The sizes of mineral grains control liberation and the subsequent separation. Automated mineralogy provides a basis for making a useful estimation of this very important parameter. But the estimation is not straightforward, partly due to the stereological aspects of the problem, and partly due to the ill-defined nature of mineral grains. The present paper gives suggestions for making these estimates. The suggestions are made on the basis of past experimental studies, supplemented by computer simulation of different shapes and sizes of grains. It is concluded that phase specific surface area provides the best "size basis" for ranking different ores. The magnitude of shape effects and size distribution effects are discussed. But the next step of predicting the extent of liberation from grain size measurements is still difficult and needs experimental support. (C) 2007 Elsevier Ltd. All rights reserved.