Six samples of tremolite of different morphological type were prepared as dusts of respirable size and used in intraperitoneal injection studies in rats. Three "asbestiform" tremolites produced mesotheliomas in almost all animals, although with significantly different tumor-induction periods. A brittle type of fibrous tremolite which, when manipulated to prepare "respirable dust," produced a sample with relatively few asbestiform fibers remaining nonetheless produced tumors in 70% of rats. Two samples of nonfibrous tremolite produced respirable dust samples containing numerous elongated fragments with aspect ratios greater than 3:1, which therefore fitted the definition of respirable fibers. Both these samples produced relatively few tumors, although one had more long "fibers" than did the brittle tremolite that produced 70% of tumors. This study has therefore demonstrated that different morphologic forms of tremolite produce dusts with very different carcinogenic potential. Carcinogenicity does not depend simply on the number of elongated particles injected, and we need to develop methods of distinguishing carcinogenic tremolite fibers from relatively innocuous tremolite dusts, with similar numbers of elongated particles of similar aspect ratios.