Biodiversity is a data-intense science, drawing as it does on data from a large number of disciplines in order to build up a coherent picture of the extent and trajectory of life on earth. This paper argues that as sets of heterogeneous databases are made to converge, there is a layering of values into the emergent infrastructure. It is argued that this layering process is relatively irreversible, and that it operates simultaneously at a very concrete level (fields in a database) and at a very abstract one (the coding of the relationship between the disciplines and the production of a general ontology). Finally, it is maintained that science studies as a discipline is able to (and should) make a significant contribution to the design of robust and flexible databases which recognize this performative character of infrastructure.