The principal magnetization of lavas of the Isachsen and Strand Fiord formations on Axel Heiberg Island is shown to predate the Eocene Eurekan Orogeny. Basalt flows of the Strand Fiord Formation, volcanigenic sandstone from the Christopher Formation, and the uppermost flows of the Isachsen Formation are normally magnetized. Reversed magnetizations are found only in the Isachsen Formation, occurring at two horizons, which, we suggest, correspond to M0 and M1 of the M sequence of marine magnetic anomalies (118–123 Ma). It is possible, therefore, that we have located, at least approximately, the base of the Cretaceous normal polarity superchron in these sections. Because inclinations are steep, the analysis of directions of magnetization is not straightforward and has been done by two methods. Method I assumes that no relative rotations have occurred amongst sample localities, and calculations on this basis show a 33 ± 24 °(P = 0.05) counterclockwise rotation with no paleolatitudinal displacement relative to North America. The rotation is in agreement with the rotation of 36 ± 8 °(P = 0.05) determined earlier from the Permian Esayoo Formation on Ellesmere Island. Analysis by method I assumes that the Esayoo and the Isachsen – Strand Fiord sampling localities on Axel Heiberg and Ellesmere Island are contained within what is essentially one large structural domain. The agreement (using method I) of paleolatitude with that of North America is consistent with standard plate reconstructions in which there is a gap of about 300 km between Greenland and Ellesmere Island. However, the dispersion of site-mean directions is greater than that expected for paleosecular variation during the Cretaceous, and therefore some of the dispersion may be attributable to relative motions amongst collecting localities. Therefore, by method II, relative rotations amongst localities are assumed to have occurred, and inclinations and declinations are analysed separately. As with method I, declinations are predominantly counterclockwise from that expected, but by method II the mean inclination (74 ± 2 °standard error) is significantly shallower than that expected (79 ± 1 °standard error). This apparent flattening is consistent with the idea that the Arctic Islands were close to Greenland in the Cretaceous and that there was no gap along Nares Strait. Hence both methods of calculation yield similar counter clockwise rotation, but each gives slightly different paleolatitudes. The latter difference cannot at present be resolved.