bad-hair day OR bad hair-day?

[RonButters at AOL.COM]

In a message dated 1/26/2000 4:29:05 AM, lynnem at COGS.SUSX.AC.UK writes:

<< The argument seems to be what kind of evidence we can use to prove the
constituency of 'bad hair day', but I think Bob Haas might be write (if he's
suggesting as I think he is) that there's some idiolectal variation here.
Perhaps some people have parsed it one way and others the other way and so
the phrase means different things to different people.  (Hey, but my way is
the correct way!  Ha!) >>

This makes good sense to me. It is clear from this seemingly endless thread
that a minority of people think of "bad hair" as a genuine compound and a
majority see  "hair day" as just a noun preceded by a noun adjunct. Since the
semantic consequences are virtually identical, there are rarely if ever any
envronments in which the competing analyses would come into conflict (the way
that, say "good-time man" and "good time-man" could).

All I have been suggesting all along is that stress patterns tend to indicate
how people parse such constructions. Take, for example, "good time man". To
indicate "good-time man," people tend to say (1 = primary stress, etc.)
"1good 2time 3man" or "2good 2time 1man." To indicate "good time-man" people
tend to say "3good 1time 2man". I have never heard anything but "3good 1hair
2day", which suggest to me that most people parse it "bad hair-day" and not
"bad-hair day". Of course, stress in English is enormously complex--and
varies dialectally.