In this article I discuss the justification of scientific change and argue that it rests on different sorts of invariance. Against this background I consider notions of observation, meaning, and regulative standards. I sketch an account of the rationale of scientific change which preserves the merits and avoids the shortcomings of the approach of Feyerabend, Hanson, Kuhn, Toulmin, and others. Each of these writers would hold that transitions from one scientific tradition to another force radical changes in what is observed, (...) in the meanings of the terms employed, and in the metastandards involved. They would claim that "incommensurable" replacement is what does, and should, occur during scientific "revolutions." Such writers, however, have trouble in comparing different scientific theories. My account permits different theories to be compared in various ways. These comparisons, I argue, are possible through appeal to "first-level" and "second-level" invariance. In this connection I argue that observation, meaning, and regulative standards should be, and in fact usually are, nontrivially invariant with respect to scientific change. (shrink)