ArticlePDF Available

Abstract and Figures

In Kinabalu National Park, Borneo we observed four colonies of the Malaysian giant ant Camponotus gigas in a primary forest. These predominantly nocturnal ants have underground nests, but forage in huge three-dimensional territories in the rain forest canopies. The colony on which our study was mainly focused had 17 nests with about 7000 foragers and occupied a territory of 0.8 ha. To improve observation and manipulation possibilities, these nests were linked at ground level by 430 m of artificial bamboo trail. A group of specialist transport worker ants carried food from `source' nests at the periphery to the central `sink' nest of the queen. Transport of food between nests started immediately after the evening exodus of the foragers. Transporter ants formed a physical subcaste among the minors and behaved according to predictions of the central-place foraging theory. Their load size was about five times that of the average forager and grew proportionally with head width. Longer distances were run by ants with greater head width and larger gross weight. Transporter ants that ran more often took heavier loads. Experiments with extra-large baits revealed that C. gigas used long-range recruitment to bring foragers from different nests to “bonanzas” at far distant places. The foraging strategy of C. gigas is based on a polydomous colony structure in combination with efficient communication, ergonomic optimization, polyethism and an effective recruitment system.
Content may be subject to copyright.
Martin Pfeier áK. Eduard Linsenmair
Polydomy and the organization of foraging in a colony
of the Malaysian giant ant
Camponotus gigas
(Hym. / Form.)
Received: 16 March 1998 / Accepted: 24 August 1998
Abstract In Kinabalu National Park, Borneo we ob-
served four colonies of the Malaysian giant ant Cam-
ponotus gigas in a primary forest. These predominantly
nocturnal ants have underground nests, but forage in
huge three-dimensional territories in the rain forest
canopies. The colony on which our study was mainly
focused had 17 nests with about 7000 foragers and oc-
cupied a territory of 0.8 ha. To improve observation and
manipulation possibilities, these nests were linked at
ground level by 430 m of arti®cial bamboo trail. A
group of specialist transport worker ants carried food
from `source' nests at the periphery to the central `sink'
nest of the queen. Transport of food between nests
started immediately after the evening exodus of the
foragers. Transporter ants formed a physical subcaste
among the minors and behaved according to predictions
of the central-place foraging theory. Their load size was
about ®ve times that of the average forager and grew
proportionally with head width. Longer distances were
run by ants with greater head width and larger gross
weight. Transporter ants that ran more often took
heavier loads. Experiments with extra-large baits re-
vealed that C. gigas used long-range recruitment to
bring foragers from dierent nests to ``bonanzas'' at
far distant places. The foraging strategy of C. gigas is
based on a polydomous colony structure in combination
with ecient communication, ergonomic optimization,
polyethism and an eective recruitment system.
Key words Camponotus áPolydomy áDispersed
central-place foraging áPolymorphism áBorneo
The problem of how to optimize individual behavior
within the dierent contexts of the ``struggle for life'' led
to the introduction of economic theories into ecology.
Among the essential problems studied with these meth-
ods are many questions concerned with resource allo-
cation in organisms (e.g., Pyke 1984). The central-place
foraging theory describes the behavior of animals gath-
ering food by starting from and returning to a central
point, their nest (e.g., Orians and Pearson 1979; Scho-
ener 1979; Cheverton et al. 1985). The problem to be
solved here is to minimize foraging costs while opti-
mizing yield. Ant colonies have complex social struc-
tures enabling them to reduce their total energy costs by
cooperative foraging (e.g., Oster and Wilson 1978).
Detailed studies of the ergonomics of foragers have
shown that polymorphism in ants is an appropriate
means to optimization. Especially well studied are the
relations between load size and body size in foragers
of harvester ants (e.g., Bailey and Polis 1987) or leaf-
cutting ants (e.g., Roces and Ho
Èlldobler 1994; Wetterer
1994). Observation of honeydew-feeding ants revealed in
some species the existence of two physical subcastes:
small honeydew gatherers and greater ``tankers'' that
transport the honeydew to the nest (e.g., Fowler 1985).
Polydomous ants are thought to achieve energetic
savings by decentralization (Ho
Èlldobler and Lumsden
1980), especially by reducing the overlap in the foraging
paths of individual workers (Davidson 1997). However,
little is known of how polydomy actually in¯uences
foraging yield (McIver 1991) and nothing is known
about the mechanisms of this ``dispersed central-place
foraging.'' Splitting of the colony may cause problems
with nest interaction. Especially in monogynous species,
polydomy may result in nests with queen and brood, and
satellite nests that contain mostly foraging workers. This
asymmetric distribution would require a steady ¯ux of
energy between the dierent nest types.
Oecologia (1998) 117:579±590 ÓSpringer-Verlag 1998
M. Pfeier (&)áK. E. Linsenmair
Theodor-Boveri-Institut der Universita
Lehrstuhl fu
Èr Tiero
Èkologie und Tropenbiologie (Zoologie III),
Am Hubland,
D-97074 Wu
Èrzburg, Germany
e-mail: pfei,
Fax: +49-931-8884352
Camponotus gigas Latreille 1802 is one of the largest
ant species of the world and a common inhabitant of
South-East Asian rain forests. The mainly nocturnal
ants live in underground nests and forage in the forest
canopy (Gault 1987; Chung and Mohamed 1993).
During the day, foraging is restricted to a few ants
searching the ¯oor of the rain forest (Tho 1981; Orr and
Charles 1994). These ants have two clearly distinguish-
able worker subcastes: minors (head width 1.6±5 mm)
and majors (head width 5±8 mm) (Orr et al. 1996;
Pfeier 1996). Although monogynous, C. gigas lives in
polydomous colonies, with only some of the peripheral
nests containing brood (more details in Pfeier 1996;
Pfeier and Linsenmair 1997).
Three main factors seem to determine the foraging
ecology and behavior of this ant: (1) its extraordinary
body size and the wide range of size classes (minors
weighed between 52 mg and 362 mg, mean = 133 mg,
SD 39, n708; majors weighed between 200 mg and
480 mg, mean 372 mg, SD 50, n89; M. Pfeier
and K.E. Linsenmair, unpublished data), (2) the modest
population size of its polydomous colonies, and (3) its
huge three-dimensional territories including clumped
and randomly dispersed food resources.
The diet of C. gigas within the studied population
consisted of 87% honeydew (or other liquid food, e.g.,
nectar of extra¯oral nectaries), 7.4% bird droppings,
1.3% other droppings, and 4% arthropods from hunting
or scavenging (n6254). Occasionally, C. gigas scav-
enges at vertebrate carcasses and gathers vertebrate
excrements (Pfeier 1996). Slightly dierent diet com-
positions (especially utilization of fungi as food) have
been reported by Orr and Charles (1994), and Levy
(1996) from Brunei.
In this study of the organization of foraging in
C. gigas, we asked the following questions. (1) Is food
transported between the nests of a polydomous colony?
(2) Does this transport function according to central-
place foraging theory, regarding travel time and/or load
size? (3) Are the dierent nests of a colony autonomous
units, or is their any coordination of recruitment
behavior among them?
Our observation plot was a 5-ha area of primary mixed dipterocarp
forest in Kinabalu National Park (KNP) near Poring Hot Spring
(District of Ranau, Sabah, Malaysia). Annual rainfall in KNP
Headquarters ranges from 2000 to 3800 mm (Kitayama 1992); in
Poring we measured a total of 3218 mm of rain in 1993. We per-
formed our ®eld work during ®ve periods totaling 2 years between
July 1991 and November 1995. Most of our observations were
conducted at night, using red-®ltered head lamps (Petzl ``Mega'',
approximately 20 lux) in order not to disturb the ants.
We observed four colonies of C. gigas with 8±17 nests. At one
of these colonies we established 430 m of bamboo bridges con-
necting the nests and foraging trees at ground level (Fig. 1), so that
ants did not have to climb up and down the trees. These bridges
followed the ants' natural trunk trails through the tree canopies.
They were readily accepted and ants abandoned their longer, ar-
boreal routes. This system oered many possibilities to observe and
manipulate the ants. The 17 nests of the most observed colony were
situated within a territory of 0.8 ha and had only about 7000 for-
agers. The colony seemed to be monogynous with nest Q being
inhabited by the queen. In our study, we focused on nests Q, E, H,
F, and T; they showed the most intense interactions and the highest
In preliminary tests we weighed the ants leaving and entering
nests Q, H, and E between 1845 and 1945 hours. We carefully
observed transport of food for several days and collected the fol-
lowing data during 260 h of observation at seven nests of the col-
ony: (a) activity of ants, (b) transport of eggs and larvae, (c) worker
transports leaving and entering the nest, and (d) input and output
of prey and honeydew.
To record the distribution of size classes during the exodus
phase, we video®lmed the exodus of the foragers (from 1730 to
1900 hours) at three nests (Q, H, T) (Panasonic NV-MS95 E, red
video light). The length of 859 ants was measured from a TV screen
with digital sliding calipers. They were compared with 60 ants that
were ®lmed at the bamboo bridges running from nest Q to other
nests at 0030 hours. We determined the error in measurement to
0.1 mm by ®lming ten screw caps with a mean diameter of 2.48 cm.
We observed a trophobiotic association and weighed 676 for-
agers gathering honeydew from a large group of Flatidae when
they climbed or left the host tree that was about 4 m away from
nest E. To study transport of food between nests, 116 ants that
Fig. 1 Map of the observed colony. The nests are marked with upper-
case letters,nestQis the nest of the queen. The bamboo trail system is
drawn with black lines,broken lines indicating `natural' paths through
the canopy that were mostly abandoned after our trail system had
been installed. HT marks foraging trees with large groups of
associated trophobionts
transported liquid food between nests Q, H, and E were caught,
measured and individually marked with small numbered plastic
tags that were glued on their thoraces. These ants were weighed
several times to determine mean gross weight and net weight of
each marked ant (electronic scales Ohaus CT 10). For quick clas-
si®cation of size groups in the ®eld, we made a template: we ®xed
the cut-o heads of ants of dierent size groups on cardboard and
covered them with cellophane. This method (B. Ho
Èlldobler, per-
sonal communication) proved to be very exact when we reexamined
samples of ants with digital sliding calipers (Mahr 16 ES). We
observed ants on ®ve nights (January/February 1994) from 1800 to
0200 hours simultaneously at nests Q, H, and E and made 615
recordings of place, time, and path of the ants. We also measured
head width, ®lling of their gasters and weight of unmarked ants
transporting food on dierent sections of the trail (Fig. 1): section
1, from nest E to nest Q (57 m); section 2, from nest H to nest Q
(30 m); section 3, from nest E to nest H (27 m); section 4, from nest
Q to nest H (30 m); section 5, from nest H to nest E (27 m). Only
those ants were included in our study that could be observed
throughout their use of the trail to be sure about their starting
point and destination.
To examine long-range recruitment, a table with a petri dish
®lled with human urine as a bait was isolated from the ground by
``tangle foot'' insect glue and connected to the runway by a bamboo
bridge. Light barriers were installed at this bridge and at other
sections of the runway system between the nests. Signals thus
triggered were ampli®ed and read into a single-board computer
(BDE-module-535, PHYTEC Messtechnik, Mainz, Germany). The
countings were performed continuously, stored in periods of 5 or
15 min on a magnetic card (Panasonic BN-MC-E) and were read
out by a laptop computer. Average trac on the runway was
controlled for several days without baiting. One experimental set-
ting was repeated at least ®ve times. To avoid habituation of the
ants we alternated dierent experiments.
Data were processed with CSS Statistica or SPSS.
Activity pattern and transport of food
Foragers of the nocturnal C. gigas left the nest within a
very narrow time window of about 50 min at around
1740 hours. Most of them climbed into the canopy. First
retrieval of food to the nests by these ants normally took
place not before 1900 hours as proved by marked for-
agers. Foragers returning to the nests during the night
generally went out a second time. Shortly before sunset,
the ants returned jointly to their nest, about 90% of
them within 30 min.
Timing of input at central nest Q showed an inter-
esting feature (Fig. 2). At 1830 hours, immediately after
termination of the exodus of foragers, we observed a
strong in¯ux of workers carrying honeydew in their
widely expanded gasters, some of them also transport-
ing bird droppings and insect prey in their mandibles.
This amount of food could not have been collected in
that activity period ± too little time had elapsed for
such successful foraging. Rather, it must have been
collected the night before (and/or during daytime) and
been stored in the peripheral nests after the nocturnal
activity phase had ended. Behavioral observations
proved this assumption. The transporter ants seemed to
dier from returning foragers by their higher load and
we suspected that they might be a larger subcaste of
minor workers. At nest H, which lay inside the hollow
trunk of a tree and had a wide opening allowing easy
inspection of a part of the nest, we often observed
several returning foragers transferring their load by
trophallaxis to one transporter ant that waited behind
the entrance.
In a weight analysis, comparing ants leaving the
nests with those entering it, from 1845 to 1945 hours,
ants leaving nest E and H were heavier than ants
entering it (Fig. 3). Our investigations showed that the
weight dierence of these workers came about because
ants leaving nests E and H at this time were mostly
transport workers carrying food to nest Q, whereas
ants leaving nest Q carried nothing. Ants entering
nests E and H, however, were only foragers returning
from their search for food in the canopy or trans-
porter ants returning from Q with empty gaster. These
results substantiated liquid food transport between the
Most of the observed food transports took place
from nest E to nests H and Q, and from nest H to nest Q
(see Fig. 1). As nest Q was the aim of transport workers
coming from dierent nests, mean total activity (MTA,
number of ants leaving and entering the nest within
15 min) during the night (1845±0500 hours) at nest Q
was much higher (109.7, SD 38.11, n138), than at
nests H, E, and F (28.4, SD 24.6, n284). The
ration of MTA during exodus (1740±1845 hours) at nest
Q (164.4, SD 152.9, n127) and at the other nests
(113.4, SD 133.8, n292) versus MTA during the
rest of the night (1845 to 0500 hours) was 1.5:1 at nest Q
and 4:1 at the other nests.
Fig. 2 Nocturnal activity of Camponotus gigas recorded at the
entrance of the main nest Q on 11 September 1992. Foragers left
the nest at sunset, and searched for food until sunrise. Input of food
started immediately after the end of the exodus phase at about 1830
hours and fell shortly before the start of the run-in. We counted only
those ants carrying food between the mandibles or in their enlarged
gasters, thus diering clearly from their unloaded nestmates
Distribution of size classes during the exodus
The mean length of all minor workers entering and leaving
the nests during our video scannings (n1306) was
17.5 mm (natural body position from above). Mean body
length of minor workers leaving nests Q, T, and
H during the exodus time was 17.27 mm (n859,
SD 1.56 mm), signi®cantly smaller than the mean
body length of transport workers that were ®lmed at the
bamboo bridges running from nest Q to other nests
from 0030±0100 hours (n60, mean 17.85, SD
1.64 mm; U-test, U28,931, Z)7.22, P<0.001).
We then focused on single nests and divided the data
of the worker ants leaving the nest during exodus time
with respect to the evening start of activity: group 1
included all ants leaving the nest until 45 min after the
start of activity at about 1745 hours, group 2 comprised
all ants leaving the nest after about 1830 hours (exactly
45 min after the start) to 1900 hours. So, group 1
comprised foragers leaving the nest during the peak time
of exodus, while group 2 contained all ants leaving the
nest after this period, which should be mostly trans-
porter ants (both groups were separated by a short
characteristic drop in activity, which could be found in
all our records (for Q, n28); see Fig. 2).
At nest Q, both groups could be discriminated with
high signi®cance (for statistics see below): group 1
comprised workers with a mean size of 17.4 mm
(n290, SD 1.63), group 2 had a mean size of
17.9 mm (n96, SD 1.77). The latter group left
nest Q after the input of food from other nests had
begun (see Fig. 2). These ants were larger than the
mean foragers and seemed to walk mostly on the trails
connecting the dierent nests of the colony. Trans-
porter ants leaving nest Q after 0030 hours (group 3)
were signi®cantly larger than group 1 ants leaving nest
Q during the exodus (17.9 vs 17.4 mm), but did not
dier from ants of group 2 [Kruskal-Wallis ANOVA
H(2, n446) 12.01, P<0.01; multiple compari-
sons with the Nemenyi test (Zar 1996): group 1 vs
group 2, P<0.01; group 1 vs group 3, P<0.01].
These results were con®rmed at nest H where ants of
group 2 were also signi®cantly larger (18.0 mm, n41,
SD 2.0) than those of group 1 (17.0 mm, n167,
SD 1.2). Transport workers leaving nest Q after 0030
hours (mean 17.9 mm) diered signi®cantly in size from
group 1 foragers, but not from group 2 [Kruskal-Wallis
ANOVA H(2, n268) 23.671, P<0.001; multiple
comparisons with the Nemenyi test: group 1 vs group 2,
P<0.001; group 1 vs group 3, P<0.001; group 2 versus
group 3, n.s.].
At nest T, all workers were rather small in size and
group 1 versus group 2 diered by only about 0.2 mm
(16.9 vs 17.1 mm, n.s.). Compared to these ants, trans-
port workers were much larger (group 3: 17.9 mm) and
diered from all ants leaving nest T during exodus
time [Kruskal-Wallis ANOVA H(2, n325) 20.15,
P<0.001; multiple comparisons with the Nemenyi test:
group 1 vs group 2, n.s.; group 1 vs group 3, P<0.01;
group 2 vs group 3, P<0.001].
Taken together these results corroborate our obser-
vations of two physical subcastes of minors in C. gigas:
the ``foragers'' and the ``transporter ants.''
Observation of foragers in a trophobiotic association
Foragers, collecting honeydew from a group of associ-
ated Bythopsyrna circulata Guerin-Mee
Âneville (Flatidae,
Homoptera) on a young Eugenia tree (Myrtaceae) 4 m
away from nest E, were weighed when entering or
leaving the tree. Their weight when climbing up the tree
was 118 mg (n310, SD 32.3). Foragers that left
the tree to carry honeydew to their nests could be sorted
into two groups: those that left the tree until 0500 hours
were heavier (143 mg, SD 44, n185) than those
that left the tree jointly in the morning, when foraging
activity stopped (126 mg, SD 29, n181; U-test,
U12,927, Z)3.77, p<0.001). Most probably,
the cessation of activity in the morning prevented the
second group from becoming completely ®lled up. Ant
loads were 25 and 8 mg, respectively.
Transport system
Weight of workers at dierent nests
In preliminary experiments, we weighed unmarked
workers after 1845 hours at nests E, H, and Q when
Fig. 3 Weight of ants leaving and entering dierent nests. These
measurements were taken after the exodus phase, so weight dierences
depend on food transport between the nests [Kruskal-Wallis ANOVA
by ranks: H(5, n769) 261.6, P<0.001; multiple comparisons
with the Nemenyi test (Zar 1996), an.s., P<0.001 for all other test
leaving or entering the nest (see Fig. 3). The dierence
between these measurements represents the mean load of
the transport workers leaving the peripheral nests and
entering central nest Q (E out, H out, Q in): 47.1 mg at
nest E, 47.5 mg at nest H, and 108 mg at nest Q. Nest Q
was at the end of this transport chain. Ants leaving Q
shortly after 1845 hours should be mostly ``empty''
transporter ants returning to their nests, because the
exodus of the foragers had already ceased and input of
food had started (see Fig. 2). Ants leaving nests E and H
at that time were loaded transporters. However, nest H
in particular was not only a ®lling station but also a
``place of transshipment,'' with many loaded transporter
ants entering it. So absolute load at nest H should have
been higher, because loaded transport workers were
found in both samples of this nest. We also may have
included some foragers in our random samples, espe-
cially at nests E and H that were situated near to the
foraging tree; this may to some extent explain the large
dierence between ``H out'' and ``Q in.''
When we measured only transporter ants, observed
moving between nests, mean gross weight was higher at
nest Q and H than at the more peripheral nests E and T.
Net weight was the same at all nests; mean load became
heavier from the periphery to the center (see Table 1).
Transporter load, transporter size, and distance
of transport
A total of 103 weight measurements were made of
unmarked carriers that ran dierent sections of the
bamboo trail system. As shown in Fig. 4, transporters
that had the longest path (from E to Q) were the
heaviest, heavier than those running only a section of
the trail. The dierence between ants that came from H
to Q and those going from E to H depended not only
on the path length, but seemed to be a result of load
transfer at nest H. Empty carriers returning to their
nests had equal weights.
The mean head width of 16 transporter ants that went
the whole way from nest E to nest Q was 4.2 mm
(SD 0.49), signi®cantly larger than that of the re-
maining 87 ants that only went a section of this course
(3.86 mm, SD 0.51; U-test, U195, Z)4.56,
P< 0.001).
In a 5-day study of 116 marked transporter ants, 21
were found at all three observed nests (Table 2, G1,
transport distance 55 m), 20 shuttled only between two
(G2, transport distance 30 or 25 m), and 75 of the
marked ants were observed at only one nest (G3). Group
1, the long-distance carriers, diered from group 3 in
their mean gross weight and also from all other marked
workers (G2+G3), however, presumably because of
their low number, not from group 2 alone. If we pooled
group 1 and 2, and tested them against 3, the gross
weights were signi®cantly dierent.
Transporter size and frequency of transports
During 38 h of observation on ®ve nights, we recorded
160 transports of liquid food that were carried out by 41
marked transport workers. Transport frequency corre-
lated with head width [Pearson correlation:
r(X,Y)0.51, r
0.26, t3.24, Bonferroni-cor-
rected P< 0.01 (see Sokal and Rohlf 1995)] and with
gross weight [Pearson correlation: r(X,Y)0.53,
0.28, t3.39, Bonferroni-corrected P< 0.01],
however not with net weight [Pearson correlation:
r(X,Y))0.04, r
0.001, t)0.17, P0.866,
Regression analysis of ant weight
Mean gross weight and net weight of marked transporter
ants depended on head width, and their carrying capacity
Table 1 Weight and load of transporter ants at dierent nests of
colony A. The ants diered signi®cantly in gross weight between Q
and E (P£0.01), Q and T (P£0.001), and H and T (P< 0.05)
[ANOVA F(3,215) 8.05, P< 0.001; Hochberg GT2 post hoc
test]. Net weight was equal at all nests [ANOVA F(2,143) 2.70,
Nest Gross weight Net weight Load
mean SD nmean SD n(mg)
(mg) (mg)
Q 274 64 94 140 34 78 134
H 255 69 78 140 26 40 115
E 224 64 20 127 29 28 97
T 212 60 27 ± ± 0 ±
Fig. 4 Weight of honeydew carriers to (and from) the central nest Q,
relative to the length of their path between the nests (marked on the
column). Those with the longest path are the heaviest, as predicted by
central-place foraging theory [nfor the dierent groups: E to Q, 16, H
to Q, 35, E to H, 14, Q to H, 18, H to E, 20; Kruskal-Wallis ANOVA
by ranks H(4, n103) 64.73, P< 0.001; multiple-comparison
testing with the Nemenyi test (Zar 1996) aP<0.05, bn.s., other
groups, P< 0.01]
grew proportionally with their size (see Fig. 5). As data of
both regressions showed replications (multiple values of
Yon each value of X), we ®rst tested them for linearity in a
one-way analysis of variance (Zar 1996) and found that
the null hypothesis of linear regression could not be
rejected: gross weight F(1)13,14 0.44, p> 0.25; net
weight F(1)10,11 1.88, p> 0.10. We calculated a
least-square ®t for gross weight of the ant:
0.45, n39, F(1,37) 30.5,
P< 0.001] and for net weight: Y)0.005 + 0.374X;
0.60, n27, F(1,25) 37.8, P< 0.001]. The
slopes of the regression lines did not dier signi®cantly in
a two-tailed t-test [Zar 1996; t(2) 62 1.37, P> 0.2,
n.s.], but elevations of both regressions diered signi®-
cantly (Zar 1996; t-test, t(2) 63 6.64, P< 0.001). This
showed that the carrying capacity of ants grew propor-
tionally with their size. We calculated a common regres-
sion coecient bc 0.509 and a common r
Mean weight of transport workers
compared to foragers
In pooling weight data of marked and unmarked
transporter ants, we obtained the following results:
mean gross weight, 255 mg (SD 68, n220); mean
net weight, 139 mg (SD 31, n143); thus mean
load: 116 mg. In 28 cases, we measured the weight of
marked ants directly before entering and after leaving
nest Q: the dierence was 137 mg (SD 49.8). The
mean head width of 222 transporter ants was 3.78 mm
(SD 0.05), signi®cantly larger (U-test, U0.0,
Z)23.29, P< 0.01) than that of the average minor
worker (3.56 mm, n853, SD 0.53; M. Pfeier and
K.E. Linsenmair, unpublished results). Foragers col-
lecting honeydew in the trophobiotic association diered
in gross and net weight from transporter ants (Fig. 6).
Dierences between nests: energetic `sinks'
and `sources' in a polydomous colony
The records of long-term observations revealed signi®-
cant dierences between `sink' nests, showing a large
input of honeydew, insects, and bird droppings, and
`source' nests exporting food (Table 3). The largest
`sink' within the colony was the nest with the queen
showing an input/output relationship of 41.6:1. How-
ever, only ants with largely extended gasters or with
food between their mandibles were classi®ed as trans-
porter ants. Foragers that arrived with smaller gasters
were not categorized as transporters, but contributed
only to `total activity'. Small loads did not pu up the
gasters, therefore we were unable to estimate the for-
agers' contribution to food input.
The queen's nest Q showed high transport rates of
workers and larvae that were carried to other nests or
brought back from there. Eggs, however, were only
carried from nest Q to other nests but never vice versa.
Within colony bidirectional transport rates of workers
and larvae were correlated with a high input/output
ratio (Spearman correlation R0.76 for worker
transport, R0.77 for larval transport, both P<
0.001). At typical `source' nests that had input/output
relations (by transporter ants) smaller than 1, we never
found brood.
Long-range recruitment
Excrement or cadavers of larger vertebrate are huge
resources that cannot be eectively exploited by single
workers. They require the cooperation of hundreds of
Fig. 5 Linear regressions of head width on ant mass, the two data sets
showing ants with full and empty gasters. Data were collected from
marked transporter ants that were observed on average 13.2 times
within 5 days. The curves in the ®gure are based on the common
regression coecient. Least-square ®t for gross weight (®lled circles):
Y0.081 + 0.509X; for net weight (open squares): Y)0.056 +
Table 2. Transport workers with dierent working sections diered
signi®cantly in their gross weight. G1 diered from G3 (P<0.05),
but not from G2 (P0.74, n.s.) (ANOVA, df 2,96, F3.69
P< 0.05, Hochberg GT2 post hoc test). G1+G2 diered from G3
(t-test, t2.56, df 97, P< 0.013) and G1 diered from
G2+G3 (t-test, t2.39, df 80, P< 0.05)
SD Gross
SD Net
SD Load
(mm) (mg) (mg) (mg)
G1 at three nests 21 3.9 0.61 284 48 148 27 136
G2 at two nests 20 3.7 0.37 264 37 135 21 129
G3 at one nest 75 3.8 0.56 240 67 130 16 110
ants. Because of its large range and its polydomous
colony structure, C. gigas seemed to be especially suited
to locate and exploit such ``bonanzas.'' However, be-
cause of polydomy, worker force is distributed among
many nests. Large resources may thus overcharge the
foraging capacity of single nests. We wanted to know
whether C. gigas had developed mechanisms to over-
come this problem by recruiting workers from more
than a single nest. To check this we worked with a
computer-recorded system of light barriers to supervise
the trunk trails at dierent sections between the nests.
Figure 7a shows the experimental setting, and a
corresponding 2-day record without experimental
recruitment is given in Fig. 8a. The correlation coe-
cients of data from bidirectional activity countings at
dierent light barriers are listed in Table 4. Correlation
without recruitment was poor and only weakly signi®-
When we oered a bait at X, workers recruited to it.
Activity at the adjacent nest (E) rose, then recruitment
spread to the next nests (®rst H, then Q). Within a pe-
riod of 15 min, 270 min after the presentation of the
bait, we counted 348 ants crossing the bridge to/from the
food table. At the same time, activity rose to 275 at nest
H and to 118 at the queen's nest. The propagation of
this activity peak is shown in Fig. 8b. Activity at
Table 3 Activity and transport of food, workers and larvae at dierent nests of the most intensively studied colony. [Only ants with bursting full gasters were classi®ed as honeydew
carriers. Output could be larger than input because (1) several entering foragers with smaller gasters (that were counted here only as a part of the total activity) gave their load to carrier
ants that were counted as `output' when leaving the nest, and (2) counting covered many periods during transport peak time, when stored food was transported.] Shown are mean values
in observation units of 15 min. All nests diered signi®cantly in mean transport rates of larvae, workers, honeydew, and nitrogenous food (Kruskal-Wallis ANOVA, H
: same transport
rates at all nests). The queen's nest Qwas the center of transport. Carriers of other nests (e.g., E,H) brought food to Q, which had only moderate total activity, but the highest input/
output relationship. In a Nemenyi test we looked for dierences between Q and all other nests. Q diered signi®cantly (P<0.05) from most other nests, non-signi®cant dierences are
marked °
Nest nin
15 min
Transport of
Input of
Output of
Input of
insects or
Output of
insects or
mean SD mean SD mean SD mean SD mean SD mean SD mean SD
Q 359 112.4 111.8 1.2 1.7 1.0 1.6 17.6 17.4 0.4 1.6 1.51 2.48 0.06 0.44 41.58
E 298 84.8 120.6 0000 6.613.7 2.1 3.8 0.23 0.63 0.20 0.68 2.97
H 230 65.7 95.3 0000 0.68 0.23 0.62 0.83
F 149 43.1 53.7 0.18 0.55 0.12 0.38 3.7 5.3 1.2 2.2 0.23 0.60 0.49 1.28 2.32
Z 35 35.8 71.3 0000 0.23 0.57°0.23 0.37
T 30 195.7 337.7 0.6°1.3 0.8 1.2 2.4 5.7 0.1°0.4 0.80 2.47 0.00°0 24.62
L 23 40.0 80.1 0000 0.63 1.6 2.2 4.7 0.21 0.92 0.67°0.25 0.30
±H(6, n1114)
H(6, n1114)
H(6, n713)
H(6, n826)
H(6, n952)
H(6, n927)
by ranks
P<0.001 P<0.001 P<0.001 P<0.001 P<0.001 P<0.001
Fig. 6 Division of labor in C. gigas. Given is the weight of the
ants in milligrams. Foragers, which gathered honeydew at the
trophobionts and brought it into nest E, about 4 m away, were the
smallest minors. Transport workers were the largest of the minor
caste ± they carried heavy loads of liquid food in their gasters.
Their mean load was about ®ve times that of the foragers. (U-tests,
signi®cance levels are corrected with the Bonferroni method
because of multiple testing: net transporters vs gross transporters
U2135, Z)13.92, P< 0.01; net foragers vs gross foragers
U18,250, Z)6.77, P< 0.01; net foragers vs net transport-
ers U11,097, Z)8.5, P< 0.01; gross foragers vs gross
transporters U3713, Z)14.2, P< 0.01; net foragers vs
gross transporters U2238, Z)18.3, P< 0.01, gross
foragers vs net transporters U13,169, Z)0.07, P0.95)
dierent sections of the trail was highly correlated
(Table 4). The long-distance recruitment was released by
single workers that laid trails between the nests. Trails
were laid by ants with full gasters returning to their
nests, and also by unloaded ants on their way to the bait.
These were sometimes followed by small groups of
workers, the typical picture of group recruitment. As we
counted the activity in both directions, we observed not
only recruitment of workers to the bait, but also trans-
port of food back to their nests. Both stopped in most
cases ®nally at central nest Q. A comparison of the
activity during our experiment (Fig. 8b) and during
normal nights (Fig. 8a) shows the very pronounced re-
sponse of the ants to our bait. This experiment proved
that several nests cooperated in foraging and a
``bonanza'' could trigger recruitment activity that spread
from the nearest to several further distant nests.
In the next assay we examined the spread and direc-
tion of recruitment further: the bait was moved to a point
between nests Q and H (Fig. 7b, Table 5). Although the
bait was very near to nest H, recruited workers came
mostly from nest Q and recruitment stopped there. Light
barriers no. 1 and no. 2, which would have indicated its
spread to other nests, showed no signi®cantly correlated
increases of activity. This shows that Q determined the
demand of food within the colony and was able to sat-
isfy this ± within near range ± by recruitment of its own
Central-place foraging theory has been applied to a wide
range of organisms that feed ospring at a central place,
or store food there for times of reduced prey availability,
e.g., birds (Leopold et al. 1996), bees (van Nieuwstadt and
Iraheta 1996), or ants (Roces 1990). Our study focused on
two of the predictions of the central-place foraging theory:
(1) the optimal central place should minimize traveling
time during resource allocation and (2) the optimal load
should increase with increasing distance of transport
(Orians and Pearson 1979; Schoener 1979).
In C. gigas we have a special case of ``dispersed
central place foraging'' (McIver 1991), because observed
colonies were polydomous. Foragers brought the food
to nests that were scattered inside the territory. A spe-
cialized group of transporter ants carried it from
`source' nests, which lay near the foraging trees, to the
`sink' nests in the center of the colony, especially to the
nest of the queen that seemed to contain most brood.
Polydomy may be ± among other reasons ± a reaction
to patch quality variation within the territory, concen-
trating foragers in patches with high food quality, or to
low food density, spreading the worker force over a
large area to more eectively gather randomly distrib-
uted resources. In C. gigas, both conditions were met, as
this ant foraged for clumped food (honeydew from as-
sociated Homoptera) as well as for randomly dispersed
food (bird droppings, insects, cadavers). In either case,
polydomy results in a reduced total travel time of all
foragers when food is transported to the central nest by
specialist transport workers that carry the loads of sev-
eral foragers. This also reduces the number of ``empty''
ants running back without load, and may optimize the
transport capacity. C. gigas transport workers became
heavier from the periphery to the central nest; this
may be partly due to better use of transport capacity
during nest-to-nest transport, when ants with smaller
loads gave their honeydew to other transport workers.
Another reason was the arrival of heavily loaded
transporter ants at nest Q that came from far distant
Fig. 7a, b The nests were con-
nected by the bamboo trail sys-
tem (shown by the continuous
line;thebroken line indicates the
natural path through the canopy).
The numbers mark the light
barriers that were connected to
the single-board computer during
the ®rst assay. aFirst part of the
recruitment experiments. We
installed a table with a bait (X)
near nest E. Measuring points 4
and 5were used alternately.
bSecond part of the recruitment
experiments. The table with the
bait was shifted to a place
between nests Qand H
The large transporter ants carried the load of ®ve
foragers, while their biomass was only 1.16 times that of
a honeydew gatherer. In addition, the energetic cost/
bene®t relationship for many ant species improves with
rising load. This was demonstrated for C. herculeanus by
Nielsen et al (1982), by Duncan and Lighton (1994) for
two species of Myrmecocystus, and by Bartholomew
et al. (1988) for Eciton. Compared to transport by for-
agers, food transport by carrier ants will reduce foraging
Ant species that are opportunistic foragers should
maximize their number of foragers (while minimizing the
costs of a single forager); in polymorphic species,
therefore, smaller workers should search for food and
alarm their larger nestmates to transport it (Carroll and
Janzen 1973). This was shown, e.g., in the leaf-cutting
Atta cephalotes, where Hubbell et al. (1980) found a
two-stage harvesting method, with smaller ants cutting
the leaves and larger ants transporting them. In the
polydomous Australian Iridomyrmex ants, foragers
show a division of labor between smaller food gatherers
and larger transporter ants (McIver 1991). We found a
similar polyethism in C. gigas. Foragers were signi®-
cantly smaller than transporter ants that run the trails
after 1845 hours. Video®lming corroborated the hy-
potheses of two physical subcastes among the minor
workers of C. gigas.
The transport system of C. gigas also ful®lled the
second prediction of the central-place foraging theory:
the gross weight of the carriers rose with larger transport
distance and increasing activity of the carriers. Longer
distances were run by larger ants. Goss et al. (1989)
Fig. 8a Two-day record of ac-
tivity at the trail system without
experimental recruitment. All
ants that ran on the bamboo trail
in both directions were counted
when passing the light barriers.
The trail sections and numbers of
the light barriers in the legend are
mapped in Fig. 7a. Because the
recorder reacted only to activity,
the period between the two noc-
turnal activity periods is short-
ened. bRecruitment to a
``bonanza.'' The arrangement of
a. The activity at the bait in-
creased in one onslaught. The
highest activity was measured at
the trail to nest E (QE), the
curves below show activities at
sections EH and HQ.Atsection
QL, activity was not correlated
with that at QE (see Table 4). On
section EH, activity was high
even after the end of recruitment,
indicating transport of food from
start (at 30 min) is a part of
transport activity after the exo-
dus phase and has nothing to do
with the experiment. Note the
dierent scale of the y-axis cf. a
showed in a modeling study that larger workers gain
most pro®t in large territories.
As our experiments revealed, the colony was organ-
ized as a unit, although polydomy persisted over years
(Pfeier 1996). The single nests were well coordinated in
their actions. Recruitment stimuli given at one nest were
easily passed on to the central nest. Long-distance re-
cruitment makes polydomy possible for species with
only moderate colony size, like C. gigas. It helps them
take advantage of the positive sides of polydomy, e.g., a
better distribution of the foragers, which therefore have
shorter paths to foraging areas and know well the exact
position of the permanent resources inside the large
territory. It also avoids most of the disadvantages of
polydomy, e.g., poor ¯ow of information, too small
worker groups, and unfavorable defense opportunities.
Within a few minutes, an alarm shifted from nest to nest.
As our observations of territory defense revealed
(Pfeier 1996), this also happened when conspeci®c en-
emies attacked colonies at the borders. Long-distance
recruitment has also been found in colonies of other
ants, e.g., the African Oecophylla longinoda (Ho
Table 4 Long-distance recruitment in C. gigas. Correlation coe-
cients of bidirectional activity countings at dierent trail sections.
Column headings indicate the numbers of the light barriers
(Fig. 7a) and the corresponding sections of the trail system. The
®rst two records are without baiting, the rest are with a bait at X.
Three records were made with a counter between nests Q and L.
However, after we observed recruitment of workers much more
often at nest F, we changed this counter to section EF. On 1 March
1994, two counters broke down because of heavy rain. The data for
22 March 1994 did not show a normal probability distribution, and
we therefore calculated the nonparametric Spearman correlation
coecient (SP). Activities of single sections were better correlated
during long-distance recruitments, and most sections also showed
correlations with activity between nest E and the bait (XE). Al-
lowing for multiple testing, we used the Bonferroni method to
correct signi®cance levels by dividing them by the number of tests
(k). Signi®cances remained high, even after correction (*P<0.05/k;
**P<0.01/k; ***P<0.001/k)
Date ns
11 March 38 n.s. n.s. ± n.s. ±±±±
12 March 39 )0.44 n.s. ± n.s. ±±±±
1994 *
4 March 48 n.s. 0. 851 ± n.s. 0.902 0.932 ± n.s.
1994 *** *** ***
7 March 24 0.740 0.855 ± 0.690 0.794 0.814 ± 0.860
1994 ** *** ** *** *** **
01 March 99 ± ± ± ± 0.663 ± ± ±
1994 ***
17 April 15 n.s. n.s. ± n.s. 0.947 n.s. ± n.s.
1994 ***
21 March 27 ± 0.858 0.839 ± 0.858 0.677 0.944 ±
1994 *** *** *** *** ***
22 March 26 ± 0.765 0.507 ± 0.643 0.637 0.825 ±
1994 SP*** SP** SP** SP*** SP**
26 March 34 ± 0.741 0.884 ± 0.759 0.498 0.756 ±
1994 *** *** *** ** ***
Table 5 Long-distance recruitment in C. gigas. The bait was sited
between nests Q and H (Fig. 7b), and shown are the correlation
coecients of dierent trail sections. Column headings as in Ta-
ble 4. The ®rst two records were made without baiting, the rest with
a bait at X. In baited assays, ants recruited in the direction of nest
Q. Ant activity at X correlated strongly with counter no. 3. Counter
no. 5 that should register transport of food to the nearer nest H was
correlated negatively or not signi®cantly, as were counters no. 1
and 2 (Bonferroni-corrected signi®cances: *P<0.05/k;**P<0.01/
k; ***P<0.001/k)
Date ns
22 April 39 0.736 n.s. n.s. n.s. n.s. n.s.
1994 ***
29 June 39 0.754 n.s. n.s. n.s. n.s. n.s.
1994 ***
18 April 47 0.4 0.429 n.s. n.s. n.s. n.s.
1994 * **
20 April 53 )0.328 0.84 )0.396 n.s. n.s. )0.357
1994 * *** ** *
21 April 36 0.464 0.929 n.s. n.s. n.s. n.s.
1994 * ***
1 July 11 n.s. 0.908 n.s. n.s. n.s. n.s.
1994 **
3 July 17 )0.706 0.602 )0.808 n.s. n.s. n.s.
1994 * ** **
and Wilson 1978); however, in C. gigas we demonstrated
interaction of nests by quantitative countings for the
®rst time in the ®eld.
Fewell et al. (1992) examined the in¯uence of dis-
tance on recruitment behavior of the South American
Paraponera clavata that lives in colonies of 2500 work-
ers. Foragers gathering food far from the nests often
give the nectar to other workers that transport it to the
nest. With growing distance, they recruit reserve workers
from the canopy that spend 98% of their time outside
the nest. This mechanism reduces recruitment time to a
quarter. It is possible that these extra-nidal worker
groups are a precursor of a polydomous nesting system
like that of C. gigas.
Most of our experiments were carried out only at one
colony; however, behavior of three other colonies of
C. gigas within our observation area seemed to be fully
consistent with the studied colony. In many observations
and minor experiments, we never found any hint of
mechanisms other than those described here. However,
due to the large experimental setting (runway system,
computer countings, large numbers of marked foragers)
and long-term observation, including up to 5 years of
data (e.g., of nest input), we were not able to observe
these colonies as thoroughly as the one on which we
concentrated our observations. For some of the prob-
lems facing behavioral studies on social insects, an in-
dividual-based assay seems to be the only possible way
of collecting data.
Acknowledgements We are grateful to Sabah Parks for the assistance
and help that allowed this study to be conducted. We speci®cally
thank Datuk Dr. Lamri Ali, Mr. Rajibi Hj. Aman, Mr. Eric Wong,
Dr. Jamili Nais, and Mr. Kasitah Karim for various kinds of support.
Special thanks to Antonia Uecker, Sabine Hussmann, Matthias
Dolek, Sani Sambuling, and Sabin Gompoyo for help in the ®eld
work, to Dipl. Ing. Gerhard Vonend for the construction of the ``ant
computer,'' and to Stephan Messner for writing the software. We
also thank Prof. Dr. Herbert Vogt for statistical advice and our re-
viewers Prof. Dr. Konrad Fiedler, Prof. Dr. Alfred Buschinger and
an anonymous reviewer for helpful comments on the manuscript.
This study was funded by the Deutscher Akademischer Aus-
tauschdienst. We also thank the Deutsche Forschungsgemeinschaft
(DFG) for granting the infrastructure for our activities within the
framework of the DFG Schwerpunktprogramm: ``Mechanismen der
Aufrechterhaltung tropischer Diversita
Bailey KH, Polis GA (1987) Optimal and central-place foraging
theory applied to a desert harvester ant, Pogonomyrmex cali-
fornicus. Oecologia 72:440±448
Bartholomew GA, Lighton JRB, Feener DH Jr (1988) Energetics
of trail-running, load carriage, and emigration in the column-
raiding army ant Eciton hamatum. Physiol Zool 61:57±68
Carroll CR, Janzen DH (1973) Ecology of foraging by ants. Annu
Rev Ecol Syst 4:231±257
Cheverton J, Kacelnik A, Krebs JR (1985) Optimal foraging: con-
straints and currencies. In: Ho
Èlldobler B, Lindauer M (eds)
Experimental behavioral ecology. Fischer, Stuttgart, pp 109±126
Chung AYC, Mohamed M (1993) The organisation and some
ecological aspects of the giant forest ant, Camponotus gigas.
Sabah Soc J 10:41±55
Davidson DW (1997) The role of resource imbalances in the evo-
lutionary ecology of tropical arboreal ants. Biol J Linn Soc
Duncan FD, Lighton JRB (1994) The burden within: the energy
cost of load carriage in the honeypot ant, Myrmecocystus.
Physiol Zool 67:190±203
Fewell JH, Harrison JF, Stiller TM, Breed MD (1992) Distance
eects on resource pro®tability and recruitment in the giant
tropical ant, Paraponera clavata. Oecologia 92:542±547
Fowler HG (1985) Alloethism in the carpenter ant, Camponotus
pennsylvanicus (Hymenoptera Formicidae). Entomol Gen
Gault D (1987) Feeding and foraging behaviour of the giant forest
ant, Camponotus gigas, in a Malaysian rain forest. Trop Biol
Newsl 52:2
Goss S, Deneubourg JL, Pasteels JM, Josens G (1989) A model of
noncooperative foraging in social insects. Am Nat 134:273±
Èlldobler B, Lumsden CJ (1980) Territorial strategies in ants
Oecophylla longinoda,Pogonomyrmex,Myrmecocystus mimicus.
Science 210:732±739
Èlldobler B, Wilson EO (1978) The multiple recruitment
systems of the African weaver ant Oecophylla longinoda
(Latreille) (Hymenoptera: Formicidae). Behav Ecol Sociobiol
Hubbell SP, Johnson LK, Stanislav E, Wilson B, Fowler HG
(1980) Foraging by bucked-brigade in leaf-cutter ants.
Biotropica 12:210±213
Kitayama K (1992) An altitudinal transect study of the vegetation
on Mount Kinabalu, Borneo. Vegetatio 102:149±171
Leopold MF, Vanelk JF, Vanheezik YM (1996) Central place
foraging in oystercatchers Haematopus ostralegus ± can parents
that transport mussels Mytilus edulis to their young pro®t from
size selection. Ardea 84A:311±325
Levy R (1996) Interspeci®c colony dispersion and niche relations of
three large tropical rain forest ant species. In: Edwards DS,
Booth WE, Choy SC (eds) Tropical rainforest research.
Kluwer, Dordrecht, pp 331±340
McIver JD (1991) Dispersed central place foraging in Australian
meat ants. Insect Soc 38:129±137
Nielsen MG, Jensen TF, Holm Jensen I (1982) Eect of load car-
riage on the respiratory metabolism of running worker ants of
Camponotus herculeanus (Formicidae). Oikos 39:137±142
Nieuwstadt MGL van, Iraheta CER (1996) Relation between size
and foraging range in stingless bees (Apidae, Meliponinae).
Apidologie 27:219±228
Orians GH, Pearson NE (1979) On the theory of central place
foraging. In: Horn DJ (ed) Analyis of ecological systems. Ohio
State University Press, Columbus, pp 155±177
Orr AG, Charles JK (1994) Foraging in the giant forest ant,
Camponotus gigas (Smith) (Hymenoptera: Formicidae):
evidence for temporal and spatial specialisation in foraging
activity. J Nat Hist 28:861±872
Orr AG, Charles JK, Yahya Hj HR, Sharebini Hj N (1996) Nesting
and colony structure in the giant forest ant Camponotus gigas
(Latreille) (Hymenoptera: Formicinae). Raes Bull Zool
Oster GF, Wilson EO (1978) Caste and ecology in the social
insects. Monogr Popul Biol 12. Princeton University Press,
Princeton, NJ
Pfeier M (1996) Sozialstruktur und Verhaltenso
Èkologie von
Riesenameisen Camponotus gigas Latreille 1802 im Regenwald
Malaysias auf Borneo. Thesis, Bayerische Julius-Maximilians-
Pfeier M, Linsenmair KE (1997) Reproductive synchronization in
the tropics: the circa-semiannual rhythm in the nuptial ¯ight of
the giant ant Camponotus gigas Latreille (Hym./Form.).
Ecotropica 3:21±32
Pyke GH (1984) Optimal foraging theory: a critical review. Annu
Rev Ecol Syst 15:523±575
Roces F (1990) Leaf-cutting ants cut fragment sizes in relation
to the distance from nest. Anim Behav 40:1181±1183
Roces F, Ho
Èlldobler B (1994) Leaf density and a trade-o between
load-size selection and recruitment behavior in the ant Atta
cephalotes. Oecologia 97:1±8
Schoener TW (1979) Generality of the size-distance relation in
models of optimal feeding. Am Nat 114:902±914
Sokal RR, Rohlf FJ (1995) Biometry ± the principles and practice
of statistics in biological research, 3rd edn. Freeman, New York
Tho YP (1981) The giant forest ant Camponotus gigas, Malaysia.
Nat Malaysiana 6:32±35
Wetterer JK (1994) Forager polymorphism, size-matching, and
load delivery in the leaf-cutting ant, Atta cephalotes. Ecol En-
tomol 19:57±64
Zar JH (1996) Biostatistical analysis, 3rd edn. Prentice Hall,
Englewood Clis, NJ
... First, it is expected to reduce food search costs by increasing the total searched area while decreasing the overlap between foragers' search paths [5,6,[25][26][27][28][29][30]. Second, dispersed central-place foraging should decrease food transport costs by reducing the average distance between food sources and the nearest nest entrance [6,23,24,28,[31][32][33][34][35][36][37][38][39][40]. Third, foraging over a larger territory is considered beneficial by allowing colonies to diversify their food sources [20,21]. ...
... Third, foraging over a larger territory is considered beneficial by allowing colonies to diversify their food sources [20,21]. Fourth, inter-nest recruitment to food sources and/or food redistribution between nests is expected to reduce the variance in the colony's foraging success over time [24,25,27,35,36,[38][39][40][41][42][43][44]. Though the potential advantages of polydomy have been repeatedly mentioned in the literature, they are supported by surprisingly little experimental evidence. ...
... The influence of polydomy on task allocation has never been formally investigated. Similarly, most support for the foraging benefits of dispersed central-place foraging derives from models [25,27,34] and observations that polydomous colonies establish new nests near stable food sources [22][23][24]33,36,[38][39][40]. Few studies have attempted to quantify the foraging efficiency of polydomous ant colonies [5,6,24,26,38,45], and even fewer used experimental manipulations to evaluate the effect of nest number on foraging efficiency [5,45]. ...
Collective foraging confers benefits in terms of reduced predation risk and access to social information, but it heightens local competition when resources are limited. In social insects, resource limitation has been suggested as a possible cause for the typical decrease in per capita productivity observed with increasing colony size, a phenomenon known as Michener’s paradox. Polydomy (distribution of a colony’s brood and workers across multiple nests) is believed to help circumvent this paradox through its positive effect on foraging efficiency, but there is still little supporting evidence for this hypothesis. Here, we showexperimentally that polydomy enhances the foraging performance of food-deprived Temnothorax nylanderi ant colonies via several mechanisms. First, polydomy influences task allocation within colonies, resulting in faster retrieval of protein resources. Second, communication between sister nests reduces search times for far away resources. Third, colonies move queens, brood and workers across available nest sites in response to spatial heterogeneities in protein and carbohydrate resources. This suggests that polydomy represents a flexible mechanism for space occupancy, helping ant colonies adjust to the environment. © 2017 The Author(s) Published by the Royal Society. All rights reserved.
... A colony has one queen (monogynous) and has a complex organisation of multiple peripheral nests around a central nest containing the queen (polydomy). In a study in Kinabalu NP (Pfeiffer & Linsenmair 1998), a single polydomous colony consisted of 7000 foragers divided over 17 nests, spanning an area of 0.8 hectares. They are common in rainforests. ...
... They are common in rainforests. Distribution Southeast-Asia (Pfeiffer & Linsenmair 1998).Tall- Both minors and majors: -Head and gaster brown, rest (alitrunk, petiole, legs and antennae) yellow; -Propodeum laterally compressed; -No (sub)erect pubescence on scape. ...
... Even though species of Crematogaster used nesting in cavities of standing plants, most species of the referred genera are well-known for being extremely generalist regarding their nesting strategies, with colonies found from the soil to the canopy of tropical environments (Baccaro et al., 2015). Future studies could answer if these nests are polydomic or monodomic, because the two patterns can be identified in tropical ants of these genera (Pfeiffer and Linsenmair, 1998;Nakano et al., 2013). In comparison to the 15 ant species found here, recorded 10 species belonging 10 genera of ants in the petioles of A. danaeifolium in Mexico, in most cases forming colonies. ...
Full-text available
Acrostichum danaeifolium, a Neotropical fern, occurs preferentially in marshy areas or at the margins of lakes and mangroves. Microlepidoptera larvae burrow through the petioles of the fern, preferentially on the non-expanded leaves. The galleries in the petiole create a new microhabitat, harboring a rich fauna of arthropods. The aim of the present study was to assess the richness of ants associated with its petiole. The study was conducted in a population of A. danaefolium from the Atlantic Forest in Rio de Janeiro state, Southeastern Brazil. Six collections were carried out every two months (2009-2010), three in the dry and three in the rainy season. The leaves were divided into three development stages: non-expanded leaves (NEL), expanded leaves (EL) and senescent leaves (SL). Seven leaves from each phase were randomly collected from seven individuals. A total of fifteen ant species were recorded. The species with the highest frequency and density in fern petioles were Camponotus crassus and Crematogaster curvispinosa. The highest ant richness and abundance was found in senescent leaves. The high number of ants found in the petioles of Acrostichum danaefolium qualifies it as a potential key species in the marshes and flooded areas where it occurs.
... Camponotus gigas feeds primarily on honeydew (Pfeiffer and Linsenmair 2000 ), much like red wood ants. However, unlike polydomous wood ants, C. gigas is monogynous (Pfeiffer and Linsenmair 1998), which is likely to profoundly affect its life history, and may make a transporter-class beneficial. This study has found that the resource redistribution through a spatially separated wood ant colony is achieved by the same behaviors used for foraging. ...
Full-text available
Resource sharing is an important cooperative behavior in many animals. Sharing resources is particularly important in social insect societies, as division of labor often results in most individuals including, importantly, the reproductives, relying on other members of the colony to provide resources. Sharing resources between individuals is therefore fundamental to the success of social insects. Resource sharing is complicated if a colony inhabits several spatially separated nests, a nesting strategy common in many ant species. Resources must be shared not only between individuals in a single nest but also between nests. We investigated the behaviors facilitating resource redistribution between nests in a dispersed-nesting population of wood ant Formica lugubris. We marked ants, in the field, as they transported resources along the trails between nests of a colony, to investigate how the behavior of individual workers relates to colony-level resource exchange. We found that workers from a particular nest “forage” to other nests in the colony, treating them as food sources. Workers treating other nests as food sources means that simple, pre-existing foraging behaviors are used to move resources through a distributed system. It may be that this simple behavioral mechanism facilitates the evolution of this complex life-history strategy.
... The connected nests of a polydomous ant colony can be represented as a transportation network, in which nests (nodes of the network) are connected by trails (edges of the network). A number of field and laboratory studies have mapped the spatial location of the nests within polydomous colonies and the trails that link them (Cherix 1980; Holt 1990; McIver 1991; Andersen and Patel 1994; Cerda et al. 1994; Federle et al. 1998; Pfeiffer and Linsenmair 1998; Dillier and Wehner 2004; Elias et al. 2005; Boudjema et al. 2006; van Wilgenburg and Elgar 2007; Buczkowski and Bennett 2008; Heller et al. 2008; Latty et al. 2011). The location of nests within a colony may play an important role in the network structure and the role of particular nests. ...
Full-text available
Efficient and robust transportation networks are key to the effectiveness of many natural systems. In polydomous ant colonies, which consist of two or more spatially separated but socially connected nests, resources must be transported between nests. In this study, we analyse the network structure of the inter-nest trails formed by natural polydomous ant colonies. In contrast to previous laboratory studies, the natural colonies in our study do not form minimum spanning tree networks. Instead the networks contain extra connections, suggesting that in natural colonies, robust-ness may be an important factor in network construction. Spatial analysis shows that nests are randomly distributed within the colony boundary and we find nests are most likely to connect to their nearest neighbours. However, the network structure is not entirely determined by spatial associations. By showing that the networks do not minimise total trail length and are not determined only by spatial associations, the results suggest that the inter-nest networks produced by ant colonies are influenced by previously unconsidered factors. We show that the transportation networks of polydomous ant colonies balance trail costs with the construction of networks that enable efficient transportation of resources. These networks therefore provide excellent examples of effective biological transport networks which may provide insight into the design and management of transportation systems.
... However, if large-scale recruitment is required to exploit a resource, then modelling predicts that polydomous colonies might be expected to lose out, because their population of potential recruits is dispersed [29]. This cost can be reduced by involving multiple nests in the recruitment process [12,30] or by recruiting from persistent foraging trails [31] . Moderate polydomy could be a form of discovery-dominance trade-off, in which having dispersed nests improves a colony's ability to find new resources (because scouts are spread relatively evenly over the foraging area) but nests are still large enough to provide enough workers to dominate a resource. ...
Full-text available
Full-text available
Ants are abundant in terrestrial ecosystems, especially in the Brazilian Cerrado, where they can play several roles at different levels of the food chain, including protection of plants against herbivores. Although there are many studies that evaluate the ant–plant interaction in the Cerrado, little is known about the natural history of most species of ants. Camponotus crassus Mayr, 1862, for example, is considered one of the main agents of plant biotic defence in Cerrado. But there are no studies specifically focused on this species, which hinders the understanding of how arthropod–plant interactions are structured in Cerrado. Here, we describe the natural history and ecology of the foraging of the C. crassus. We conducted the study from January 2013 to December 2014 in 10 quadrants of 40 m² to measure: the abundance, density and distribution of nests, location of the nests, the internal structure of the nests, the daily foraging of workers out of the nest, the food items they collect and the existence of territoriality and dominance of the workers on the soil and vegetation. We found 18 nests, 13 in the soil and 5 in hollow trunks on the ground with variable internal structures. The distribution of nests is aggregate, with density of 0.045 nests/m² and average distance of 3.73 m between nests. The foraging activity occurs on the daytime during the rainy and dry season. Extrafloral nectar and honeydew were the resources most collected, comprising 83.33% of the resources in the rainy period and 30% in the dry period. Camponotus crassus is a dominant species, especially on vegetation, although it also forages on the soil. This is the first study to evaluate in detail the natural history and foraging ecology of C. crassus, a diurnal, aggressive and territorial ant that mainly forage climbing onto the plants.
Full-text available
Identifying the boundaries of a social insect colony is vital for properly understanding its ecological function and evolution. Many species of ants are polydomous: colonies inhabit multiple, spatially separated, nests. Ascertaining which nests are parts of the same colony is an important consideration when studying polydomous populations. In this paper, we review the methods that are used to identify which nests are parts of the same polydomous colony and to determine the boundaries of colonies. Specifically, we define and discuss three broad categories of approach: identifying nests sharing resources, identifying nests sharing space, and identifying nests sharing genes. For each of these approaches, we review the theoretical basis, the limitations of the approach and the methods that can be used to implement it. We argue that all three broad approaches have merits and weaknesses, and provide a methodological comparison to help researchers select the tool appropriate for the biological question they are investigating.
Full-text available
Apiculture in Brazil is quite profitable and has great potential for expansion because of the favorable climate and abundancy of plant diversity. However, the occurrence of pests, diseases, and parasites hinders the growth and profitability of beekeeping. In the interior of the state of São Paulo, apiaries are attacked by ants, especially the species Camponotus atriceps (Smith) (Hymenoptera: Formicidae), which use the substances produced by Apis mellifera scutellata (Lepeletier) (Hymenoptera: Apidae), like honey, wax, pollen, and offspring as a source of nourishment for the adult and immature ants, and kill or expel the adult bees during the invasion. This study aimed to understand the invasion of C. atriceps in hives of A. m. scutellata. The individuals were classified into castes and subcastes according to morphometric analyses, and their cuticular chemical compounds were identified using Photoacoustic Fourier transform infrared spectroscopy (FTIR-PAS). The morphometric analyses were able to classify the individuals into reproductive castes (queen and gynes), workers (minor and small ants), and the soldier subcaste (medium and major ants). Identification of cuticular hydrocarbons of these individuals revealed that the eight beehives were invaded by only three colonies of C. atriceps; one of the colonies invaded only one beehive, and the other two colonies underwent a process called sociotomy and were responsible for the invasion of the other seven beehives. The lack of preventive measures and the nocturnal behavior of the ants favored the invasion and attack on the bees.
The epiphytic Bird's Nest Fern (Asplenium nidus complex) has a large basket-shaped rosette that accumulates leaf litter. We investigated the role of these ferns in supporting invertebrate populations in the primary lowland dipterocarp forest of Danum Valley, Sabah, Malaysia. Ferns were divided into three size classes: large (rosette diameter >60 cm), intermediate (30-60 cm), and small (<30 cm). Seven hectares of forest were surveyed: the canopy had a mean density of 30 large ferns/ha and 20 intermediate ferns/ha. Six large and five intermediate ferns were removed from the crowns of Parashorea tomentella (Dipterocarpaceae) at heights between 39 and 52 in. The largest ferns had fresh weights of ca 200 kg. The mean animal abundance in large and intermediate ferns was 41,000 and 8000, respectively. Termites and ants represented at least 90 percent of the abundance in these ferns. Of the I I ferns, 4 contained a nest of Hospitalitermes rufus (Nasutitermitinae), while another contained a nest of an undescribed species of Hospitalitermes. An additional 56 small ferns were removed from the low canopy (2-6 m above the forest floor), of which only I contained a termite nest (Nasutitermes neoparvus). These results suggest that Bird's Nest Ferns contain ca 0.5 million termites/ha and contribute almost one ton (dry mass) of suspended soil and plant material/ha. Five of the trees containing large ferns were fogged immediately before the removal of the ferns. From these samples we were able to estimate the total number of animals in each tree crown. When each estimate was added to the abundance in each fern, the results suggested that a single large fern may contain from 7 to 93 percent of the total number of invertebrates in the crown. Although these results must be treated with caution because of the small sample size, they have important implications for studies of canopy invertebrates.
Full-text available
The nest of a colony of Camponotus gigas (Latreille) located in a hollow log was fumigated and systematically opened. 784 minor workers, 45 major workers and 2 alate males were retrieved, as well as 157 larvae and 34 eggs. Although the queen was not located the presence of eggs and young larvae strongly suggests that it was a queenright colony and the queen was lost during the opening of the nest. Several species of other organisms were found living within the nest, including isopods, crickets (Orthoptera; Gryllidae), cockroaches (Blattodea; Blattidae), an unidentified earwig (Dermaptera) and larvae of a pyralid moth (Lepidoptera; Pyralidae). During two 24 hour observation periods an average of l33 ants were seen to leave or enter the nest, mostly at night, suggesting that only a small proportion of workers are involved in foraging at any given time. Marked ants from this colony were seen to enter another nest nearby, which was probably a satellite of the first nest sin~e activity there ceased soon after the latter was destroyed.
Full-text available
Presents a model of foraging that applies to social insects foraging without recruitment, cooperation, or communication in the search for or retrieval of food. It simulates a colony's foraging via a series of differential equations that quantify the forager activity, the food-source dynamics, the interactions of foragers and food sources, and the colony's energy budget for foraging, defined as the number of nonforagers that the foragers can feed. At the colonial level, the influence of the number of foragers and the size of the foraging area is examined; the colony's social structure, in which only a small proportion of workers forage, greatly limits its potential benefit, leading to the definition of a maximum socially compatible foraging benefit. This is measured in relation to forager characteristics (size and the possession or lack of a memory of food-source location) and to food-source characteristics (size and duration of availability). Foragers without memory obtain their highest benefit when the food sources equal in size their load-carrying capacity, but those with memory can exploit sources several times larger with nearly maximum benefit. There is always on forager size that achieves a maximum benefit greater than that of other sizes. Increasing the sources' duration of availability, or decreasing the competition for them, increases this size. Larger foragers achieve their maximum benefit with larger foraging areas and lower foraging densities. For sources of intermediate size and long availability, a large forager without memory has a higher benefit than a small one with memory; with short availability, the reverse is true. It is predicted that social insects foraging without recruitment or cooperation tend to group into 2 classes. One is characterized by large colonies, foragers, foraging areas, and prey, low foraging densities, and aggressive territorial behavior; the other class is the inverse. -Authors
Full-text available
The foraging behaviour of the giant ant Camponotus gigas , was studied in primary lowland rainforest in Brunei. Nests, found mainly in soil at the base of trees or in fallen timber, were present at a density of 14 per hectare. Foragers were active nearly 24 h per day but much higher levels of activity were recorded at night. By day major and minor workers foraged singly or in small groups on the ground, and by night they ascended in a steady stream into the canopy. It is suggested that the canopy is avoided during the day due to high temperatures and low humidity, and the ground is avoided at night due to increased predatation risks. Overall,
Full-text available
If natural selection shaped foragers of eusocial insects to be efficient food harvesters, leaf-cutting ants should be suited to test optimal foraging predictions because they can reach their own decisions about the prey (leaf) size to be retrieved through their cutting behavior. In a test of the prediction of enhanced selectivity with increasing distance, worker ants foraging at tables 1 and 5 m from their nest cut larger food fragments further from the nest. For each distance, there was a significant regression between the weight of the fragment and that of the ant. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Full-text available
Assumptions behind optimal foraging theory are: 1) an individual's contribution to the next generation (its fitness) depends on behaviour during foraging; 2) there should be a heritable component of foraging behaviour (the actual innate foraging responses or the rules by which such responses are learned), and the proportion of individuals in a population foraging in ways that enhance fitness will tend to increase over time; 3) the relationship between foraging behaviour and fitness is known; 4) evolution of foraging behaviour is not prevented by genetic contraints; 5) such evolution is subject to 'functional' constraints (eg related to the animal's morphology); and 6) foraging behaviour evolves more rapidly than the rate at which relevant environmental conditions change. The review considers recent theoretical and empirical developments, dealing with behaviour of animals while they are foraging (but ignoring the timing of and the amount of time allocated to such behaviour). Ideas considered include risk aversion and risk proneness; optimal diet; optimal patch choice; optimal patch departure rules; optimal movements; and optimal central place foraging. Means of evaluating optimal foraging theory are discussed. -P.J.Jarvis
In numbers and biomass, ants (Hymenoptera, Formicidae) often dominate arthropod faunas of tropical rainforest canopies. Extraordinary ant abundance is due principally to one or a few species able to tap the high productivity of canopy foliage by feeding on plant and homopteran exudates. Prior studies of nitrogen isotopic ratios show that exudate‐feeders derive much of their nitrogen (N) by processing large quantities of N‐poor, but carbohydrate (CHO)‐rich, exudates. CHOs in excess of those that can be coupled with protein for growth and reproduction (postulated as the colony's first priorities) may be directed at little cost and some profit to functions that increase access to limiting protein. High dietary CHO:protein ratios for exudate‐feeders appear to subsidize ‘high tempo’ foraging activity, defence of absolute (level III) territories, and production of N‐free alarm/defence exocrine products that enhance ecological dominance in contests with other ants. Among organisms (e.g. plants and Lepidoptera) symbiotic with ants, CHO:protein ratios of ant rewards may control both the identities of ant associates and the quality of ant‐rendered services. Dietary ratios of CHO:protein play an important and previously unrecognized role in the ecology and evolution of ants generally. Modifications of worker digestive systems in certain ant subfamilies and genera represent key innovations for handling and processing large volumes of liquid food. The supreme tropical dominants are species released from nest site limitation and able to place their nests in the vicinity of abundant exudate resources. Polydomy appears to be typical of these species and should produce energetic savings by taking colony fragments to the resource.
We measured the energy costs of locomotion and load carriage in two species of the honeypot ant, Myrmecocystus. The diurnal forager, Myrmecocystus mendax, ran voluntarily in a running tube respirometer at 40° C. After running for about an hour, foragers were removed and given access to honey water, which they drank. The energy cost of carrying this internal load was then measured. The nocturnalforager, Myrmecocystus mexicanus, ran voluntarily in the running tube at 30° C. These ants could not be persuaded to collect honey water. The minimum costs of transport (MCOT)-160 and 105 J kg⁻¹ m⁻¹ for M. mendax (6 mg) and M. mexicanus (14 mg), respectively-were substantially below values estimated from current allometric equations relating MCOT to body mass. Using our voluntary locomotion regime, we found the Y-axis intercept of the relation between metabolic rate and running speed to be equivalent to standard metabolic rate for all species and load conditions. The cost per unit mass of internal load carriage in M. mendax (the smallest animal to date for which we have data on load carriage costs) was equivalent to that of body mass carriage. Internal load carriage therefore does not appear to offer cost savings compared with external load carriage.
If pursuit or provisioning time is unrelated to prey size, the following size-distance properties are expected for a variety of prey-handling functions. 1. The best prey size increases with distance. 2. For given values of profit (measured as energy per unit time), profit changes less with distance, the larger the prey. 3. The range of sizes taken shifts toward larger sizes, as distance increases. If pursuit or provisioning time increases with prey size, these effects can be reversed.