This paper develops an approach to the scientific realism debate that has three main features. First, our approach admits multiple criteria of reality, i.e., criteria that, if satisfied, warrant belief in the reality of hypothetical entities. Second, our approach is experiment-based in the sense that it focuses on criteria that are satisfied by experiments as opposed to theories. Third, our approach is local in the sense that it focuses on the reality of particular kinds of entities. We apply this approach to a case that many philosophers have debated, namely, Jean Perrin's work on atoms and molecules. We provide a novel account by arguing that Perrin's work warranted a minimal belief in the reality of atoms and molecules as unobservable, discrete particles by satisfying a criterion of reality that we call experimental determination of number per unit. By doing so, he confirmed Avogadro's hypothesis, but he did not confirm other key constituents of the atomic theories involved. We argue that our account of Perrin's work is preferable to several other accounts, and we use this as a reason in support of our approach to the realism debate more generally.