>Followup...
>
>I just renamed the folder containing the copy of the database that I am testing. Now, __RI_DELETE_projvisits correctly finds both the parent and child tables, sees the conflict and prevents the delete.
>
>Any thoughts on why it looked elsewhere for the child table?
>
>Thanks
>Jim

It may be a bug in RI's code - with USE command it opens the first table in finds and the other database happens to be first in the search path (this is what I suspect may be happening).