SOMATIC mutation has been considered a likely initiating step in chemical carcinogenesis1–3, chiefly because of the close correlation between the mutagenic and carcinogenic activity of various chemicals4–6. Study of cells from patients with disease predisposing to a high incidence of cancer, such as xeroderma pigmentosum and Fanconi's anaemia, has suggested that defective DNA repair is correlated with this predisposition to malignant transformation. When cells are transformed by 3-methylcholanthrene7, 4-nitroquinoline-1-oxide (4NQO)8, or X irradiation9–11, division is necessary soon after treatment to fix the transformation: if treated cells are unable to grow the transformation is not fixed. These results could be explained if DNA damage induced by a carcinogen is converted into a stable and replicable structural change only by means of DNA replication before the intervention of error-free repair mechanisms. The model of mutagenesis12–14 based on evidence obtained with micro-organisms suggests that errors introduced by an error-prone postreplication repair mechanism are the principal source of mutation. Furthermore, caffeine inhibits “postreplication repair” (presumably an error-prone mechanism) but not excision repair (presumably error-free) in rodent cells15–19. I now report that caffeine reduces the transformation frequency of mouse cells treated with 4NQO. This supports the somatic mutation theory of cell transformation.