Pollination syndromes are defined as suites of floral traits, which have evolved repeatedly across angiosperms in adaptation to distinct functional pollinator groups. These syndromes were developed to classify flowers under a functional ecological perspective by their most efficient pollinator, irrespective of their phylogenetic relationship. Several studies show strong support for the concept of pollination syndromes while others raised concerns about their reliability. Melastomataceae offer an ideal system to test the concept of pollination syndromes, since many different pollinators and specialized floral morphologies have been reported for the family. The large, pantropically distributed family is dominated by buzz-bee pollination (95.5% of species), where pollen grains are released from tubular, poricidal anthers through vibrations. In eight (of 23) tribes, shifts to different vertebrate pollinators (i.e. bats, rodents, hummingbirds, passerine birds, flowerpiercers) have occurred. Although only studied at a small scale up to now, these shifts seem to associate with marked changes in floral functional traits such as reward type, mechanism of pollen release, and corolla shape. Shifts among major functional pollinator groups (e.g., bees and hummingbirds), are generally regarded as a major source of increased floral disparity (morphological diversty). We recorded 44 pollination-relevant functional floral traits across 411 species (260 with pollinator observations, 151 without pollinator observations), spanning the whole family (at least 1 species per genus). We focused on system-specific functional floral traits (e.g., appendage shape, structure of thecal wall) but also include traits traditionally used in pollination syndromes (e.g., corolla shape, corolla color). We used machine learning algorithms to identify pollination syndromes for species with empirically documented pollinators and employed these trained models to predict pollinators for species with unobserved pollinators. We ran ancestral character state estimation to determine directionality in pollinator shifts in Melastomataceae, and tested whether pollinator shifts mirror shifts in selection regimes using Ornstein-Uhlenbeck models (OU-models). Furthermore, we contrasted floral disparity among the different syndromes, biogeography, and tribes to evaluate the relative role of pollinator shifts in generating floral diversity. Our results indicate strong support for four well differentiated pollination syndromes within Melastomataceae: (i) widespread buzz-bee pollination; (ii) pollination by mixed assemblages of nectar-foraging vertebrates such as birds, bats, and rodents; (iii) pollination by food-body-foraging passerines; and (iv) generalized systems. Pollination syndromes in Melastomataceae can be discriminated by six system specific floral traits of which reward type and pollen release mechanism are the two most important traits to distinguish the different syndromes. In addition, we found that the “buzz-bee” and “nectar-foraging vertebrate” syndromes are highly diverse and pollinator shifts contribute to floral disparity, which is not directly correlated with clade size. One explanation for the high morphological diversity within the “buzz-bee” syndrome might be the high diversity of bees that buzz-pollinate Melastomataceae. We reconstructed the “buzz-bee” syndrome as ancestral from which all other syndromes evolved repeatedly. Finally, we expect to find several subsyndromes within the “buzz-bee”, the “nectar-foraging vertebrate”, and the “generalist” syndromes as indicated by the PCoA and the output from the OU-models.