This paper describes the process by which we have constructed a corpus of heterogeneous texts about non-human animals. It aims to contribute both methodologically – in respect of the challenges of compiling a thematic corpus – and substantively – in relation to the identification of some features of discourse about animals. Having introduced the research project and its guiding questions, the article describes the principles of data selection and the procedures used in analysis. We highlight the methods we devised both to avoid the potential circularity associated with pre-determined search terms, and to overcome the limitations of a relatively small corpus containing a wide range of relevant vocabulary. We go on to report some initial findings on the most frequent animal naming terms and adjectives describing them, including a small case study of the adjectives ‘live’ and ‘dead’. The article concludes by indicating the ways in which the iterative methods we have employed are open to further extension, and points to some methodological and substantive implications of this enterprise.