Summary One hundred and one histologically confirmed gastric cancer patients in Gwynedd, North Wales, were matched by sex, age and social class to two hospital inpatients without cancer. Seventy-seven of the gastric cancer cases were also matched, using the same criteria, to a patient with a confirmed cancer of a different site (excluding oesophagus). A (Stocks, 1936(Stocks, , 1937(Stocks, , 1939 and although the disease is declining in the United Kingdom as a whole, the incidence in some administrative districts of Gwynedd (North-west Wales) is still substantially higher than the national average. This fact is strikingly displayed in mapped form by Howe (1970) and by Gardner et al. (1983).The high incidence in North Wales has never been satisfactorily explained although the possibility that environmental and/or dietary factors are involved has been investigated by a number of different authors (Stocks, 1957;Davies & Wynne Griffith, 1954;Howe, 1979). Furthermore, population migration studies among the Welsh (Armstrong et al., 1983) and other races (Haenszel, 1961) would seem to exclude genetic influences as a major factor in aetiology.Many parts of Gwynedd have in the past formed 'island communities' with static populations who could thus be exposed throughout life to local environmental carcinogens. Bracken (Pteridium aquilinum) is widely distributed and one of the most successful weeds worldwide (Fenwick, 1988). It is estimated that almost 7% of Wales is occupied by bracken, and in Gwynedd the coverage is substantially greater, exceeding 20% land cover in parts of the county (Taylor, 1985).Many investigators have demonstrated the carcinogenic potential of bracken in a variety of animal species. Chronic bovine enzootic haematuria, which may be followed by bladder cancer, has been reported among cattle from many parts of the world (Pamukcu et al., 1967;Jarrett et al., 1978). In all cases there is good evidence that the animals have been affected as a consequence of grazing upon growing bracken or of eating cut bracken used as bedding; and bracken feeding experiments have confirmed the association (Pamukcu et al., 1967). The earliest experimental work was performed on rats who were given bracken fronds in their diet for 2 months; all the animals succumbed to ileal adenocarcinoma (Evans & Mason, 1965). The same workers obtained bladder tumours in guinea pigs, but in mice the most frequent malignancies produced were leukaemias and gastric carcinoma (Evans, 1984). The same malignancies can also be produced by feeding bracken spores to these animals (Evans, 1986). Among the features of bracken carcinogenicity of special interest observed in these mouse experiments are the vulnerability of the young weanling animal and the relatively long latent period before gastric tumours develop (Evans, 1987