Gap junctions provide direct intercellular communication by linking adjacent cells with aqueous pores permeable to molecules up to 1 kDa in molecular mass and 8-14 A in diameter. The identification of over a dozen connexins in the mammalian gap junction family has stimulated interest in the functional significance of this diversity, including the possibility of selectivity for permeants as seen in other channel classes. Here we present a quantitative comparison of channel permeabilities of different connexins expressed in both HeLa transfectants (rat Cx26, rat Cx32 and mouse Cx45) and Xenopus oocytes (rat Cx26 and rat Cx32). In HeLa cells, we examined permeability to two fluorescent molecules: Lucifer Yellow (LY: anionic, MW 457) and 4',6-diamidino-2-phenylindole, dihydrochloride (DAPI, cationic, MW 350). A comparison of the kinetics of fluorescent dye transfer showed Cx32, Cx26 and Cx45 to have progressively decreasing permeabilities to LY, but increasing permeabilities to DAPI. This pattern was inconsistent with selection based on physical size of the probe, nor could it be accounted for by the differences between clones in the electrical conductance of the monolayers. In Xenopus oocytes, where electrical and dye coupling could be assessed in the same cells, Cx32 coupled oocytes showed an estimated 6-fold greater permeability to LY than those coupled by Cx26, a comparable result to that seen in HeLa cells, where an approximately 9-fold difference was seen. The oocyte system also allowed an examination of Cx32/Cx26 heterotypic gap junction that proved to have a permeability intermediate between the two homotypic forms. Thus, independent of the expression system, it appears that connexins show differential permeabilities that cannot be predicted based on size considerations, but must depend on other features of the probe, such as charge.