The microsomal epoxide hydrolase (mEH) and cytochrome P450s catalyze the sequential formation of carcinogenic metabolites. According to one algorithm for predicting the membrane topology of proteins, the human, the rabbit, and the rat mEH should adopt a type II topology. The type II topology is also predicted by a recently established neuronal network which is trained to recognize signal peptides with very high accuracy. In contrast to these predictions we find, based on N-glycosylation analysis in a cell-free and in a cellular system, that the membrane anchor of human, rat, and rabbit mEH displays a type I topology. This result is correctly predicted by the positive inside rule in which negatively charged residues, the distribution of which differs in the mEH membrane anchor of these species, have only a modulating role for the membrane topology of proteins. However, our results demonstrate that this role is not strong enough to force the mEHs into a type II topology, not even in the case of the rabbit mEH, in which the only positively charged residue in the C-terminal part of the topogenic sequence is flanked by five negatively charged residues.