Thanks Gavin for a nice article. It seemed to me back in the 90s and seems to me today still that the distinction between statistical account and ethnographic account is a consequence of poor-quality thinking. Statistics are a very good way to interpret large datasets dealing with phenomenon that have low variance. But a “biographic” account (or rather, a great many such accounts) is a more suitable way to deal with a phenomenon that has high variance. In other words, statistical analysis of pulsars is valid but statistical analysis of psychopaths won’t be helpful; the former display a set of limited variables while the latter may display a very wide range of different variables.
Knowing one’s level of abstraction and the number of independent variables associated with that level of abstraction guide the mechanisms we use to understand (or at least attempt to describe) the phenomenon in question. Imagining that one technique alone serves all cases (“let’s do statistics for everything!”) is mere simple-mindedness that perhaps reflects an insecurity in the researchers themselves; the trappings of “hard science” compensating perhaps for some intellectual vacuity in the research itself.