Given the high frequency of ras oncogene activation in several common human cancers, its signal pathways are an important target for novel therapy. For practical reasons, however, these have been studied mainly in the context of transformation of established fibroblast cell lines, whereas ras acts at an earlier stage in human tumorigenesis and predominantly on epithelial cells. Here we have developed a more directly relevant model - human primary thyroid epithelial cells - which are a major target of naturally-occurring Ras mutation, and in which expression of mutant Ras in culture induces clonal expansion without morphological transformation, closely reproducing the phenotype of the corresponding tumour in vivo. Transient or stable expression of mutant H-ras (by scrapeloading or retroviral infection) at levels which stimulated proliferation induced sustained activation and translocation of MAP kinase (MAPK) in these cells. Inhibition of the MAPK pathway at the level of MAPKK, by expression of a dominant-negative mutant or by the pharmacological inhibitor PD98059, efficiently blocked the proliferative response. Conversely, selective activation of MAPK by a constitutively-active MAPKK1 mutant failed to mimic the action of Ras and, although this was achievable with activated Raf, micro-injection of anti-ras antibodies showed that this still required endogenous wild-type Ras function. In contrast to recent results obtained with a rodent thyroid cell line (WRT), therefore, activation of the MAPK pathway is necessary, but not sufficient, for the proliferogenic action of mutant Ras on primary human thyroid cells. These data emphasize the unreliability of extrapolation from cell lines and establish the feasibility of using a more representative human epithelial model for Ras signalling studies.