Actually, "\xA0" is a "non-breaking space" in "Latin1" (ISO 8859-1). I would expect an app like MS-Word to produce a byte like this if it were "preceded" (or "followed", depending on your notion of byte order) by a null byte, indicating UTF-16 encoding. But then I'd expect perl to complain about null bytes (but maybe it wouldn't -- I haven't tried). Or, if Word was using Latin1 as its output encoding (but that seems unexpected).