Science education researchers agree about the importance of evidence in science practices such as argumentation. Yet, disagreements and ambiguities about what counts as "evidence" in science classrooms pervade the literature. We argue that these ambiguities and disagreements can be viewed as falling along three fault lines: (i) the source of evidence, specifically, whether it must be first‐hand; (ii) whether "evidence" must always be empirical; and (iii) the extent to which evidence is inferred, and what degree of inference transforms "evidence" into something else. In this paper, after showing how these three fault lines manifest in the literature, we argue that these three dimensions of disagreements and ambiguities are not confined to research and research‐based curricula; they are also salient in teachers' classroom practice, as illustrated by a dramatic, multiday debate between a mentor teacher and her teacher intern. After establishing the salience of the three fault lines in both research and practice, we explore whether Next Generation Science Standard (NGSS) can provide a resolution to the teachers' debate and to the disagreements/ambiguities in the literature. Our analysis reveals that NGSS reproduces rather than resolves those three fault lines—but in doing so, it invites a resolution of a different type. Instead of providing a single, precise, context‐independent definition of "evidence," NGSS implicitly reflects a defensible view that what counts as "evidence" depends on the epistemic aims of the practices in which the students are engaged. This implied context‐dependency of what counts as good evidence use, we argue, could be made explicit in an addendum document clarifying aspects of NGSS. Doing so would provide valuable guidance to teachers, teacher educators, and researchers. [ABSTRACT FROM AUTHOR]