The breakdown of positivism involved a focus on perspective, and with that focus came attention to difference in most academic fields. Translation studies is no exception, and in the last fifty years, most schools of translation theory and practice have moved away from prescriptive and normative stances on translation, viewing translation instead as a one-to-many process in which translators must make choices. If this is true, then difference is inescapable in translation and the question of similarity is a problem. This paper argues that rather than taking the form of an invariant core, similarity between a text and its translations takes the form of a cluster of "family resemblances". One way to understand these family resemblances is to see translation as a metonymic process in which each translator privileges specific parts of a source text in the transfer process.