Any division between scientific practice and a metalevel of the methods and goals of science is largely a false dichotomy. Since a priori, foundationist or logicist approaches to normative principles have proven unequal to the task of representing actual scientific practice, methodologies of science must be abstracted from episodes in the history of science. Of course, it is possible that such characteristics could prove universal and constant across various eras. But, case studies show that they are not in anything beyond (...) the strictures applied to everyday, commonsense reasoning (e.g., a requirement of noncontradiction in a deductive argument). Hence, even if some presently-on-offer methodology or description of past scientific practice were adequate, it need not remain so for current (‘frontier’) areas of science. For this reason, it is important to examine recent episodes in, say, high-energy physics. Results from case studies of several episodes in that field are used to argue that successful practice leads scientists to countenance essential changes in the methodological framework at the levels of the criteria employed in judging theories (i.e., what counts for an explanation and what are canons of rationality) and of the goals of science. *Partial support for this research was provided by the History and Philosophy of Science Program of the National Science Foundation under grants Nos. SES-8606472 and SES-8705469. A preliminary version of this paper was given at an HPS seminar at King's College, London University in May 1988. Helpful comments and useful criticisms were made by several colleagues, especially Ernan McMullin, Heinz Post and Simon Saunders (none of whom are to be held responsible for or necessarily even in agreement with the views expressed here.). (shrink)