A simple method is derived for combining the results from m groups of paleomagnetic directions or poles where the groups may have widely differing numbers and where the resultant vectors for the individual groups are not known. Numerical experiments with Fisher distributions show that, given a reasonable number of group means, surprisingly reliable values for R can be obtained even if the group precisions vary widely. An important advantage of this method is that one does not have to consider whether each group by itself is sufficient to average secular variation. Even a single VGP can be included as one of the groups. As an example the method is applied to combining the Cretaceous poles for North America. ¿ American Geophysical Union 1995