Aim Local-scale processes at species distribution margins can affect larger-scale distribution dynamics, but are rarely studied. The objective of this research was to elucidate the nature of distribution limits by studying the comparative structure, dynamics and environmental associations of breeding bird populations at their distribution margin. We hypothesized that climate is principally responsible for setting distribution limits, whereas biotic habitat features are more strongly associated with distribution patterns within the range.Location Southern California, USA.Methods During 2005-2007 we studied the distribution patterns of breeding birds in three study areas, each spanning a low-elevation (200-1800 m) desert scrub-tochaparral gradient. We used logistic regression with hierarchical partitioning to assess the independent effects of environmental variables (e.g. climate versus habitat) on distributions. We tested for shifts in the relative importance of these environmental variables in determining distribution limits versus within-range patterns, and we also compared higher-and lower-elevation groups of species.Results Distribution patterns were highly variable among species, but were remarkably static over the three study areas and 3-year study period. Across species, habitat floristic variables performed relatively well at explaining distribution patterns. For higher-elevation species (chaparral birds), climate was relatively important in setting their lower distribution limits, and there was a shift to a greater importance of biotic habitat (mainly habitat structural variables) for determining within-range patterns. Relationships were more mixed for lowerelevation species (desert scrub birds), but with respect to distribution limits, biotic habitat variables tended to be more important relative to climate than we observed for chaparral birds.Main conclusions Along this warm, arid elevational gradient, higher-elevation chaparral birds are more limited by climate at their lower margin than are lowerelevation desert birds at their upper margin, suggesting that climate plays a strong role (relative to other values) in excluding non-desert birds from desert. However, given the strong differences among species, predictive distribution models will need to be individually tailored, and for most species biotic habitat variables were of greater importance than climate in determining limits. This research highlights the usefulness of studying environmental relationships at distribution margins and the importance of considering biotic relationships in forecasting distribution shifts under changing climates.