Volcanogenic halogens, in particular bromine, potentially play an important role in the ozone depletion of the atmosphere. Understanding bromine behaviour in magmas is therefore crucial to properly evaluate the contribution of volcanic eruptions to atmospheric chemistry and their environmental impact. To date, bromine partitioning between silicate melts and the gas phase is very poorly constrained, with the only relevant experimental studies limited to investigation of synthetic melt with silicic compositions. In this study, fluid/melt partitioning experiments were performed using natural silicate glasses with mafic, intermediate and silicic compositions. For each composition, experiments were run with various Br contents in the initial fluid (H₂O-NaBr), at T - P conditions representative of shallow magmatic reservoirs in volcanic arc contexts (100-200 MPa, 900-1200°C). The resulting fluid/melt partition coefficients (D_Br^f/m)are: 5.0 ± 0.3 at 1200°C -100 MPa for the basalt, 9.1 ± 0.6 at 1060°C - 200 MPa for the andesite and 20.2 ± 1.2 at 900°C - 200 MPa for the rhyodacite. Our experiments show that D_Br^f/m increases with increasing SiO₂ content of the melt (as for chlorine) and suggest that it is also sensitive to melt temperature (increase of D_Br^f/m with decreasing temperature). We develop a simple model to predict the S-Cl-Br degassing behaviour in mafic systems, which accounts for the variability of S-Cl-Br compositions of volcanic gases from Etna and other mafic systems, and shows that coexisting magmatic gas and melt evolve from S-rich to Cl-Br enriched (relative to S) upon increasing degree of degassing. We also report first Br contents for melt inclusions from Etna, Stromboli, Merapi and Santorini eruptions and calculate the mass of bromine available in the magma reservoir prior to the eruptions under consideration. The discrepancy that we highlight between the mass of Br in the co-existing melt and fluid prior to the Merapi 2010 eruption (433 and 73 tons, respectively) and the lack of observed BrO (from space) hints at the need to investigate further Br speciation in 'ash-rich' volcanic plumes. Overall, our results suggest that the Br yield into the atmosphere of cold and silicic magmas will be much larger than that from hotter and more mafic magmas.