Ash-tensing in *ANAE* (was: The duration of /ae/ and /ai/)

[Damien Hall]

A point of clarification:  Randy said:

'Labov et al. seem to have ignored that function words
can be accented or not, affecting the tensing of ash: "I can do it." (not
tensed because it is reduced) vs. "Yes, I can." (tensed because it is
accented).'

No, they didn't ignore that.  As far as I'm aware, none of Labov's work on the
quality of any vowel analyses unstressed tokens, for that very reason.  For
this particular case, I've actually written (under his direction) formal tests
to measure ash-tensing in that very word, and the approved sentences were as
follows:

- It's very difficult to get a good cheesesteak, but at _________ you can.
- These days, Coke cans are made of __________.

So, stressed tokens of *can* (n.) and *can* (v.).

(Obviously, the idea of the sentences was for the informant to concentrate on
filling in the blank and hopefully not realise that we didn't care in the
slightest what they filled it in with.  (Well, unless you actually like
cheesesteaks and wanted to get a Philly native's recommendation, which I did.)
In this way, they would produce a natural token of stressed *can*, and wouldn't
stress it too much because they knew that was what we were looking for.)

This leaves aside the interesting question of the U-shaped trajectory of /ae/
for some speakers.  I believe it's true for *ANAE* that only one point of each
vowel was measured, but there are two points to make about that:
- *ANAE* does describe the phenomenon of 'Northern breaking' (and Southern, but
that's not what we're talking about here):  the 'breaking' of a phonemic
monophthong into a diphthong (or maybe even a triphthong:  I'm not familiar
enough with *ANAE* to say, off the top of my head).
- In other work, Labov (et al) certainly do take account of the fact that
trajectories can vary.  For steady-state vowels, or vowels which phonemically
have a steady state, the measurement technique Labov recommends is to measure
at the point of inflection of a formant, since for any given token this
represents its closest approach to the target that the speaker was aiming for
(whether the target is reached or not is a different question, and of course
the target isn't the same from speaker to speaker).  If both F1 and F2 are
inflected, the point of inflection of F1 should usually be measured, as height
is often more perceptually prominent than degree of fronting.

Damien Hall
University of Pennsylvania

------------------------------------------------------------
The American Dialect Society - http://www.americandialect.org