I also went for answer 2 for the reasons as laid down by vk-kirov - I think in retrospect the wording 'what type of index (if any) will SQL Server make to support the FOREIGN KEY constraint' makes it pretty clear that seperately creating a unique constraint should not be considered as SQL creating a supporting index. So I guess I need to let Hugo off. <grumps>Still too much about the semantics though - bah!</grumps>