Parabiotic associations between tropical ants: equal partnership or parasitic exploitation?


Correspondence author. E-mail:


1. The huge diversity of symbiotic associations among animals and/or plants comprises both mutualisms and parasitisms. Most symbioses between social insect species, however, involve social parasites, while mutual benefits have been only suspected for some parabiotic associations – two colonies that share a nest.

2. In the rainforest of Borneo, we studied parabiotic associations between the ants Crematogaster modiglianii and Camponotus rufifemur. Parabiotic nests were regularly found inside hollow tree trunks, most likely initiated by Cr. modiglianii. This species frequently nested without its partner, whereas we never found non-parabiotic Ca. rufifemur nests. We experimentally investigated potential benefits, potential interference competition for food (as a probable cost), and foraging niches of both species.

3. The two species never showed aggressive interactions and amicably shared food resources. However, Cr. modiglianii had a wider temporal and spatial foraging range than Ca. rufifemur, always found baits before Ca. rufifemur and recruited more efficiently. Camponotus rufifemur probably benefited from following pheromone trails of Cr. modiglianii. In turn, Ca. rufifemur was significantly more successful in defending the nest against alien ants. Crematogaster modiglianii hence may profit from its partner’s defensive abilities.

4. In neotropical parabioses, epiphytes grown in ‘ant-gardens’ play a crucial role in the association, e.g. by stabilization of nests. Hemiepiphytic Poikilospermum cordifolium (Cecropiaceae) seedlings and saplings frequently grew in the entrances of parabiotic nests in Borneo, obviously dispersed by the ants. In cafeteria experiments, both parabiotic ants carried its elaiosome-bearing seeds into the nest. However, P. cordifolium does not provide additional nest space, contrasting with neotropical ant-gardens.

5. The parabiotic association appears beneficial for both ant species, the main benefits being nest initiation by Cr. modiglianii and interspecific trail-following (for Ca. rufifemur), and, in turn, nest defence by Ca. rufifemur (for Cr. modiglianii). However, Ca. rufifemur seems to be more dependent on its partner than vice versa.


Symbiotic interactions between different organisms are ubiquitous and include species of all five kingdoms (e.g. Ollerton 2006). Social insects represent particularly attractive symbiotic partners, as their long-lived colonies offer large, predictable resources and protection against enemies or environmental disturbance. Hence, although social insect colonies are heavily defended against intruders, it is not surprising that they often represent whole ecosystems in themselves. They are frequently inhabited by various species of invertebrates which overcome the colony’s defence against other species and depend more or less strictly on this symbiosis. Guests of social insect colonies are especially diverse in ant and termite colonies (Kistner 1979; Hölldobler & Wilson 1990), and various invertebrate species live in bee nests, including social parasites (Michener 2000). In ants, guest species include parasites that prey on their hosts’ brood or workers, and commensals that feed on waste or simply seek shelter and protection in the ant nest (Kistner 1979; Schultz & McGlynn 2000; Pierce et al. 2002; Geiselhardt, Peschke & Nagel 2007). Intriguingly, even whole ant colonies can live in nests of other ant species, ranging from loosely associated, facultative commensals or cleptoparasites to highly specialized social parasites (Kaufmann et al. 2003; Huang & Dornhaus 2008).

Very few cases world-wide, however, are known where two ant species live in an association that appears symmetric, without obvious parasitic or exploitative interactions. Forel (1898) first described such an association (between Dolichoderus and Crematogaster) in the Colombian rainforest and named it ‘parabiosis’. By introducing this term, he indicated that this was a new kind of association – it was not clear whether parabioses were mutualistic, commensalistic or parasitic. However, as the two species shared a nest and tolerated each other, parabioses differed from other associations where, e.g. two ant species nest closely together but do not share nests (e.g. Czechowski 2004). Parabioses are largely confined to associations between species of Crematogaster and either Camponotus, Dolichoderus, Odontomachus or Pachycondyla in South American and Southeast Asian rainforests (Orivel, Errard & Dejean 1997; Menzel, Linsenmair & Blüthgen 2008b). Often, epiphytic plants inhabit these nests (‘ant-gardens’), and they are important mutualists because their roots are crucial for nest stability. The nests may include further associated species, such as trophobionts and other insect guests (Corbara, Dejean & Orivel 1999; Kaufmann & Maschwitz 2006). Since Forel’s first report, the ecological character of neotropical ant-garden parabioses has been debated (e.g. Swain 1980; Davidson 1988; Dejean et al. 2000). It remains largely unresolved whether they represent a case of social parasitism, commensalism or mutualism, although a recent study tentatively suggested a mutualistic interaction (Vantaux et al. 2007). In Southeast Asia, parabioses have been poorly studied to date, so even less is known about reciprocal costs and benefits in these associations.

Mutualism is defined as an interaction between two species that conveys a net benefit to both partners. Thus, the benefits each species gains from its partner outweigh the costs of the interaction. The study of mutualistic relationships involves analysing the costs and benefits either party incurs through its partner. These are usually quantified in terms of reproduction, survival or growth. Many mutualistic systems allow experimental manipulation in situ, e.g. exclusion of one partner, to estimate its impact on the other party. Using this approach, mutualistic benefits have been studied in various systems, e.g. between cleaner fish and their clients (Grutter 1999), plants and seed-dispersers (Levey, Silva & Galetti 2002), ants and trophobiotic aphids (Stadler & Dixon 2005), ants and myrmecophytes (Heil & McKey 2003), plants and pollinators (e.g. Kearns & Inoue 1993; Klein, Steffan-Dewenter & Tscharntke 2003) and mycorrhizal fungi and their hosts (Johnson, Graham & Smith 1997). In parabiotic associations, these costs and benefits are more difficult to estimate. As both partners are eusocial and the colonies are long lived, benefits in terms of growth, reproduction and survival are hard to quantify. In addition, removing or excluding one of the partners from a nest is practically impossible without severely affecting the other.

In this study, we investigated parabioses of Crematogaster modiglianii and one of two Camponotus (Myrmotarsus) species in the lowland rainforest of Borneo. These parabiotic nests are often associated with the hemiepiphyte Poikilospermum cordifolium. Experimental manipulation of the nests is additionally hampered as they are usually located in living trees in mature (and often protected) rainforests. We thus focused on potential costs and benefits that were experimentally accessible outside the nest and in laboratory experiments.

One of the most probable costs of being parabiotic is increased food competition, which is likely to result in aggressive resource defence. In turn, food competition may be reduced by niche differentiation, e.g. spatial and temporal foraging range, different food preferences, or a dominance-discovery trade-off. Possible benefits that could be conveyed by the parabiotic lifestyle include joint nest defence, reciprocal provision of nest space, food exchange via trophallaxis and mutual brood care. Furthermore, parabiotic ants could benefit from the epiphytes through provision of nest space or nutrition (e.g. in neotropical ant-gardens, Vantaux et al. 2007), while the plants themselves may benefit from seed dispersal and herbivore protection offered by parabiotic ants (Heil & McKey 2003). We conducted a series of experiments series to evaluate these hypotheses.

Materials and methods

Study site and ants

Research was carried out at the Danum Valley Conservation Area between August and December in the years 2006–2008. The area (5°N, 117°50′E) is one of the major remaining patches of lowland dipterocarp rainforest in Sabah (Malaysian Borneo). It has a typical equatorial rainforest climate with a mean annual temperature of 26·9 °C and a yearly rainfall of 2700 mm.

We studied parabiotic associations between Crematogaster (Paracrema) modiglianii Emery 1900, which is monomorphic and measures c. 2–3 mm, and two Camponotus species, both of which are polymorphic with body lengths between 5 and 13 mm (Fig. 1). The Camponotus partner was chiefly Ca. (Myrmotarsus) rufifemur Emery 1900 (34/37 cases) or, in three cases, most probably Ca. irritabilis (Smith) 1857 (identification by Seiki Yamane). Additional parabioses between the same species were studied in Gunung Mulu National Park (Sarawak) and the rainforest around Sepilok (Sandakan, Sabah). Voucher specimens of all ant species are deposited at the Forest Research Center in Sepilok, Sabah (Malaysia) and at the Department of Zoology, University of Würzburg. An overview of the hypotheses tested and the experiments performed is given in Table 1.

Figure 1.

 (a) Crematogaster modiglianii and Camponotus rufifemur workers jointly tending Cicadellidae trophobionts. (b) Crematogaster modiglianii worker carrying a Poikilospermum cordifolium seed. Note the pronounced elaiosome. (c) The two parabiotic species amicably sharing a tuna bait. (d) Crematogaster modiglianii and Ca. rufifemur during trophallaxis. (e) Mounting behaviour of Cr. modiglianii towards a Ca. rufifemur major. (f) The two species at a common nest entrance on the nest tree; note the P. cordifolium sapling growing in the entrance hole. Photos (a–e) by F. Menzel, (f) by N. Blüthgen.

Table 1.   Overview of the research questions and hypotheses addressed in this study
Ant–ant interactions
  1. Note that for a number of hypotheses, the tested predictions are necessary but not sufficient.

Food competition and competitive avoidance through foraging niche differentiation
(1) There is aggressive interference competition at food baits
Test: Observation of interspecific aggression during bait experiments; test for correlation between Camponotus rufifemur presence and number of Crematogaster modiglianii foragers
(2) The foraging niche of the two species differs regarding diurnal foraging activity, foraging ranges and/or food preferences
Test: Bait experiments (diurnal/nocturnal activity, honey/tuna), nest distance assays, survey of trophobioses and EFNs
(3) Dominance-discovery trade-off: one species recruits faster than the other (in combination with interference competition)
Test: Recruitment efficiency surveys
Interactions within the nest that might be beneficial for one or both species
(1) The species care each other’s brood
Test: Brood care assays
(2) Mutual food exchange via trophallaxis
Test: Test for interspecific trophallaxis via stained food
Reciprocal provision of nest space
(1) Parabiotic nests differ in relative worker and brood abundance of each species
Test: Nest dissections; nest survey
(2) Parabiotic ants are specialized on certain trees
Test: Nest survey (tree species, diameter etc.)
Nest defence
(1) Ca. rufifemur is a better nest defender than Cr. modiglianii
Test: Nest defence assays
Ant–plant interactions
Benefits to Poikilospermum cordifolium provided by the two ant species
(1) P. cordifolium can establish at parabiotic nests
Test: Nest survey (presence of epiphytes); survey of turnover and longevity of P. cordifolium
(2) The ants protect P. cordifolium epiphytes from herbivores
Test: Comparison of percent herbivory damage of P. cordifolium epiphytes at trees inhabited by parabioses and at uninhabited trees
(3) One or both ant species retrieve P. cordifolium seeds to the nest and leave them close to the entrance
Test: Cafeteria experiments
Benefits to Ca. rufifemur and/or Cr. modiglianii provided by P. cordifolium
(1) The elaiosome-bearing seeds are attractive to the ants
Test: Cafeteria experiments
(2) Extrafloral nectar of P. cordifolium contains essential amino acids
Test: Analysis of amino acids in nectar droplets

Foraging and recruiting behaviour

Food competition and potential foraging niche differentiation were studied using bait experiments. We attached plastic platforms (c. 8 × 8 cm) directly to the nest tree of seven parabiotic nests. Ten platforms per colony were provided with either honey or tuna baits and the workers present on the platform were counted 1–2 h later. These surveys were conducted during the day (between 11:30 am and 4:30 pm) or at night (between 8 pm and 11:30 pm). We obtained data from 82 platforms (of a total of 170) where at least one parabiotic species had recruited. Using generalized linear models (GLMs; with Gaussian error distribution), we calculated the effects of bait and time of day on the abundance of each species present at the platform. For Cr. modiglianii, the presence of Ca. rufifemur at the bait was included as an additional factor. Using this setup, we tried to detect interference competition, which would be visible through aggressive interactions at the baits and bait monopolization. Furthermore, the setup allowed investigation of possible niche differences in diurnal foraging activity and coarse food preferences between protein and fat-containing tuna and carbohydrate-containing honey. For further evidence on food preferences, we surveyed foragers of both ant species at trophobioses and extrafloral nectaries in the vicinity of two parabiotic nests.

As Ca. rufifemur is notably larger than Cr. modiglianii, it might possess a larger foraging range and find baits faster. We therefore studied recruitment in relation to nest distance. We spanned horizontal pairs of string (Ø 2 mm) from parabiotic nest trees (= 7 nests) to surrounding trees in heights up to 4 m. Plastic platforms (c. 6 × 6 cm, = 95) were attached to these strings between 0 and 5·5 m from the nest tree. These platforms were provided with baits between 7:30 and 10 pm and checked for ants 30–70 min later. As baits, we used either honey or tuna in equal frequencies. The experiment was carried out for 4–5 times per parabiosis. From these data, we calculated the maximum foraging distance (per nest) for each species and recorded the number of baits tended by the two respective species in relation to nest distance. At two parabiotic nests, we studied recruitment efficiency by applying honey or tuna baits to leaves close to a parabiotic nest (1–3 m distance). We then recorded the number of workers at the baits every 2 min for 30–45 min. A total of six diurnal and four nocturnal replicates was performed.

Nest architecture and nest sites

As a potential benefit of living in parabiosis, one species might use nest space, i.e. wood cavities or carton shelters, which is provided by the partner species. We therefore thoroughly dissected 10 parabiotic nests to find out how closely the two species interact within the nest and whether one of the species might use nest space provided by the other. We recorded nest architecture, the areas each species occupied, abundance of the two species, and the presence of brood, alates, and guest species. The 10 nests were located in living trees (diameter at breast height, d.b.h. 5–15 cm; = 7), a liana, a dead tree and a log. Eight non-parabiotic Cr. modiglianii nests were dissected for comparison (one in a living tree and seven in dead branches). In addition, we surveyed 37 parabiotic nests around Danum Valley Field Center (DVFC). These were located in hollow, living tree trunks, lianas or (in three cases) in dead logs on the forest floor. We recorded nest tree species, d.b.h., number of nest entrances and the presence of epiphytes and carton material at the nest entrances. Similarly, 13 non-parabiotic Cr. modiglianii nests were surveyed, whereas no non-parabiotic Ca. rufifemur nests were found.

Behavioural interactions within the nest

Within the nest, the two species might profit from reciprocal brood care. To verify this hypothesis, worker groups of either species (from the same parabiotic nest) were put in a box together with brood of their own and the other species (= 2 per species). The brood was uncovered, but we offered a nearby shelter of small wooden pieces. After c. 1 h, we checked whether the workers had carried the brood under the shelter.

We tried to detect interspecific trophallaxis between the two species using worker groups. These were kept in the DVFC laboratory in fluon-coated plastic boxes on a tuna and honey diet for several days or weeks. Food dyed with Rose Bengale (Chroma, Münster/Germany) was provided to a worker group of only one species (Cr. modiglianii: = 3; Ca. rufifemur: = 2). Two days later, the other species was introduced and the dyed food removed. After another 2 days, we dissected the gasters of the second species and checked whether the digestive tracts were stained with Rose Bengale.

Nest defence

One of the two ant species might also profit from the other’s nest defence against enemies. Therefore, we estimated the ability of each species to defend the nest against intruders. We confronted nine Ca. rufifemur–Cr. modiglianii parabioses with living workers of three different ant species (Myrmicaria sp., Crematogaster inflata Smith 1857, and Dolichoderus thoracicus Smith 1860). All three are among the dominant ants in the rainforest around DVFC. While Myrmicaria sp. is strictly nocturnal, the other two species are active day and night, Cr. inflata being more active at night and D. thoracicus being more active during the day. In haphazard order, each of the three intruders was held with forceps at the nest entrance for 3 min or until it was killed. For each test, we recorded which of the parabiotic species was present at the nest entrance, the number of bites of either parabiotic species towards the intruder, as well as which species killed the intruder. The tests were conducted at nine parabiotic nests with six replicates both at night (after 6:30 pm) and during the day (before 6 pm).

Interactions with Poikilospermum epiphytes

Seedlings and larger individuals of the hemiepiphytic P. cordifolium (Barg-Petr.) Merr. (Cecropiaceae) frequently grow at the entrances of parabiotic nests (Fig. 1f). This plant may profit from the ants through seed dispersal or herbivore defence. We therefore surveyed growth and mortality of P. cordifolium individuals at parabiotic nests in several consecutive field seasons (March 2006 to October 2008). The plants were mapped up to a height of around 4 m at 15 parabioses. For all plants with the largest leaf longer than 3 cm, we estimated percent herbivory damage to each leaf, and, for comparison, included five individuals not associated with parabiotic nests.

To investigate attractiveness of P. cordifolium diaspores to each parabiotic species compared to those of the congeneric, syntopic P. suaveolens, and to compare retrieval rates of diaspores of these two species by the two ants, we conducted cafeteria experiments with seeds and perianths of both Poikilospermum species. In Poikilospermum, the fleshy, bright blue perianths are persistent and cover the ripe seeds. On plastic platforms at the nest entrance, we offered both seeds (with their elaiosomes, see Fig. 1b) and perianths of the two Poikilospermum species. Four pieces per item were provided. As control, four pieces of rice (grains cut into three pieces each) were offered as they were of similar weight (5·9 ± 1·4 mg, mean ± SD of 40 pieces) and thus a comparable food item. For the following 30 min, we recorded the number of pieces carried into the nest as well as the number of workers at each of the five different items (as an alternative measure of attractiveness). The seeds used for the experiments weighed 3·1 ± 0·8 mg (P. cordifolium) and 5·3 ± 0·8 mg (P. suaveolens, both = 36). Perianths weighed 7·3 ± 5·4 and 11·4 ± 4·7 mg, respectively (both = 28). For three parabiotic nests, eight replicates each were carried out at night under red light. To lure out enough workers, tuna was offered prior to the experiment in a separate plastic cap (Ø 2·5 cm) that was removed at the onset of the experiment. We compared the attractiveness among the four food items (thus, excluding the controls) by analyzing the number of retrieved pieces (for both species separately) and (for Cr. modiglianii) the number of workers present at the food items 5 min after the start of the experiment. The impacts of the variables ‘food item’ and ‘colony’ were estimated using GLMs with binomial error distribution and chi-squared tests (retrieved pieces) and GLMs with Gaussian error distribution and F tests (number of workers).

Poikilospermum cordifolium leaves are often visited by Cr. modiglianii workers. The leaves produce small nectar droplets on their upper surface, which may be an important food source for ants. Using an amino acid analyser (Biotronik LC3000, Maintal, Germany), we therefore measured free amino acids in nectar droplets of three P. cordifolium individuals (see Appendix S1, Supporting Information for further details).


Foraging and recruiting behaviour

Both Ca. rufifemur and Cr. modiglianii were attracted to tuna and honey baits on platforms directly at the nest tree. The presence of Ca. rufifemur did not affect the number of Cr. modiglianii workers at the baits (GLM: F1,78 = 0·75, = 0·39, Fig. 2a), thus there was no evidence of interference competition or even resource monopolization. Aggression between the two species was never observed (see also Menzel et al. 2008b). Ca. rufifemur largely ignored the much smaller Cr. modiglianii; however, sometimes single Cr. modiglianii workers were antennated very intensely. Ca. rufifemur workers were significantly more abundant at tuna baits than at honey baits (F1,80 = 9·7, = 0·0025) but almost only recruited at night (F1,79 = 16·2, = 0·0001, Fig. 2b). In contrast, Cr. modiglianii was slightly but significantly more abundant at honey baits (F1,80 = 4·0, = 0·048) and recruited during day and night (F1,79 = 1·2, = 0·27, Fig. 2c).

Figure 2.

 (a) Number of Crematogaster modiglianii workers at tuna baits depending on the presence or absence of Camponotus rufifemur. Data for honey baits are similar. (b) Number of Ca. rufifemur and (c) number of Cr. modiglianii workers recruiting to honey or tuna baits during day or night. Numbers of replicates are given above each plot. ***Significant at < 0·001.

Crematogaster modiglianii, often together with Ca. rufifemur, regularly tended trophobionts. These included two Coccoidea species under carton shelters at parabiotic nests, various membracid and cicadellid nymphs (both Cicadelloidea) and a lycaenid caterpillar (Lepidoptera) (Fig. 1a). They also foraged together at carrion and extrafloral nectaries (e.g. of Mallotus miquelianus (Scheff.) Boerl., Euphorbiaceae, or Diospyros toposioides King & Gamble, Ebenaceae).

In the nest distance assays, Cr. modiglianii foraged at baits further distant from the nest than Ca. rufifemur. The maximal foraging distance of Cr. modiglianii (per parabiotic nest) was significantly higher than that of Ca. rufifemur (paired t-test: t6 = 7·02, = 0·0004; Fig. 3a). Moreover, the proportion of baits attended by Ca. rufifemur only (as opposed to those attended by Cr. modiglianii or both species) decreased with nest distance (Fig. 3b). This relation was similar for both honey and tuna baits. Crematogaster modiglianii was very effective in finding newly placed baits. The workers found them within 15 min in all of the 10 recruitment surveys, with 22·8 ± 8·0 foragers at the baits after 15 min (mean ± SE, Fig. 3c). In contrast, Ca. rufifemur only approached the baits in three cases at dusk or at night, and only reached them 3–40 min after Cr. modiglianii.

Figure 3.

 (a) Maximal foraging distance per parabiotic colony for Crematogaster modiglianii and Camponotus rufifemur (both = 7 colonies) ***significant at < 0·001. (b) Proportion of baits attended by Cr. modiglianii, Ca. rufifemur, or both, in relation to nest distance. (c) Example of the recruitment progress of Cr. modiglianii and Ca. rufifemur to a tuna bait.

Nest architecture and nest sites

Among the 10 parabiotic nests we opened, only two – located in living tree trunks – contained brood of both species. In one case, Ca. rufifemur workers and brood occupied the whole central cavity (which was compartmentalized with carton material), whereas Cr. modiglianii workers and brood only occurred in a small side compartment and in a bracket fungus close to a nest entrance. In the other nest, Ca. rufifemur and Cr. modiglianii brood (as well as Cr. modiglianii alates) was found in multiple separate, walnut-sized compartments within the trunk. Crematogaster modiglianii also settled in finely hollowed areas around openings (which were too narrow for Ca. rufifemur). Both species kept their brood separate but very close to each other, often separated by only a few centimetres. Likewise, the workers usually stayed among conspecifics albeit close to the partner species. The remaining eight parabiotic nests (also located in trees, lianas or logs) contained workers, brood, and sometimes alate or dealate queens of Cr. modiglianii, as well as Ca. rufifemur workers, but no Ca. rufifemur brood. In three cases, fewer than 20 Ca. rufifemur workers but up to several thousand Cr. modiglianii workers were found in these nests. We frequently found ‘shelters’ in hollow logs, lianas, dead branches or carton-covered twigs that contained workers of Cr. modiglianii or both species, but no brood. Cr. modiglianii workers, but not Ca. rufifemur, also readily moved into hollow Uncaria stalks (inner diameter c. 4 mm) offered close to existing nests and sometimes also stored brood in these stalks.

One very populous non-parabiotic Cr. modiglianii nest was located in the trunk of a small Diospyros tree (Ebenaceae). It consisted of several unconnected cavities in the trunk, which probably stemmed from activities of a wood-boring insect, and contained workers, brood, alates and several dealate queens. The other seven non-parabiotic Cr. modiglianii nests we dissected were located in dead branches that were entangled in the understorey and contained workers and sometimes brood.

Several newly eclosed imagines of the wood-boring beetle Apriona flavescens Kaup 1866 (Cerambycidae) were found inside parabiotic nests, suggesting that cavities made by their larvae may serve as starting points for colony foundations. Further guest species of parabiotic nests included few (up to five) imagines of Tenebrionidae (cf. Tetraphyllus sp.) and Scarabaeidae, larvae of Scarabaeidae and Scirtidae, Myrmecophilus sp. (Myrmecophilidae, Grylloidea), Psychodidae larvae (Diptera), and less than 100 individuals each of the ant genera Leptogenys, Pristomyrmex and an unidentified termite species.

Most (32/37) of the parabiotic nests of Cr. modiglianii and Ca. rufifemur occurred in hollow, living trees (Fig. 1f). Nest trees were 5–32·2 cm in d.b.h. (median: 9·8 cm, = 30) and belonged to c. 24 species from 15 families. The families most commonly represented among nest trees were Euphorbiaceae (e.g. Baccaurea spp.) and Myrtaceae (exclusively Syzygium spp.), each represented by seven of 32 identified, living nest trees. Three further parabiotic nests were found in dead logs or branches on the forest floor and two were located in lianas in the understorey (e.g. Uncaria ferrea DC, Rubiaceae). The nests had up to six entrances between 0 and 400 cm above the soil, which were partly covered with carton material in 10/37 nests. Most of them were used by both species together. No differences between parabioses with Ca. rufifemur and Ca. cf. irritabilis were found. We never found Ca. rufifemur nesting without its parabiotic partner. In contrast, 13 non-parabiotic Cr. modiglianii nests were found. They were located in small trees (d.b.h. 4–6 cm, = 6) or dead branches entangled in the understorey (= 6) or on the forest floor (= 1). Further Cr. modiglianii worker groups were found in carton shelters around small, living twigs or within dead branches.

Behavioural interactions within the nest

Trophallaxis between the two species, initiated via solicitation by Ca. rufifemur, was observed several times in laboratory colonies (Fig. 1d). Via stained food, food transfer from Cr. modiglianii to 10 out of 17 Ca. rufifemur workers was detected in one out of three worker groups but not in the others. We did not detect food transfer in the opposite direction. In our assays, we never observed interspecific brood care. Each species only carried its own brood under the shelter and ignored brood or pupae of the partner.

Nest defence

In the nest defence assays, both Ca. rufifemur and Cr. modiglianii usually killed the intruder ants that were held at the nest entrances (88% of all assays). Crematogaster modiglianii repulsed intruders by spreadeagling them, i.e. several workers grabbed the intruder’s legs or antennae with their mandibles and pulled backwards, leading to the intruder’s death after some time. In successful cases, 7·5 ± 3·1 workers (mean ± SD) had seized the intruder within 3 min. In contrast, most attacks by Ca. rufifemur only involved one or two Ca. rufifemur workers. They bit the intruder (3·9 ± 2·9 bites per successful repulse) and often killed it within less than 30 s.

The chances of a successful repulse were clearly higher when both species were at the nest entrance, compared to assays where only Cr. modiglianii was present (Fig. 4a). This effect was significant for the intruder species Myrmicaria sp. and Cr. inflata (Fisher’s exact test: = 0·00096 and 0·023, respectively), but not for D. thoracicus (= 0·095). In contrast, Ca. rufifemur alone defended intruders as successfully as did both species together (= 1 for D. thoracicus and Myrmicaria sp.; Cr. inflata was always repulsed when Ca. rufifemur was present, irrespective of presence or absence of Cr. modiglianii).

Figure 4.

 (a) Proportion of successful repulses of ant intruders in relation to the presence of either parabiotic species at the nest. White: intruder not repulsed; grey: intruder repulsed. ***< 0·0001; *< 0·05, < 0·1, ‘n.s.’> 0·1, 1no test possible as all intruders were repulsed. (b) Identity of attacking species when both were present and the repulse was successful. White: Camponotus rufifemur attacks, grey: both attack, black: Crematogaster modiglianii attacks. Numbers or replicates are given behind each bar.

In seven out of nine colonies, Ca. rufifemur majors were nearly always present at the nest entrances during both day and night. The probability of successful repulse of an intruder, as well as the presence of either species at the nest entrance, did not differ between diurnal and nocturnal experiments (all three Fisher’s = 1). When both species were present (64% of all assays), it was usually Ca. rufifemur who attacked the intruder (75% of cases, Fig. 4b). Both D. thoracicus and Cr. inflata were observed to attack and kill single Cr. modiglianii workers, but were always killed by Ca. rufifemur. Myrmicaria often sprayed venom and therefore was less frequently attacked by both parabiotic species.

Interactions with Poikilospermum epiphytes

The hemiepiphyte P. cordifolium was found at 22 of the 37 parabiotic nests in Danum Valley (Fig. 1f). Individuals chiefly occurred as seedlings (two-cotyledon stage; 59·0%) or saplings (34·9%), with stalks up to 10 cm long and leaves up to 8 cm long, and grew in carton-covered (9/22 cases) or blank nest entrances (13/22 cases). Often, they grew in high densities, with small carton patches (c. 5 × 3 cm) around a nest entrance being inhabited by one to six seedlings. At one nest, a carton patch of c. 40 × 5 cm carried 68 P. cordifolium seedlings. Most of them, however, failed to establish over a longer period. Fifty-one of 77 individuals beyond the seedling stage had died or disappeared after 1 or 2 years. Poikilospermum cordifolium seedlings or saplings also grew in four non-parabiotic Cr. modiglianii nests, including ones in dead logs or branches. Beside seedlings and saplings, several parabiotic nest trunks carried large P. cordifolium individuals with leaves of up to 60 cm length; at one nest, the whole tree was overgrown by a large P. cordifolium (>10 m high). Poikilospermum cordifolium individuals on active parabiotic nests suffered slightly, but not significantly less herbivory than on presently unoccupied trees (Fig. 5a; Wilcoxon test: = 15, = 0·36, N1 = 9, N2 = 5). Poikilospermum cordifolium also grew at nest entrances of Diacamma sp. and Cr. inflata. The congeneric P. suaveolens was sometimes found at nests of Crematogaster difformis Smith 1857 but never at parabioses. At Gunung Mulu, parabiotic nests were sometimes associated with Poikilospermum oblongifolium. Other epiphytes (e.g. Polypodiaceae, Piperaceae) were irregularly found growing on nest trees but never in the nest entrances.

Figure 5.

 (a) Mean % of leaf area (per tree) destroyed by herbivores of Poikilospermum cordifolium epiphytes on trees inhabited by parabioses and uninhabited trees. (b and c) Retrieval of seeds and perianths of P. cordifolium and Poikilospermum suaveolens, and rice controls by (b) Camponotus rufifemur and (c) Crematogaster modiglianii. The three different symbols refer to three different parabiotic nests.

In the cafeteria experiments, both Ca. rufifemur and Cr. modiglianii regularly retrieved seeds (Fig. 1b) and perianths of P. cordifolium and P. suaveolens but only few pieces of rice that served as control. We found no significant preference among the four Poikilospermum items (GLM excluding rice controls: inline image ≤ 0·63,  0·89 for both ant species). For Ca. rufifemur, the rate of retrieval differed strongly among the three parabiotic nests (GLM: inline image = 22·2, < 0·0001, Fig. 5b), which relates to unequal Ca. rufifemur abundance at these nests. Crematogaster modiglianii workers rarely carried the offered items to the nest (Fig. 5c), but all four items (excluding the rice controls) quickly attracted significant numbers of Cr. modiglianii foragers. On average, 6·1 ± 0·6 Cr. modiglianii workers but only 0·2 ± 0·05 Ca. rufifemur workers (mean ± SE, both = 96) were foraging at each item 5 min after start of the experiment. The Cr. modiglianii abundance significantly differed among the three colonies (GLM: F2,90 = 4·16, = 0·019) but not among the four food items (F3,92 = 1·45, = 0·23). After the experiments, Poikilospermum seeds were sometimes found in the carton material near the nest entrances.

Nectar droplets of P. cordifolium leaves regularly contained all essential amino acids except for methionine and tryptophan, as well as up to 12 non-essential ones (see Appendix S1, Supporting Information).


Parabiosis is an unusually close association between two ant species that share a common nest. Besides the nest, the parabiotic ants Cr. modiglianii and Ca. rufifemur even share food resources without aggression and often engage in mutual communication, e.g. antennating and trophallaxis. It is largely unknown whether a parabiosis confers benefits to both partners, or whether one species exploits the other. The present study therefore tries to evaluate costs and benefits which the two species incur from the parabiotic way of life. An overview of the alleged mutual benefits among Cr. modiglianii, Ca. rufifemur and the epiphytic P. cordifolium is given in Fig. 6.

Figure 6.

 Overview of the mutual benefits among Crematogaster modiglianii, Camponotus rufifemur and Poikilospermum cordifolium. Thick arrows highlight the alleged key benefits assumed by the authors. (E) Benefit experimentally shown; (O) benefit inferred from observations and therefore tentative.

Foraging niches of the two species

Camponotus rufifemur and Cr. modiglianii shared food resources, e.g. baits or trophobioses, without aggression or other evidence of interference competition. This ‘food-sharing’ is highly unusual, given that most ant species, including neotropical parabiotic ants, aggressively monopolize high-quality food resources (Swain 1980; Blüthgen & Fiedler 2004; Blüthgen, Mezger & Linsenmair 2006). Exploitation competition may be reduced for Cr. modiglianii: it was active during day and night and had a significantly wider foraging range than the mainly nocturnal Ca. rufifemur. Moreover, Ca. rufifemur was significantly more abundant at tuna (which contain proteins and fats) than at honey baits (which contain chiefly carbohydrates). In contrast, Cr. modiglianii more abundantly foraged at honey baits, extrafloral nectaries, and tended trophobioses (see also Blüthgen et al. 2006).

Crematogaster modiglianii quickly recruited conspecifics to newly discovered baits, whereas Ca. rufifemur never reached baits before Cr. modiglianii. Experimental evidence confirmed that Ca. rufifemur follows trails of Cr. modiglianii but not vice versa (T. Pokorny, unpublished data; Menzel 2009). The exploitation of another species’ pheromone trails is an example of ‘olfactory eavesdropping’ or ‘informational parasitism’ (Adams 1990; Nieh, Barreto & Contrera 2004). It represents an important benefit Ca. rufifemur derives from its partner, and a possible cost for Cr. modiglianii in terms of exploitation competition during times when both species are active.

Camponotus rufifemur several times solicited Cr. modiglianii for trophallaxis. Seidel (1994) similarly reports interspecific trophallaxis solicited by Camponotus in a Camponotus–Crematogaster parabiosis in Western Malaysia. This behaviour may be advantageous for Camponotus, but its importance is difficult to judge.

Nest architecture and the foundation of parabiotic nests

Among the parabiotic nests we opened, only two contained brood of both species. They were separated from each other, and the two species did not care each other’s brood. Thus, parabiotic nests clearly differ from social parasites where host and parasite brood are communally cared for (Hölldobler & Wilson 1990). The remaining nests contained workers and often queens and brood of Cr. modiglianii, but only workers of Ca. rufifemur. In addition, monospecific shelters with only Cr. modiglianii workers (and sometimes brood) were found, suggesting that this species is polydomous.

These observations suggest that Cr. modiglianii initiates nests, which are subsequently colonized by Ca. rufifemur. Like other Crematogaster species (Tschinkel 2002; Longino 2003), Cr. modiglianii readily colonizes hollow structures such as certain lianas (e.g. Uncaria sp.) or cavities made by wood-boring insects (such as the cerambycid beetle Apriona flavescens). In many cases, Cr. modiglianii most likely actively expands these cavities like other extensively wood-excavating Crematogaster species (e.g. Cr. difformis, N.B. personal observation). As Ca. rufifemur has been shown to follow artificial trails of Cr. modiglianii (T. Pokorny, unpublished data; Menzel 2009), it may reach Cr. modiglianii nests by following their trails. This hypothesis accounts for the occurrence of large Cr. modiglianii nests with only few Ca. rufifemur workers present. Camponotus rufifemur may initially profit from cavities built by Cr. modiglianii, but also provide additional nesting space when excavating wood itself. Each ant species may hence profit from the other’s cavity-building activities, albeit at different stages of the parabiotic nest.

Parabiotic nests were mostly located in the trunks of living trees, with a broad range of both tree species and trunk diameter. Non-parabiotic Cr. modiglianii nests were sometimes located in living trees, but often in smaller, dead branches that hung in the understorey. Only few nests were in branches on the forest floor, probably because these decay quickly. That parabiotic nests occurred mainly in living trees and only rarely in dead branches may hence be due to the larger spatial requirements of both species nesting together. Moreover, since Ca. rufifemur, like almost all congeneric species, has no metapleural glands (Hölldobler & Wilson 1990), it may be more vulnerable to microbial infections and thus avoid nesting in rotting wood, unlike Cr. modigianii.

Defensive abilities of Ca. rufifemur

In our nest defence assays, the presence of Ca. rufifemur workers significantly raised the probability of successfully repulsing ant intruders, compared to when Cr. modiglianii alone was present. Although mostly nocturnal, Ca. rufifemur also defended the nest entrances during the day. Crematogaster modiglianii probably profits from its partner’s nest defence although it is capable of repulsing intruders as well. This species may also profit from Ca. rufifemur’s defence of food resources. Other dominant ants in the same habitat, e.g. Dolichoderus thoracicus, Cr. inflata and Cr. difformis, are individually stronger than Cr. modiglianii, but inferior to Ca. rufifemur (Menzel et al. 2008b, unpublished data). In the absence of the latter, these species may thus displace Cr. modiglianii from food sources.

Interactions with Poikilospermum epiphytes

Poikilospermum cordifolium grew in the nest entrances of many parabioses and non-parabiotic Cr. modiglianii nests. In cafeteria experiments, both ant species were attracted to its seeds and perianths and carried them into the nest. Thus, they dispersed the seeds and placed them into suitable sites. This represents an important benefit for P. cordifolium although seedlings and saplings at parabiotic nests suffered a high mortality. As a second benefit to P. cordifolium, the parabiotic ants might deter herbivores, as in numerous other plant–ant associations (Heil & McKey 2003). However, as P. cordifolium plants at parabioses did not show significantly less leaf damage than those on trees without ant nests, this potential effect was not confirmed in our study.

The ants, in turn, benefit from nutritious extrafloral nectar of P. cordifolium, which contains most essential amino acids, and probably from nutrients of the elaiosome-bearing seeds and perianths (Gammans, Bullock & Schönrogge 2005). In ant-garden parabioses of the neotropics and on the Malay peninsula, the epiphytes stabilize the nest and are thus crucial for nest maintenance (Yu 1994; Weißflog 2001; Kaufmann 2002). In both regions, the Camponotus partner plants their seeds and thus provides an indirect benefit to Crematogaster via the epiphytes. This benefit is confined to nutrition in the parabioses studied here, since they are located within trunks, and P. cordifolium does not provide any additional nest space.

The congeneric, syntopic Poikilospermum suaveolens was never found at parabiotic nests although its seeds were as attractive to ants as those of P. cordifolium. In contrast to the parabiotic ants, P. suaveolens seems to prefer open, sun-exposed sites such as gaps or the canopy layer (F.M. personal observation). We therefore suggest that the absence of P. suaveolens at parabiotic nests is not due to differential ant preferences but rather to different habitat requirements.

Comparison with neotropical parabioses

Neotropical parabioses are ant-gardens, i.e. they consist of free-hanging carton nests stabilized by the roots of epiphytes which grow in the carton. Usually, they are inhabited by Camponotus femoratus Fabr. 1804, which initiates the ant-gardens by actively planting epiphyte seeds into the carton material. Its parabiotic partner, Crematogaster limata agg. (including Cr. levior, Longino 2003), is too small to carry epiphyte seeds and joins the ant-garden at a later stage (Davidson 1988; Orivel & Dejean 1999). Neotropical parabioses are presumably mutualistic as well, albeit for partly different reasons (Vantaux et al. 2007). Similar to Southeast Asian parabioses, in neotropical ones Ca. femoratus seems to take advantage of its Crematogaster partner’s foraging activity by following its trails. Crematogaster, however, is usually chased away from resources and thus suffers a cost from its partner’s informational parasitism (Swain 1980; Davidson 1988). However, Ca. femoratus also delivers benefits to Cr. limata agg. by defending the nest (Wheeler 1921) and, most importantly, by providing nest space through the construction of ant-gardens (Davidson 1988; Orivel & Dejean 1999).

Conclusion: Southeast Asian parabioses – a mutualistic association?

Our studies suggest that the parabiotic association is beneficial for both ant species, as well as for P. cordifolium (Fig. 6). Camponotus rufifemur benefits from Cr. modiglianii’s information on food sources and possibly from interspecific trophallaxis. Possible costs that Cr. modiglianii may experience from having its food sources exploited by Ca. rufifemur may be outweighed by the latter’s defence of nest and food resources. Hence, the mutual services in the parabiosis are essentially nutrition and protection. These are among the services most commonly traded between mutualists, e.g. in ant–plant protection mutualisms or trophobiotic ant–aphid interactions (Bronstein & Barbosa 2002; Heil & McKey 2003; Stadler & Dixon 2005). It remains unknown how important these benefits actually are for the parabiotic ants, and how they translate into reproductive success. Moreover, depending on local conditions such as enemy pressure and food or nest site availability, the net outcome of costs vs. benefits may be highly variable (Pontin 1978; Bronstein 1994; Johnson et al. 1997; Bronstein & Barbosa 2002). However, Ca. rufifemur appears more dependent on its partner than vice versa since we never found non-parabiotic nests of this species, whereas Cr. modiglianii frequently nested without its partner.

How the parabiotic life form evolved is so far unknown. It may have evolved from trail-sharing or other, less intimate associations. In the parabiosis studied here, interspecific trail-following by Ca. rufifemur may be a key factor in the association. By following Cr. modiglianii trails, Ca. rufifemur reaches not only food sources but also Cr. modiglianii nests. Provided that Ca. rufifemur queens also follow Cr. modiglianii trails, this may be a mechanism by which parabiotic nests are initiated. Notably, both species possess unusual cuticular hydrocarbons that favour interspecific tolerance, which may have been important in the evolution of the parabiotic lifestyle (Menzel, Blüthgen & Schmitt 2008a; Menzel et al. 2008b). Moreover, Cr. modiglianii produces so far unidentified cuticular compounds which reduce aggression in Ca. rufifemur (Menzel 2009). Thus, it seems plausible that preadaptations to facilitate parabiotic associations include an ability to perceive and follow heterospecific trails, as well as a nestmate recognition system that allows high interspecific tolerance.


Ants were kindly identified by Seiki Yamane (Kagoshima University, Japan) and Heike Feldhaar (University of Würzburg, Germany). Wolfgang Schawaller (Staatl. Museum für Naturkunde Stuttgart) and Arthur Chung identified the beetle guests, the myrmecophilous crickets were determined by Sigfrid Ingrisch and the hemipteran trophobionts by Dirk Mezger. Tree species were identified by Bernadus Bala Ola (Danum Valley Field Center). Heike Feldhaar provided valuable ideas for interpretation of the data. We thank Waltraud Wetzel, Michael Staab and Christian Peter for their valuable assistance in the field. We are also grateful to the Danum Valley Management Committee (DVMC) and Malaysian Economic Planning Unit (EPU) for permission to conduct research, the Royal Society’s South East Asia Rainforest Research Programme (SEARRP) and the staff at Danum Valley Field Center, especially Bernadus ‘Mike’ Bala Ola and Alex Karolus, for logistic support and help in the field. Arthur Chung (Forest Research Center, Sepilok) kindly supported our work as local collaborator. Furthermore, we would like to thank the Sarawak Forestry Department for research permission in Gunung Mulu National Park and Brian Clark for logistic help at Mulu. Two anonymous reviewers gave valuable suggestions to improve the paper. This research was supported by the Sonderforschungsbereich ‘Mechanisms and Evolution of Arthropod Behaviour’ (SFB-554) of the German Research Foundation (DFG) and complies with current laws of Malaysia and Germany. F.M. was supported by a doctoral fellowship provided by the Bayerische Graduiertenförderung (BayEFG) and the Studienstiftung des deutschen Volkes.