I have argued elsewhere that the status attributed to string theory and cosmic inflation by exponents of those theories may be best understood by extending the concept of theory confirmation. The proposed extension admits a mode of meta-empirical confirmation, which relies on meta level observations about the research process that lie outside the theory’s intended domain. In the present talk, I will compare this step to an earlier extension of the concept of confirmation that happened in the early 20th century where very similar considerations were of central importance. 19th century physics understood confirmation to be limited to hypotheses about observable objects. Therefore, even committed atomists like J.C. Maxwell at the time took atomism to be unconfirmable as a matter of principle. In the early 20th century, the mounting conceptual and evidential support for atomism led to the emergence of the wider concept of confirmation familiar today. I will discuss three episodes in the history of the endorsement of atomism that demonstrate how, in course of this development, meta-empirical assessment moved from being a controversial mode of reasoning in the later 19th century towards being absorbed into the emerging new notion of empirical confirmation in the early 20th century.