Henkel and Franklin (1998) present a series of well-designed experiments in support of the conclusion that memory for the source of an item is affected by the similarity between the item and other information in memory. Their principle analyses use an empirical measure of source memory that is a variant of a measure evaluated by Murnane and Bayen (1996). We point out an important assumption that underlies the use of this measure and question additional arguments and analyses that Henkel and Franklin offer in support of their conclusions. The problems discussed illustrate the need for careful consideration of the technical characteristics of measures and the theoretical assumptions on which measures rest when one is conducting research on source memory.