Thanks for all the help Mark and Bill. I think I've got it sorted now. I did some testing on the server this morning and discovered it was choking just before the UPDATE, on a replace statement that tested ok in development.

In answer to your question, Bill, the only difference between my development/testing environment and the server is the quantity of data; I find it easier to test with less than 1.5 million records getting in the way! But in fact that is why the error didn't show up before; for once my test data was cleaner than the live data.

Many thanks again for all the help.

Nanci