It appears the maximum precision of the NUMERIC type is scale 8 for precision <=10 and then (scale+precision) <=18 for precision > 10. Is that correct? I'm running 64bit on windows, would another platform allow an increase in scale? e.g. create table tnum ( a numeric (10,5), a2 numeric (10,6), a3 numeric (10,7), a4 numeric (10,8), b numeric (11,5), b2 numeric (11,7), c numeric (12,2), c2 numeric (12,5), c3 numeric (12,6), d numeric (16,2), e numeric (17,0), e2 numeric (17,1), f numeric (18,0) );