ABSTRACT
We begin with the problem of causality, a central debate in the social sciences and despite the number of contributions a seemingly intractable one.2 The problem, and that of inductive knowledge to which it is closely linked, revolves around the leap of faith that is required to move from the observation that x is related to or correlated with y to the statement that x causes y. Statistical tests allow us to say with an exact certitude whether a particular relationship exists; they do not allow us to draw any conclusions about the causal status of that
relationship. Thus we might discover that unsafe sex is correlated with, say, perceived HIV status, but to move from that observation to the assertion that the one causes the other is not straightforward. Such a statement is possible only after a theoretical link is established between the two phenomena. Epidemiology (the medics’ social science) and, it has to be said, much contemporary American sociology have always relied heavily on statistical techniques, but ridden relatively lightly on the theory which allows such techniques to inform statements about cause. The emergence of the computer and the consequent wide availability of powerful statistical techniques have ensured that any data set will be beaten into submission by multiple logistic regression, log-linear analysis or whatever the preferred instrument is. Unfortunately, this automatic recourse to number crunching can all too easily take the place of sustained critical thought. Thus papers report that such and such a list of variables is associated with unsafe sex (or HIV antibody status or progression to AIDS), with the implication that these are in some sense causal. This may or may not be the case.
