Archaeological data are shadowy in a number of senses. They are notoriously incomplete and fragmentary, and the sedimented layers of interpretive scaffolding on which archaeologists rely to constitute these data as evidence carry the risk that they will recognize only those data that conform to expectation. These epistemic anxieties further suggest that, once recovered, there is little prospect for putting "legacy" data to work in new ways. And yet the "data imprints" of past lives are a rich evidential resource; archaeologists successfully mine old data sets for new insights that redirect inquiry, often calling into question assumptions embedded in the scaffolding that made their recovery possible in the first place. I characterize three strategies by which archaeologists address the challenges posed by legacy data: secondary retrieval and recontextualization of primary data, and the use old data in experimental simulations of the cultural past under study. By these means, archaeologists establish evidential claims of varying degrees of credibility, not by securing empirical bedrock but through a process of continuously building and rebuilding provisional empirical foundations.