Habitat loss, specifically deforestation, is the biggest threat to global biodiversity. Despite this, we still have relatively little understanding of how ecological communities vary in their response to deforestation. where the most sensitive communities are, and what drives this sensitivity. In this thesis, I answer these questions, amongst others, at the global and local scale. In Chapter 2 and Chapter 3 I investigate the causes of global variation in community sensitivity, using range maps of terrestrial vertebrates to identify latitudinal patterns in the distribution of forest species, and community datasets to find thresholds in responses of communities to deforestation. In Chapter 4 and Chapter 5 I explore variation in community sensitivity to forest loss at the local scale, focusing on changes in species interactions caused by forest loss, and responses of communities to fragmentation. In Chapter 2 I found that proportions of forest species decreased from the equator to the poles, even once variables such as forest cover were accounted for. This finding was corroborated in Chapter 3 where I identified that communities at the equator were more sensitive to deforestation than those at the poles. I additionally found that communities in highly disturbed regions were less sensitive than those in undisturbed regions. At the local level I found that species interactions altered as deforestation occurred, and that they were most negative in intermediately forested areas. I further found that the spatial configuration of deforestation can impact community responses. A few large patches maintain community composition, while many small patches increase species richness. Overall the results from my thesis suggest that communities do vary in their response to deforestation and that communities in undisturbed areas close to the equator are most sensitive.