In flotation processes, gas dispersion plays a critical role for particle collection (recovery) and froth mass transport (selectivity). To evaluate this effect at industrial scale, the bubble surface area flux (S-B) is typically estimated from the plant superficial gas rate (J(G)) and bubble size distribution (BSD) measurements, where BSD is represented by the Sauter mean diameter (D-32). However, the same S-B value can be obtained from different combinations J(G) and D-32. Also, the same D-32 value can be obtained from different BSDs. For this reason, it is necessary to determine the overall BSD to adequately evaluate the gas dispersion. Commercial sensors can obtain representative and reliable samples of the BSD in flotation equipment at pilot scale. Nevertheless, the accuracy of the mean, standard deviation and D-32 parameter estimations at industrial scale strongly depends on the cluster and large bubbles detection and quantification. In this study, BSDs were measured in several flotation plants using the McGill bubble viewer device together with the USM-IMA software. Data were recorded from 10 m(3) to 300 m(3) mechanical cells in rougher, cleaner and scavenger copper/molybdenum flotation circuits. Using the USM-IMA semi-automatic software, which is able to detect the size of single, clustered and large bubbles from recorded images, a strong dependence was observed between the Sauter mean diameter and the standard deviation of the population in the range of D-32 = 0.9-5.0 mm and J(G) = 0.5-2.0 cm/s. This result indicates the existence of a gas dispersion pattern in mechanical cells; thus, the BSD can be described with a unique parameter. This pattern represents a powerful tool for mechanical flotation cell modelling (design) and diagnosis (control). (C) 2011 Elsevier Ltd. All rights reserved.