>I played around with changing the order of the tables >in the FROM clause and in the use_merge hint. It does >change the order in the plan, but the logical reads come >to the same. Arul A merge join is like a zipper - you take two (sorted) rowsets and read through both of them concurrently. So (in principle, in a simple case) there's no difference between merge-joining A to B or merge joining B to A; I'm intrigued that there's no table access to CUS_ACC. Does that contain only the ACC_ID column, whose value we can derive from the PK_CUS_ACC index? Or is it index-organised? Nigel