I see this. When I tested wrote and tested this I am fairly sure I used QA (not sure if was 7 or 2000 thou) and against multiple 7 instances but it would have been pre SP4 on SQL 7. If I get a chance I will probe into what changed the effect on SQL 7 and when. But at the time of the articles write up that was the effect you would get. Just one additional reason IS NULL is a better choice because if there was a fundamental chage in evaluation then your expected results may not be what you got after whatever changed this. It'll probably be a month before I can test thou as I am in the middle of a big project.