Fifty years ago, Baker and Fedorov proposed that the high species diversity of tropical forests could arise from the combined effects of inbreeding and genetic drift leading to population differentiation and eventually to sympatric speciation. Decades of research, however have failed to support the Baker–Fedorov hypothesis (BFH), and it has now been discarded in favor of a paradigm where most trees are self-incompatible or strongly outcrossing, and where long-distance pollen dispersal prevents population drift. Here, we propose that several hyper-diverse genera of tropical herbs and shrubs, including Piper (>1,000 species), may provide an exception. Species in this genus often have aggregated, high-density populations with self-compatible breeding systems; characteristics which the BFH would predict lead to high local genetic differentiation. We test this prediction for five Piper species on Barro Colorado Island, Panama, using Amplified Fragment Length Polymorphism (AFLP) markers. All species showed strong genetic structure at both fine- and large-spatial scales. Over short distances (200–750 m) populations showed significant genetic differentiation (Fst 0.11–0.46, P < 0.05), with values of spatial genetic structure that exceed those reported for other tropical tree species (Sp = 0.03–0.136). This genetic structure probably results from the combined effects of limited seed and pollen dispersal, clonal spread, and selfing. These processes are likely to have facilitated the diversification of populations in response to local natural selection or genetic drift and may explain the remarkable diversity of this rich genus.