Recent developments in machine vision, applied to froth flotation, show the potential for the use of the measurement of froth surface descriptors to provide an improvement to the operation of flotation cells and banks. One such descriptor, froth colour, has been traditionally assumed to relate to the grade of the concentrate recovered. This work sets out to test this assumption, and was performed using a machine vision system, SmartFroth. This system included a camera hood with built in LED lighting, a colour calibration patch and a CIE 'Lab' colour space conversion algorithm, which all served to reduce the effect of ambient lighting on the colour measurements. The effectiveness of these techniques was evaluated by comparing colour measurements of dry synthetic ore under various lighting conditions. It was found that the influence of ambient lighting had not been completely eliminated. Batch flotation tests were conducted on a sulphide ore, from the Kennecott Utah Copperton Concentrator, and established that froth colour does not account for entrained gangue, which is barely visible on the froth surface. Thus, froth colour alone cannot be used to accurately predict concentrate grade. (C) 2008 Elsevier Ltd. All rights reserved.