The in vitro metabolic ketone reduction of warfarin and its 4'-analogues acenocoumarol (4'-nitrowarfarin) and 4'-chlorowarfarin has been investigated using microsomal and cytosolic fractions of several species. Both subcellular fractions showed ketone reductase activity. The cytosolic fractions, in most species, exhibited strong substrate stereoselectivity as well as product stereoselectivity, i.e. the R(+)enantiomer was preferred as a substrate to be reduced mainly to the RS alcohol. Phenobarbital and methylcholanthrene induced cytosolic ketone reductase activity in the rat 5- to 11-fold and 3- to 7.4-fold, respectively. The microsomal fractions also showed substrate and product stereoselectivity. Contrary to the cytosolic fractions a general pattern for substrate and product stereoselectivity could not be seen. Stereoselectivity seemed species dependent (e.g. sheep vs bovine and pig). Rat liver microsomes showed practically no ketone reductase activity. Induction by phenobarbital or methylcholanthrene resulted in only a slight rise, if any, in rat microsomal ketone reductase activity. Both cytosolic and microsomal ketone reductases proved to be NADPH dependent. Substitution of the 4'-hydrogen of warfarin resulted in a change in reduction rates and in some cases, even in a change in substrate or product stereoselectivity. The data indicate that microsomal ketone reductases are different from cytosolic ketone reductases. Prelog's rule for product stereoselectivity of metabolic ketone reduction, when applied to the ketone reduction of the warfarin analogues did not agree with all data presented here.