On 08/02/13 04:02, Steve Hayes wrote: > On Thu, 07 Feb 2013 20:01:28 +0000, Richard Smith <richard@ex-parrot.com> > wrote: > >> I've been considering how to record civil registration of births and >> deaths in my own genealogy database. Currently, if I discover a birth >> registration in, say, Q1 1867, then I enter a birth event for Jan-Mar >> 1867 and a source citation to the UK GRO index (or to FreeBMD sa >> appropriate). However, this is wrong as the person could have been born >> in late Dec 1866 and the birth registered (perfectly properly) a week or >> two later, as happened with my great grandmother. And it's certainly >> possible that the birth was earlier still and registered late -- just >> because (in the UK) there was a legal requirement to register births >> within 42 days doesn't mean it necessarily always happened like that. >> >> But I'm now thinking that I should consider 'birth registration' to be a >> separate event to 'birth', much as I consider 'baptism' to be separate >> to 'birth'. I can then say, quite correctly, that the birth >> registration occurred in Jan-Mar 1867, and I can use this to estimate a >> birth in Nov 1866 - Mar 1867. This also allows the 'birth >> registration' event to have different participants to the 'birth' event. >> For example, a birth registration (in the UK, at least) has an >> informant who is often but not always a parent. > > You don't say what your "own genealogy database" is. It's a database and a collection of scripts I've developed myself. So my question isn't how I should do it in a specific piece of software, but rather, given a clean slate, how would you wish to record it? My basic event data model is the one in http://vocab.org/bio/0.1/ though I'm not adverse to deviating from it if there's a need to. (Though, to be honest, it's so general that most of the time I'm asking myself which bits I don't want to use, rather than which bits I need to extend.) At the moment, I'm thinking that the way forward is to add a 'Registration' event (perhaps with a sub-type 'BirthRegistration') and link it as a followingEvent from a 'Birth' event. Let me try to give a concrete example. I find my great grandmother's birth registration on FreeBMD: Susan BARFOOT, b reg. Q1 1867, South Stoneham, Hants. [vol.2c, p.4] This is evidence that someone called Susan BARFOOT was born, but not necessarily that the birth occurred in Q1 1867 (it didn't) or in the South Stoneham registration district (though in this case it did). It's the registration that occurred in that quarter and that place. It is, however, evidence that the birth occurred no later than Q1 1867. So I would create a data structure that looks something like this: [Person: name="Susan BARFOOT"] | V life event: type=birth | [Event: type=Birth] | V following event: type=registration | [Event: type=Registration; date=1867-Q1; place="South Stoneham R.D."] I may choose to add an estimated date to the Birth event itself, but if I do, I don't lose track of the fact that the registration date is nominally exact, and any estimate of a date of birth is just that: an estimate. Arguably, the Birth event there is redundant in that any event involving a person must be a followingEvent from the birth. If someone appears in a phone book, then it's evidence that they were born somewhere, somewhen. But I think there is a difference. The registration of a birth is closely related to the birth itself, and if you consider other forms of civil registration, e.g. of deaths, then the the associated death event does convey genuinely new information. Anyway, I'd be interested to hear any views on this strategy. > But beware if you upload anything to Ancestry.com with imprecise > place names, as it will change them to something that it has in its > database, which may well be on another continent. Yes, I've seen that before, and it's one reason why I no longer upload trees to Ancestry. To make matters worse, Ancestry's list of place names is missing the parish of Fawley, Hants, which is one of the most commonly used parishes in my database, and given half a chance will auto-correct it to Fawley, Berks. Richard

Hi Richard, > This is evidence that someone called Susan BARFOOT was born, but not > necessarily that the birth occurred in Q1 1867 (it didn't) or in the > South Stoneham registration district (though in this case it did). It's > the registration that occurred in that quarter and that place. It is, > however, evidence that the birth occurred no later than Q1 1867. > > So I would create a data structure that looks something like this: > > [Person: name="Susan BARFOOT"] > | > V life event: type=birth > | > [Event: type=Birth] > | > V following event: type=registration > | > [Event: type=Registration; date=1867-Q1; place="South Stoneham R.D."] If I were to create an evidence based model, which I once hope to incorporate in Gramps, I would probably reverse this, meaning that the model starts with the evidence/source as a top level object, like this: [Source: title="Susan's birth record"; date=1867-Q1; ...] | V | [Event: type=Birth; date=...] | V | [Person: name="Susan BARFOOT"; role=child] [Person: name="... BARFOOT"; role=father] [Person: name="..."; role=mother] And if it were a baptism instead, I would add witnesses to the person list too. The event persons (or personas) listed here are not the individuals in my tree, but separate entities that can be linked to individuals in my conclusions, i.e. the tree in Gramps, or whatever software you use. The individuals can thus be seen as event person containers, that don't have much attributes of their own but are merely ways to group event persons to what you assume to be true individuals. The idea behind this is that keeping the persons in the registration part of the model is the best way to separate evidence and conclusions. cheers, Enno