Oecologia (2002) 131:137Ð144 DOI 10.1007/s00442-002-0874-z

COMMUNITY ECOLOGY

Andreas Floren á Alim Biun á K. Eduard Linsenmair Arboreal as key predators in tropical lowland rainforest trees

Received: 24 September 2001 / Accepted: 2 January 2002 / Published online: 14 February 2002 © Springer-Verlag 2002

Abstract Ants numerically dominate the canopy fauna on the ground and lower vegetation (Hölldobler and of tropical lowland rain forests. They are considered to Wilson 1990), and fogging studies demonstrate that they be key predators but their effects in this regard have only also dominate in the canopy (Erwin 1983; Stork 1991; rarely been studied on non-myrmecophytes. A conspicu- Floren and Linsenmair 1997; Wagner 1997; Adis et al. ously low abundance of less mobile, mainly holometa- 1998). A conspicuously low abundance of less mobile bolous like larvae corresponds arthropods in the trees, mainly larvae of holometabolous with dominance, while hemimetabolous highly taxa like Lepidoptera- or Coleoptera, which contribute mobile nymphs occur regularly and in large numbers in less than 1% to a maximum 3% of all arthropods, corre- the trees. This is in contrast to the temperate regions lates with ant dominance. In contrast, highly mobile where ants are mostly lacking on trees and holometabol- nymphs of hemimetabolous taxa provide, on average, ous larvae are frequent. In this study we experimentally 17% of individuals to a community (Floren and Linsenmair measured ant predation in the trees by offering caterpil- 1997,1998a; Stork 1991; Wagner 1997). This is in con- lars as baits. Fifty-four ant species were tested, of which trast to the temperate regions where tree communities 46 killed caterpillars and carried them away to their nests contain only few ants and numerous less mobile arthro- while only eight species ignored the offered larvae. pods (e.g. Horstmann 1976/1977; Southwood et al. 1982; Insecticidal knockdown fogging of ten trees after finish- Simandl 1993; Wagner 1996; A. Floren unpublished ing the prey experiments showed that on average 85% of data). These results suggest that the arboreal ants strongly ant individuals per tree were predacious. With the analy- influence the composition of the fauna in trees sis of another 69 foggings and meticulous observations of tropical lowland forests by exerting a high predation in many other trees this suggests that arboreal ants are pressure. This might also explain in part the extraordi- responsible for the low abundance of less mobile arthro- narily high dynamics in species composition in tree- pods in tropical lowland rain forest canopies. Ant preda- associated communities (expressed as fast species turnover) tion was significantly lower in a disturbed forest indicat- as found for example in daily re-fogging experiments of ing that human disturbance induces a change in the func- individual trees (Floren and Linsenmair 1997, 1999, tional interactions in these ecosystems. unpublished data; see also Stork 1991). Myrmecophilous and myrmecophytic plants make use of ants as protective Keywords Anthropogenic disturbance á Canopy á agents against herbivores by attracting ants with food Community structure á Ecosystem function á Fogging and/or shelter in specialized plant structures (Fiala and Maschwitz 1992; Fiala et al. 1994; Fonseca and Ganade Introduction 1996; Bronstein 1998; Heil et al. 2000, 2001; Oliveira et al. 1987; Letourneau and Dyer 1998). Because such trees Formicidae are considered to be the most abundant and are only a minority in primary forests, their special rela- most important predators in tropical lowland rain forests tionships provide us with no explanation of the general scarcity of less mobile arthropods in tropical lowland rain A. Floren (✉) á K.E. Linsenmair forest trees. Until now, however, no close attention has Department of Ecology and Tropical Biology, been paid to the effects that ant predation pressure exert Biozentrum, Am Hubland. University Wuerzburg, 97074 Wuerzburg, Germany on the composition of arthropod communities in canopies e-mail: [email protected] of the large majority of trees that are neither myrmecophil- Tel.: +49-931-8884376, Fax: +49-931-8884352 ous nor myrmecophytes. A. Biun This is the first study to measure ant predation in Sabah Parks, Peti Surat 10626, 88806 Kota Kinabalu, Malaysia tropical lowland rain forest trees by offering caterpillars 138 as prey and by quantifying attack rates of various ant tree were tested. If ant individuals did not behave in a uniform species. By fogging ten trees after finishing the prey- way, the behaviour most often observed was used to categorize them. However, inconsistent results were only recorded in a few offering experiments we were able to assess the pro- cases. The caterpillars showed either no recognizable reaction or a portion of predacious to non-predacious ants in these clear defence and/or escape behaviour (strong beating of the body communities. in order to shake off the ants or spinning down from the branch or leaf on a thread of silk). In these experiments we assume that ant behaviour in response Materials and methods to the baits does not differ from those of ants which forage in a tree and accidentally meet a caterpillar. This assumption corre- sponds with the results of earlier studies (Goetzke 1993; Floren Study site and Linsenmair 1997; Berghoff 1998). In order to test whether ants that were attracted by tuna baits reacted differently compared Field work was done in January and February 2000 in the tropical to ants that met a larva accidentally, individual caterpillars were lowland rain forest of Kinabalu National Park, substation Poring put on leaves and observed until they were encountered by an ant. Hot Springs, in Sabah, Malyasia (6¡2.75′N, 116¡42.2′E). Details about the study area have been published elsewhere (Floren and Linsenmair 1997, 2000). Insecticidal fogging In order to determine relative proportions of predacious to non- Tree species chosen for the prey experiments predacious ants in individual communities, ten trees out of the same sample were fogged after the prey experiments were For carrying out the prey experiments we selected 45 trees of the finished. Because abundances of most ant species are represented lower canopy stratum, between 20 and 30 m high. Twenty-six trees in approximately correct proportions in the fogging samples, it is belonged to the species Aporusa lagenocarpa (Euphorbiaceae)*, possible to roughly assess Ð based on the comparison with the tuna the focal tree of our earlier studies (Floren and Linsenmair 1997, bait data Ð the ratio of predatory to non-predatory ant specimens in 2000), and ten were Depressa nervosa trees (Guttiferae). Further- a particular tree. Comprehensive information about the method of more, ants were tested that lived in the following trees: Barringtonia insecticidal fogging is given in Floren and Linsenmair (1997, scortechinii* (two trees), B. gitingensis (Lecythidaceae), Cryptoc- 2000, 2001) and Adis et al. (1998). Since we have no indication of arya sp. (Lauraceae), Payena lucida (Sapotaceae), Aporusa the existence of tree-specific ant species (Floren and Linsenmair maingayi (Euphorbiaceae)*, Dacryodes laxa (Burseraceae)*, 1997, 2000), the selection of the trees for fogging followed practical Palaquium rostratum (Sapotaceae)* and Ochanostachys amentacea criteria. Trees should be fully grown and should grow in similar (Olacaceae)*. Ten trees of the species marked with an asterisk habitats, as regards for example edaphic factors, degree of shading, were fogged after the prey experiments finished (see “Insecticidal etc. (Floren and Linsenmair 1998b). Tree structural parameters, fogging” below). such as the amount of dead wood, should be similar and tree cover by lianas and epiphytes should be low. No myrmecophilous nor myrmecophytic trees were investigated [however, a highly special- The prey species ized ant species of the Cladomyrma () which lived in a liana growing in a tree was tested in the prey experi- In most ecosystems, larvae of Lepidoptera are important herbi- ments]. The ten trees selected for fogging were four A. lagenocarpa, vores (e.g. Koptur 1984; Smiley 1987; Letourneau et al. 1993). two B. scortechinii, and the other trees given above. For our experiments we chose caterpillars of a species, genus Onebala, which were found in leaf-rolls of Hibiscus shrubs. No other species of Lepidoptera larvae were found in Investigations in a disturbed forest greater abundance. Onebala specimens are not hairy and we have no indication that they are chemically defended (C. Schulze, Earlier investigations indicated that ant predation, and thus their personal communication). Caterpillars of two size classes were effects on arthropod communities, might differ between primary used in the experiments: (1) large individuals of 2Ð2.5 cm size, and disturbed forests. These investigations were carried out in and (2) small larvae of about 1 cm. three forests of different disturbance level. They were clear cut, used some years for crop planting, abandoned, and left to natural regeneration 5, 15, and 40 years ago. From each forest ten con- The prey-offering experiments specific trees were fogged. Details are given in Floren and Linsenmair (1998b) and Floren et al. (2001a). Three years ago, Because a number of ant individuals per species was needed for these forests were completely destroyed by a fire. the prey experiments, ants were attracted to the trees with pieces In order to test whether ant predation differed between the of tuna bait that were placed on leaves and most branches within primary and a disturbed forest we included a strongly disturbed reach of the stem. We have shown earlier that most arboreal ant forest in our study. It was around 500 m from the primary forest, species are generalist feeders that can be attracted with tuna baits. separated by private gardens and the entrance area of the National In fact, this method allows one to map the ant nests of a tree Park. The forest was used as pasture for water buffaloes. The crown (for details see Floren and Linsenmair 1997, 2000). The oldest trees were around 40 years old and reached 20 m height. ants had already discovered 80% of the baits after 15 min. A large Only a single canopy layer, not yet closed, had formed. Ant caterpillar was then placed carefully ca. 10 cm from the bait. After species found on 19 non-myrmecophytic or non-myrmecophilous an ant had discovered the caterpillar, the reactions of both ant and trees were used for the experiments. These were Chionanthus caterpillar were recorded. In the case that the ants did not attack a pluriflorus (Oleaceae), Guioa pleuropteris (Sapindaceae), Glochidium large larvae, a small caterpillar was offered a few minutes later in sp. (Euphorbiaceae), Ficus septica (Moraceae) and Artocarpus a new trial using new ant individuals. After ants had encountered communis (Moraceae). the caterpillars, the following three situations were distinguished: (1) ants which directly attacked and killed the caterpillars which carried the prey away into their nests, (2) ants which did not react Results to large caterpillars but attacked small larvae, (3) ants which ignored caterpillars. Each experiment consisted of five replicates with the same ant species in which new ant individuals always In the primary forest a total of 188 prey experiments on took part. Between two and seven of the larger ant colonies per 45 trees were carried out with 54 ant species. Ant species 139 Fig. 1 Attack rates of arboreal ant species on experimentally offered caterpillars. Black bars represent relative proportion of test results (n=188). White bars represent number of species per behavioural category (n=54). For more information see text

which had several nests in a tree or which occurred in were collected by fogging (Table 1). The number of spe- different trees might have been tested several times cies varied between seven and 20 per tree, however, because it was not always possible to asses whether they without considering the rare species (those found with belonged to the same species. The results of the prey- less than five individuals) which represented 67 species offering experiments are summarized in Fig. 1. In 96 trials (46.9%) of all species collected. Of the more common 76 (51.1%) the ants immediately attacked large caterpillars species (53.1%), 54 (37.8%) were tested in the prey (≥2 cm) and in another 76 tests (40.4%) they did not experiments, including all large colonies. This corresponds react to large larvae but attacked small ones (≤1 cm). Only to an average of 48.9% (SD 10.8) of the more common in 16 tests (8.5%) did the ants not react at all to the prey. ant species found on an individual tree. Because these These frequencies differ highly significantly from those species had established larger colonies they also repre- expected on the basis of the null hypothesis that both sented most of the ants of a fogged community, namely kinds of behaviour were equally common (χ2=55.319; 85.7% (SD 11.4). The average number of species that P<0.001). If species numbers are considered, then 29 ant attacked large caterpillars per tree was 19.4% (SD 8.4, species (53.7%) attacked large caterpillars while only median 17.7) and they contributed 37.1% (SD 33.8, eight species (14.8%) ignored them completely. If the median 14.3) of all individuals to a community. After first two behavioural categories are combined then there adding ants that attacked only small caterpillars, the mean were 46 species (85.2%) and 172 tests (91.5%) in which abundance of predatory ant individuals per community at least small caterpillars were attacked by the ants. rose to 79.7% (SD 12.5, median 82.6), which differs Table 1 gives the taxonomic composition of the ant species highly significantly from the abundance of non-predatory used in the prey experiments. In the nests of three of the ants, that contributed on average 6.0% (SD 5.0, median eight ant species which did not attack caterpillars (one 5.4; Mann Whitney U-test, P<0.001). species each of the genera Crematogaster, Monomorium, In order to test whether ants that were attracted with and Oligomyrmex) we found scale (Coccoidea, baits behaved differently to those which randomly Homoptera). The remaining non-predacious species nested encountered a larvae, we placed individual caterpillars in stem cavities or dead wood and we do not know on leaves and waited until they were found by an ant. whether they hosted associated coccoids or not. These trials were difficult to perform since the caterpillars moved quickly away in most cases. In 18 cases an encounter was documented. In four cases caterpillars The insecticidal foggings were grasped and carried away by a large Polyrhachis ant that foraged alone. With the exception of two cases, Ants contributed between 44.5% and 66.4% of arthro- in which ant and caterpillar ignored each other, caterpillars pods to these tree communities. In total, 143 species fled immediately. 140 Table 1 Differences in the number of predatory ant Subfamily Genus Primary forest Disturbed forest species in a primary and a disturbed forest Number of Tested as Number of Tested as species non-predacious species non-predacious

Myrmicinae Crematogaster 13 3 3 2 Monomorium 31 0 Oligomyrmex 41 0 Pheidole 21 Pheidologeton 01 Tetramorium 10 Cladomyrma 10 Meranoplus 10 Pristomyrmex 01 Formicinae Camponotus 10 2 4 3 Polyrhachis 521 Oecophylla 01 Echinopla 20 Lepisiota 10 Paratrechina 20 Dolichoderus 411 Technomyrmex 211 Tapinoma 21 1 1 Philidris 011 Ponerinae Diacamma 11 Pseudomyrmecinae Tetraponera 033 Total 54 8 21 13 Species number collected 143 by fogging

Fig. 2 Correlation between the relative proportion (Rel. prop.) of ants and caterpillars in trees of a primary and three disturbed regenerating forests of 5, 15, and 40 years of age. Ten trees from each forest were fogged. Primary forest trees, trees of the 40-year-old forest, trees of the 15-year-old forest, trees of the 5-year-old forest

The prey experiments in the disturbed forest 2001). This was most clearly demonstrated by less mobile larvae of Lepidoptera (Floren and Linsenmair In an earlier study we found that primary forest commu- 1999). Figure 2 shows that their numbers significantly nities changed significantly on the ordinal level in dis- increased in the disturbed forests where ants had turbed secondary forests (Floren and Linsenmair 1999, strongly decreased (Spearman’s rank correlation, ρ=Ð0.733, 141 Table 2 Abundance of ants and caterpillars in primary forest and Primary forest Disturbed forest (years of regrowth) three forests of different distur- bance level. Ten conspecific trees 5 Years 15 Years 40 Years from each forest were fogged % Ants per tree community 57.21±10.94 21.00±14.59 48.37±10.05 30.89±7.51 % Caterpillars per tree community 0.88±0.59 8.81±6.77 4.29±1.84 4.49±2.89 a Standardization of data for Range of caterpillars sampled per tree 21Ð68 50Ð480 70Ð185 31Ð445 a crown projection of 1 m2 and Standardized numbers of caterpillarsa 2.23±0.53 10.32±4.35 6.55±3.29 11.31±6.10 a leaf cover of 100%

Fig. 3 Comparison of ant predation in a primary and in a disturbed forest, measured by attack rates on caterpillars; 188 prey experiments were carried out in the primary forest, 28 in the disturbed forest

P<0.001, r2=0.537). In contrast to the primary forest, the the average 46% of predatory ants per community in the relative proportion of caterpillars varied largely in the disturbed forest did not differ significantly from the 54% disturbed forests. Table 2 gives more details for the for- of non-predatory ants (Wilcoxon test, P=0.933). These ests investigated. With the exception of the 15-year-old differences in the proportion of ant predation in both forest, relative proportions of ants were significantly forests were Ð not surprisingly Ð also reflected in the lower in the disturbed forests than in the primary forest species composition (Table 1). Relative proportions of (Mann Whitney U-tests after Bonferoni correction, non-predacious ant species increased from 14.8% in the P<0.001, for the comparison of the primary forest with primary forest to 61.9% in the disturbed forest. This was the 15-year-old forest, P=0.123). In contrast, numbers of most obvious in the species-rich genera Camponotus caterpillars were always significantly higher in the dis- (Formicinae) and Crematogaster (Myrmicinae). In turbed forests (P<0.001) while they did not differ among contrast to the primary forest, were no Tetraponera spp. each other. This is also demonstrated by comparing (Pseudomyrmecinae) occurred on the ten trees fogged, absolute numbers of caterpillars (data were standardized three species were found in the disturbed forest and all according to 1 m2 and 100% leaf cover, U-tests ignored the offered prey caterpillars. P<0.001). However, many caterpillars were found again in the 15- and 40-year-old forest types although ants had become the numerically dominant group again. In order Discussion to explain high numbers of caterpillars we hypothesized that ants in the disturbed forests were less predacious Ants numerically dominate the canopy of tropical lowland than in the primary forest (Floren and Linsenmair 1999, rain forests worldwide (Erwin 1983; Stork 1991; Floren 2001). This hypothesis was tested in a preliminary study. and Linsenmair 1997; Wagner 1997; Adis et al. 1998). In From 19 trees we tested 21 ant species in 28 prey our study area, they provide on average 60% of all experiments. Figure 3 compares ant predation in the pri- arthropods of a tree community (Floren and Linsenmair mary and in the disturbed forest. The average proportion 1997). It is generally assumed that ants exert a high of predatory ants per tree in the primary forest differed predation pressure on the arthropod fauna, however, highly significantly from the proportion of non-predatory evidence for this comes almost exclusively from ants (Wilcoxon signed rank test, P<0.001). In contrast, myrmecophilous and myrmecophytic plants (e.g. Koptur 142 1984; Smiley 1987; Letourneau et al. 1993; Gaume et al. 1. Many of the small ant colonies in the trees were not 1997; Heil et al. 2000, 2001) while studies on trees that tested, and we have no reason to assume that these are not associated with ants are rare (Memmott et al. species would behave in a different way to those used 1993). In this study we tested whether ants also exert a in the prey experiments. high predation pressure on arthropods on non-myrmeco- 2. Many predatory species which were not found nesting philous and non-myrmecophytic trees. We quantified ant in a study tree remained unconsidered, like army ants behaviour on the basis of experimental data in standard- or big species with large foraging areas like Polyrhachis ized prey-offering experiments in the trees and found spp. (Dejean et al. 1994). that the majority of ant species (46 of the tested 54 species) 3. The abundance of stem-inhabiting species, of which attacked and killed the caterpillars and carried them many were aggressive predators, was certainly away to their nests. The few non-predacious species underestimated from the fogging samples because the (eight of 54) were all relatively small (between 1 mm insecticidal fog does not affect ants hidden in stem and 3 mm) and at least the carton-nesting species among cavities. them contained scale insects (Homoptera) cultivated for 4. Nocturnal species were not taken into consideration. nutritional supply. Fogging ten trees subsequent to the 5. Highly specialized predatory ant species which do not prey-offering experiments showed that predatory ants prey on caterpillars were not considered. always represented more than two-thirds of ant individuals in a tree. Dominance of predacious ants in the canopy High ant-predation pressure raises the question of where was further confirmed by the results of the sorting of do particularly highly diverse Lepidoptera of tropical another 69 fogging samples from this forest which con- forests develop. Probably, the lower vegetation is of tained many of the tested predacious species (A. Floren, greater significance where caterpillars can be found unpublished data) and also by two studies on the organi- regularly (own observations). We know of no detailed zation of arboreal ant communities (Götzke 1993; Berghoff investigation, however, which addresses this question. 1998). These results strongly indicate that ants indeed Occasional observations showed that ants also preyed exert a high predation pressure on the arthropod fauna on other species of offered caterpillars, independently of and that there are probably only few trees, at least in the whether they were covered by hairs or not (unpublished rain forest investigated by us, on which less mobile ar- observations). thropods may be safe from predacious ants. The omni- In order to test whether ants that were attracted with presence of predacious ants is very likely to be the rea- baits behaved differently to those which randomly son for the pronounced rarity of less mobile arthropods encountered a larvae, we placed individual caterpillars of holometabolous taxa (mainly larvae), which only pro- on leaves and waited until they were found by an ant. vide around 1% of arthropod individuals in most com- When they met, the caterpillar immediately fled. In four munities. That the low numbers of caterpillars in the rain cases they were killed by individually foraging Polyrhachis forest trees are not a methodological artefact is further- workers. These results support the interpretation that more suggested by the absence of lepidopteran parasito- arboreal primary forest ants are highly predacious. Besides ids among the Ichneumonidae () (Horstmann caterpillars, it is well known that ants also prey inten- et al. 1999). In contrast, the nymphs of hemimetabolous sively on other arthropods like termites (Oliveira et al. Hemiptera and Orthoptera, which are mostly very mobile 1987) or leaf-mining insects (Memmott et al. 1993). It and can usually evade foraging ants, occur regularly in has been demonstrated in temperate forests that ant preda- the trees, contributing around 17% of arthropods to a tion varies during the course of the seasons (Horstmann community (Floren and Linsenmair 1997, 1999; Floren 1975, 1976/1977). We have, however, no indications et al., 2001b). These results contrast with those of the from foggings and observations in the trees carried out temperate regions where arboreal ants are lacking in during different times in the year that such changes most trees while less mobile arthropods occur in high occur in the forest investigated (Floren and Linsenmair numbers (e.g. Horstmann 1976/1977; Simandl 1993; 2000; Horstmann et al. 1999). Wagner 1996; A. Floren, unpublished data). However, ground-living ants can also enter trees in great numbers in the temperate regions, like species of Formica, and Comparing ant predation pressure in primary influence communities in a similar way as in the tropics and disturbed forests by reducing particularly less mobile arthropods, while some honey dew-producing species profit from the Our fogging studies in differently disturbed forest types presence of the ants (Laine and Niemelä 1980; Ito and showed that caterpillars can occur in high numbers in Higashi 1991; Whittaker 1991; Mahdi and Whittaker trees although ants are numerically dominant. Hypotheti- 1993; Karhu 1998; Horstmann 1999). cally we argued that there might be fewer predatory ant These results support our earlier finding that most of species in disturbed forests (Floren and Linsenmair the arboreal ant species are generalist feeders which 1999). A first test of this hypothesis was carried out in a opportunistically use every food item available (Floren strongly disturbed forest which was comparable to the and Linsenmair 1997, 2000). However, in our study, ant forest where the cited studies were performed. The original predation was still rather underestimated because: study area was completely destroyed by a fire in 1998. 143 Despite the comparatively small sample size the data in- the African equatorial forest (Hymenoptera: Formicidae, dicate that ant predation is indeed distinctly lower than Formicinae). Sociobiology 23:293Ð313. Erwin LE (1983) Beetles and other insects of tropical forest in the primary forest [based on our earlier results (Floren canopies at Manaus, Brazil, sampled by insecticidal fogging. and Linsenmair 2001) we estimate that the 21 species Tropical rainforest: ecology and management. Blackwell, tested in the prey experiments represent most of the Oxford, pp 59Ð75 common ant species in this forest type; compare Ta- Fiala B, Maschwitz U (1992) Domatia as most important adaptations in the evolution of myrmecophytes in the paleotropical tree ble 1]. This is most probably a consequence of the strong genus Macaranga (Euphorbiaceae). Plant Syst Evol 180: biotic and abiotic changes which caused a complete 53Ð64 change in ant communities (Floren and Linsenmair 1999, Fiala B, Grunsky H, Maschwitz U, Linsenmair KE (1994) 2001). For example, ant communities in the primary for- Diversity of ant-plant interactions: protective efficancy in est could not be distinguished from random communi- Macaranga species with different degrees of an ant association. Oecologia 97:186Ð192 ties, while communities in the disturbed forest showed a Floren A, Linsenmair KE (1997) Diversity and recolonisation clearly deterministic pattern (Floren et al. 2001a). dynamics of selected arthropod groups on different tree One might argue that the differences in arthropod com- species in a lowland rain forest in Sabah, Malaysia with munities and in the extent of ant predation are not a conse- special reference to Formicidae. In: Stork NE, Adis JA, Didham RK (eds) Canopy arthropods. Chapman and Hall, quence of the disturbance but a result of the investigation London, pp 344Ð382 of two different canopy strata (the lower canopy of a mul- Floren A, Linsenmair KE (1998a) Non-equilibrium communities tilayered primary forest and the monolayered canopy of of Coleoptera in trees in a lowland rain forest of Borneo. the disturbed forest). There are, however, several points Ecotropica 4:55Ð67 Floren A, Linsenmair KE (1998b) Diversity and recolonisation of that provide arguments against such an assumption. arboreal Formicidae and Coleoptera in a lowland rain forest in Among the 41 fogged primary forest trees were eight trees Sabah, Malaysia. Selbyana 19:155Ð161 which were directly exposed to the sun, including a high- Floren A, Linsenmair KE (1999) Changes in arboreal arthropod canopy tree (Aglaia sp., Melicaeae), but these communi- communities along a disturbance gradient. Selbyana 20: ties were not different from those of the shaded trees (A. 284Ð289 Floren A, Linsenmair KE (2000) Do ant mosaics exist in pristine Floren, unpublished data). In addition, communities of ten lowland rain forest? Oecologia 123:129Ð137 trees of the high canopy that were fogged by Stork (1991) Floren A, Linsenmair KE (2001) The influence of anthropogenic in Brunei showed no differences to the communities disturbances on the structure of arboreal arthropod communities. fogged by us. In contrast, differences in communities of Plant Ecol 153:153Ð167 Floren A, Freking A, Biehl, M, Linsenmair KE (2001a) Anthropo- the disturbed forests were clearly recognizable. We have genic disturbance changes the structure of arboreal tropical ant therefore no reason to assume that differences in tree com- communities. Ecography 24:547Ð554 munities were mainly due to the selection of habitat. The Floren A, Riede K, Ingrisch S (2001b) Diversity of Orthoptera reported differences in the proportion of predacious ants from Bornean lowland rainforest trees. Ecotropica 7:33Ð42 Fonseca CR, Ganade G (1996) Asymmetries, compartments and indicate that a functional change of the influence of arbo- null interactions in an Amazonian ant-plant community. real ants on the arthropod fauna occurs in disturbed for- J Anim Ecol 65:339Ð347 ests. The conditions under which such a change takes Gaume L, McKey D, Anstett MC (1997) Benefits conferred by place are currently being investigated. “timid” ants: active anti-herbivore protection of the rainforest tree Leontodoxa africana by the minute ant Petalomytrmex phylax. Oecologia 112:209Ð216 Acknowledgements We thank the director of Sabah Parks, Datuk Götzke A (1993) Ameisenzönosen ausgewählter tropischer Ali Lamri, for the permission to work in the Kinabalu Park. For Baumkronen: Struktur, Diversität und Ressourcennutzung der their participation in the field work we are very much indebted to Gemeinschaft. Department of Animal Ecology and Tropical Michael Matzat, Konstans Wells and Stefan Otto. We thank Chris- Biology, Julius-Maximilians-University, Wuerzburg tian Schulze who identified the caterpillars used for the prey ex- Heil M, Koch T, Hilpert A, Fiala B, Linsenmair KE (2000) periments and Thomas Wagner, Martin Heil, Alain Dejean and an Extrafloral nectar production of the ant-associated plant, anonymous referee for critical comments on the manuscript. Fi- Macaranga tanarius, is an induced, indirect, defensive nancial support for this study came from the German Science response elicited by jasmonic acid. Proc Natl Acad Sci USA Foundation (DFG), Li 150/1-4. 98:1083Ð1088 Heil M, Fiala B, Maschwitz U, Linsenmair KE (2001) On benefits of indirect defence: short-and long-term studies of antiherbivore protection via mutualistic ants. Oecologia 126:395Ð403 References Hölldobler B, Wilson EO (1990) The ants. Belknap, Cambridge, Mass. Adis J, Harada Y, Fonseca C R, Paarmann W, Rafael JA (1998) Horstmann K (1975) Zur Regulation des Beuteeintrages bei Arthropods obtained from the Amazonian tree species Waldameisen (Formica polyctena Foerster). Oecologia 22: “Cupiuba” (Goupia glabra) by repeated canopy fogging with 57Ð65 natural pyrethrum. Acta Amazonica 28:273Ð283 Horstmann K (1976/1977) Waldameisen (Formica polyctena Berghoff S (1998) Strukturierungsmechanismen arborealer Ameisen- Foerster) als Abundanzfaktoren für den Massenwechsel des zönosen in einem Tieflandregenwald in Sabah, Malaysia, Eichenwicklers Tortrix viridana L. Z Angew Entomol 82: Borneo. Department of Animal Ecology and Tropical Biology. 421Ð235 Julius-Maximilians-University, Wuerzburg Horstmann J (1999) Der Einflu§ der kleinen Waldameisen (Formica Bronstein JL (1998) The contribution of ant-plant protection to polyctena Foerster) auf die Zusammensetzung und Diversität our understanding of mutualism. Biotropica 30:150Ð161 arborikoler Arthropodengemeinschaften in Eichen. Department Dejean A, Lenoir A, Godzinska EJ (1994) The hunting behavior of Animal Ecology and Tropical Biology, Bayerische Julius- of Polyrhachis laboriosa, a non-dominant arboreal ant of Maximilians-University, Wuerzburg 144 Horstmann K, Floren A, Linsenmair KE (1999) High species- vegetation: ants as potential antiherbivore agents. Oecologia richness of Ichneumonidae from the canopy of a Malaysian 74:228Ð230 rain forest. Ecotropica 5:1Ð12 Simandl J (1993) Canopy arthropods on Scots pine: influence of Ito F, Higashi S (1991) An indirect mutualism between oaks and season and stand age on community structure and the position wood ants via aphids. J Anim Ecol 60:463Ð470 of sawflies (Diprionidae) in the community. For Ecol Manage Karhu KI (1998) Effects of ant exclusion during outbreaks of a 62:85Ð98 defoliator and a sap-sucker on birch. Ecol Entomol 23:185Ð194 Smiley JT (1987) Heliconius caterpillar mortality during establish- Koptur S (1984) Experimental evidence for defense of Inga ment on plants with and without attending ants. Ecology (Mimosoideae) saplings by ants. Ecology 65:1787Ð1793 65:1787Ð1793 Laine KJ, Niemelä P (1980) The influence of ants on the survival Southwood TRE, Moran VC, Kennedy EJ (1982) The richness, of mountain birches during an Oporinia autumnata (Lep. abundance and biomass of the arthropod communities on Geometridae) outbreak. Oecologia 47:39Ð42 trees. J Anim Ecol 51:635Ð649 Letourneau DK, Dyer LA (1998) Density patterns of Piper ant-plants Stork NE (1991) The composition of the arthropod fauna of and associated arthropods: top-predator trophic cascades in a Bornean lowland rain forest trees. J Trop Ecol 7:161Ð180 terrestrial system? Biotropica 30:162Ð169 Wagner T (1996) Artenmannigfaltigkeit baumkronenbewohnender Letourneau DK, Arias F, Jebb M (1993) Coping with enemy-filled Arthropoden in zentralafrikanischen Wäldern, unter besonderer space: herbivores on Endospermum in Papua New Guinea. Berücksichtigung der Käfer. Rheinische Friedrich-Wilhelm- Biotropica 25:95Ð99 University, Bonn Mahdi T, Whittaker JB (1993) Do birch trees (Betula pendula) Wagner T (1997) The beetle fauna of different tree species in grow better if foraged by wood ants? J Anim Ecol 62:101Ð116 forests of Rwanda and east Zaire. In: Stork NE, Adis JA, Memmott J, Godfray HCJ, Bolton B (1993) Predation and Didham RK (eds) Canopy arthropods. Chapman and Hall, parasitism in a tropical herbivore community. Ecol Entomol London, pp 344Ð382 18:348Ð352 Whittaker JB (1991) Effects of ants on temperate woodland trees. Oliveira PS, Silva AF, Martins AB (1987) Ant foraging on extra- In: Huxley CR, Cutler DF (eds) Ant-plant interactions. Oxford floral nectaries of Qualea grandiflora (Vochysiaceae) in cerrado University Press, Oxford, pp 67Ð79