If we had enough data then this statistical approach would undoubtedly sort out these things, and a lot of problems are arising precisely because we haven't got enough documents for the statistical approach to be wholly valid. I know you can calculate levels of probability and so forth, but to establish this really clearly we want a lot more information than we have actually got available. This is surely our major problem that we are still at the very limits at which you can use a technique of this sort. - John Chadwick
