Recent investigations of demersal fish communities in deep (less than 50 m) rugged habitats have considerably increased our knowledge of the factors that influence the assemblage structure of fishes across mesophotic to deep-sea depths. Although habitat types influence deepwater fish distribution, whether different rugged seafloor features provide functionally equivalent habitat for fishes is poorly understood. In the northeastern Caribbean, numerous rugged seafloor features (e.g., seamounts, banks, canyons) punctuate insular margins, and thus create a remarkable setting in which to examine demersal fish communities across various seafloor features. Also in this region, several water masses are vertically layered in the water column, creating strong stratification layers corresponding to specific abiotic conditions. In this study, we examined differences among fish assemblages across seafloor features (e.g., seamount, canyon, bank/ridge) and water masses at depths ranging from 98 to 4060 m in the northeastern Caribbean. We conducted 26 remotely operated vehicle (ROV) dives across 18 sites, yielding 156 species; 42% of which had not been previously recorded from particular depths or localities in the region. While fewer species were observed at seamounts than at other habitats in the NE Caribbean, assemblage structure was similar among habitat features. Thus, similar to seamount studies in other regions, seamounts in the Anegada Passage do not harbor distinct communities from other rugged, topographic features. Species assemblages, however, differed among depths, with zonation generally corresponding to water mass boundaries in the region. High species turnover occurred at depths less than 1200 m, and is driven by changes in water mass characteristics including temperature (4.8-24.4 ºC) and dissolved oxygen (2.2-9.5 mg per l). Our study demonstrates the importance of water masses in shaping community structure of benthic fauna, while considerably adding to the knowledge of mesophotic and deep-sea fish biogeography.