> It is a long time since I read that part of my copy of Knuth, and > it currently resides in Canada, while I reside in Germany, but I > seem to recall that he did give a reference for his algorithm. I have copy of the Soundex algorithm, translated to Rexx from a PL/I program in VMSHARE's NOTE SOUNDEX. The original PL/I code was done by Ezra Goldman of the University of Illinois at Chicago. He attributes the "secret algorizm" to an IBM publication (GC20-1707-0) "Identification Techniques". His code is dated 1978, so if you're a pack rat, you might have a copy in your manual library. > I assume sounmds was a typing error! Would SOUNDEX consider it > equivalent to sounds? :-) Yes: Soundex('sounds') = 'S532' Soundex('sounmds') = 'S532' That's the whole idea, right? Ross Patterson Rutgers University