Ant gardens (AGs) are specialized ant-plant associations where arboreal ants build their carton nests in association with epiphytes that use the carton as a substrate. Most of the epiphytes are planted by ants; therefore, seed selection by ants is a key driver of the epiphyte composition of AGs. However, deterministic post-dispersal factors, such as the surrounding environmental conditions and plant succession, may also influence epiphyte composition. Here we ask whether epiphyte composition on a local scale is associated with dispersal constraints, local environmental conditions (light availability, number of branches and nest height) or AG successional stage. We sampled all epiphyte species in 18 AGs formed by Camponotus femoratus and Crematogaster levior in Central Amazon, Brazil. AGs were located within a range of 1 km and at a maximum of 20 m from the edges of a dirt road within a primary forest. Epiphytic composition showed strong spatial structure, decreasing in similarity with increasing distance. Environmental conditions and AG successional stage were not related to AG floristic composition, suggesting a key role of stochastic processes related to seed dispersal. A combination of seed abundance and attractiveness in neighbouring AGs seems to drive the higher similarity in epiphyte composition among closer AGs.

Several studies have shown that qualitative (binary) ant-plant networks are highly nested in tropical environments, in which specialist species (with fewer interactions) are connected with generalists (with the most interactions) in cohesive subgroups. Interactions occur in both qualitative and quantitative networks, however, how their frequency may structure the nestedness in ecological networks involving these organisms is, we believe, unknown. Based on this perspective, we used nestedness analysis to address the effect of interaction frequency on ant-plant networks (n = 14 networks). Unlike binary networks, quantitative networks are often significantly non-nested. In addition, species with a higher interaction frequency have a higher number of links, indicating that these species are possibly more abundant and/or competitive. Moreover, different biological parameters can change the nature of ant-plant interactions, as a plant can be a good resource for one ant and a ‘bad’ resource for another. Thus, this suggests a new perspective for the study of interaction networks in the tropics, since species with lower interaction frequency are not necessarily subsets of species with higher frequency, and consequently generate the non-nested pattern in quantitative networks.

The relationship between Camponotus sp. ants and the carnivorous pitcher plant, Nepenthes bicalcarata, from Borneo were investigated. The ants nest in the hollow tendrils of the plant, and feed on large prey items caught by the pitchers. These are hauled from the pitcher fluid by the ants and consumed. When large prey items are absent, the ants feed upon mosquito larvae which inhabit the pitcher fluid. The accumulation of excess prey in pitchers can lead to putrefaction of the contents, and disruption of the pitcher's digestive system. Experiments on the possible benefits of the ants' behaviour to the plant showed that the accumulation of excess prey and the putrefaction of the pitcher contents were significantly reduced in the presence of the ants. The accumulation of ammonia in pitchers was unaffected by the presence of the ants when small prey were added to pitchers, but was reduced significantly in the presence of ants when large prey were added. It is suggested that Camponotus sp. and N. bicalcarata have a mutualistic association, from which the ants obtain food and nesting sites and, in return, prevent damaging accumulation of excess prey in the pitchers.

Ants are important agents of seed dispersal and they can play a fundamental role in the establishment and survival of plants (Beattie 1985). Leaf-cutting or fungus-growing ants (Myrmicinae: Attini) are dominant elements of the ant fauna throughout the neotropics. They collect different types of vegetative and reproductive plant material that serve as substrate for underground fungus gardens,which are the only food source for their larvae (Stradling 1991).

The poor availability of suitable substrate and nutrients strongly limits the distribution and growth of vascular epiphytes in lowland rain forests (Benzing 1990, Nieder et al. 2000). In some epiphyte species nutrition may be assisted by adventitious roots that grow into animal debris in plant cavities such as domatia and bromeliad tanks (Huxley 1980). For epiphyte species lacking these modifications, animals may nevertheless play a substantial role by providing a large proportion of the limited substrate in lowland forests (Catling 1995, Longino 1986). Such associations between epiphytes and nutrient/substrate-providing animals may often be non-specific and commensalistic (Davidson & Epstein 1989, Longino 1986), while highly evolved mutualistic associations occur in the case of ant gardens which are very abundant in neotropical forests (Huxley 1980, Kleinfeldt 1986, Ule 1901). Ant gardens typically are densely inhabited by different epiphytes from various plant families whose seeds or fruits are attractive to the ants and carried into the nest (Davidson 1988). In addition, ants have been suggested to play a role in protection and nutrition of ant-garden epiphytes (Kleinfeldt 1978, 1986). Ants may benefit from epiphytes through increased nest stability (Yu 1994) or nutrition via extrafloral nectaries, fruit pulps or seed arils (Davidson 1988, Kleinfeldt 1986). In this study,we compare the nutrient quality of such ant gardens with other similar substrates rarely inhabited by epiphytes, namely nests and galleries of ants and termites.