> Sorry, I completely missed the boat on the solution. Here's what you > should look into. Convert each word in your English database to its > number pad equivalent. Then perform the number search on that column. I thought about that, but if I use my suggested approach, then the varchar comparison is a simple equality comparison and I am uncertain if doing a string equality will be significantly faster than a numerical one. You also got me thinking, however. Perhaps I could index the words very simply by adding an index for the first letter. Then I would need only to search that subset, based on the first letter of my string. I could always do this to two levels, but I think one would help and would be easy to manage. Thanks.