Ball-mills are used widely for secondary grinding. Loveday (2010) reported on laboratory tests in which small pebbles (7-25 mm) were used in various proportions with balls. The optimum proportion of pebbles, by Volume, was found to be about 25%. Substantial savings in power and ball consumption were achieved, with no loss in grinding capacity. However, continuous pilot-plant tests were disappointing, because the consumption of small pebbles was too high. The idea of using a mixture of balls and pebbles, at a mill speed suitable for ball-milling, was revisited in this investigation, using a normal spectrum of pebble sizes (19-75 mm). Batch tests in a pilot-scale mill (0.57 m diameter) were used to compare ball-milling to various ball/pebble mixtures. The mill power was measured online by monitoring lateral torque on a freely suspended motor and gearbox. Initial tests were done using pebbles from previous tests on gold ore, in combination with balls, to mill silica sand (0.61.5 mm) The size distributions of the balls and the pebbles were calculated to simulate steady-state addition of balls (37.5 mm) and partly-rounded pebbles (19-75 mm). There was a reduction in pebble consumption, as expected, when using larger rounded pebbles, to about 6% of total production. The grinding capacity, when using a mixture containing 25% pebbles, was the same as that with balls alone, resulting in a 13% saving in energy and an implied saving in ball consumption of 25%. It was concluded that the use of a composite load for secondary grinding is a very attractive option. Further tests were done, using samples from a platinum mine, namely the feed to a secondary mill (ball-mill) and rounded pebbles from a primary AG mill. Grinding capacity was maintained over the range 0-30% pebbles, by volume, with savings in energy and ball consumption increasing progressively. At 25% pebbles, the saving in energy was 18% and the pebble consumption was about 7% of total production. The theory of pebble wear was developed during this investigation, for application to batch grinding tests. The size distribution of the pebbles was measured after each test and the theory was used to calculate the pebble wear rate (mm h(-1)) and the specific wear rate (kg h(-1) kg(-1)) for all screen fractions. The wear rate of pebbles in the presence of balls was higher than that for pebbles alone. When the proportion of pebbles in the mill charge was low (up to 30%), the larger pebbles wore away significantly faster than expected, based on a constant surface wear rate. This could be due to size segregation in the mill. (C) 2016 Elsevier Ltd. All rights reserved.