Chapter 3 considers different approaches to data collection. Three case studies are included. The first study involves a purpose-built corpus of news articles about obesity. We focus on theoretical considerations attending to corpus design, as well as practical challenges involved in processing texts provided by repositories such as LexisNexis to make them amenable to corpus analysis. The second study focuses on how corpus linguists might work with existing datasets, in this case, transcripts collected by research collaborators conducting ethnographic research in Australian Emergency Departments. We discuss the ways in which data collected for the purposes of different kinds of analysis is likely to require some pre-processing before it becomes suitable for corpus-based analysis. The third study is concerned with the creation of a corpus of anti-vaccination literature from Victorian England. We discuss the challenges involved in sourcing historical material from existing databases, selecting a principled set of potential texts for inclusion, and using optical character recognition (OCR) software to convert the texts into a format that is appropriate for corpus tools.