Various studies in science education have concluded that successful science learning sometimes consists in students having two or more incompatible representations regarding a phenomenon. Specifically, a pluralist perspective acknowledges that such representational plurality is normal and even beneficial for the individual. Our working hypothesis in this paper is that in order for such plurality to be indeed functional, it must be adequately integrated into a cognitive structure responsible for its management. Following Cosmides and Tooby (Cosmides and Tooby, Sperber (ed), Metarepresentations in an evolutionary perspective, Oxford University Press, 2000), we use the concept of metarepresentation for this purpose. A metarepresentation is a structure that includes both a representation and various information about this representation, called tags. We focus on two kinds of tags: (1) scope tags, which are cognitive elements responsible for specifying the circumstances in which a representation can profitably be used; and (2) qualificative tags, which are judgments about the properties and value of a representation (e.g., simplicity, understandability, usefulness). We argue the concepts of metarepresentation and tags (or their equivalent) are required to better understand the nature of expertise in situation of representational plurality. In our view, humans have large minds: they can acquire a repertoire of incompatible representations and use it efficiently in various situations. Expertise does not lie solely in the mastering of one scientific representation, but in the capacity to operate adequately the repertoire one possesses.