Whoops, didn't notice my mistake with the DWORD and QWORD definitions until I read your posting.
Yes, the 80x86 contains byte-level instructions, so in that sense my definition seems off. The definition I was using was the definition as it applied to mainframe and minicomputers. Microcomputers are somewhat of a different sort of beast, though its development roughtly parallels that of the mainframe and minis (at least when it comes to OS architecture). Won't be surprised when meaning of byte gets changed along the way too, especially when ASCII gets replaced with something like Unicode. Speaking of character sets... I cringe every time I think about problems with EBCDIC to ASCII conversions going awry, or dealing with processing CDC 6/12 character set (though this pretty much prepared me for dealing with oddities of character encodings used for Asian languages such as with JIS, Shift-JIS, and EUC for Japanese text).