A single Eastern Kentucky high volatile A bituminous coal was prepared into 24 polished pellets, each of which was point counted at 0.4-mm intervals by a single operator. Examination of the resulting data, both in paired sequences of 500 counts from individual pellets and in randomly extracted 1000-point sets, shows that a certain natural variation exists, converging on a "most likely absolute difference" for this particular coal. All coals will vary, but it is instructive to consider the precision expected in counting a complex material before considering the reproducibility expected from multiple operators counting the same coal. (C) 2000 Elsevier Science B.V. All rights reserved.