Fragmentation of natural landscapes is a pervasive process in the world. Common models predict coherent change in assemblages, with less numerous species becoming locally extinct first, then species of intermediate abundance, and so forth. Relative-abundance distributions should change systematically in landscapes characterized by greater change. Such a predictable sequence of change is not evident in the avifaunas of landscapes of central Victoria, Australia, where relative-abundance patterns in more affected landscapes bear little resemblance to reference distributions. I provide two sets of analyses of relative-abundance distributions: (1) analyses that do not depend on the identity of individual species and (2) abundance spectra, which use ordered lists of species ranked by species' commonness in reference systems. While abundance spectra change dramatically in smaller remnants, relative-abundance distributions change little, suggesting that the "reorganization" of abundances occurs over ecological time frames. The dispersal-limited multinomial is a flexible distribution that may fit many data sets yet be unrelated to assumptions (species neutrality) and processes (fixed total numbers of individuals) of the unified neutral theory. A more complete understanding of human impacts at landscape scales must include capacities to predict those species that will be advantaged by change, as well as those that will be disadvantaged. © 2007 by The University of Chicago. All rights reserved.