It is almost certainly the case that many populations have always existed as metapopulations, leading to the conclusion that local extinctions are common and normally balanced by migrations. This conclusion has major consequences for biodiversity conservation in fragmented tropical forests and the agricultural matrices in which they are embedded. Here we make the argument that the conservation paradigm that focuses on setting aside pristine forests while ignoring the agricultural landscape is a failed strategy in light of what is now conventional wisdom in ecology. Given the fragmented nature of most tropical ecosystems, agricultural landscapes should be an essential component of any conservation strategy. We review the literature on biodiversity in tropical agricultural landscapes and present evidence that many tropical agricultural systems have high levels of biodiversity (planned and associated). These systems represent, not only habitat for biodiversity, but also a high-quality matrix that permits the movement of forest organisms among patches of natural vegetation. We review a variety of agroecosystem types and conclude that diverse, low-input systems using agroecological principles are probably the best option for a high-quality matrix. Such systems are most likely to be constructed by small farmers with land titles, who, in turn, are normally the consequence of grassroots social movements. Therefore, the new conservation paradigm should incorporate a landscape approach in which small farmers, through their social organizations, work with conservationists to create a landscape matrix dominated by productive agroecological systems that facilitate interpatch migration while promoting a sustainable and dignified livelihood for rural communities.