the game. Players usually do not arrive at particular positions by accident or luck; they are there because of the complex interlocking activities that produce the game. Ethnography tries to do just that: describe the apparently messy and complex activities that make up social action, not to reduce their complexity but to describe and explain it.7 This is what makes ethnography a demanding approach: it is not enough (not by a very long shot) to follow a clear, pre-set line of inquiry and the researcher cannot come thundering in with pre-established truths. The procedure is what Hymes (1980: 89) calls ‘democratic’: ‘a mutual relation of interaction and adaptation’ between ethnographers and the people they work with, ‘a relation that will change both’. That too is counter-hegemonic.
We now come to a tricky issue, one that has plagued many researchers facing supervisors and colleagues steeped in a more positivistic tradition of science: representativeness. What exactly do ethnographic data reveal? What sort of relevance do they have for ‘society’? How confidently can you make generalisations from your data?
A first and elementary point is this. Ethnography is an inductive science, that is: it works from empirical evidence towards theory, not the other way around. This has been mentioned several times already: you follow the data, and the data suggest particular theoretical issues. Ethnography, thus, belongs to a range of other scientific disciplines in which induction rather than deduction is the rule – history, law and archaeology are close neighbours. Inductive sciences usually apply what is called the case method: a methodology in which one uses case analyses to demonstrate theory. In the words of Lee Shulman (1986: 11):
A case, properly understood, is not simply the report of an event or incident. To call something a case is to make a theoretical claim – to argue that it is a ‘case of something’, or to argue that it is an instance of a larger class.
Your data become cases of such larger categories by applying theoretical models to them; theory is the outcome of a theorisation of your data, you ‘theorise them into a case’, so to speak. To turn to Shulman again: ‘Generalisation does not inhere in the case, but in the conceptual apparatus of the explicator’ (1986: 12).
This is an important point: generalisation is perfectly possible, and it depends on the theoretical apparatus that you bring to bear onto your data. Thus, in a situation in which your data are classroom observations about response behaviour by pupils, your data can be framed in, for instance, a Marxist perspective in which social class distinctions are central issues. Your analysis of the data will then focus on features in the data that speak to social class distinctions, and your generalisations will be about such class issues. If you frame your data in a cognitive-psychological theoretical approach, the data will be analysed accordingly and your generalisations will be about cognitive processes you observe in response behaviour.
Such things, of course, do not occur just at the end of your trajectory. You have explored theoretical frameworks prior to starting your fieldwork, and many of the choices mentioned here have been more or less determined by your particular research preparation and the formulation of your research goals. You usually know beforehand whether you will use a Marxist or a cognitive-psychological framework for your work, and these choices have influenced the design of your fieldwork and, of course, the particular kinds of data you have collected. The important point here is, however, methodological: generalisation is perfectly possible, because your data instantiate a case, and such a case belongs to a larger category of cases. The unique and situated events you have witnessed can and do indeed reveal a lot about the very big things in society.
The case method, as said, is typical for inductive sciences, and especially in legal studies the case method is dominant, also in teaching law. The interesting thing, however, is that it in turn builds upon a much older tradition, which Carlo Ginzburg (1989) calls the ‘evidential or conjectural paradigm’: evidential because it uses (inductive) empirical facts as its point of departure, ‘conjectural’ because these facts are seen as probably meaning this-or-that. The facts generate hypotheses that can then be verified. This paradigm is epitomised by Sherlock Holmes, who was able to deduce more insights from a cigarette butt left in an ashtray than his rival police inspector could by deploying his elaborate (deductive) criminal investigation tactics. But it is also epitomised in clinical medicine, where the surgeon first searches for small symptoms (‘clues’) that can then be conjecturally related to a larger category – the disease – and then be treated with drugs or other means. Thus the surgeon spots a rash on your arms, a swollen liver and a yellowish colour in your eyes, they hypothetically connect this to hepatitis, and then administer drugs to fight hepatitis. The surgeon’s hypothesis will be proven correct when the drugs are effective and the symptoms disappear.
Ginzburg finds ancient roots for this paradigm in divination – where the divinator would examine small things in order to predict big things – and he nicely summarises the case:
the group of disciplines which we have called evidential and conjectural (...) are totally unrelated to the scientific criteria that can be claimed for the Galilean paradigm [in which individual cases do not count – JB & DJ]. In fact, they are highly qualitative disciplines, in which the object is the study of individual cases, situations, and documents, precisely because they are individual, and for this reason get results that have an unsuppressible speculative margin; just think of the importance of conjecture (the term itself originates in divination) in medicine or in philology, and in divining. (Ginzburg, 1989: 106)
History, philology, psychoanalysis, archaeology, medicine, law, art history: these are the companions of ethnography in a long and venerable tradition of scientific work. In fact, every truly social science falls in this category. Chomsky’s linguistics was an attempt to bring the study of language – a social science, evidently – into the orbit of Galilean science. To Chomsky and his followers, linguistics would be a deductive science in which individual performance had no place, because individual cases could never invalidate the generalisations made from theory. In other social sciences as well, we have seen how strong the appeal of a deductive Galilean model of science was. The effect has been that the existence, and the validity, of this evidential and conjectural paradigm has been largely forgotten. Yet, it is the methodological basis for generalisation in ethnography, and it is a very firm basis.
Notes
(1) The following sections are based on a paper called ‘Ethnography as counter-hegemony’; International Literacy Conference, Cape Town 2001, downloadable from http://www.kcl.ac.uk/education/wpull.html.
(2) The journal Ethnography (launched in 2000) testifies to the impact of ethnography in a wide range of social sciences. An important, and frequent, contributor to the journal was Pierre Bourdieu, operating alongside sociocultural and linguistic anthropologists and microsociologists. Bourdieu’s own take on fieldwork and ethnography was exemplified in a special issue of Ethnography in 2004 (Wacquant, 2004).
(3) Cf. Hymes (1980: 89): ‘The earliest work that we recognize as important ethnography has generally the quality of being systematic in the sense of being comprehensive’.
(4) It may be interesting to point out that this view has percolated contemporary pragmatics. In the introduction to the Handbook of Pragmatics (Verschueren, 1995), pragmatics is defined as a functional perspective on language and communication. Verschueren refers, significantly, to Sapir (1929) as a source of inspiration for this view.
