Abstract
It is shown that the apparent systematic discrepancies between values of the exchange integral J derived from different classes of experiments are removed when the angular dependence of J is taken into account.