Here are the figures Jim...
SET DELETED ON
Union 0.356 secs
Single 22.3 secs

SET DELETED OFF
Union 0.01 secs
Single 20.9 secs

This is in line with what you are saying, but the join even more spectacularly outperforms the "OR" query with SET DELETED OFF!!

>Hi David,
>
>That was the point of the article (or at least one of them) - when you have no deleted records you have the worst case (performance-wise) when DELETED is ON *and* you have a key on Deleted().
>
>You can check this out quickly by simply running the SQL with DELETED set OFF to see if there is a difference.
>
>Prior discussions on this led the believers to conclude that one should *test* using a key on Deleted() (assuming DELETED is ON) to verify that the query is otherwise fully optimized but them remove the key for production purposes.
>
>Cheers,
>
>Jim N
>
>
>>G'day Jim,
>>I haven't looked at the article again, you could be right but I'd be surprised. My tables had no deleted records.
>>
>SNIP