A number of problems in mineral processing demands that the distribution of material over composition and size ranges be described. The description of the progressive liberation of ore components with grinding is one such problem. Coal particle distribution as a function of particle size and density is another common problem. Invariably, the methods of analysis of mineral mixtures lead to a description of the materials in terms of discrete size and composition classes, but subsequent calculations using the material description benefit from the use of an interpolated version of the original data. The interpolation must be positive and smooth and it should be possible to recover the original data from the interpolated form of the material description. At least, the interpolation function should have the properties of a probability density function, that is, the area or volume under the interpolator must be unity. A robust means of computing such interpolations, using mineral liberation data, is described herein.