There is now strong evidence suggesting that interactions between plants and their species-specific antagonistic microbes can maintain native plant community diversity. In contrast, the decay in diversity in plant communities invaded by nonnative plant species might be caused by weakening negative feedback strengths, perhaps because of the increased relative importance of plant mutualists such as arbuscular mycorrhizal fungi (AMF). Although the vast majority of studies examining plant-soil feedbacks have been conducted in a single habitat type, there are fewer studies that have tested how the strength and direction of these feedbacks change across habitats with differing dominating plants. In a fragmented montane agricultural system in Colombia, we experimentally teased apart the relative importance of AMF and non-AMF microbes (a microbial filtrate) to the strength and direction of feedbacks in both native and nonnative plant species. We hypothesized that native tree species of forest fragments would exhibit stronger negative feedbacks with a microbial filtrate that likely contained pathogens than with AMF alone, whereas nonnative plant species, especially a highly invasive dominant grass, would exhibit overall weaker negative feedbacks or even positive feedbacks regardless of the microbial type. We reciprocally inoculated each of 10 plant species separately with either the AMF community or the microbial filtrate originating from their own conspecifics, or with the AMF or microbial filtrate originating from each of the other nine heterospecific plant species. Overall, we found that the strength of negative feedback mediated by the filtrate was much stronger than feedbacks mediated by AMF. Surprisingly, we found that the two nonnative species, Urochloa brizantha and Coffea arabica, experienced stronger negative feedbacks with microbial filtrate than did the native forest tree species, suggesting that species-specific antagonistic microbes accumulate when a single host species dominates, as is the case in agricultural habitats. However, negative feedback between forest trees and agricultural species suggests that soil community dynamics may contribute to the re-establishment of native species into abandoned agricultural lands. Furthermore, our finding of no negative feedbacks among trees in forest fragments may be due to a loss in diversity of those microbes that drive diversity-maintaining processes in intact tropical forests.