A statistical method is developed to evaluate stratigraphic correlations that are based on the similarity of paleomagnetic directions. It involves comparison of the probabilities that such directions could have arisen by either simultaneous or random samplings of the ancient magnetic field. To calculate these probabilities one must estimate the effects of paleomagnetic errors and paleosecular variation, repsectively. The method is tested successfully on sections of 8--10 Grande Ronde Basalt (Columbia River Basalt Group) flows at Sentinel Gap and Umtanum Ridge, which are separated by only 25 km and contain two flows thought to be the same on geological grounds. More significantly, for widely separated sections at Sentinel Gap and Butler Canyon the analysis again shows that simultaneous magnetization is a much more likely explanation of the general similarity of paleomagnetic directions than chance agreement of ancient field directions. The simplest interpretation of this result is that many of the flows at the two sites are the same, implying that series of Grande Ronde flows extended over distances of several hundred kilometers.