Abstract The brittle behaviour of ceramic materials makes imperative the development of accurate and reproducible methods of measuring their resistance to fracture. To this end, a European round robin was set up to investigate the relative merits of five different methods of fracture toughness testing. Of these the single edge notch bend — saw cut (SENB-S) method seemed to deliver the most reproducible results, both within and between laboratories. However, it has been observed empirically that if notches are cut too thick, the values of fracture toughness determined are systematically too high. An explanation and a theoretically based relationship to describe this behaviour are presented. It is suggested that this effect results from the interaction of the stress field around the notch tip and defects related to the microstructure or machining damage. Measured data from a number of materials seem to correlate well with the theory. It is shown that if correct values of fracture toughness are to be determined with the SENB-S method, the notch width must be of the order of the size of the relevant microstructural or machining-induced defects (e.g. large pores and weak grain boundaries).