In message <479FB9D5.4090501@oucs.ox.ac.uk> Sebastian Rahtz <sebastian.rahtz@oucs.ox.ac.uk> writes:
> James Cummings wrote
> > And if Bart was there as well?  How would I say Marge and Bart were 'the simpsons' in question attending an event?  
> yes, fair point. I still dont think the multiple @ref flies, though,
> because of the ambiguity of what it means.

I recognise that this is the problem.  The choices seem to be:

a) Leave it how it is, people do whatever and feel the TEI is flawed
b) Leave it how it is, but eventually give some suggestions on how to deal with all the cases you and I have mentioned
c) Change ref (say) to allow multiple URIs but stipulate what this means (alternatives, or multiple people all present, or whatever), and provide guidance for the other possibilities.
d) Change it some other way and provide guidance.

For my case, I'll muddle alone the best I can.  I'll probably add an new attribute  which takes multiple whitespace separated words for them to put it in with the understanding that these are all a one-to-one relationship with the people in this group name.

I suppose another change which would suit my use-case needs would be to allow personGrp to contain person.  Thus you can reference the 'multiple people as one group', but record information about the different people in that group.  But that really isn't the point of that element (since if the people are able to be distinguished they aren't really a personGrp). In my use case, the data will be used by statisticians drawing network relationships between shifting groups of people and distinct meetings.  Thus, that every individual is counted as an individual is paramount, even if they are recorded as 'the smiths' in the source text.

-James