>Thanks, Naomi
>
>Your method works fine, but, if I understand correctly, it won't work with SQL Server 2000, so Sergey'd method would be helpful there.
>Is that correct?

Yes, it's correct, but I would recommend derived table approach over correlated subquery.

Please visit the blog I gave you a reference to. It discusses all the methods with pros and cons. And in the part 2 you may see some measurement and you can also test performance for your cases - in fact, I appreciate additional test cases for that blog improvement.