Before the early 1980s, much attention had been given in the social survey literature to the analysis and interpretation of data, but much less to the problems of constructing the individual datum. Yet without good work at datum level a good data set cannot be produced, and without good data no useful analyses and interpretations may be made. What do we mean by ‘survey data’? What are ‘good’ data, and how do we recognise them?Originally published in 1984, Nicholas Bateson tackles t...