Eastern deciduous forests are undergoing major changes in species composition and diversity. The causes of these changes remain poorly understood and controversial because few studies have investigated more than one mechanism at a time. We conducted a large-scale experiment to simultaneously test three key hypotheses about factors driving recent changes in forest composition and diversity in mixed mesophytic forests in West Virginia. 1. The fire hypothesis: Periodic fires promote diversity and the regeneration of a suite of fire tolerant species. 2. The browsing hypothesis: Over-browsing by deer prevents palatable species from establishing or entering the sapling size class. 3. The gap hypothesis: intermediately shade-tolerant species require large canopy gaps to regenerate. We are tested these hypotheses using a factorial design where we are experimentally manipulating fire, browsing by deer, and canopy gap formation. We report on findings 5 yr after treatment. Browsing is the primary mechanism suppressing regeneration of diverse forests at our site and likely elsewhere where deer populations are above 20 km^-1. Effects of fire and gap creation were in large part masked by deer browsing. We suggest that the most effective strategy in regenerating a diverse forest is to exclude deer and create canopy gaps. Fire by itself should not be used as a management tool as it mainly promotes species like Betula spp. and Acer pensylvanicum that are not part of the existing, diverse canopy.