Epidemiological research during the past 40 years has demonstrated with increasing clarity that amphibole asbestos fibres--crocidolite, amosite and tremolite--are more carcinogenic than chrysotile. A smaller number of well-controlled studies using lung burden analyses, while adding to the specificity of this conclusion, have shown that amphibole fibres also differ from chrysotile in being far more durable and biopersistent in lung tissue. Analyses of mesothelioma and lung cancer in a large cohort of Canadian chrysotile miners and millers have recently shown that the low-level presence of fibrous tremolite in these mines, rather than the chrysotile, may well be responsible. The high risk of lung cancer, but not of mesothelioma, in the chrysotile textile industry remains anomalous and cannot be explained in this way. These various findings are directly relevant to the choice of the experimental methods which should be used for screening man-made fibres for industrial use. Although it is clear that biopersistence is a major determinant of cancer risk in animals, and perhaps also in man, other factors affecting the biological activity of mineral fibres may also be important.