Habitat fragmentation is a leading cause of extinction, with effects that may be particularly pronounced in tropical ecosystems. However, little is known regarding the demographic mechanisms underlying changes in abundance in fragmented landscapes. Using six years of demographic data collected from Ͼ6600 individuals of the Amazonian understory herb Heliconia acuminata, we calculated population growth rate () in experimentally isolated 10-ha forest fragments, 1-ha forest fragments, and continuous forest. We then used life-table response experiment analyses to elucidate the mechanisms responsible for observed differences in . On average, ഠ 1.05 in continuous forest, while ഠ 1 in both 1-ha and 10-ha fragments. However, while the differences in between 10-ha fragments and continuous forest were largely attributable to the negative contribution of stage-speciﬁc fertility rates, reduced in 1-ha fragments was due to both reductions in reproductive rates and changes in the rate of plant growth. Our results show that similar reductions in in fragments of different sizes can be driven by distinct demographic mechanisms. Without comprehensive demographic data, attempts to mitigate the decline of populations in fragmented landscapes could be unsuccessful because they might be focusing on inappropriate demographic targets.