I think a 16-bit float type would require compiler revisions as opposed to doing something within the present type classes. This is similar to how Java would benefit from an unsigned byte primitive type for processing images, etc., whereas Haskell already has Word8. -- Regards, Casey