Understanding local and global extinction is a fundamental objective of both basic and applied ecology. Island biogeography theory (IBT) and succession theory provide frameworks for understanding extinction in changing landscapes. We explore the relative contribution of fragment size vs. succession on species' declines by examining distributions of abundances for 18 plant species declining over time in an experimentally fragmented landscape in northeast Kansas, U.S.A. If patch size effects dominate, early-successional species should persist longer on large patches, but if successional processes dominate, the reverse should hold, because in our system woody plant colonization is accelerated on large patches. To compare the patterns in abundance among patch sizes, we characterize joint shifts in local abundance and occupancy with a new metric: rank occupancy-abundance profiles (ROAPs). As succession progressed, statistically significant patch size effects emerged for 11 of 18 species. More early-successional species persisted longer on large patches, despite the fact that woody encroachment (succession) progressed faster in these patches. Clonal perennial species persisted longer on large patches compared to small patches. All species that persisted longer on small patches were annuals that recruit from the seed bank each year. The degree to which species declined in occupancy vs. abundance varied dramatically among species: some species declined first in occupancy, others remained widespread or even expanded their distribution, even as they declined in local abundance. Consequently, species exhibited various types of rarity as succession progressed. Understanding the effect of fragmentation on extinction trajectories requires a species-by-species approach encompassing both occupancy and local abundance. We propose that ROAPs provide a useful tool for comparing the distribution of local abundances among landscape types, years, and species.