Introduction

One of the more notorious of Garfinkel's methodological recommendations, exemplified in his early ‘breaching experiments’, was the importance of rendering the familiar strange in order to make visible the work necessary to sustain the common understandings and practical reasoning that is the basis of the social order (Garfinkel, 1967). For Garfinkel, of course, no activity could, in principle, be exempt from this proposal, least of all sociology itself. The radicalness of Garfinkel's recommendations is manifested and celebrated in the contributions to this volume and, of course, elsewhere. Our concern in this chapter is to develop in depth a methodological issue which is deeply rooted in the human sciences and which arises from out of the general concern with measurement as considered by Mike Lynch in the preceding chapter: the structure of evidence and inference as exhibited in that family of techniques known by a name popularised, though in condemnation, by Blumer, as ‘variable analysis’ (Blumer, 1956). These techniques encompass not only the standard means of data collection in social research, such as questionnaires of all types, interviewing, the organisation of the inferential basis of these in survey designs and the principles of sampling, but also the sophisticated, and increasingly esoteric arsenal of statistical analysis. For most practising sociologists, and social researchers more generally, variable analysis is the familiar set of methods of social research against which all other approaches have to be judged and weighed.