There is a large and apparently causal link between life-events and depression. At least half the depressed women we studied had a recent life-event of aetiological importance. However before this is discussed it is essential to deal with some of the methodological issues surrounding the establishment of causality in non-experimental research. It is not just a matter of statistics and technical issues - although these can be involved. Methodological considerations start from the fact that it will usually be possible to reach alternative conclusions about the same set of results: methodology concerns anything that enables the researcher to rule out explanations that compete with the one favoured. Returning to the investigation of life-events and sore throats in children, perhaps it was not emotional upset but change in routine brought about by the crisis that led to the raised rate of illness? This is an explanation that still assumes that there is a causal link. More fundamentally the whole idea of a causal link might be rejected; it might be suggested, for instance, that the regular fortnightly questioning conveyed to parents what the study was about and some parents, in order to be helpful, made an effort to think of a crisis to 'give' to the investigators once they knew their child had a sore throat. And that this and not a causal link explains the association between life-events and sore throats. Investigators have to do what they can to anticipate such objections. The first, for instance, might be met by excluding from the analysis events likely to have led to changes in a child's routine (e.g. admission of mother to hospital) and demonstrating that the association still held with the remaining events (e.g. the child witnessing a bad road accident). In response to the second, more fundamental, objection it can be pointed out that the association holds not only for overt illness but for the acquisition of strep throats

without overt illness. Since parents could not have been aware of such changes in their child the main results become more plausible. These are typical methodological arguments and illustrate how the issues can usually be appreciated without special statistical or· mathematical knowledge. One further point should be borne in mind. It is unnecessary to demonstrate that a methodological objection is correct, only that it is a reasonable possibility. It was only possible, not obviously true, that the parents of children with sore throats had made an effort to remember events to 'give' the investigator. The objection is plausible enough to deserve some response.