Amazonian forests are vulnerable to a wide range of human threats, such as selective logging and forest fires. The capacity of Amazonian forests to recover from human-disturbances depends, among other factors, on the persistence of frugivory interactions leading to essential ecological functions, like seed dispersal. Although important, little is known about frugivory interactions in the Amazonian forests. Most studies focus on a single or a limited set of similar species; neither studies include both arboreal and terrestrial frugivores. Moreover, most studies have not studied the impact of human-disturbances. We investigated the impacts of selective logging and forest fires on frugivory interactions in Amazonian forests. We sampled interactions at the community level, surveying arboreal and terrestrial frugivores across 17 forest transects with variable disturbance histories. We found that undisturbed forests held a significantly higher richness of species and interactions distinct from those of interactions human-disturbed forests. Logged forests burned 17 years previously held substantially lower richness of species and interactions, and the interaction composition was almost entirely dissimilar to those of undisturbed forests. Logged (unburned) and logged forests burned three years previously also differed substantially in their frugivory interaction composition, but overall richness and frugivory interactions richness was similar. However, neither selective logging nor forest fires changed the structural properties of frugivory networks, which are highly modular, moderately specialised and poorly connected and nested. β-diversity of plant and frugivore species as well as their interactions was high among all transects, mainly due to the high spatial turnover. Our study provides the first empirical evidence of the negative effects of fires combined with selective logging on frugivory interactions in tropical forests, highlighting the long time-scales required to evaluate impacts and reestablish ecological processes.

The ecological factors driving specialisation in species interaction networks along environmental gradients at large spatial scales are poorly understood. Although such drivers can have synergistic impacts, previous work has mainly assessed effects of network type and the abiotic environment separately. We conducted a meta-analysis of existing network data to assess the interactive effects and relative importance of these drivers of specialisation in ant-plant networks at global scales. We collated 74 ant plant networks from 1979–2023, categorised into four network types: plants that provide ants nesting sites (myrmecophytes); plants that provide only food sources (myrmecophiles); plants for which ants disperse seeds (myrmecochories); plants on which ants forage only (foraging). We explored how network specialisation varies between interaction types with elevation, latitude, and anthropogenic disturbance. We used a standard measure of network specialisation (H2’), tested whether standardising this against network null models influenced results (H2’ z-score), and measured phylogenetic network specialisation (dsi*). We found that H2’ was strongly affected by habitat disturbance, elevation and interaction type in a manner congruent with previous work, However, these effects disappeared once H2’ was standardised (H2’ z-score). The disappearance of these effects indicates that previous results may relate to variation in network structure rather than specialisation. This is supported by the existence of correlations between network species richness/weighted connectance and H2’. Phylogenetic network specialisation (dsi*) was greater for myrmecophytes than for other three network types. This probably relates to closer co-evolution between partners in myrmecophytic network. Phylogenetic network specialisation did not vary significantly with elevation, latitude or anthropogenic disturbance. Our results demonstrate that ant-plant network types, in this case relating to strength of mutualistic interaction, is the main driver of network specialisation, and that previously reported impacts of latitude, elevation and anthropogenic habitat disturbance are likely to have been mediated mediated via correlations with network size.