The increasing use of research metrics in the monitoring and evaluation of research has led to the expansion of research information infrastructure, as indicated by the rising number of national databases for research output and research information systems. At the same time, debates on requirements of national databases for research output are typically guided by simplistic assumptions about what an infrastructure is and how it shapes our understanding of different phenomena. We present an alternative framing of databases wherein infrastructure and context are seen as mutually constitutive. Drawing upon qualitative data on 12 national databases in Europe, we investigate the role that databases for research output play in research metrics and consequently in our understanding of research activities. Specifically, we bring to the surface database features that are typically left implicit and yet are crucial for understanding these databases in context. A key outcome of this study is an identification of two database logics--New Public Management and Enlightenment--that underpin database designs and the organization of database work. These logics do not merely guide practical database work (e.g. by determining the features of highest priority), but through the outcomes of that work they both enact and distribute a particular conception of what research is. In this article, we show how the two logics manifest and how they are related to each other. Situating these findings in the landscape of developments in research information infrastructure, we discuss implications for research evaluation and the understanding of research in general.