Interactions between weaver ants Oecophylla smaragdina, homopterans, trees and lianas in an Australian rain forest canopy


  • Nico Blüthgen,

    Corresponding author
    1. Lehrstuhl Tierökologie I, Universität Bayreuth, D-95440 Bayreuth, Germany
      Nico Blüthgen, Lehrstuhl Tierökologie I,Universität Bayreuth, D-95440 Bayreuth, Germany. E-mail:
    Search for more papers by this author
  • Konrad Fiedler

    1. Lehrstuhl Tierökologie I, Universität Bayreuth, D-95440 Bayreuth, Germany
    Search for more papers by this author

Nico Blüthgen, Lehrstuhl Tierökologie I,Universität Bayreuth, D-95440 Bayreuth, Germany. E-mail:


  • 1Tritrophic interactions between the weaver ant Oecophylla smaragdina (Fabricius) (Hymenoptera: Formicidae), plants and honeydew-producing trophobionts (Hemiptera: Sternorrhyncha and Auchenorrhyncha, Lepidoptera: Lycaenidae) were studied in a rain forest canopy in Northern Queensland, Australia.
  • 2Most commonly attended trophobionts by O. smaragdina at this study site were Coccidae ( Coccus sp., Milviscutulus sp.) and Membracidae ( Sextius sp.), followed by Toxoptera aurantii (Aphidae), Planococcus citri (Pseudococcidae), Icerya sp. (Margarodidae), an unidentified species of Eriococcidae, Austrotartessus sp. (Cicadellidae), and lycaenid butterfly larvae ( Anthene seltuttus , Arhopala centaurus group).
  • 3Most trophobionts were highly polyphagous, and trees and lianas from many plant species and families acted as homopteran hosts. However, lianas were found to play a key role. First, the majority (68%) of aggregation sites was found on lianas, especially on the legumes Entada phaseoloides and Caesalpinia traceyi , and secondly, per capita ant visitation rate ( VR ) at coccoids was significantly higher on lianas compared to trees. In total, VR to homopterans was 64% higher on lianas.
  • 4Sites of ant–homopteran aggregations were regularly replaced by new locations on fresh plant growth. The mean longevity of nests of this polydomous ant species was 131 days, of individual aggregation sites with membracids 54 days and with coccoids 130 days.
  • 5Our results suggest that plant-specific differences in suitability for honeydew production (especially the availability of lianas) and the availability of preferred trophobionts have a strong influence on the vigour of Oecophylla colonies.


Weaver ants of the genus Oecophylla are among the most dominant and important ants in tree canopies of the humid tropics of Africa (O. longinoda (Latreille)), as well as in South-East Asia, Australia and western Pacific islands (O. smaragdina (Fabricius)). In Australia, the distribution of the latter species is confined to tropical forests or woodlands in the northern and north-eastern parts of the country (Lokkers 1986). Oecophylla ants establish huge colonies comprising several tens or hundreds of thousands of workers. Their polydomous colonies cover territories that include the crowns of several trees (Hölldobler 1983; Peng, Christian & Gibb 1998). Weaver ants use their silk-producing larvae to build nests from leaves spun together (Hölldobler & Wilson 1990). Leaf ‘pavilions’, where ants and trophobionts are sheltered, but no brood is raised, are constructed in the same way (Way 1963).

Oecophylla colonies play a key role in the ecosystems in which they occur. First, their competitive dominance over many other ant species affects the entire arboreal ant community. Oecophylla colonies defend mutually exclusive territories against conspecific colonies or competing ant species, while permitting co-occurrence of certain other ant species ( Hölldobler 1983 ). Secondly, weaver ants are substantial predators of other arthropods of various sizes ( Hölldobler 1983 ; Dejean 1990 ), resulting in significant reductions in various insect pests (Australia: Peng, Christian & Gibb 1997 ; Malaysia: Way & Khoo 1991 ). Thirdly, trophobiotic interactions between weaver ants and honeydew-producing insects are abundant and diverse. They involve various homopterans and lycaenid butterflies ( Way 1963 ; Hölldobler & Wilson 1990 ), some of which are specific to weaver ants ( Seufert & Fiedler 1996 ; Eastwood & Fraser 1999 ). All those effects may potentially translate into increased plant performance as result of reduced herbivory ( Messina 1981 ).

While a strong impact of weaver ants on the canopy flora and fauna seems undisputed, we do not yet know which factors control abundance and distribution of these ants in natural rain forest canopy, and the role of food resources has not been addressed so far. Prey is generally scarce in the rain forest canopy in relation to highly abundant ants (Stork 1991; Floren & Linsenmair 1997). Resources derived from plant sap (particularly honeydew) are more abundant and predictable and may explain the success of large arboreal ant colonies with their high energetic demands (Tobin 1994; Davidson 1997; Blüthgen et al. 2000). One should therefore expect that composition, abundance and distribution of the herbivorous trophobionts, which in turn may depend on the local flora, will influence the success of Oecophylla colonies (bottom-up effects).

Most studies on predation and trophobiosis by weaver ants have been performed in plantations, but the patterns may be substantially different in more complex and species-rich natural ecosystems. The coastal Australian rain forest is characterized by a particularly high abundance of lianas (Tracey 1982). Lianas may therefore play an important role in providing plant-derived sources, such as honeydew or extrafloral nectar, for huge weaver ant colonies. However, their importance to weaver ants has never been addressed explicitly, and is therefore the subject of the current investigation.

The objectives of our study were to:

  • 1investigate the effect of plant life forms (trees vs. lianas) and composition of homopteran fauna on the intensity of trophobiosis with Oecophylla ;
  • 2analyse spatial and temporal dynamics of honeydew use by these highly dominant ants; and
  • 3compare the host plant specificity of nests and honeydew feeding sites.

Materials and methods


This study was carried out using the Australian Canopy Crane in Far North Queensland, Australia (16°07′S, 145°27′E, 80 m a.s.l.). The study area contains lowland rain forest characterized by a high abundance of lianas and an average canopy height of 25 m (complex mesophyll vine forest, Tracey 1982). Average rainfall is about 3500 mm per year, 60% of which occurs in the wet season between December and March, and the mean daily temperature is ranging from 22 °C (July) to 28 °C (January) (Turton, Tapper & Soddell 1999). During the study period between September 1999 and May 2001, the forest was in an early stage of recovery from cyclone ‘Rona’ in February 1999, which damaged large parts of the canopy. The canopy crane was 48·5 m tall, with a jib length of 55 m, which allowed us to directly examine the canopy in a 0·95-ha forest area. Results reported here relate to this area covered by the jib and its immediate surroundings.


Using the canopy crane, homopterans and tending ants were examined on 30 trees with Oecophylla nests, including six tree crowns where a complete count was attempted and repeated during different months in order to reveal temporal changes.

‘Aggregation sites’ (AS) were defined as ant–homopteran sites on a single host plant species which included at least one homopteran and one ant showing characteristic trophobiotic behaviour, and which was separated from the next such spot by at least 20 cm. For each AS, we counted homopteran nymphs and adults greater than about 1 mm length and ants in their immediate vicinity. We also recorded the host tree or liana species on which homopterans were feeding. Counts at each AS were completed within c. 3 min. Overall, counts were spread over the period from 8 a.m. to 5 p.m., thus covering much of the daylight hours. Pavilions sheltering homopterans were carefully opened for inspection, but not nests.

The momentary per capita visitation rate (VR) was defined for each aggregation site as:


Estimates of the number of homopterans on the tree by extrapolation of counts on a few twigs (Blüthgen et al. 2000) were avoided, as variability between branches and between trees and lianas was very pronounced. Due to the relative openness of the canopy in the study plot after the cyclone, large parts of 11 tree crowns were easily accessible for observations, so surveys were regarded as more or less complete. However, numbers of ants and homopterans are only ‘snapshots’ of the ant colony activity – they neither include ants that have been running between feeding sites and nests during the census, nor do they cover visits of homopterans on adjacent unaccessible trees. Also, diurnal changes in visitation patterns could not be addressed by this means of recording.

Homopterans collected from numerous AS were identified by taxonomic specialists (see Acknowledgements) in order to obtain information about the specificity of their relationship with ants. Because Coccoidea (including Coccidae, Pseudococcidae, Eriococcidae, Margarodidae) were neither collected from all AS nor distinguished in the field, they were pooled and referred to as ‘coccoids’ in the following analyses.


Oecophylla nests in the study area were recorded in regard to host tree, identity of leaves used and diameter of the nest [estimated to the nearest 5 cm; ‘diameter’ in oval nests = (length + width)/2]. A complete survey of the site was not attempted.

On six trees, locations of active nests and AS were labelled with plastic tape and checked 1–3 times after a time span (t) of 19–204 days (n= 9 surveys of nest sites vs. 12 surveys of AS).

The mean establishment rate (E) and abandonment rate (A) of one nest or AS was obtained as:

E  =  n /Σ (number of nests or AS not previously recorded/ t ),
A  =  n /Σ (number of previously active nests or AS that were abandoned/ t ).

Assuming a constant ‘mortality’ (i.e. proportional abandonment) (M) of nests and AS over time, the (more informative) mean longevity of active nests or AS (L) was calculated as:

L  =  n /Σ ( M )


M  = (number of abandoned nests or AS/number of previously active nests or AS)/ t .


VR was normalized for one-way ancova by log-square-root transformation. Normality of VR was confirmed using Kolmogorov–Smirnov tests, and variance homogeneity by Levene tests. Number of homopterans (untransformed) was used as the covariate. For statistical comparison of VR between trees and lianas only those trees were considered that carried lianas in the crown, while other tests were performed on all trees. Labelling of the location of AS on six trees showed that only a minority of AS (total 18%) were identical between different surveys. The high dynamics and the relatively short life span of homopteran nymphs in relation to intervals between surveys led us to pool AS and VR from all surveys.



During the entire study, no Oecophylla colony was found that did not attend homopterans, and all major trails of this ant in the canopy lead to aggregation sites with homopterans (AS).

A total of 490 AS was recorded on the 30 trees studied (including repeated censuses on six trees). Most AS involved membracids or coccoids, while associations with aphids, cicadellids, and lycaenids were uncommon (Table 1). The median number of individuals per AS was highest for coccoids (22) and aphids (11), followed by membracids (6), cicadellids (1·5) and lycaenids (1). Mixed aggregations of coccoids and membracids (4 AS), or coccoids and lycaenids (1 AS), were rare. Homopteran species are listed in Table 2 (note that all membracids, aphids and cicadellids represented a single species each).

Table 1.  Distribution of aggregation sites (AS) of Oecophylla smaragdina tending different homopterans and lycaenids on trees and climbing plants. Data are considered for a total of 30 trees (including eight trees without lianas). Numbers following the tree species correspond to tree labels of the crane site. Cell entries are numbers of AS with the number of investigated tree canopies in brackets. Total numbers of AS and trees are smaller than pooled numbers where different homopterans or food plants co-occurred on the same AS vs. the same tree, respectively. Species of homopterans and lycaenids are listed in Table 2
Homopteran host plant species1Coccoidea2MembracidaeAphidaeCicadellidaeLycaenidae3Total
  • 1

    Plant families: Apocynaceae (APOC), Caesalpiniaceae (CAES), Convolvulaceae (CONV), Flagellariaceae (FLAG), Lauraceae (LAUR), Menispermaceae (MENI), Mimosaceae (MIMO), Moraceae (MORA), Myristicaceae (MYRI), Myrtaceae (MYRT), Sapindaceae (SAPI), Proteaceae (PROT).

  • 2

    Only partly collected. C = Coccus sp. (Coccidae), Co = Coccidae (immature), E = Eriococcidae, M = Milviscutulus sp. (Coccidae), P = Planococcus citri Risso (Pseudococcidae), I = Icerya sp. (Margarodidae).

  • 3

    Ar = Arhopala centaurus group, An = Anthene seltuttus (Röber) (lycaenid on #1004 not collected).

Trees143 (10)  3 (2)6 (4)2 (2)3 (3)156 (15)
 Acmena graveolens L.S. Smith   (MYRT) #3032  4 (1) M  –  4  (1)
 Cardwellia sublimis F. Muell.   (PROT) #3028 #4025 #5027 11 (3) C  –2 (2) 13  (3)
 Cryptocarya hypospodia F. Muell.   (LAUR) #7066 13 (1) Co  –2 (1) 15  (1)
 Endiandra microneura C.T. White   (LAUR) # 2002 68 (1) C M E  –1 (1)1 (1) Ar 69  (1)
 Endiandra cf. monothyra B.P.M.   Hyland (LAUR) #1004 13 (1) C M  2 (1)1 (1) 16  (1)
 Myristica insipida R.Br.   (MYRI) #1059  7 (1) M  1 (1)  8  (1)
 Synima cordierii Radlk.   (SAPI) #7094  –  –1 (1) An  1  (1)
 Syzygium cormiflorum B.P.M. Hyland   (MYRT) #5034 19 (1) C M  – 19  (1)
 Syzygium sayeri B.P.M. Hyland   (MYRT) #3023  5 (1)  –  5  (1)
 Unident. #1023  –  –1 (1)  1  (1)
 Unident. #3034  –  –  2 (1)  2  (1)
 Unident. #7072  2 (1)  –  2  (1)
 Unident. #7093  1 (1)  –  1  (1)
Climbing plants 49 (15)289 (17)334 (22)
  Caesalpinia traceyi L. Pedley   (CAES)  3 (3) C196 (4)199  (4)
  Entada phaseoloides Merrill   (MIMO) 28 (9) C P 90 (13)114 (14)
  Merremia peltata Merrill (CONV)  8 (4) M  1 (1)  9  (4)
  Melodinus australis Pierre (APOC)  5 (1) M  5  (1)
  Stephania japonica Miers (MENI)  1 (1)  1  (1)
  Ficus pantoniana King (MORA)  1 (1) I  1  (1)
  unident.  2 (1)  2 (2)  4  (3)
  Flagellaria indica Linn. (FLAG)  1 (1)  1  (1)
Total plants188 (19)288 (17)6 (4)2 (2)3 (3)490 (30)
Table 2.  Trophobiotic partners of Oecophylla smaragdina in this study
  • 1

    Genus in need of revision (Max Day, personal communication).

CoccidaeCoccus sp. Milviscutulus sp.
PseudococcidaePlanococcus citri Risso
MargarodidaeIcerya sp.
AphidaeToxoptera aurantii (Boyer de Fonscolombe)
MembracidaeSextius cf. ‘ kurandae1
CicadellidaeAustrotartessus sp.
LycaenidaeAnthene seltuttus (Röber)
 Arhopala centaurus group

Homopteran taxa were very unequally distributed among host plant species and life forms (Table 1). Nearly all AS with membracids (99%) were found on lianas, especially Entada phaseoloides and Caesalpinia traceyi. In contrast, most coccoids (74% of AS) were consuming sap from trees. The few AS with aphids, cicadellids and lycaenids were found exclusively on trees. The unequal distribution of AS with coccoids vs. membracids on trees vs. climbing plants was highly significant (χ2 = 296·1; d.f. = 1; P < 0·0001). Many coccoid AS (82 of 188 AS) occurred on trees without suitable lianas in the crown area. Considering only those trees that carried lianas, distribution of coccoid AS across plant growth forms was much more even (61 vs. 45 AS, respectively) and did not significantly deviate from an equal distribution (χ2 = 1·4; d.f. = 1; P= 0·24).

Between 10 and 2115 homopterans (median: 455) were counted on each of 11 tree crowns where we were able to examine most parts of the crown (n = 26 surveys). These homopterans were tended by a similar number of Oecophylla workers (median: 449, range: 20–1218). Numbers of homopterans and attended ants per tree were significantly correlated (Pearson's r = 0·79; P < 0·001). Additional coccoids were usually tended inside the nests, but were not counted here as this would have required nest destruction. A closer inspection of a single nest that had been previously abandoned by Oecophylla revealed 895 scale insects. Numbers of homopterans and ants were highly variable between trees and different surveys (months), although no clear seasonal pattern could be found and ranking of trees was not stable (Fig. 1a). It is unclear how much these results were affected by post-cyclone succession. Ants attended homopterans more continuously on lianas throughout the year, while homopteran attendance occasionally dropped to zero on trees (Fig. 1b,c).

Figure 1.

(a–c). Total number of Oecophylla smaragdina workers attending homopterans outside their nests during five surveys between October 1999 and January 2001: (a) on six trees incl. lianas (2–5 surveys per tree). Tree numbers (see Table 1) are: #1004 (open squares), #2002 (filled squares), #3023 (open circles), #5027 (open triangles), #5034 (filled circles) and #7066 (filled triangles); (b) on liana vs. tree as homopteran host plant on #1004 and (c) on tree #5027. Smooth curves between the data are plotted for the purpose of readability.

Homopterans varied strongly in aggregation size. There was a significantly higher number of coccoids (mean ± SEM: 59·1 ± 9·2; n = 182) than membracids per AS (10·5 ± 0·9; n = 286) (Mann–Whitney U = 11007, P < 0·0001).

Ant visitation rate (VR) was negatively correlated with the number of homopterans (H) present at an AS. On a log-log plot (Fig. 2), the relationship was linear. The negative regression was significant for both AS with coccoids (Hc) (log VR = −0·22 log Hc + 1·4; r2 = 0·39; n = 174; P < 0·0001) and membracids (Hm) (log VR = −0·25 log Hm + 1·5; r2 = 0·23; n = 277; P < 0·0001). There was no significant difference between the regression slopes for these two homopteran taxa (t = 0·84; d.f. = 447; P = 0·40). These findings lead us to incorporate H as covariate in the following analysis of the effect of plant life form on VR, with data from coccoids and membracids combined.

Figure 2.

Relationship between ant visitation rate ( VR ) and total number of homopterans per aggregation site, plotted on log-log scale. Coccoids are represented by open circles, membracids by closed squares.

Ant recruitment to trophobionts was significantly affected by host plant life form. Considering only trees that carried lianas, VR was 64% higher to homopterans that were feeding sap from lianas compared to trees (Fig. 3). This effect was highly significant (ancova: F1;367 = 10·6, SSeffect = 0·97, SSerror = 0·05; P = 0·0016), using H as covariate. The effect on VR was slightly weaker, but still highly significant when only coccoids were considered (52% higher on lianas), when all trees were included irrespective of liana presence (45% higher on lianas), or when the covariate was not included.

Figure 3.

Visitation rate ( VR = number of Oecophylla smaragdina workers per individual homopteran), compared between lianas vs. trees (where lianas available) as homopteran hosts (mean ± SEM). Mean values vary significantly between plant life-forms ( ancova , see text).

Mean VR (± SEM) to coccoids on lianas was 1·3 (± 0·2). There was no difference between the two legume species (E. phaseoloides and C. traceyi) and non-legume lianas (ancova: F1;35 = 0·33, SSeffect = 0·01, SSerror = 0·03; P = 0·57). On two coccoid host trees without lianas (which were excluded from the preceding analysis), VR was higher on one tree (Endiandra microneura; mean VR ± SEM = 2·0 ± 0·3; n= 68) and lower on the other (Cryptocarya hypospodia; 0·6 ± 0·1; n = 13). Among trees with lianas, mean VR at coccoids was 1·2 (± 0·4; n = 19) on one Syzygium cormiflorum and lower on the remaining six trees (mean VR ranging between 0·5 and 0·8). Per capita ant recruitment to membracids varied significantly between the legumes E. phaseoloides (3·2 ± 0·4; n = 79) and C. traceyi (2·2 ± 0·1; n = 191) (ancova: F1;267 = 5·3, SSeffect = 39·6, SSerror = 7·4; P = 0·02), but was generally higher compared to coccoids.

Locations of Oecophylla–homopteran aggregations were highly dynamic. The mean (± SEM) establishment rate of new AS per tree was one AS per 6·8 (± 1·7) days vs. 2·6 (± 1·2) days (coccoids vs. membracids, respectively), while one AS per 15·9 (± 5·3) days vs. 3·6 (± 1·2) days was abandoned. Aggregations with coccoids were less dynamic than those with membracids. Mean longevity of AS with coccoids was 130 (± 14) days and only 54 (± 10) days with membracids. The difference in ‘mortality’ (M) of AS with coccids and membracids was significant (t = 2·6; d.f. = 18; P < 0·05; M log-square-root transformed).

Oecophylla constructed pavilions with leaves that were woven together with larval silk in a similar way as nests. Number of pavilions per tree varied (mean: 4·6; range 0–18; n = 26 surveys on 11 trees). Pavilions were usually less than 10 cm in diameter, and therefore considerably smaller than nests of mature colonies. Between one and 15 leaves (median: 3; counted for n = 59 AS) were incorporated in each pavilion. They were found on young foliage of trees and lianas where many homopterans were tended by ants, often several meters away from the nearest ant nest. Plants with large leaves or leaflets (> 5 × 5 cm) were used more frequently for pavilion construction ( E. phaseoloides : 34% of AS; other lianas: 33%, trees: 36%) than C. traceyi (15%), a vine with relatively small leaflets (< 2 × 3 cm). Forty-one per cent of AS with coccoids and 19% of AS with membracids were sheltered within pavilions. Both factors were interrelated, as most membracids occurred on C. traceyi (see Table 1 ). On one occasion some ant larvae were found inside the pavilion, which are necessary for the production of silk. Some pavilions were repaired by ants after damage during the census. Pavilions were also abandoned by ants during maturation of the plant tissue in the same way as nonsheltered AS, and mean (± SEM) longevity was 108 (± 15) days (longer than mean longevity of AS with membracids, similar to AS with coccoids).


Oecophylla nests were found on 39 trees from over 18 species and eight families of plants (six trees unidentified). Nests were woven from tree or liana leaves (tree leaves only: 24 trees, tree and liana leaves: 11 trees, liana leaves only: four trees). A mixture of tree and liana leaves was used either in separate nests or incorporated in the same nest. The most commonly used liana was Merremia peltata (Convolvulaceae). Most common host trees were Acmena graveolens and Syzygium sayeri (Myrtaceae) (each species represented by four trees inhabited by Oecophylla ), Endiandra microneura (Lauraceae), Cardwellia sublimis (Proteaceae), Argyrodendron peralatum (Sterculiaceae), and Myristica insipida (Myristicaceae) (three trees each). All these common hosts were among the 18 most common and largest trees in the crane plot (each with at least 10 trees of d.b.h. ≥ 10 cm). Sizes of leaves or leaflets utilized by Oecophylla were ‘normal’, ranging from c . 5 × 8 cm to 20 × 20 cm (upper end: M. peltata ). The mean height of the 39 Oecophylla host trees was 23·2 m (± 1·1 m SEM), and significantly higher than the mean for the remaining trees (15·2 ± 0·2 m; 667 trees of d.b.h. ≥ 10 cm; Mann–Whitney U  = 3429, P < 0·0001). Many examples of common trees that were not recorded as hosts for Oecophylla nests were either bearing very large or tough foliage that was obviously unsuitable for nests, e.g. Alstonia scholaris R. Br. (Apocynaceae) and palms ( Normanbya normanbyi L.H. Bailey, Licuala ramsayi Domin), or they were relatively small understorey trees (height < 15 m), e.g. Cleistanthus myrianthus Kurz (Euphorbiaceae) and Antirhea tenuifolia B.D. Jacks (Rubiaceae).

Between one and five nests were found on each tree (median: 3), which were 10–50 cm (median: 30 cm, n= 34) in diameter. Nests were frequently abandoned by Oecophylla and replaced by new nests. New nests appeared at a rate of one nest per 56 (± 11) days (mean ± SEM), and nests were abandoned at a similar rate (52 ± 11 days). Mean nest longevity was 131 (± 21) days.



Our results confirm that for omnivorous weaver ants, apart from preying or scavenging on arthropods and harvesting extrafloral nectar, honeydew derived from various homopteran partners plays a crucial role. Honeydew and nectar may even represent the key resource to compensate the high energy requirements, especially the carbohydrate-driven metabolism of adult ants (Davidson 1997) of very large Oecophylla colonies in the canopy.

Honeydew production might exceed most floral and extrafloral nectar sources in terms of quantity as well as quality. Indirect evidence supporting this notion comes from the observation that the total number of Oecophylla workers tending homopterans was much higher than those tending nectaries at the same time (N. Blüthgen, unpublished data). In addition, at our Australian study site a much higher proportion of Oecophylla workers returned to their nest with expanded gaster (filled honey-crop) than with prey carried between mandibles (unpublished data). However, these counts could be misleading if haemolymph from insect prey were also carried in the honey-crop (important in red wood ants: Horstmann 1974).

As was recorded elsewhere (Way 1963), Australian Oecophylla ants attend a broad spectrum of homopterans and some lycaenids. Hence, trophobiotic associations are highly unspecific, yet non-random from the ants’ perspective. Many lycaenids attended by weaver ants (including the two species found here, Anthene seltuttus and one representative of the Arhopala centaurus group) are obligate myrmecophiles that associate exclusively with O. smaragdina (Eastwood & Fraser 1999). Another potential candidate for an obligate myrmecophile is the membracid Sextius cf. ‘kurandae’. Sextius species are commonly myrmecophilous (Buckley 1983; Day 1999). Both Planococcus citri and Toxoptera aurantii are facultative myrmecophiles that may or may not associate with ants (Way 1963; Bigger 1993; Mary Carver, personal communication). None of the trophobionts was found with any other ant species in the study area (except some Icerya and unidentified coccoids in the understorey). On the other hand, five ant species were very commonly found attending different homopterans that were not attended by Oecophylla. Both lycaenid species (Braby 2000) and most homopterans attended by Oecophylla in this study are highly polyphagous. Such opportunistic feeding may be a common character of many trophobiotic partners of Oecophylla (see Fiedler & Maschwitz 1989). Membracids (Sextius cf. ‘kurandae’), scale insects (Coccus, Milviscutulus), margarodids (Icerya sp.), citrus mealybugs (Planococcus citri) and aphids (Toxoptera aurantii) were each recorded on several host plant families in this study and/or elsewhere, the latter two species being worldwide pests in plantations (Way 1963; Carver 1978; Bigger 1993; Ben-Dov 1994). The cicadellid Austrotartessus was found on two Cardwellia sublimis trees, although food plants other than eucalypts and ant associations are largely unrecorded for the entire subfamily Tartessinae (Evans 1981).

Ant-attended homopterans were highly abundant, with a median of 455 homopterans and 449 attending ants per tree in this study. These values are slightly higher than those found in a Venezuelan forest canopy involving 16 ant species (median: 54 membracids or 300 coccoids per tree; Blüthgen et al. 2000), but considerably smaller than the values reported by Dejean et al. (2000) from a rain forest canopy in Cameroon (3–7 × 105 coccoids and the same number of attending Crematogaster depressa ants per tree). In Oecophylla, a high proportion of coccoids, perhaps even the majority, is attended inside the nest (see also Way 1963).

Nests and homopteran aggregation sites (AS) are highly dynamic and become replaced after a few months. Nests, pavilions and AS with coccoids had a similarly short durability in this study (108–131 days), and AS with membracids were even more short-lived (54 days). The high turnover may be a response to maturation of plant tissue and concomitantly diminishing honeydew productivity or quality (see Douglas 1993), so that fresh tissue is preferably colonized. Membracids are more mobile than largely flightless coccoids in this regard. We did not investigate whether the dynamics of AS can be attributed to replacement (e.g. mortality and predation) of individual homopterans, to their mobility, or to transport by ants. On some occasions, however, Oecophylla workers were observed to carry Sextius nymphs between their mandibles from one AS to another. Transport of trophobionts by O. smaragdina and O. longinoda is common (Way 1963; Fiedler & Maschwitz 1989).


The silk-woven pavilions described here (see also Way 1963) are probably analogous to the ‘barrack nests’ described by Hölldobler (1983) from an area not far from our study site. However, while Hölldobler (1983) suggested that their function was to house-guard ants along territorial borders as a defence force against intruders, their clustered and temporally restricted distribution on favourable feeding sites for homopterans (young shoots of trees and especially lianas), interspersed between open ant-homopteran aggregations rather than being concentrated along territorial margins or strategically important locations, indicates their relation with trophobiosis. We therefore conclude that their main use is based on trophobiotic interactions − thus they act as ‘dairies’ instead of ‘barracks’.

Such ‘dairies’ were also observed in O. smaragdina colonies in Malaysia, where they harboured lycaenid caterpillars and membracids (e.g. Fiedler & Maschwitz 1989), and seem to be a common characteristic of these ants. They may be functionally analogous to pavilions built by other ants using silk, debris or carton (e.g. Maschwitz, Dumpert & Schmidt 1985; Liefke et al. 1998; Blüthgen et al. 2000). All these structures may protect trophobionts and ants against rain and other environmental influences, as well as against competitors or parasites (Way 1963; Gibernau & Dejean 2001).


Weaver ant nests occurred on a broad spectrum of plant species. Many, perhaps the majority, of plant species with foliage of normal density and softness, including all common trees or lianas, seem suitable for nest building. Most trees inhabited by Oecophylla carried lianas, often with large amounts of foliage in the crown. Some liana species (especially the legumes E. phaseoloides and C. traceyi) were obviously preferred over trees as feeding ground for ant-tended homopterans. It is difficult to disentangle the effects of life form (lianas) and taxonomy (legumes) in this study, as none of the legume trees (e.g. the common species Castanospermum australe A. Cunn. & Fraser) hosted ant-tended homopterans. The preference for legume lianas does not seem to be caused primarily by host plant restrictions on the homopterans involved, because all common homopterans were highly polyphagous.

Instead, we believe that this preference indicates two non-exclusive factors:

  • 1A higher quantitative honeydew output of homopterans on lianas. This may be one of the major causes for the higher ant visitation rate to homopterans on lianas compared to trees (including effects within coccoids and a higher VR at membracids which are more abundant on lianas).
  • 2A higher honeydew quality. Legumes are especially frequent hosts for ant-tended homopterans, possibly as an effect of nitrogen fixation through symbiotic Rhizobium bacteria in root nodules. For lycaenid caterpillars in America and Australia, Pierce (1985 ) found that myrmecophily is particularly common on nitrogen-fixing legumes, although this pattern was challenged by a more detailed analysis ( Fiedler 1995 ).

Lianas may have a more continuous production of new vegetative tissue than trees, which are more seasonal (Putz & Windsor 1987; but see Hegarty 1990). The availability of fresh tissue for homopterans may be crucial for a high quality and/or quantity of honeydew flow, and thus be often higher on lianas than on trees. Consequently, ant–homopteran interactions in this study were found to be more continuous and intense on lianas than on trees, where they may temporally disappear. The high mobility and turnover of the ant–homopteran aggregations observed in our study may be a potent strategy for maximizing honeydew yield by tracking resources in space and time within the ants’ territory.

Territoriality in weaver ants may thus be primarily caused by the necessity to maintain a sufficient supply of honeydew. The availability of honeydew may also affect the size of territories: larger territories may be required where the density of high-quality homopteran substrates is low or temporally restricted. The size of O. smaragdina territories varies considerably from a single tree to over 20 major trees (Hölldobler 1983). Territories in the relatively open, liana-rich forest of the study site are restricted typically to relatively few (often 2–3) tree crowns. In South-East Asia, this ant species is much more common in disturbed forests or along forest edges, where again lianas are more prevalent (unpublished observations). Forest disturbance through cyclones is not uncommon in northern Australia and is usually followed by rapid regrowth of trees and lianas. The study site represents a typical recovering forest after the severe impact of a cyclone during February 1999. This succession may benefit the nutrition of Oecophylla colonies, although immediate short-term impacts of cyclones on Oecophylla nests are severe (Begg 1977). We therefore expect that there are close relationships between forest succession, density of homopteran populations, ant territoriality and territory size, but such conclusions await carefully designed studies, as multiple factors may be involved.

If ants protect trees efficiently against herbivory, the presence of suitable lianas may cause a substantial benefit for the host tree through the attraction of high numbers of resident or visiting ants that indirectly benefit the tree. At the same time, costs of harbouring large numbers of homopterans (loss of phloem sap, risk of infections) are kept to a minimum as long as homopterans occur preferentially on lianas. It would be interesting to investigate whether this pattern could potentially counterbalance costs of liana loads and shading on trees (Stevens 1987). Climbing plants (particularly legumes) could also be considered in management of plantations where Oecophylla is successfully used as biological control agent (Way & Khoo 1991; Peng et al. 1997), but where plants may be much more susceptible to homopteran-transmitted infections than in natural forest systems, so that the net effect of ant–homopteran interactions is often viewed as detrimental (James, Stevens & O’Malley 1997).

Our results emphasize that trophobiotic associations with honeydew producing herbivorous insects are a key component in the foraging of Oecophylla ants. There is much variation in the intimacy of relationship towards various honeydew producers. Moreover, trophobiotic aggregations on lianas turned out to be more predictable over the course of the year and received higher per-capita attendance through ants. Due to the predaceous capacity of weaver ants, these patterns have the potential to contribute to ecosystem-wide effects such as the balance between trees and lianas. It remains to be uncovered which characters of honeydew quantity and quality are responsible for the patterns documented here, and how interactions with other ants and nutrient sources contribute to determining spatial and temporal representation of these dominant arboreal ants.


We thank Nigel Stork (Rainforest CRC) for continuous support and initiating the canopy crane project, and two anonymous reviewers for constructive criticism of an earlier manuscript draft. Coccoids were identified by John Donaldson (DPI Indooroopilly), aphids by Mary Carver (ANIC Canberra) and membracids by Max Day (ANIC Canberra). Trees were identified by Andrew Small and Bob Jago (Cairns). Identification of lianas was supported by Bruce Gray and Bernie Hyland (CSIRO Atherton). Martin Freiberg (University of Ulm) kindly provided tree height measurements. The Studienstiftung des deutschen Volkes provided a scholarship to NB.