We study compartmentalization in a Mediterranean pollination network using three different analytical approaches: unipartite modularity (UM), bipartite modularity (BM) and the group model (GM). Our objectives are to compare compartments obtained with these three approaches and to explore the role of several species attributes related to pollination syndromes, species phenology, abundance and connectivity in structuring compartmentalization. BM could not identify compartments in our network. By contrast, UM revealed four modules composed of plants and pollinators, and GM four groups of plants and five of pollinators. Phenology had a major influence on compartmentalization, and compartments (both UM and GM) had distinct phenophases. Compartments were also strongly characterized by species degree (number of connections) and betweenness centrality. These two attributes were highly related to each other and to phenophase duration. Differences among compartments in abundance were only apparent with GM. We attribute this to the fact that abundance is strongly correlated with Degree, and the GM algorithm is particularly powerful at discriminating species based on degree. On the other hand, the role of pollination syndrome-related features in compartmentalization mostly emerged with UM. Only UM compartments differed in corolla length and pollen production. Both UM and GM compartments differed in their pollinator spectra. We found inconsistent reciprocity between plant attributes and pollinator spectra, thus it is difficult to conclude compartments follow clear-cut syndromes. Also, both UM and GM identified a compartment composed of pollinators with long activity periods that acted as connectors, linking all compartments providing cohesiveness to the network. © 2012 The Authors.