A theoretical treatment of forgetting is presented which uses scaled variables, provides a new definition of independence of rate of forgetting from initial level of learning, and provides an appropriate statistical test of that independence based on functional measurement methodology. It is shown that Loftus's (1985a) horizontal interaction method leads to conclusions that are at odds with existing theory, and his conclusion that forgetting is slower for higher degrees of original learning is weakened thereby although it is consistent with his method. Arguments are given for preferring the method presented here. Slamecka and McElree's (1983) approach is rejected, but their conclusion that forgetting rate does not depend on original learning is supported.I am indebted to Donald Bambcr, Geoffrey Loftus, Sandy Pollatsek, Keith Rayner, Norman Slamecka, and two anonymous reviewers for helpful comments on earlier drafts.