The Janzen–Connell (JC) hypothesis provides a conceptual framework for explaining the maintenance of tree diversity in tropical forests. Its central tenet—that recruits experience high mortality near conspecifics and at high densities—assumes a degree of host specialization in interactions between plants and natural enemies. Studies confirming JC effects have focused primarily on spatial distributions of seedlings and saplings, leaving major knowledge gaps regarding the fate of seeds in soil and the specificity of the soilborne fungi that are their most important antagonists. Here we use a common garden experiment in a lowland tropical forest in Panama to show that communities of seed-infecting fungi are structured predominantly by plant species, with only minor influences of factors such as local soil type, forest characteristics, or time in soil (1–12 months). Inoculation experiments confirmed that fungi affected …