Morphological trait matching between species affects resource partitioning in mutualistic systems. Yet, the determinants of spatial variation in trait matching remain largely unaddressed. Here, we generate a hypothesis that is based on the geographical distributions of species morphologies. To illustrate our hypothesis, as a study system we use hummingbirds in the tropical Andes. Hummingbirds with specialized morphologies (i.e. long or curved bills) may forage on flowers that are inaccessible to hummingbirds with generalized bill morphologies (i.e. small-to-medium-sized bills with no curvature), yet the vast majority of hummingbirds have generalized bill morphologies. Thus, we propose that trait matching across space is determined by the distribution of morphological specialists. In the Andes, we observe the richness of specialized hummingbird morphotypes to peak at high and low elevations. Therefore, we hypothesize that trait matching should be most influential in predicting pairwise interactions at high and low elevations. We illustrate our hypothesis by field observations along an elevational gradient in Podocarpus National Park (Ecuador). Using Bayesian hierarchical modelling of interaction frequencies in combination with network analyses, we found that hummingbirds at high and low elevations contributed to resource partitioning by foraging on morphologically close-matching flowers. Moreover, at high and low elevations, hummingbirds with specialized morphologies showed a stronger tendency to visit close-matching flowers than morphological non-specialists did. In contrast, at mid-elevations, hummingbirds were not attracted to morphologically close-matching flowers. These results suggest that the spatial distribution of specialized morphotypes determines trait matching and the partitioning of interactions within hummingbird-plant communities.