Studies with 58 polynuclear aromatic hydrocarbons have shown that to repress demethylation of dimethylnitrosamine (DMN) in rat liver, the hydrocarbons must satisfy specific requirements of molecular geometry regarding size, shape, and coplanarity. Expressing the molecular size of these planar compounds by the two-dimensional area occupied, the size for maximal repressor activity ranges between about 85 and 150 A2. In addition to being within the correct molecular size range the hydrocarbons must have an elongated-rather than compact-molecular shape; circularly shaped and/or highly symmetrical hydrocarbons, such as coronene, triphenylene, ovalene, and tetrabenzonaphthalene, have very low activity or are inactive, in spite of being in the optimum size range. Coplanarity of the molecule is a critical requirement; thus, the potent carcinogen, 9,10-dimethyl-1,2-benzanthracene, is inactive as repressor of DMN-demethylase synthesis. Two exceptions, fluoranthene and benzol[ghi] fluoranthene, showed significant induction of DMN-demethylase. The molecular size distribution of hydrocarbons that repress the DMN-demethylase shows a mirror-image relationship with respect to the earlier reported molecular size requirement for indcution of azo dye N-demethylase. Compounds other than hydrocarbons also show the mirror-image relationship in the sense that pregnenolene-16alpha-carbonitrile, alpha- and beta-naphthoflavone, and Aroclor 1254 (known to be inducers of various mixed-function oxidases) are strong repressors of DMN-demethylase. Aminoacetonitrile, a strong inhibitor of carcinogenesis by DMN, is also a potent repressor of DMN-demethylase. The enzyme is inhibited by pretreatment of the animals with cobaltous chloride, an inhibitor of the synthesis of cytochrome P-450. Pregnenolone-16alpha-carbonitrile and 3-methylcholanthrene, despite their similarity of action on DMN-demethylase, have different effects on azo reductase, which is repressed by the former and induced by the latter compound.