Discussions about social science digital archives have tended to address methodological, technological, ethical and legal issues such as what digital tools to use; what metadata to collect; how to curate numerical or interview data; and how to make archived materials available in ethical and legal ways. These discussions take place on the assumption that these archival practices are ontologically independent from archival materials. Following Derrida (1995), we want to explore the relationship between archival practices and archival documents on the assumption that 'archivization produces as much as it records the event' (Derrida 1995:17). On this approach, archival practices are understood as non-innocent, culturally and historically-specific practices that, in the act of 'preservation', help make specific 'memories' at the expense of others (Barad 2007, Derrida 1995, Foucault 1972. In this paper we take up this issue in relation to the curation of social science quantitative research data. We conceptualise data curation practices in broad terms as including specific practices-e.g. cleaning up of datasets; data anonymisation; streamlining interviews; data storing, categorising and visualisation; data search tools; etc.-as well as a wider range of knowledge, ethical, legal, political and economic practices these practices are entangled with-e.g. field-and discipline-specific knowledge-making practices; national/international data management and curation policies, practices and guidelines; ethical guidelines provided by professional bodies; countryspecific data protection, copyright, and information sharing legislation; etc. For the purposes of this paper we focus on three specific data curation practices-data cleaning, data anonymisation and metadata preparation-and investigate the ontological processes through which these practices help constitute the survey data they ostensibly archive.