Sustainable management strategies for tropical forest ecosystems require a detailed understanding of the distribution of tropical forest species, limitations on distribution, as well as occurrence patterns in the absence of limitations. In the Americas, many tropical palms (Arecaceae) are locally abundant keystone species that provide a number of non-timber forest products. Here we focus on 10 such species (Aphandra natalia, Attalea phalerata, Euterpe oleracea, Iriartea deltoidea, Mauritia flexousa, Oenocarpus bataua, Phytelephas aequatorialis, Phytelephas macrocarpa, and Phytelephas seemannii) found in western South America. This study aimed to determine: (i) what variables control potential species distribution, (ii) whether their distribution is spatially constrained and if so, (iii) how their potential ranges compare to their actual ranges. We built species distribution models using Maxent software with three groups of predictor variables: C (climatic), N (non-climatic environmental), S (spatial), and combinations thereof: C, CN and CNS. The CNS predictor variable combination gave superior predictive ability based on median Area Under the Curve (AUC) values and was consistent with available range maps. In contrast, individual C and CN predictor variables gave inferior AUC values and showed less similarity to range maps. We concluded that the CNS model that combined climatic, non-climatic, and spatial variables best represented the actual distribution of the 10 palm species, whereas C and CN models approached their potential distribution. Precipitation during the driest quarter and annual precipitation were the most important predictors of the potential distribution, whereas temperature of the coldest month, soil, and the Normalized Difference Vegetation Index (NDVI) were less important. We found that the actual distributions of palm species were spatially constrained, and that the spatial variables largely coincided with the dispersal barriers of the Andean Cordillera and past climatic fluctuations. The estimated potential range varied from nearly equal to the actual range, to up to 0.55×106km2 (55.1×106ha) larger than the actual range, depending on model, threshold settings, and species. We thus concluded that current palm resources in the forests of western South America have major potential for ecological expansion. These results are relevant to current palm management and resource conservation planning, as well as future strategies that will have to address climate change and increasing human disturbance.