1. We examined riverine desert fish assemblages in the Chihuahuan Desert, USA at multiple spatial scales of similarity to assess long-term changes to assemblage distinctiveness, identify individual species responsible for changes, and determine the importance of geographic context and species resolution in interpreting patterns of change. 2. We used a well-documented historical data set on fish distribution and abundance, and recent collections of fishes that provided a paired analytical design across 36 localities spanning nearly three decades of time. Patterns of faunal homogenization and differentiation were assessed at basin-wide, sub-basin and river-reach scales with species occurrence and relative abundance data. Individual species responses were examined to identify the drivers of assemblage change across time. 3. Patterns of similarity varied across spatial scales and produced seemingly incongruous trends in assemblage similarity across time. Patterns of assemblage distinctiveness depended on the spatial extent of the analyses, the geographical structuring of the fish assemblages, and whether occurrence or relative abundance data were used. These dependencies led to interesting and conflicting patterns of homogenization and differentiation. The Rio Grande sub-basin showed strong homogenization with convergence between upstream and downstream reaches that corresponded to declining water quality and quantity from the Rio Conchos in Mexico. In contrast, the Pecos River sub-basin showed strong differentiation between upstream and downstream reaches that corresponded to the successful colonization and spread of the non-native gulf killifish (Fundulus grandis) in the highly degraded upper reach. Spatial variability in fish assemblages and their degree of change from historical conditions were largely dependent on anthropogenic modifications to the flow regime and variability in the success of invasive gulf killifish in the basin. 4. The use of species occurrence or abundance data, and the spatial scale of analysis are crucial choices in studies of faunal homogenization and differentiation, and we have demonstrated how these choices lead to variable results for our study system. Our multi-scale approach and examination of individual species responses identified the ultimate drivers of these differences and illustrated the importance of scale-dependent effects and geographical context on patterns of assemblage distinctiveness, especially with regard to species invasion, species loss and abundance shifts.