A diamond classification system based on the physical features of diamond as a function of size is described. Crystal form, crystal angularity, crystal regularity determine the major morphological divisions of the classification, but transparency or opacity, colour, the number of inclusions and surface features allow further subdivision. The system has been used to define uniquely samples of diamonds from three kimberlite diatremes, Premier, Finsch, and Koffyfontein, and an alluvial source, Dreyers Pan. In the cases of the diatreme occurrences variations in diamond crystal shape versus size are explained in terms of a primary crystal population of octahedra and twinned crystals (macles) subsequently altered by secondary processes in the kimberlite magma. Variations in colour are discussed in terms of aggregation of nitrogen into platelets in the diamond crystal lattice as the diamond grew. Characteristic features of alluvial diamonds from Dreyers Pan allow some conclusions to be drawn relating to the nature of the original kimberlite(s) from which the diamonds are derived. The usefulness of the classification scheme as a framework into which other quantitative diamond studies can be fitted, is outlined.