Compared to water alone, the addition of frother produces a narrower distribution and finer bubble size. An empirical model is presented for bubble size as a function of frother type using a normalized concentration. The model provides a benchmark for assessing operations. There is no agreed mechanism of how frothers act to reduce bubble size. Prevention of coalescence is a common explanation and this action is illustrated. The case against the 'coalescence model' alone is presented and a breakup mechanism is introduced based on pilot cell and single bubble studies. Phenomena associated with bubble shape, velocity and surface flows are used to illustrate the force associated with surface tension gradients. Breakup mechanisms resulting from this force are proposed. The concept is extended to high salt concentrations which also promote small bubble formation. (C) 2008 Elsevier Ltd. All rights reserved.