Our knowledge of biodiversity responses to fragmentation comes from human-driven fragmentation processes. However, evaluating forest patches with natural edge can produce useful models to understand how the micro-environmental gradient between forest edge and interior shapes the plant communities, regardless habitat loss. Here, we tested how forest edges and interiors differ in soil characteristics, communities’ structure and diversity metrics, and the extent to which these differences are related to forest size and shape of patches. We present data of ten tropical montane cloud forest patches surrounded by native non-forest vegetation (highland grasslands) of southeast Brazil. In each forest, we established five plots right on the edge and five randomly distributed over the forest interior. Our dataset covers a total of 5495 trees, and the main findings showed that patches with natural edges are also influenced by the edge effect, affecting the structure, species composition and soil variables. We found that with the increase in distance from the forest edge, the soils became more acid, with greater accumulation of organic matter, phosphorus, silt and sand, in contrast, the content of total exchangeable bases and clay decreased. When accounting for the effect of patch size and shape in the relationships between forest habitat and the structural community metrics, tree density and upper 95-percentile stem diameter tended to be higher in larger patches. In contrast, aboveground biomass and median stem diameter were positively related to patch shape in forest interiors. In addition, species composition showed marked differences between habitats, but rarefied species richness and Fisher's diversity index were higher, in the interior, only in large patches. These results indicate that besides differences between forest habitats, with increasing patch size and circularity, the differences become more pronounced. We would like to highlight that in conservation and management decisions, peculiarities of each habitat must be taken into account, with techniques that assess the surrounding matrix (man-made or natural) and that consider the edge and the interior as distinct habitats.