The statistical procedure advocated by Harker & Parker (1945) for the estimation of true triple junction angles in annealed aggregates of opaque material, is questioned in the light of more recent data from transparent sections suggesting that the true angle may itself have a natural variation even at equilibrium. A computer simulation technique is used to show that the mean of the true triple junction angle can be estimated (with varying precision) from the frequency distribution of apparent angles provided that the natural variation of the true angle does not have a standard deviation greater than about 20°.
Theoretically, the standard deviation of the true angle variation can also be deduced from the apparent angle distribution but it is shown that this is only feasible if the standard deviation of the true triple junction angle is greater than 10°.
Published apparent angle distributions do not conform to the theoretical distribution and it is suggested that they are similar to those that would be obtained if triple junctions making low angles with the section plane are not measured. This restricts the information that can be obtained and questions the validity of some results.