Answers to the questions "which rocks contain diamonds?" and "which indicator minerals should diamond explorers look for?" are no longer so easily obtained, now that diamonds are known to occur in lamproites and various types of lamprophyres, as well as their traditional source of kimberlites. One way of isolating those factors (host-rock"geochemistry, geographical or tectonic location, xenolith minerals, etc.) which are most critical in determining the diamond potential of a given pipe, is a detailed and critical re-examination of the huge volumes of accumulated data on kimberlites, lamproites and related rocks. A large analytical database to this end has been set up on an Apple Macintosh II microcomputer using the quasi-relational database management system 4th Dimension. Currently, it contains data for up to 90 major and trace elements, isotopic ratios and contained mineral compositions for some 4,000 rock specimens from all over the world. Many experiences in setting up this database are probably relevant to other geologists setting up their own systems: they include determining the relationships between data items; the advantages of codes over free text in representing categorical variables such as rock types; the need for rigorous error trapping and data validation; and the problem of missing data values and ambiguities thereof. The working system is now being used in a rigorous statistical assessment of the criteria which distinguish diamondiferous from non-diamondiferous rocks.