The anticarcinogenic effect of vitamin A2 (dehydroretinol and 3-hydroxyretinol) compounds was studied and compared with that of vitamin A1 (retinoic acid, retinol and retinal) and carotenoids (lutein and β-carotene) in the benzo[a]pyrene (B(a)P)-induced forestomach tumour model of female Swiss mice in vivo. Tumour growth and gross tumour incidence observed after the administration of B(a)P (eight doses of 1 mg, twice weekly for 4 weeks) and retinoids/carotenoids (2·5 and 4·7 μm per animal per d, 2 weeks before, during and 2 weeks after B(a)P) showed that the groups supplemented with lutein and 3-hydroxyretinol produced the best results in inhibiting tumour growth and had low tumour incidence compared with the control group given B(a)P only (P<0·05). Weights recorded after the different treatments showed that the β-carotene-supplemented group exhibited maximum weight gain, followed by retinal, retinol, retinoic acid, lutein, dehydroretinol and 3-hydroxyretinol. These results indicate that the anticarcinogenicity of the compounds is not related to the vitamin A biopotencies. Vitamin A2 compounds having half the biopotency of the vitamin A1 compounds were seen to be anticarcinogenic. Again, among the carotenoids, lutein, having 50 % less biopotency, showed more significant results than β-carotene. Thus it is imperative to conclude that the low animal growth achieved with these compounds has a correlation with the highest suppression of tumour occurrence in the present experiment. Therefore, the daily consumption of foods having high content of lutein and vitamin A2 should be given due importance and weight in further studies.