Island ecosystems are vulnerable to the impacts of invasive non-native species. Non-native plant invasions often modify the structure and compromise the stability of recipient communities. To mitigate or revert the effects of invasive species on island biota, various strategies of invasive species control and ecological restoration (e.g. removal of non-native plant species) are being used on many islands worldwide. To develop and implement appropriate management strategies, it is critical to quantitatively assess the effects of the management intervention on ecosystem structure and functioning. This thesis uses a network approach and an established large-scale restoration experiment in a natural island ecosystem to investigate the longer-term (>5 years) responses of native mutualistic (seed-dispersal) and antagonistic (pre-dispersal seed predation and plant-pollinator antagonist) interactions to the removal of non-native plant species (i.e., restoration). The research was conducted on eight mountaintops (4 restored and 4 unrestored) of ca. 1 ha on the island of Mahé, Seychelles, in the Western Indian Ocean. The thesis contains four data chapters based on four independent investigations. The first chapter investigates the role of inselberg forest remnants and frugivorous birds on the natural regeneration of surrounding invaded forests. We show that native forest remnants, by providing a source of native seeds for dispersal and recruitment, can facilitate the natural regeneration of adjacent invaded forests mediated by native frugivores, which may promote the resistance of these patches and their surroundings to plant invasions. The second chapter explores the seasonal variation in the impact of non-native plant species on frugivory and seed dispersal networks in restored and unrestored plant communities. Our findings highlighted marked variations in the effects of non-native plants on seed dispersal interactions with seasonality and plant invasion. Frugivore foraging behaviour changed with the availability of resources in invaded, but not in restored plant communities. Plant invasion therefore has a direct negative impact on frugivores and native seed dispersal, and restoration can buffer against seasonal variation in resource constraints. In the third chapter I use interaction networks on pre-dispersal seed predation to investigate the effects of restoration on seed predator insect assemblages. Restored communities supported a more diverse seed-predator insect community, which suggested that the homogenisation process of the ecosystem caused by plant invasions is in parts revertible. Restoration therefore appears to preserve the native seed-predator insects and their interactions with native plants. Finally, I experimentally quantified the impact of two invasive ant species on the pollination and plant reproduction of four native plant species and assessed whether these effects differed with restoration. Ants contributed marginally to pollination of the native plants, but aggressive behaviour by invasive ants reduced flower visitations by flying pollinators, thereby potentially negatively affecting plant reproductive success. These effects were similar in restored and unrestored sites. This thesis provides clear evidence on the importance of restoring native remnant forest patches, such as those on mountaintops on the island of Mahé, to maintain native plant communities and the functioning of these unique ecosystems through the conservation of seed dispersal interactions, insect diversity and pollination. These findings can be used to guide future management strategies that aim to achieve the conservation and restoration of native biodiversity and ecosystem functioning.