John,

<< Ahh.  Now I see where the results are coming from - I couldn't work it
before. If this means that correlated sub queries are executed on a row by
row basis for aggregate outer queries then I guess there is some scope for
optmization (geeky interest rather than a request <bg>). >>

Yeah, I think that EDB is just simply thinking that it doesn't need to
execute the sub-query as an aggregate operation, and so it is basically
combining a row-by-row evaluation with an aggregate evaluation, or
vice-versa.

--
Tim Young
Elevate Software
www.elevatesoft.com