Ian Goddard wrote: > In the example which I quoted it's fortunate that the PR also includes > dates of birth so the situation is clear.  But what if it only had the > baptismal dates?  Your fuzzy rule would tell us that all 5 children were > born within a year. No, the rule was that "If a person was baptised on some date, then, in the absence of evidence to the contrary, he or she was probably born within the previous year", so we'd infer that all 5 children were *probably* born within a year. However, you'd probably have another similar fuzzy rule saying something like "Within any one year period, a woman probably gives birth to no more than two children." Again, exceptions are possible (triplets; twins and a single child in rapid succession), but they're not common. But in any good reasoning system like this, these fuzzy rules would be configurable. So if you were researching a family in a culture where it is common practice to baptise many children together, you'd be able to disable the rule, or tweak it to suit the situation. > I can, however, see the utility.  I frequently come across the situation > where there are multiple branches of the same family, often with > contemporary fathers of the same name, no mother's name given, and > sometimes having children baptised with the same name.  It would be very > useful to have a program to generate the possible alternative sets of > families. Yes, that's exactly the type of application I have in mind. Another is spotting when new evidence changes existing theories. Sometimes finding a new baptism is enough to rethink whether some distant cousin is one family or two, and changing that can have knock-on effects elsewhere as you have to rethink how the branches fit together and allocate marriages and burials to them. > And in this neck of the woods it be even better to have it > generate lists of children whose baptisms have been missed because the > minimum number of families is greater than the number of candidate fathers. Again, a good idea, and something that a computer is ideally suited to. But the really simple applications are valuable too. For example, I'm envisaging data entry for baptisms as a spreadsheet-like interface, say with columns for date, name, sex, father and mother (though they'd be configurable). (I certainly don't want one dialogue box per person as most applications have!) Once you've transcribed them from the parish register, it would be nifty if it could auto-group them into families. Often there are lots of children that partition into very obvious sets of siblings, and having the them automatically group together allows you to focus on the less obvious ones. We'll never get to a stage where the whole reasoning process is automated -- and where would the fun be if we did? -- but that's no reason not to allow a computer to automate the parts that can easily be automated. Richard