This study uses a corpus of just under two billion words from one nineteenth-century British newspapers, the Liverpool Mercury, to explore shifting attitudes to slavery in Britain in the nineteenth century in the context of a port city that benefitted from the trade. In doing so, explore three methodological issues –how to explore concepts in large corpora, how to do this over time and how to deal with poor quality data. Our approach to the study of concepts through time uses a new approach to looking at word usage change over time, Usage Fluctuation Analysis (McEnery, Brezina and Baker, 2019). Our exploration of the issue of poor quality data is motivated by the variable quality of the OCR texts which constitute our nineteenth-century newspaper corpus data. Problems in data quality bedevil work on large scale text collections of historic material. In this paper we will show that collocation, the core technique of UFA, can be used on such data if appropriate settings are chosen that minimise the problems arising from poor quality electronic text, permitting the exploration of corpus data at scale.