Austral Ecology (2007) 32, 93–104 10.1111/j.1442-9993.2007.01744.x Ant mosaics in a tropical rainforest in Australia and elsewhere: A critical review NICO BLÜTHGEN1* AND NIGEL E. STORK2† Department of Animal Ecology and Tropical Biology, University of Würzburg, Biozentrum, D-97074, Germany (Email: [email protected]);and 2Rainforest Cooperative Research Centre at James Cook University, Cairns, Queensland, Australia 1 Abstract The concept of ‘ant mosaics’ has been established to describe the structure of arboreal ant communities in plantations and other relatively simple forest systems. It is essentially built upon the existence of negative and positive associations between ant species plus the concept of dominance hierarchies. Whether this concept can be applied to ant communities in more complex mature tropical rain forests has been questioned by some authors. Here we demonstrate that some previous attempts to prove or disprove the existence of such ant mosaics sampled by knockdown insecticide canopy fogging in near pristine tropical forests may have been thwarted by poor statistical power and too coarse spatial resolution, and the conclusions may be highly dependent on ant species and forest stratum selected for the study. Moreover, the presence or absence of ant mosaics may be driven by the density of suitable resources. We use an intensively studied ant community in the lowland rainforests of North-East Queensland, Australia to outline processes that may lead to ant mosaic patterns, reasoning that competition for highly predictable resources in space and time such as honeydew and nectar is a fundamental process to maintain the mosaic structure. Honeydew and nectar sources, particularly their amino acids, are of crucial importance for nourishment of arboreal ant species. We use canopy fogging data from the same site in Australia and from two mature rainforests in South-East Asia to compare spatial avoidance and co-occurrence patterns implied by ant mosaics. Significant negative and positive associations were found among the three most abundant ant species in each dataset. Several problems with such spatial analyses are discussed, and we suggest that studies of ant mosaics in complex rainforest communities would benefit from a more focused approach on patterns of resource distribution and their differential utilisation by ants. Key words: ant mosaic, Australia, competition, dominant ant, South-East Asia, tropical rainforest. THE ANT MOSAIC CONCEPT In the early 1970s, entomologists studying ants associated with cocoa and other plantation crops, initially in West Africa and later elsewhere in the world, demonstrated a hierarchical structure to ant communities in such plantations (Room 1971; Majer 1972, 1976a,b,c, 1993; Leston 1973; Room & Smith 1975; Room 1975; Taylor 1977; Jackson 1984b; Majer et al. 1994). They observed that two or more species of ants were typically much more abundant than any other species and that the areas of activity of these ‘dominant’ ant species in the trees did not overlap. The patchwork distribution these ants formed was termed ‘ant mosaic’ (Leston 1973). Such dominant ant species maintain large colonies and are characterised by their high abundance and activity density, but *Corresponding author. † Present address: Department of Natural Resource Management, University of Melbourne, Melbourne,Vic. 3010, Australia. Accepted for publication December 2006. © 2007 Ecological Society of Australia numerical dominance in ants is often (but not always) associated with behavioural dominance in terms of a high inter- and intraspecific aggression and potential to defend resources and territories against other ants (Hölldobler 1983; Hölldobler & Wilson 1990; Majer 1993; Andersen 1995; Davidson 1997). Interference competition between ants can be intense (Jackson 1984a; Fellers 1987; Savolainen & Vepsäläinen 1988; Andersen 1992; Andersen & Patel 1994). Ant mosaics may thus largely reflect ongoing or past competition between the dominant ants involved, leading to mutual exclusion of the dominant ants’ territories, thereby avoiding frequent direct confrontations which may be costly. Majer (1976c) showed how the experimental removal of a colony of a particular dominant ant species leads to replacement by a neighbouring colony from a different species, this providing strong evidence for competitive release. Species-specific microhabitat requirements may play an additional role in shaping the ants’ distribution (Leston 1973; Majer 1976c; Andersen 1995; Way & Bolton 1997). Besides mutual exclusion between dominants, nearly all ant mosaics described to date also 94 N. B L Ü T H G E N A N D N. E . S TO R K included patterns of coexistence, particularly among non-dominant or submissive ant species within each dominant’s territory. Both patterns may lead to a complex situation of significantly negative and positive associations between species pairs (Room 1971; Leston 1975; Taylor 1977; Majer 1993; Majer et al. 1994). A number of studies from all tropical continents have contributed to our understanding of ant mosaics, which appear to be a nearly universal feature of tropical plantations. Ant mosaics have also been noted in unmanaged ecosystems, such as mangrove forests (Adams 1994) and secondary tropical forests (Leston 1978; Dejean et al. 1994). Ant mosaics have been the subject of several reviews (Jackson 1984a; Majer 1993; Dejean & Corbara 2003) and yet only a few studies have examined ant mosaics in the canopy of primary rainforest. These include the work of Dejean and colleagues using the canopy raft and sled in Cameroon (Dejean et al. 1999; Dejean & Gibernau 2000; Dejean et al. 2000), among other studies (Blüthgen et al. 2000). Recently, the existence of ant mosaics in plantations and, particularly, in mature rainforests has been questioned (Floren & Linsenmair 2000; Ribas & Schoereder 2002).The former authors investigated a highly diverse canopy fauna in a mature rainforest in Sabah, Borneo. They made extensive collections of arboreal ants from 19 individual Aporusa or Xantophyllum trees in the lower canopy using knockdown insecticide fogging (Floren & Linsenmair 1997).The collection of unwanted insects from taller neighbouring trees was carefully prevented by covering the sampled trees with sheets (see also Adis et al. 1999). Their statistical analysis of the resulting data failed to significantly demonstrate effects of mutual exclusion between ant species (Floren & Linsenmair 1997, 1998, 2000; Floren et al. 2001). Aside from these analytical results, extensive observations in these relatively small crowns indicated that resource monopolisation by dominant ants was uncommon on these trees (Floren & Linsenmair 2000; A. Floren, pers. comm. 2006). Floren and Linsenmair (2000) thus concluded that the diverse and complex ant communities in rainforest canopies are more likely structured by stochastic processes rather than by competitive effects that are the heart of ant mosaic theory. Moreover, Ribas and Schoereder (2002) used a different statistical approach based on a ‘checkerboardedness’ index (Stone & Roberts 1990) to reanalyse several previous studies on ant mosaics from plantations and secondary forests and failed to significantly confirm a structured distribution in most of these studies. They suggested that spatial distribution patterns may not be different from expectations based on null models and may not necessarily imply competition between these species. Subsequently, Floren and Linsenmair (2005) also applied the ‘checkerboardedness’ index to reanalyse their data from the mature rainforest ant community and found that it did not deviate significantly from random associations. The presence or absence of ant mosaics in plantations and forests may be important to understand the general role of interspecific competition in communities of variable degrees of disturbance and complexity. Moreover, ant mosaics may also play a fundamental role for the structure of the arthropod community as a whole and resulting ecosystem functions. In plantations, certain dominant ants have been shown to play a crucial role in pest control, while others may be less effective or even have detrimental net effects (Way 1953, 1963; Way & Khoo 1991; Dejean et al. 1997b). Moreover, it has been suggested that each dominant ant species may be associated with a different arthropod fauna, including competitors, prey or trophobiotic partners tended for honeydew (Bigger 1993). Such assemblages may vary among dominant ant species and be relatively distinct (Room & Smith 1975; Majer 1976a). Arthropod assemblages may even change towards the typical associates of a new dominant ant species following an experimental removal of the nests of a previous dominant ant (Majer 1976b). The presence and identity of dominant ants may thus affect the distribution of other arthropods, either directly or indirectly, for example, via predation, competition or via mutualistic interaction. Such effects have been rarely examined in natural forest ecosystems. For example, dominant ants have been shown to prevent colonisation of several common lycaenid caterpillars on a tree species in Malaysia and tolerate only the presence of specialised obligate myrmecophile species (Seufert & Fiedler 1996). For African savanna trees, Mody and Linsenmair (2004) have shown that the complex arthropod assemblage changed in a predictable way when ants were excluded, and the effect on damage by herbivores differed significantly among different dominant species. Hence, the ants’ specific impact on arthropods potentially translates into plant performance and may thus provide a keystone function for a variety of ecosystem processes. The question addressed in our paper is whether the current sampling methods, sampling scales and classical analytical methods used by most researchers are adequate to detect the existence of ant mosaics with sufficient power and reliability. We first illustrate an example of an intensively studied ant community which we suggest exhibits a clear ant mosaic pattern in a relatively simple tropical forest system in North-East Queensland, describing mechanisms that were found to be important for competition in this community.We then compare spatial patterns implied from canopy fogging at the same site with those from more complex rainforest communities elsewhere and critically discuss the methodology, selection of sampling units, characteristics and putative pitfalls associated with analyses of ant mosaics used to date. © 2007 Ecological Society of Australia ANT MOSAICS IN TROPICAL RAINFORESTS CASE STUDY IN NORTH-EAST QUEENSLAND In the coastal lowland rainforest in Cape Tribulation (Daintree, North-East Queensland, Australia) we used the canopy crane (Stork 2007) to study processes that shape the distribution of ants on plants. The rainforest in this region is regularly disturbed by cyclones, resulting in a relatively open forest canopy not taller than 25–40 m. The 1-ha crane site has 92 species of tree above 10 cm d.b.h. and about 30 species of vines (see Laidlaw et al. 2007). Fourteen of the tree species, 10 species of climbing plants, and six understorey shrubs possess extrafloral nectaries, all of which were most commonly attended by ants (Blüthgen & Reifenrath 2003). Australian rainforests maintain a large proportion of the continent’s fauna and flora, but are comparatively species-poor in relation to South-East Asian forests (for ants, see Brühl et al. 1998; Shattuck 1999; Majer et al. 2001). A total of 66 species of ants have been collected in the canopy and understorey at the canopy crane site and nearby (Blüthgen & Fiedler 2004b; Blüthgen et al. 2004b; see Appendix i for an updated list). Additional ant species have been found at ground level, although no extensive search for ants associated with the ground layer, leaf litter and dead wood has been made. Two species of ants, Oecophylla smaragdina (Formicinae) and Anonychomyrma gilberti (Dolichoderinae), are numerically dominant in the canopy and understorey. Both species have large colonies that forage on several neighbouring trees and are clearly mutually exclusive. At territorial borders, extended combats between these two species were observed on three occasions, when a number of workers were killed and, on some trees, O. smaragdina replaced A. gilberti (N. Blüthgen, unpubl. obs. 2000). Colonies of A. gilberti nested exclusively in Syzygium ‘erythrocalyx’ (Myrtaceae) tree trunks and attended cicadellids and extrafloral nectaries on the same and on several neighbouring trees (Blüthgen et al. 2004b). Oecophylla smaragdina built leaf nests in tree crowns and attended large aggregations of hemipterans in the upper canopy, particularly on vines (Blüthgen & Fiedler 2002), as well as several extrafloral nectar sources in all strata (Blüthgen et al. 2004b). Both ant species are also highly predatory, although honeydew and nectar provide a large part of their nitrogen and carbon budgets (Blüthgen et al. 2003). Oecophylla smaragdina was particularly common on honeydew and nectar sources with the highest nutrient quality in terms of amino acids and carbohydrates, where it often monopolizes these resources and effectively excludes its competitors (Blüthgen & Fiedler 2004a). On trails and poorer nectar sources, it often co-occurred with certain other ant species, including several species of the genera Crematogaster, Echinopla and Polyrhachis, but never © 2007 Ecological Society of Australia 95 with A. gilberti (Blüthgen et al. 2004b). Anonychomyrma gilberti itself was only associated with a small number of ant species, particularly Polyrhachis spp. (Blüthgen et al. 2004b). Thus, a specific spectrum of submissive ants was tolerated by each dominant ant, a pattern that has also been reported as integral part of ant mosaics elsewhere (Dejean & Corbara 2003). STATISTICAL PITFALLS Historically, in most ant mosaic studies, the association between a pair of species is commonly tested by c2-statistics or Fisher’s exact tests (Leston 1973, 1975; Room & Smith 1975; Jackson 1984b; Dejean et al. 1994; Majer et al. 1994). The association is represented in a 2 ¥ 2 contingency table, with two rows for one species being present or absent, and two columns for the other species present or absent, respectively. The total of all cell entries then equals the total number of sampling units (trees or plots), hence including units where none of the two species was present. A critical issue of contingency table analysis is the statistical independence of count data – an assumption frequently violated by spatial autocorrelations or genetic relatedness of ant colonies though processes like colony budding. These pairwise tests were often applied to many combinations of ant species, without any corrections for multiple comparisons in order to deal with an increasing global risk of type I error (i.e. falsely rejecting the null hypothesis). In contrast, Floren and Linsenmair (2000) examined 63 ant species that occurred on three or more trees, and tested the co-occurrence of all 1953 combinations between these species, using c2-tests with a Bonferroni correction for the number of tests applied. However, the methodological drawback of this method lies in a high risk of a type II error (Cabin & Mitchell 2000; Moran 2003), particularly since the individual tests are clearly non-independent (Gotelli & Ellison 2004). Note that there are more sensitive techniques than (non-sequential) Bonferroni correction, most notably ‘false discovery rate’ (Bejamini & Hochberg 1995), but they do not entirely reduce the problem. This risk of incorrectly accepting the null hypothesis becomes excessively large in combination with a small sample size. Even for a hypothetical case of a complete mutual exclusion between two highly common ant species on the 19 trees sampled by Floren and Linsenmair (e.g. each of the ant species occurred on eight trees without any overlap), this statistic would not yield a significant effect. Species were included in the analysis if they were found on at least three trees, although a complete exclusion between species that occurred on five trees each would not cause a significant result in a c2-test, even before correcting for multiple comparisons. 96 N. B L Ü T H G E N A N D N. E . S TO R K Given the low statistical power, the lack of any significant positive or negative associations in their analysis is not surprising. One way to at least partly overcome this problem is to limit the number of comparisons based on a priori selection criteria. In true ant mosaics, territorial exclusiveness is limited to dominant ant species with large colonies, while many other (subdominant and nondominant) ants typically coexist with these dominants (Majer 1993; Dejean & Corbara 2003). Analyses may therefore be restricted to dominant species in each dataset, and hence to the most meaningful comparisons (Leston 1975; Jackson 1984b; Davidson et al. 1989; Bigger 1993). We are aware that any selection is problematic, as behavioural dominance is a subjective criterion as long as detailed insights and behavioural bioassays are unavailable for all species, and behavioural dominance is not necessarily correlated with colony size and colony abundance. For our analysis below, we selected the three most abundant species (in terms of the total number of individuals) that occurred in three or more sampling units (i.e. funnels, trees, or plots, respectively). Two studies (Ribas & Schoereder 2002; Floren & Linsenmair 2005) have used a statistic derived from studies of island geographical distribution to test ant mosaic patterns, namely a randomisation of the ‘checkerboardedness’ or C-score index (Stone & Roberts 1990). Like several other techniques that allow a comparison of co-occurrence patterns with null models (see Gotelli & Graves 1996; Gotelli 2000), this technique reduces the complex data of positive, neutral and negative associations between many pairs of species into a single value rather than showing the variation of associations. A higher C-score than expected by a null model suggests that negative associations predominate in the assemblage. Both studies (Ribas & Schoereder 2002; Floren & Linsenmair 2005) applied this method to entire ant communities, as well as to reduced datasets from which rare species or otherwise classified ‘non-dominants’ have been removed. They tested whether C-scores were higher than random, implying that negative associations outweigh positive associations in a community structured by ant mosaics. However, this assumption may not hold for numerous ant mosaics described, where the number of significant positive associations reached or even exceeded the number of negative associations between abundant species (Room 1971; Leston 1975; Taylor 1977; Majer 1993; Majer et al. 1994). Therefore, C-scores for ant communities may be expected to be similar to, or even lower than, those of null models if positive interactions are common. Results may thus strongly depend on the restriction of the dataset to true ‘dominants’ (but see Ribas & Schoereder 2002), and on the abundance level and criteria used for this selection. Both types of analyses, the 2 ¥ 2 association tables and C-score randomisation as used above, reduce the information of species abundance to binary data of presence or absence. This approach will be mostly conservative for detecting negative associations, since it does not account for the numerical effect observed within areas where different dominant ants overlap, that is, at least one species is usually not abundant (Majer 1972; Jackson 1984b; our Fig. 1). Furthermore, the inclusion of all sampling units lacking any of the two species into the 2 ¥ 2 association analysis (but not into C-score) may be conservative for testing exclusiveness. There has been a controversial debate about the validity and properties of different approaches to detect species exclusion in the context of island biogeography (Grant & Abbott 1980; Gilpin & Diamond 1982; Manly 1995; Gotelli & Graves 1996; Gotelli 2000). These approaches, and several other techniques of parametric-free statistics and simulations, may also include fruitful techniques to analyse ant mosaics (Floren & Linsenmair 2000, 2005), if the mixture of negative and positive associations is appropriately controlled for and if these techniques are unaffected by unequal sampling effort between species-poor and species-rich systems. Moreover, an inclusion of abundance data into such analyses may provide a better statistical power. DETECTING ANT MOSAICS USING CANOPY FOGGING While observational evidence and the distribution pattern of ants on plants indicate a pronounced ant mosaic at the study site in Queensland, and the underlying mechanisms at least partly understood (resource quality), the question remains whether spatial effects of this mosaic are visible in canopy fogging data. For this purpose, we used unpublished data from canopy knockdowns in the same forest performed in 2000 (R.L. Kitching and coworkers, unpublished 2000). Three 10 ¥ 10 m plots were installed underneath several tree crowns in the forest, each with 20 halfmetre square collecting trays (Majer et al. 2001). In total, 29 ant species were collected in these samples, sorted by one of us (N. Blüthgen) and compared with specimens from the same site that were identified with the help of several taxonomists (see Acknowledgements). We use 2 ¥ 2 contingency tables and correlations between number of workers to illustrate patterns implying mutual exclusion or positive associations, two methods which we will critically discuss below, and restrict our analysis to the three most abundant species. Oecophylla smaragdina and A. gilberti were the most abundant ants in the fogging samples, corresponding to the canopy fogging data from two other sites nearby (Majer et al. 2001).These species represented 34% and 27%, respectively, of the total 342 individuals and © 2007 Ecological Society of Australia ANT MOSAICS IN TROPICAL RAINFORESTS 97 Fig. 1. Abundance of the three most abundant ant species in three canopy fogging samples from mature rainforests from (a) North-East Queensland, Australia, (b) Sulawesi, and (c) Brunei. Number of ant workers shown for each trap (a, b) or tree (c). Note that most points are distributed along the axes or close to the axes, while areas between axes are largely empty, thus only one species was abundant in a tray.The commonly co-occurring Oecophylla smaragdina and Crematogaster cf. fusca (a) provide the most prominent exception. occurred in 30 and 18, respectively, of the 60 trays.They only occurred together in two trays. The exclusive occurrence of these species is highly significant in a Fisher’s exact test (Table 1a). In each of the two trays where O. smaragdina and A. gilberti were collected together, only a single O. smaragdina worker was caught © 2007 Ecological Society of Australia (Fig. 1a). Hence, there was no case where both were common in the same tray. Consequently, there was a highly significant negative correlation between the number of O. smaragdina and the number of A. gilberti per tray (Spearman rank rS = -0.78, P < 10-6, n = 46 trays with at least one of both species). 98 N. B L Ü T H G E N A N D N. E . S TO R K Table 1. Co-occurrence of the three most abundant ant species in each of the fogging samples from (a) North-East Queensland, Australia, (b) Sulawesi and (c) Brunei, and sampling units analysed were trays (a, b) or trees (c) (a) Australia, 60 trays Oecophylla smaragdina Anonychomyrma gilberti Crematogaster cf. fusca (b) Sulawesi, 350 trays Crematogaster sp.3 Dolichoderus sp.1 Iridomyrmex sp.1 (c) Brunei, 10 trees Myrmicaria sp.B Crematogaster sp.B Myrmicaria sp.A Ants 117 93 38 n 30 18 24 Ants 581 501 304 n 46 41 64 Ants 1310 789 488 n 7 4 3 O. smaragdina 30 – 2 (–) (P = 0.0001) 18 (+) (P = 0.003) Crematogaster sp.3 36 – 15 (+) (P < 0.0001)† 15 (+) (P = 0.20)‡ Myrmicaria sp.B 7 – 1 (–) (P = 0.033) 0 (–) (P = 0.008) A. gilberti 18 – 2 (–) (P = 0.004) Dolichoderus sp.1 41 – 14 (+) (P = 0.12)§ Crematogaster sp.B 4 – 3 (+)¶ Maximum overlap that would yield a significant negative association: †1 tray, ‡3 trays, §2 trays. ¶ Numbers of occurrences too low to test for significant positive or negative association. ‘Ants’ denotes the total number of ant workers collected, n the total number of sampling units in which each species was recorded, and each interaction cell provides the number of sampling units where both species overlap. (+) or (–) denote whether this number was higher or lower than expected by chance, respectively. Significance level shown for Fisher’s exact test, two-tailed. The third most common ant in this sampling, Crematogaster cf. fusca (11% of the individuals), rarely overlapped with A. gilberti, but was frequently sampled together with O. smaragdina in the same tray. Fishers exact test confirmed a significantly negative association between C. fusca and A gilberti and a significant positive association between C. fusca and O. smaragdina (Table 1a). A negative correlation was found between the number of C. fusca versus A. gilberti workers in a tray (rS = -0.81, P < 10-6, n = 40), while abundances of C. fusca and O. smaragdina were not significantly correlated (rS = -0.18, p = 0.28, n = 36) (Fig. 1a). These fogging data were consistent with direct observations at nectar sources and trails, where O. smaragdina and C. fusca were commonly found together, but neither A. gilberti with O. smaragdina nor A. gilberti with C. fusca (Blüthgen et al. 2004b). These three ant species were also the most abundant ones in the surveys of nectar plants (Blüthgen et al. 2004b), and among the seven most abundant ones on sugar baits in the same area (Blüthgen & Fiedler 2004b). Therefore, behavioural dominance and mutual aggression may be associated with abundance in some cases (O. smaragdina and A. gilberti), but not in others (C. fusca). Can such distribution patterns also be detected in canopy fogging samples from more species-rich communities that are much less understood? We used the same approach to analyse two datasets from canopy fogging in Brunei (northern Borneo) and the Indonesian island of Sulawesi. In the former study, 10 semiisolated trees ranging in height from 30 to 75 m in a mature lowland dipterocarp forest in Brunei were fogged and the falling insects were collected on a single large sheet under each tree (total sheet area under 10 trees = 200 m2). The faunal composition, guild structure, body size species/abundance relationships, and factors affecting the similarity of the faunas in the different trees sampled have been published elsewhere (Stork 1987, 1991; Morse et al. 1988). In the second canopy fogging study in Sulawesi, 350 catching trays (each 1 m2) were placed haphazardly in three 12 ¥ 12 m plots (lowland forest data only), where the crowns of the trees often overlapped (Stork & Brendell 1990). Totals of 98 and 77 morphospecies, respectively, were identified by Barry Bolton (Natural History Museum, London) from the Brunei and Sulawesi samples. The former study is analysed at the level of individual trees and the latter at the level of individual trays (data from plots pooled). The results are shown in Table 1. In Brunei (Table 1c), two species combinations revealed a significant negative association.The remaining third association seemed to be positive, but could not be tested given the low sample size of trees (Table 1c). The analysis of the fogging trays in Sulawesi (Table 1b) showed a relatively large overlap between the three most abundant species (one combination was significantly positive). However, this example also illustrates the low statistical power for detecting negative associations in such a community, since each of the three species was only found on a relatively small proportion (12–18%) of the 350 fogging trays analysed. The threshold for a significantly negative effect in Fisher’s exact test is very low: for instance, only an overlap in one or none of the trays would yield a significant exclusion between the two most abundant species (Table 1b). The entire analysis is therefore highly © 2007 Ecological Society of Australia ANT MOSAICS IN TROPICAL RAINFORESTS constrained by the abundance of the species in relation to the number of sampling units (if trays lacking any of the two species were excluded from the analysis, all three associations would have been significantly negative). For the Brunei data, there is a negative correlation between the abundances of both Myrmicaria species (rS = -0.79, P < 0.01, n = 10), while correlations between each of these with Crematogaster were not significant (P ⱖ 0.22) (Fig. 1c). For the data from Sulawesi, abundances of all three species were strongly negatively correlated (all rS ⱕ -0.51, P ⱕ 10-6, n ⱖ 72 trays in which at least one species of each pair occurred). Therefore, while abundance data clearly suggest negative associations between the three dominant ant species in Sulawesi (Fig. 1c), this is not evident in the more conservative approach based on presence–absence data (Table 1b). Any analysis based on presence–absence may clearly underestimate the degree these species avoid each other, since occasions where a single worker co-occurs with its putative competitors are evaluated as co-occurrence in the same way if both species were common (Jackson 1984b). Correlations between the number of ants per tray represent a finer scale, but may be more difficult to interpret as putative effect of competition in relation to null models: a simple patchy distribution of workers in social insect (irrespective of competitive interactions) may also lead to negative correlations of worker abundances if the colony density is comparatively low. This problem is less evident, but not absent in the cruder contingency table analyses. SPATIAL SCALE OF MOSAICS Social insects such as ants usually show a highly patchy distribution when they are mapped at a small spatial scale, especially when nests or resources with mass recruitment are involved.The above analyses of the ant distribution patterns in Australia and Sulawesi were based on single 1-m2 trays. This unit certainly represents only a small fraction of typical ant territories, and especially so for the dominant territorial species that create the ant mosaic (if it does exist). On the other hand, insects captured by each tray may involve a vertical column of several plant layers which, in complex forest canopies, could include different compartments of ant mosaics (Majer 1993). Such columns in pristine forest can be up to 70 m high, as in the case of some of the Bornean trees, whereas in cocoa plantations, mangroves or secondary forests they may be only 5–20 m high. Hence, the chance of several dominant species being found in different strata may be increased in mature forest. Floren and Linsenmair (2000) solved this problem by restricting their analyses to trees with relatively small crowns in the subcanopy layer and by using protective sheets against upper forest layers. © 2007 Ecological Society of Australia 99 The main statistical problem of an analysis of single trays (as in Fig. 1a,b and Table 1a,b) is that neighbouring ones may frequently not represent independent units (ant colonies). Data from each tray are thus prone to spatial autocorrelation, limiting the expressiveness of tests based on trays if not adequately controlled by null models. Correlations between the abundance of one species in each spatial unit versus that of another species as used above (see also Majer 1972) may thus have to be interpreted with caution and can not be seen as direct evidence for competition. In contrast to single trays, the spatial resolution at the level of individual tree canopies (all trays pooled) may be too crude to detect exclusive territories, at least in complex forest canopies. Despite this crude resolution, two species pairs in Brunei (Table 1c, Fig. 1c) show a strong pattern of mutual avoidance on the level of trees, but other cases involving patterns on smaller scales may be more difficult to detect. Small isolated tree crowns in plantations, as in the classic ant mosaic studies, usually represent small spatial units which ant colonies may be able to monopolize, sometimes by defending the trunk or major branches (Hölldobler & Lumsden 1980). However, in several cases, even such small tree crowns may harbour nests or areas of activity of two dominant ant species. Such ‘codominants’ on trees in plantations were usually separated vertically or had different activity schedules or food niches (Room 1971; Majer 1972, 1976a; Leston 1973; Way & Bolton 1997). Simultaneous existence of different dominants may be even more pronounced in large crowns of rainforest trees where lianas may connect neighbouring crowns and lead to structurally complex, intermingled canopy layers. For instance, in an Amazonian rainforest, different dominant ant species and their trophobionts were found on different branches of the same large tree crown (Blüthgen et al. 2000). This spatial complexity is much reduced in small-crowned isolated forest trees which may be more comparable to crowns in some tree plantations (Floren & Linsenmair 2000). On the other hand, colony territories of some dominant ant species may even comprise a number of adjacent trees (e.g. Hölldobler 1983; Bigger 1993; Majer 1993). This suggests that detailed mapping of the distribution of dominant ant colonies within an area, in conjunction with adequate spatial statistics, may be a more appropriate approach to analyse ant mosaics than selecting some non-neighbouring trees as in the study by Floren and Linsenmair (2000). WHICH ANT SPECIES ARE CONSIDERED DOMINANT, AND FOR WHICH RESOURCES DO THEY COMPETE? Of course, the restriction of the analysis above to the three most abundant ant species is rather arbitrary. 100 N. B L Ü T H G E N A N D N. E . S TO R K There is no clear a priori definition of dominance that can be applied to an unknown ant community so far. While experienced myrmecologists have little difficulty in assigning a species to be dominant, subdominant or submissive in intensively studied ant communities (e.g. Hölldobler 1983; Andersen 1995; Dejean et al. 2000), more objective criteria would be needed to disentangle the relationship between dominance and other attributes. For example, numerically abundant species may not necessarily show a high level of interspecific aggression (Davidson 1998). Davidson (1997) suggests that the most dominant arboreal ants are characterised by a modified proventriculus that enables them to effectively harvest honeydew and other fluids, associated with a high level of carbohydrate feeding, carbohydrate-based defence and high ‘tempo’ activity (see also Davidson et al. 2004; Davidson 2005). Such species usually come from a restricted number of predictable phylogenetic groups, involving genera or species that are dominant both in plantations as well as in mature rainforests; examples include Crematogaster, Dolichoderus and some Camponotus (worldwide), Oecophylla (palaeotropics) and Azteca (neotropics) (Majer 1993; Davidson 1997). The high taxonomic similarity between dominants in tropical plantations versus forest canopies suggests that similar ecological interactions may be found in these ecosystems. Most dominant arboreal ants are omnivorous, but feed extensively on honeydew in trophobiotic associations with hemipterans or lycaenids in plantations (Room & Smith 1975; Jackson 1984b; Bigger 1993) or natural ecosystems (Davidson 1997; Dejean et al. 2000; Fiedler 2001). Ants usually monopolize such trophobionts by effectively excluding their competitors (Dejean et al. 1997a; Blüthgen et al. 2000, 2004b, 2006). Honeydew and nectar may provide spatiotemporally relatively constant, renewable rewards for which ants may compete (Jackson 1984b; Blüthgen et al. 2000; Hossaert-McKey et al. 2001). In contrast, prey and food items collected by opportunistic scavengers may be typically more scattered, less predictable, and usually replenish following a brief period of consumption. Therefore, if territorial defences aim to directly defend resources (rather than spaces that may potentially include resources), ant territoriality may be expected to be particularly pronounced for qualitatively and quantitatively rewarding honeydew and nectar sources. Among such liquid food sources, many honeydew sources tend to be more productive, contain more amino acids, and are spatially more concentrated compared with most extrafloral and floral nectar sources (Blüthgen et al. 2004a; N. Blüthgen pers. obs. 2000). Many dominant canopy ants obtain a large proportion of their nitrogen requirements for larval growth from amino acids in honeydew and/or nectar (Blüthgen et al. 2003; Davidson et al. 2003). These features also apply to mound-building wood ants (Formica) (e.g. Savolainen & Vepsäläinen 1988), suggesting that high levels of trophobioses are indeed characteristic for many territorial dominant ants worldwide. Studies in mature rainforests have shown that anthemipteran aggregations can be abundant (Blüthgen et al. 2000; Dejean et al. 2000). On all 10 upper canopy trees sampled in Brunei, hemipterans were abundant, contributing to 7.0–26.0% of the arthropod individuals per tree (Stork 1991). These hemipterans were dominated by cicadellids and membracids with an average of 47 species per tree (N. Stork, unpubl. data 1982). Many species from both taxa are strongly tended by ants for honeydew (Delabie 2001; Blüthgen et al. 2006). Hemipterans in the canopy investigated in Sulawesi contributed between 2.8% and 12.3% of the individuals in different plots (Stork & Brendell 1990). These figures certainly underestimate the abundance of hemipterans, since several sessile hemipterans such as scale insects whose stylets are firmly inserted into the plant tissues are poorly captured in canopy foggings or not at all (Dejean et al. 2000). However, these taxa can play a major role as honeydew producers and are commonly attended by ants in rainforests (Dejean et al. 2000; Blüthgen et al. 2006). Aggregations of anttended hemipterans often occur in the upper canopy or in forest gaps (Blüthgen et al. 2000, 2006), so light availability might play an important role. Plant productivity affects the quality and quantity of honeydew, and both homopterans and ants might actively seek the most productive parts of the plant, with ants being able to actively transport some homopterans to their feeding sites (Way 1963). Consequently, hemipterans in forest canopies are typically confined to the outermost twigs on trees and climbing plants where the stem and leaf tissue is relatively young (Blüthgen et al. 2000; Blüthgen & Fiedler 2002). Floren and Linsenmair (1997, 2000), studying trees of the lower canopy that were always exceeded in height by trees of the upper canopy, found no spatio-temporally constant trophobiosis between ants and hemipterans. Only 1.7– 11.6% of the individuals in their samples were hemipterans (Floren & Linsenmair 1997), less than in the Brunei data from Stork (1991). The absence of pronounced trophobiotic associations as defendable resources may indeed be a major reason why ant mosaics could be non-existent. In summary, lower canopy trees may have much fewer rewarding and stable resources for ants to compete for than the upper canopy, and hence ant mosaics may be less evident or absent in the lower canopy than the upper canopy. While few studies have indicated the importance of honeydew and nectar use by ants in rainforest canopies by canopy observations (Blüthgen et al. 2000, 2004b; Dejean et al. 2000) or indirectly by stable isotope analysis (Blüthgen et al. 2003; Davidson et al. © 2007 Ecological Society of Australia ANT MOSAICS IN TROPICAL RAINFORESTS 2003), honeydew productivity and availability has not been quantified in upper and lower rainforest canopies. OUTLOOK Overall, we suggest that effects of competition between ants in species rich communities may need to be investigated with higher spatial resolution and statistical power. Until tests with an appropriate spatial scale and statistical power are applied, it will be impossible to decide whether patterns such as ant mosaics do not exist or are just more difficult to detect in complex communities compared with plantations. Our data from Brunei strongly suggest the presence of ant mosaics in the upper forest canopy. Moreover, while the term ‘ant mosaic’ as defined by Leston (1973) implies a merely static spatial pattern, such spatial patterns are based on behavioural or ecological processes and may be relatively dynamic. Spatial distribution is only a rather indirect measure or consequence of complex competitive interactions and may not necessarily reveal underlying structural mechanisms of competitive hierarchies in ant communities (see Jackson 1984a). Therefore, an increased focus on the ants’ behavioural interactions (Hölldobler 1979, 1983; Adams 1994; Mercier et al. 1998), colonisation abilities (Greenslade 1971; Yu et al. 2001) or interaction with natural resources (Davidson 1997; Hossaert-McKey et al. 2001) may be more fruitful than a sole focus on static spatial distribution patterns of workers or colonies. A survey of honeydewproducing hemipterans and associated ants in different strata of mature forests would be required to evaluate the potential of stable resources as a basis for ant mosaics. Bait experiments may provide a good picture of ant activity and competitive interactions, combined with a much higher statistical power than usually provided by spatial mapping (Andersen 1992; Andersen & Patel 1994; Yanoviak & Kaspari 2000; Floren et al. 2002; Blüthgen & Fiedler 2004b; Gibb & Hochuli 2004), although the criticism that the quality and spatial concentration of many typical baits may not always reflect natural resources may be valid (Kay 2002; Ribas & Schoereder 2002). Moreover, experimental manipulations may be a very fruitful approach to study competition in general (Schoener 1983), specifically in ant communities where a dominant ant can be removed to study the effects on competitive release (Majer 1976c; Perfecto 1994; Gibb & Hochuli 2004). Finally, a clearer definition and better understanding of ‘dominance’ in ants and its relationship to colony size and local abundance is needed to allow more precise predictions about structural patterns in ant communities. © 2007 Ecological Society of Australia 101 ACKNOWLEDGEMENTS We thank Alan Andersen, Brian Heterick, Rudy Kohout, Hanna Reichel and Steve Shattuck for their help with identification of Australian ants, Barry Bolton for identification of the South-East Asian ants, and Roger Kitching and his team for providing the ants from their fogging samples in Australia. We thank Konrad Fiedler for his continuous advice during the field work in Australia and helpful comments on the manuscript, Rob Dunn and Jonathon Majer for providing insightful comments, and K. Eduard Linsenmair and Andreas Floren for stimulating this discussion as well as providing helpful comments on earlier versions of the manuscript. Field work in Australia was kindly supported by the Deutsche Forschungsgemeinschaft (Fi 547/9-1), by Fuchs Oils (Mannheim/Sydney) through the Australian Canopy Crane company, and by a doctoral fellowship of the Studienstiftung des deutschen Volkes to N.B. REFERENCES Adams E. S. (1994) Territory defense by the ant Azteca trigona: maintenance of an arboreal ant mosaic. Oecologia 97, 203–8. Adis J., Basset Y., Floren A. & Hammond P. M. & Linsenmair K. E. (1999) Canopy fogging of an overstorey tree – recommendations for standardization. Ecotropica 4, 93–7. Andersen A. N. (1992) Regulation of ‘momentary’ diversity by dominant species in exceptionally rich ant communities of the Australian seasonal tropics. Am. Nat. 140, 401–20. Andersen A. N. (1995) A classification of Australian ant communities, based on functional groups which parallel plant life-forms in relation to stress and disturbance. J. Biogeogr. 22, 15–29. Andersen A. N. & Patel A. D. (1994) Meat ants as dominant members of Australian ant communities: an experimental test of their influence on the foraging success and forager abundance of other species. Oecologia 98, 15–24. Benjamini Y. & Hochberg Y. (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B 57, 289–300. Bigger M. (1993) Ant–homopteran interactions in a tropical ecosystem. Description of an experiment on cocoa in Ghana. Bull. Entomol. Res. 83, 475–505. Blüthgen N. & Fiedler K. (2002) Interactions between weaver ants (Oecophylla smaragdina), homopterans, trees and lianas in an Australian rainforest canopy. J. Anim. Ecol. 71, 793– 801. Blüthgen N. & Fiedler K. (2004a) Competition for composition: lessons from nectar-feeding ant communities. Ecology 85, 1479–85. Blüthgen N. & Fiedler K. (2004b) Preferences for sugars and amino acids and their conditionality in a diverse nectarfeeding ant community. J. Anim. Ecol. 73, 155–66. Blüthgen N. & Reifenrath K. (2003) Extrafloral nectaries in an Australian rainforest – structure and distribution. Aust. J. Bot. 51, 515–27. Blüthgen N., Verhaagh M., Goitía W., Jaffé K., Morawetz W. & Barthlott W. (2000) How plants shape the ant community 102 N. B L Ü T H G E N A N D N. E . S TO R K in the Amazonian rainforest canopy: the key role of extrafloral nectaries and homopteran honeydew. Oecologia 125, 229–40. Blüthgen N., Gebauer G. & Fiedler K. (2003) Disentangling a rainforest food web using stable isotopes: dietary diversity in a species-rich ant community. Oecologia 137, 426–35. Blüthgen N., Gottsberger G. & Fiedler K. (2004a) Sugar and amino acid composition of ant-attended nectar and honeydew sources from an Australian rainforest. Austral Ecol. 29, 418–29. Blüthgen N., Stork N. E. & Fiedler K. (2004b) Bottom-up control and co-occurrence in complex communities: honeydew and nectar determine a rainforest ant mosaic. Oikos 106, 344–58. Blüthgen N., Mezger D. & Linsenmair K. E. (2006) Anthemipteran trophobioses in a Bornean rainforest – diversity, specificity and monopolisation. Insect Soc. 53, 194–203. Brühl C. A., Gunsalam G. & Linsenmair K. E. (1998) Stratification of ants (Hymenoptera: Formicidae) in a primary rain forest in Sabah, Borneo. J. Trop Ecol. 14, 285–97. Cabin R. J. & Mitchell R. J. (2000) To Bonferroni or not to Bonferroni: when and how are the questions. Bull. Ecol. Soc. Am. 81, 246–8. Davidson D. W. (1997) The role of resource imbalances in the evolutionary ecology of tropical arboreal ants. Biol. J. Linn. Soc. 61, 153–81. Davidson D. W. (1998) Resource discovery versus resource domination in ants: a functional mechanism for breaking the trade-off. Ecol. Entomol. 23, 484–90. Davidson D. W. (2005) Ecological stoichiometry of ants in a New World rain forest. Oecologia 142, 221–31. Davidson D. W., Snelling R. R. & Longino J. T. (1989) Competition among ants for myrmecophytes and the significance of plant trichomes. Biotropica 21, 64–73. Davidson D. W., Cook S. C., Snelling R. R. & Chua T. H. (2003) Explaining the abundance of ants in lowland tropical rainforest canopies. Science 300, 969–72. Davidson D. W., Cook S. C. & Snelling R. R. (2004) Liquidfeeding performances of ants (Formicidae): ecological and evolutionary implications. Oecologia 139, 255–66. Dejean A. & Corbara B. (2003) A review of mosaics of dominant ants in rainforests and plantations. In: Arthropods of Tropical Forests: Spatio-Temporal Dynamics and Resource Use in the Canopy (eds Y. Basset, V. Novotny, S. E. Miller, R. L. Kitching) pp. 341–7. Cambridge University Press, Cambridge. Dejean A. & Gibernau M. (2000) A rainforest ant mosaic: the edge effect (Hymenoptera: Formicidae). Sociobiology 35, 385–401. Dejean A., Akoa A., Djieto-Lordon C. & Lenoir A. (1994) Mosaic ant territories in an African secondary rain forest (Hymenoptera: Formicidae). Sociobiology 23, 275–92. Dejean A., Bourgoin T. & Gibernau M. (1997a) Ant species that protect figs against other ants: result of territoriality induced by a mutualistic homopteran. Ecoscience 4, 446–53. Dejean A., Djieto-Lordon C. & Durand J. L. (1997b) Ant mosaic in oil palm plantations of the southwest Province of Cameroon: impact on leaf miner beetle (Coleoptera: Chrysomelidae). J. Econ. Entomol. 90, 1092–6. Dejean A., Corbara B. & Orivel J. (1999) The arboreal ant mosaic in two Atlantic rain forests. Selbyana 20, 133–45. Dejean A., McKey D., Gibernau M. & Belin M. (2000) The arboreal ant mosaic in a Cameroonian rainforest (Hymenoptera: Formicidae). Sociobiology 35, 403–23. Delabie J. H. C. (2001) Trophobiosis between Formicidae and Hemiptera (Sternorrhyncha and Auchenorrhyncha): an overview. Neotrop. Entomol. 30, 501–16. Fellers J. H. (1987) Interference and exploitation in a guild of woodland ants. Ecology 68, 1466–78. Fiedler K. (2001) Ants that associate with Lycaeninae butterfly larvae: diversity, ecology and biogeography. Divers. Distrib. 7, 45–60. Floren A. & Linsenmair K. E. (1997) Diversity and recolonization dynamics of selected arthropod groups on different tree species in a lowland rainforest in Sabah, Malaysia with special reference to Formicidae. In: Canopy Arthropods (eds J. A. N. E. Stork & R. K. Didham) pp. 344–81. Chapman & Hall, London. Floren A. & Linsenmair K. E. (1998) Diversity and recolonization of arboreal Formicidae and Coleoptera in a lowland rain forest in Sabah, Malaysia. Selbyana 19, 155–61. Floren A. & Linsenmair K. E. (2000) Do ant mosaics exist in pristine lowland rain forests? Oecologia 123, 129–37. Floren A. & Linsenmair K. E. (2005) The importance of primary tropical rain forest for species diversity: an investigation using arboreal ants as an example. Ecosystems 8, 559–67. Floren A., Freking A., Biehl M. & Linsenmair K. E. (2001) Anthropogenic disturbance changes the structure of arboreal tropical ant communities. Ecography 24, 547–54. Floren A., Biun A. & Linsenmair K. E. (2002) Arboreal ants as key predators in tropical lowland rainforest trees. Oecologia 131, 137–44. Gibb H. & Hochuli D. F. (2004) Removal experiment reveals limited effects of an behaviorally dominant species on ant assemblages. Ecology 85, 648–57. Gilpin M. E. & Diamond J. M. (1982) Factors contributing to non-randomness in species co-occurrences on islands. Oecologia 52, 75–84. Gotelli N. J. (2000) Null model analysis of species co-occurrence patterns. Ecology 81, 2606–21. Gotelli N. J. & Ellison A. M. (2004) A Primer of Ecological Statistics. Sinauer Associates, Sunderland. Gotelli N. J. & Graves G. R. (1996) Null Models in Ecology. Smithsonian Institution, Washington. Grant P. R. & Abbott I. (1980) Interspecific competition, island biogeography and null hypotheses. Evolution 34, 322– 41. Greenslade P. J. M. (1971) Interspecific competition and frequency changes among ants in Solomon Islands coconut plantations. J. Appl. Ecol. 8, 323–52. Hölldobler B. (1979) Territories of the African weaver ant (Oecophylla longinoda (Latreille)). A field study. Z. Tierpsychologie 51, 201–13. Hölldobler B. (1983) Territorial behavior in the green tree ant (Oecophylla smaragdina). Biotropica 15, 241–50. Hölldobler B. & Lumsden C. J. (1980) Territorial strategies in ants. Science 210, 732–9. Hölldobler B. & Wilson E. O. (1990) The Ants. Harvard University Press, Cambridge. Hossaert-McKey M., Orivel J., Labeyrie E., Pascal L., Delabie J. H. C. & Dejean A. (2001) Differential associations with ants of three co-occurring extrafloral nectary-bearing plants. Ecoscience 8, 325–35. Jackson D. A. (1984a) Competition in the tropics: ants on trees. Antenna 8, 19–22. Jackson D. A. (1984b) Ant distribution patterns in a Cameroonian cocoa plantation: investigation of the ant mosaic hypothesis. Oecologia 62, 318–24. © 2007 Ecological Society of Australia ANT MOSAICS IN TROPICAL RAINFORESTS Kay A. (2002) Applying optimal foraging theory to assess nutrient availability ratios for ants. Ecology 83, 1935–44. Laidlaw M., Kitching R., Goodall K., Small A. & Stork N. E. (2007) Temporal and spatial variation in an Australian tropical rainforest. Austral Ecol. 32, 10–20. Leston D. (1973) The ant mosaic – tropical tree crops and the limiting of pests and diseases. PANS (Pest Articles News Summaries) 19, 311–41. Leston D. (1975) The ant mosaic: a fundamental property of cocoa farms. In: Proceedings of the 4th International Cocoa Research Conference, St Augustine, Trinidad January 8–18, 1972. pp. 570–581. Government of Trinidad and Tobago. Leston D. (1978) A neotropical ant mosaic. Ann. Entomol. Soc. Am. 71, 649–53. Majer J. D. (1972) The ant mosaic in Ghana cocoa farms. Bull. Entomol. Res. 62, 151–60. Majer J. D. (1976a) The ant mosaic in Ghana cocoa farms: further structural considerations. J. Appl. Ecol. 13, 145–55. Majer J. D. (1976b) The influence of ants and ant manipulation on the cocoa farm fauna. J. Appl. Ecol. 13, 157–75. Majer J. D. (1976c) The maintenance of the ant mosaic in Ghana cocoa farms. J. Appl. Ecol. 13, 123–44. Majer J. D. (1993) Comparison of the arboreal ant mosaic in Ghana, Brazil, Papua New Guinea and Australia – its structure and influence on arthropod diversity. In: Hymenoptera and Biodiversity (eds J. LaSalle & I. D. Gauld) pp. 115–41. CAB International, Wallingford. Majer J. D., Delabie J. H. C. & Smith M. R. B. (1994) Arboreal ant community patterns in Brazilian cocoa farms. Biotropica 26, 73–83. Majer J. D., Kitching R. L., Heterick B. E., Hurley K. & Brennan K. E. C. (2001) North-south patterns within arboreal ant assemblages from rain forests in eastern Australia. Biotropica 33, 643–61. Manly B. F. J. (1995) A note on the analysis of species co-occurrences. Ecology 76, 1109–15. Mercier J. L., Dejean A. & Lenoir A. (1998) Limited aggressiveness among African arboreal ants (Hymenoptera: Formicidae) sharing the same territories: the result of a co-evolutionary process. Sociobiology 32, 139–50. Mody K. & Linsenmair K. E. (2004) Plant-attracted ants affect arthropod community structure but not necessarily herbivory. Ecol. Entomol. 29, 217–25. Moran M. D. (2003) Arguments for rejecting the sequential Bonferroni in ecological studies. Oikos 100, 403–5. Morse D. R., Stork N. E. & Lawton J. H. (1988) Species number, species abundance and body length relationships of arboreal beetles in Bornean lowland rain forests trees. Ecol. Entomol. 13, 25–37. Perfecto I. (1994) Foraging behavior as a determinant of asymmetric competitive interaction between two ant species in a tropical agroecosystem. Oecologia 98, 184– 92. Ribas C. R. & Schoereder J. H. (2002) Are all ant mosaics caused by competition? Oecologia 131, 606–11. Room P. M. (1971) The relative distributions of ant species in Ghana’s cocoa farms. J. Anim. Ecol. 40, 735–51. Room P. M. (1975) Diversity and organization of the ground foraging ant faunas of forest, grassland and tree crops in Papua New Guinea. Aust. J. Zool. 23, 71–89. Room P. M. & Smith E. S. C. (1975) Relative abundance and distribution of insect pests, ants and other components of the cocoa ecosystem in Papua New Guinea. J. Appl. Ecol. 12, 31–46. © 2007 Ecological Society of Australia 103 Savolainen R. & Vepsäläinen K. (1988) A competition hierarchy among boreal ants: impact on resource partitioning and community structure. Oikos 51, 135–55. Schoener T. W. (1983) Field experiments on interspecific competition. Am. Nat. 122, 240–85. Seufert P. & Fiedler K. (1996) The influence of ants on patterns of colonization and establishment within a set of coexisting lycaenid butterflies in a south-east Asian tropical rain forest. Oecologia 106, 127–36. Shattuck S. O. (1999) Australian Ants: Their Biology and Identification. CSIRO Publishing, Collingwood. Stone L. & Roberts A. (1990) The checkerboard score and species distributions. Oecologia 85, 74–9. Stork N. E. (1987) Arthropod faunal similarity of Bornean rain forest trees. Ecol. Entomol. 12, 219–26. Stork N. E. (1991) The composition of the arthropod fauna of Bornean lowland rain forest trees. J. Trop Ecol. 7, 161– 80. Stork N. E. (2007) Australian tropical forest canopy crane: New tools for new frontiers. Austral Ecol. 32, 4–9. Stork N. E. & Brendell M. J. D. (1990) Variation in the insect fauna of Sulawesi trees in season, altitude and forest type. In: Insects and the Rain Forests of South East Asia (Wallacea) (eds W. J. Knight & J. D. Holloway) pp. 173–90. Royal Entomological Society of London, London. Taylor B. (1977) The ant mosaic on cocoa and other tree crops in Western Nigeria. Ecol. Entomol. 2, 245–55. Way M. J. (1953) The relationships between certain ant species with particular reference to biological control of the coreid, Theraptus sp. Bull. Entomol. Res. 44, 669–91. Way M. J. (1963) Mutualism between ants and honeydew producing Homoptera. Annu. Rev. Entomol. 8, 307–44. Way M. J. & Bolton B. (1997) Competition between ants for coconut palm nesting sites. J. Nat. Hist. 31, 439–55. Way M. J. & Khoo K. C. (1991) Colony dispersion and nesting habits of the ants, Dolichoderus thoracicus and Oecophylla smaragdina (Hymenoptera: Formicidae), in relation to their success as biological control agents on cocoa. Bull. Entomol. Res. 81, 341–50. Yu D. W., Wilson H. B. & Pierce N. E. (2001) An empirical model of species coexistence in a spatially structured environment. Ecology 82, 1761–71. Yanoviak S. P. & Kaspari M. (2000) Community structure and the habitat templet: ants in the tropical forest canopy and litter. Oikos 89, 259–66. APPENDIX I Ant species collected from the canopy and under-storey vegetation at the Australian canopy crane site (Cape Tribulation, North-East Queensland) and surrounding forests Subfamilies: DOL, Dolichoderinae; FOR, Formicinae; MYR, Myrmicinae; PON, Ponerinae; PSE, Pseudomyrmecinae. Anonychomyrma gilberti (DOL), Camponotus (Colobobsis) macrocephalus (FOR), Ca. sp.1 (macrocephalus gp.), Ca. sp.3, Ca. sp.5 (both novaehollandiae gp.), Ca. sp.4 (macrocephalus gp.), Ca. sp.6 104 N. B L Ü T H G E N A N D N. E . S TO R K (gasseri gp.), Ca. sp.7 (extensus gp.), Ca. vitreus, Cardiocondyla wroughtonii (MYR), Crematogaster aff. fusca (MYR), Cr. aff. pythia, Cr. sp.3, Echinopla australis (FOR), Heteroponera sp.1 (PON), Leptomyrmex unicolor (DOL), Monomorium cf. intrudens (MYR), Mo. fieldi var. laeve nigrius, Mo. floricola, Odontomachus cephalotes (PON), Oecophylla smaragdina (FOR), Pachycondyla sp.1 (PON), Paratrechina minutula gp. (FOR), Pa. vaga gp., Pheidole nr. impressiceps (MYR), Ph. aff. platypus, Ph. sp.1, Ph. sp.4, Ph. sp.5, Ph. sp.11, Ph. sp.13, Ph. cf. athertonensis, Pheidologeton affinis (MYR), Podomyrma sp1 (MYR), Polyrhachis ‘Cyrto 03’, Po. ‘Cyrto 06’, Po. ‘Cyrto 08’, Po. ‘Cyrto NB5041’, Po. yorkana (all Cyrtomyrma), Po. (Hagiomyrma) thusnelda, Po. barretti, Po. cupreata (both Hedomyrma), Po. (Myrma) foreli, Po. (Myrmatopa) lobokensis, Po. (Myrmhopla) mucronata, Po. (Myrmothrinax) delicata, Rhopalomastrix rothneyi (MYR), Rhoptromyrmex wroughtonii (MYR), Rhytidoponera kurandensis (PON), Rh. purpurea, Rh. spoliata, Rh. sp.1, Solenopsis sp.1 (MYR), Strumigenys guttulata (MYR), Tapinoma melanocephalum (DOL), Ta. minutum gp., Technomyrmex aff. albipes (DOL), Te. quadricolor, Te. sp.3, Tetramorium bicarinatum (MYR), Tet. insolens, Tet. validiusculum, Tetraponera nitida (PSE), Turneria bidentata (DOL), Vollenhovia sp1 (MYR), Vombisidris australis (MYR). Underlined species names or name additions are corrections from previous publications (Blüthgen et al. 2003; Blüthgen & Fiedler 2004a,b; Blüthgen et al. 2004b) by Alan Andersen (pers. comm. 2004). © 2007 Ecological Society of Australia
© Copyright 2026 Paperzz