The two main assembly processes claimed to structure plant communities are habitat filtering and competitive interactions. The set of species growing in fire-prone communities has been filtered in such a way that species without fire-persistence traits have not successfully entered the community. Because plant traits are evolutionarily conserved and fire traits are correlated with other plant traits, communities under high fire frequency should not include all possible trait combinations, and thus the morphospace occupation by species in these communities should be lower than expected by chance (underoccupied). In contrast, communities under low fire frequency would lack the filtering factor, and thus their underoccupation of the morphospace is not expected. We test this prediction by comparing the morphospace occupation by species in communities located in the western Mediterranean Basin, five of them subject to high fire frequency (HiFi) and four to low fire frequency (LowFi). We first compile a set of morphological and functional traits for the species growing on the nine sites, then we compute the morphospace occupation of each site as a convex hull volume, and finally, to assert that our results are not a product of a random branching pattern of evolution, we simulate our traits under a null model of neutral evolution and compare the morphospace occupation of the simulated traits with the results from the empirical data. The results suggest that, as predicted, there is a clear differential morphospace occupation between communities under different fire regimes in such a way that the morphospace is underoccupied in HiFi communities only. The simulation of a neutral evolutionary model does not replicate the observed pattern of differential morphospace occupation, and thus it should be attributed to assembly processes. In conclusion, our results suggest that fire is a strong community assembling process, filtering the species that have fire-persistent traits and thus assembling phenotypically and phylogenetically clustered communities with vacant zones in the morphospace.