Leston 1973a; H?lldobler &
Wilson 1990). This higher activity should be reflected
in greater abundance in the traps. Correspondingly,
88% of the species in the present study that were ob-
served to monopolize baits (15 of 17 species), were
among the most abundant species caught in the traps
and were captured in over 50% of all study trees.
Studies with insect traps (as with insecticidal knock-
down) also yield an unknown number of ?tourists? or
?transient species? (sensu Stork 1987), i.e. taxa not
genuinely associated with the trees in which they were
trapped. An example for such transient species are the
ten species of army ants (Ecitoninae), or the leaf cut-
ters Atta cephalotes and Acromyrmex octospinosus
(Table 1): these are obligate ground nesters obviously
only foraging in the study trees. If the ?tourist? fraction
is large, subtle patterns in species diversity might easily
be blurred. Our in situ observations reduced this prob-
lem. During only 82 hours of observation, we docu-
mented local nesting sites for 40 of the 91 species col-
lected during one year (i.e., over almost 8800 hours)
of trapping. Only the rarer species and the ?transients?
(e.g. the ten army ant species) were not recorded dur-
ing direct observations (Table 1). Overall, the results
of the observations confirmed the outcome of the trap
survey, both in terms of the (non-detectable) influence
of epiphytes on the ant fauna and the lack of an ant
mosaic.
370 Stuntz et al.
Basic Appl. Ecol. 4, 4 (2003)
Table 5. Dominant ant species in Annona glabra. Dominance was inferred
when a species occupied over 50% of all baits in a given tree at the end of
an observation period.
Species dominant on n trees
Azteca cf. trigona 2
Azteca forelii 4
Camponotus (Myrmeurynota) sp. 7 (cf. linnaei) 2
Camponotus atriceps 1*
Dolichoderus bispinosus 1
Monomorium floricola 1
Odontomachus ruginodis 3*
Odontomachus bauri 1
Paratrechina sp. 1 1
Paratrechina sp. 2 4
Pheidole punctatisssima 1
Solenopsis zeteki 3
* species active mostly nocturnally
Ant mosaics. Since Leston (1973a) originally de-
scribed the phenomenon of an ?ant mosaic?, there have
been numerous accounts of mosaic-structured ant
communities (reviewed in H?lldobler & Wilson
1990). However, most of the information on these
highly deterministic and predictable communities
comes from locations with a somehow impoverished
fauna, e.g., African and Brazilian cocoa farms (Room
1971; Leston 1973b; Majer 1976), mangroves (Cole
1983; Adams 1994), or tropical Australia (Fox & Fox
1982; Majer 1982), which is known for its little di-
verse ant fauna (Majer 1990).
The ant fauna studied here was not mosaic-like
structured. Dominant ants neither had a set of favored
subordinates typical for a mosaic nor did they consis-
tently exclude other dominants yielding mutually ex-
clusive territories. In contrast, Berghoff, Zotz & Lin-
senmair (unpubl. observ.) found mosaic-like struc-
tured ant assemblages on Annona glabra trees bearing
the epiphytic orchid Caularthron bilamellatum
(Rich.f.) Schult. This orchid is a true myrmecophyte,
providing nesting space in its hollow pseudobulbs and
nutrition from extrafloral nectaries. According to
Jackson (1984), ant mosaics are often established
around a predictable food source. While none of our
study epiphytes supplied extrafloral nectar,
Caularthron guaranteed a year-round supply of nour-
ishing exudates. As nectar is the main diet of most
canopy dominants (Leston 1973a; Kaspari 1993), the
orchid?s nectaries may provoke vital interspecific com-
petition. In contrast, the absence of such strong deter-
ministic forces might allow a less hierarchic, more
stochastic array of ant species. H?lldobler & Wilson
(1990) noted the worldwide tendency that true domi-
nants occur only in regions where faunas as a whole
are small (boreal Europe, small islands, orchards). If
species poorness is a prerequisite for the establishment
and maintenance of ant mosaics, their occurrence is
unlikely in very species-rich ecosystems like the
Malayan rainforest canopy (Floren & Linsenmair
2000) or the one we studied.
Epiphytes and ants
Epiphytes dwelling in the crowns of the study trees
had no effect on the latter?s arboreal ant fauna. Diver-
sity measures were not influenced by the type or
amount of epiphytes in the respective crown, neither
was ant abundance (Table 2?4, Fig. 2). There is a
wealth of information about positive (although often
facultative) interactions between ants and ant-plants
in the tropics (e.g., Schimper 1888; Wheeler 1942;
Janzen 1974). The present study investigated whether
non-myrmecophilic epiphytes also contribute to ant
diversity by increasing the structural heterogeneity of
the canopy habitat and by providing shelter from cli-
matic extremes and predators. However, only 29% of
the observed ant colonies nested inside the non-myrme-
cophilic epiphytes Tillandsia and Vriesea (Fig. 3),
whilst 48% preferred dead wood as substrate. This
observation strongly disagrees with Richards? (1996)
notion that epiphytes provide the chief nesting sites for
arboreal ants in tropical rainforests. Ants readily use
the available infrastructure provided by epiphytes in
tropical canopies, but are opportunistic in the use of
alternatives: most of the species nesting within epi-
phytes were also present on trees free of epiphytes,
where they mostly used dead wood. A similarly oppor-
tunistic use of nesting options in epiphytic bromeliads
is reported by Bl?thgen et al. (2000) for a Venezuelan
lowland forest.
Conclusion
Epiphytes had no significant influence on either ?-di-
versity nor ?-diversity of ants. Several of the most
abundant ant species showed typical dominant behav-
ior in the study trees, but the lack of interspecific asso-
ciations suggests that the community as a whole was
not arranged in a well-organized ant mosaic. We con-
clude that non-myrmecophilic epiphytes in tropical
tree crowns, although readily used as nesting sites and
shelter, do not influence local or between-habitat di-
versity of ants. Instead, ants seem to be highly oppor-
tunistic with respect to their host plants.
Acknowledgements. We thank John T. Longino (Ever-
green State College, Washington, USA) and Philip S. Ward
(National Museum of Natural History, Washington, USA),
who identified the voucher specimens to species level.
Jonathan Gonzalez and Alejandro Almanza helped in
mounting and pre-sorting the specimens. This study was
funded by the German Academic Exchange Service (DAAD),
the European Science Foundation (ESF) and the Deutsche
Forschungsgemeinschaft via the Graduiertenkolleg of the
Department of Botany, Universit?t W?rzburg.
References
Adams ES (1994) Territory defense by the ant Azteca trigo-
na: maintenance of an arboreal ant mosaic. Oecologia 97:
202?208.
Adis J, Harada AY, da Fonseca CRV, Paarmann W, Rafael JA
(1998) Arthropods obtained from the Amazonian tree
species ?Cupiuba? (Goupia glabra) by repeated canopy
fogging with natural pyrethrum. Acta Amazonica 28:
273?283.
Benzing DH (1990) Vascular Epiphytes. Cambridge Univer-
sity Press, Cambridge.
Do epiphytes shape arboreal ant communities? 371
Basic Appl. Ecol. 4, 4 (2003)
Bl?thgen N, Verhaagh M, Goit?a W, Bl?thgen N (2000) Ant
nests in tank bromeliads ? an example of non-specific in-
teraction. Insectes Sociaux 47: 313?316.
Bolton B (1994) Identification Guide to the Ant Genera of
the World. Harvard University Press, Cambridge.
Br?hl CA, Mohamed M, Linsenmair KE (1999) Altitudinal
distribution of leaf litter ants along a transect in primary
forests on Mount Kinabalu, Sabah, Malaysia. Journal of
Tropical Ecology 15: 265?277.
Cole BJ (1983) Assembly of mangrove ant communities: pat-
terns of geographical distribution. Journal of Animal
Ecology 52: 339?347.
Cotgreave P, Hill MJ, Middleton DAJ (1993) The relation-
ship between body size and population size in bromeliad
tanks. Biological Journal of the Linnean Society 49:
367?380.
Croat TB (1978) Flora of Barro Colorado Island. Stanford
California Press, Stanford.
Davidson DW, Epstein WW (1989) Epiphytic associations
with ants. In: L?ttge U (ed) Vascular Plants as Epiphytes.
Springer Verlag, Berlin, pp 200?233.
Dejean A, Gibernau M (2000) A rainforest ant mosaic: the
edge effect (Hymenoptera: Formicidae). Sociobiology 35:
385?401.
Dejean A, Olmsted I, Snelling RR (1995) Tree-epiphyte-ant-
relationships in the low inundated forest of Sian Ka?an
Biosphere Reserve, Quintana Roo, Mexico. Biotropica
27: 57?70.
Erwin TL (1983) Tropical forest canopies: the last biotic
frontier. Bulletin of the Entomological Society of America
29: 14?19.
Fiala B, Linsenmair KE, Maschwitz U (1994) Diversity of
ant-plant interactions in south-east Asian rain forests. An-
drias 13: 169?178.
Floren A, Linsenmair KE (1994) Zur Diversit?t und
Wiederbesiedlungsdynamik von Arthropoden auf drei
Baumarten in einem Regenwald in Sabah, Malaysia. An-
drias 13: 23?28.
Floren A, Linsenmair KE (1997) Diversity and recoloniza-
tion dynamnics of selected arthropod groups on different
tree species in a lowland rainforest in Sabah, Malaysia
with special reference to Formicidae. In: Stork NE, Adis J,
Didham RK (eds) Canopy Arthropods. Chapman & Hall,
London, pp 344?381.
Floren A, Linsenmair KE (2000) Do ant mosaics exist in
pristine lowland rain forests? Oecologia 123: 129?137.
Fowler HG, De Souza ALB, Delabie JHC (1998) Wasmannia
spp. (Hym., Formicidae) and insect damages to cocoa in
Brazilian farms. Journal of Applied Entomology 122:
339?341.
Fox MD, Fox BJ (1982) Evidence for interspecific competi-
tion influencing ant species diversity in a regenerating
heathland. In: Buckley RC (ed) Ant-Plant Interactions in
Australia. Dr. W. Junk, The Hague.
H?lldobler B, Wilson EO (1990) The Ants. Springer Verlag,
Berlin.
Jackson DA (1984) Ant distribution in a Cameroonian
cocoa plantation: investigation of the ant mosaic hypothe-
sis. Oecologia 62: 318?324.
Janzen DH (1974) Epiphytic myrmecophytes in Sarawak:
mutualism through the feeding of plants by ants. Biotropi-
ca 6: 237?259.
Kaspari M (1993) Body size and microclimate use in
neotropical granivorous ants. Oecologia 96: 500?507.
Leigh EG Jr., Windsor DM (1982) Forest production and
regulation of primary consumers on Barro Colorado Is-
land. In: Leigh EG Jr., Rand AS, Windsor DM (eds) The
Ecology of a Tropical Forest. Smithsonian Institution
Press, Washington DC, pp 111?122.
Leston F (1973a) Ants and tropical tree crops. Proceedings
of the Royal Entomological Society of London, ser. C 38:
1.
Leston D (1973b) Ecological consequences of the tropical
ant mosaic. Proceedings of the Seventh International
Congress of the International Union for the Study of So-
cial Insects: 235?242.
Linsenmair KE (1990) Tropische Biodiversit?t: Befunde und
offene Probleme. Verhandlungen der Deutschen Zoologi-
schen Gesellschaft 83: 245?261.
Longino JT, Colwell RK (1997) Biodiversity assessment
using structured inventory: capturing the ant fauna of a
tropical rain forest. Ecological Applications 7:
1263?1277.
Longino JT, Nadkarni NM (1990) A comparison of ground
and canopy leaf litter ants (Hymenoptera: Formicidae) in
a neotropical montane forest. Psyche 97: 81?93.
Ludwig JA, Reynolds JF (1988) Statistical Ecology, a Primer
on Methods and Computing. John Wiley and Sons, New
York.
Magurran A (1988) Ecological Diversity and its Measure-
ment. Princeton University Press, Princeton.
Majer JD (1976) The maintenance of the ant mosaic in
Ghana cocoa farms. Journal of Applied Ecology 13:
123?144.
Majer JD (1982) Ant-plant interactions in the Darling
botanical district of Western Australia. In: Buckley RC
(ed) Ant-Plant Interactions in Australia. Dr. W. Junk, The
Hague, pp 45?61.
Majer JD (1990) The abundance and diversity of arboreal
ants in northern Australia. Biotropica 22: 191?190.
Majer JD (1994) Arboreal ant community patterns in
Brazilean cocoa farms. Biotropica 26: 73?83.
Majer JD, Delabie JHC (1999) Impact of tree isolation on
arboreal and ground ant communities in cleared pasture
in the Atlantic rain forest region of Bahia, Brazil. Insects
sociaux 46: 281?290.
Majer JD, Recher HF, Ganesh S (2000) Diversity patterns of
eucalypt canopy arthropods in eastern and western Aus-
tralia. Ecological Entomology 25: 295?306.
Montgomery GG (1985) Impact of vermilinguas (Cyclopes,
Tamandua: Xenarthra = Edendata) on arboreal ant popu-
lations. In: Montgomery GG (ed) The Evolution and Ecol-
ogy of Aramadillos, Sloths, and Vermilinguas. Smithsoni-
an Institution, Washington, D.C., pp 351?363.
Richards PW (1996) The Tropical Rain Forest. An Ecologi-
cal Study. Cambridge University Press, Cambridge.
Richardson BA (1999) The bromeliad microcosm and the
assessment of faunal diversity in a neotropical forest.
Biotropica 31: 321?336.
372 Stuntz et al.
Basic Appl. Ecol. 4, 4 (2003)
Room PM (1971) The relative distributions of ant species in
Ghana?s cocoa farms. Journal of Animal Ecology 40:
735?751.
Schimper AFW (1888) Die Wechselbeziehungen zwischen
Pflanzen und Ameisen im tropischen Amerika. Botanische
Mitteilungen aus den Tropen 1: 1?98.
Southwood TRE (1978) Ecological Methods. Chapman &
Hall, London.
Stork NE (1987) Guild structure of arthropods from Bornean
rain forest trees. Ecological Entomology 12: 69?80.
Stuntz S, Simon U, Zotz G (1999) Assessing potential influ-
ences of vascular epiphytes on arthropod diversity in trop-
ical tree crowns. Selbyana 20: 276?283.
Stuntz S, Simon U, Zotz G (2002a) Rainforest aircondition-
ing: the moderating influence of epiphytes on the micro-
climate in tropical tree crowns. International Journal of
Biometeorology 46: 53?59.
Stuntz S, Ziegler C, Simon U, Zotz G (2002b) Structure and
diversity of the arthropod fauna within three canopy epi-
phyte species in Central Panama. Journal of Tropical
Ecology 18: 161?176.
Stuntz S, Simon U, Zotz G (2003) Arthropod seasonality in
tree crowns with different epiphyte loads. In: Basset Y,
Novotny V, Miller SE, Kitching R (eds) Arthropods of
Tropical Forests ? Spatio-Temporal Dynamics and Re-
source Use in the Canopy. Cambridge University Press,
Cambridge (in press).
Tobin JE (1995) Ecology and diversity of tropical forest
canopy ants. In: Lowman MD, Nadkarni NM (eds) Forest
Canopies. Academic Press, San Diego, pp 129?147.
Wilson EO (1971) The Insect Societies. The Belknap Press of
Harvard University Press, Cambridge.
Yanoviak SP, Kaspari ME (2000) Community structure and
the habitat templet: ants in the tropical forest canopy and
litter. Oikos 89: 259?266.
Zotz G, Dietz H, Bermejo P (1999) The epiphyte community
of Annona glabra on Barro Colorado Island, Panama.
Journal of Biogeography 26: 761?776.
Do epiphytes shape arboreal ant communities? 373
Basic Appl. Ecol. 4, 4 (2003)