More than you ever...

[Dennis R. Preston]

OK. Since there are competing arguments for why the "ph" in "amphitheater"
is often rendered with a /p/ (rather than an /f/), let's see how they shake
out.


I. I find no alternative words to examine. That is, there is nothing else
in the "common" English vocabulary with the structure

XVM+fv+'x'

+ means "syllable boundary"

XVM means any segment plus vowel plus /m/; caps means primary stress ("am")

fv means /f/ plus any vowel ("phi"); lower case means weakest stress

x means any syllable; ' ' means weak stress (but stronger than the
preceding
        syllable ("the")


II. Let's have a more detailed look at the procedures. If the underlying
structure I propose above is correct, whether regressive or progressive
assimilation is at work is (at least here) a by-product of stress. That is,
the stressed syllable is likely to be the "assimilator" and the unstressed
(or more weakly stressed) syllable the "assimilated." In this case, /æm/
(the first symbol here should be the low-front vowel) is the assimilator,
/f@/ (/@/ = schwa) the assimilated. (Note that the underlying /i/ of "phi"
has already laxed to schwa (/or /I/ in some dialects) by virtue of its
presence in a weakly-stressed syllable.

If this story is true, then I still like my earlier contention that the
first two of the features associated with /m/ (bilabial and stop) will
transfer (progressive assimilation) to /f/, converting it to /p/.

III. What about two other facts raised in recent messages?

A. What happens if we assume that the /æ/ of "am" not only nasalizes but
that the nasal is deleted (a common process)?

B. What happens if spelling has caused an underlying reanalysis of the word
so that it "has" a /p/ not an /f/, i.e.:

XVM+pi+'x'

C. What is the interplay between these two possibilities?

OK, one by one.

First, if underlying /f/ is changed to /p/ (as a result of spelling, as I
suggested in an earlier message we should be on the lookout for), then the
product would be the same as the rule I gave - /ÆMp at .../ (overlooking the
nasalization of /æ/, which I have capped to show major stress).

This would not be the first case in which a similar product was gotten to
by different processes. Of course, both can be "true." In fact, it is
simply an empirical question as to which underlying form a speaker has.
(Some may have both, although probably in different "dialects.") How would
you find out? One (not sure-fire) way is to have speakers pronounce items
very slowly (or syllable by syllable). For example, one way to discover
what the "underlying" vowel is in the words "a" and "the" is to ask people
to say "He's not A man; he THE man" (which, since it is idiomatic, is not
too hard to get many people to do). In this way, whether or not the
stressed for of "a" is /@/ or /ey/ and the stressed form of "the' is /th@/
or /thiy/ will jump right out at you. Of course, elicitation problems
abound here. If you give a "normal speed" representation, how will you know
that that has not primed the slow speech response? Problematic, but not
debilitating.

Second, suppose, as suggested in another message, the "am" becomes simply
/æN/ (by which I mean "nasalized æ," with no following nasal segment,
although I assume that the vowel is compensatorily lengthened).
Nasalization of the preceding vowel (and subsequent loss of nasal,
especially when a consonant follows, as it does here) is a common process.

This gets tricky.

First, if the individual has rerepresented the underlying form with a /p/,
something very interesting follows. Here we go:

/ÆM+p at .../ --> /ÆNM+p at .../ (nasalization of æ)
/ÆNM+p at .../ --> /ÆN:+p at .../ (deletion of nasal segment, compensatory
lengthening
                                of nasalized vowel)
/ÆN:+p at .../ --> /ÆNM+p at .../ (transitional nasal insertion - if you don't
believe
                                this rule, try nasalizing the /æ/ in "apple." I
                                bet you say "ample," not just "apple" with the
                                vowel nasalized. - Removal of lengthening.)

In short, I claim the form /ÆN:+p at .../ is very unlikely, and, strange to
say, the output is still the same.

Second, suppose the underlying form preserves the historical /f/:

/ÆM+f at .../ --> /ÆNM+f at .../ (nasalization of æ)
/ÆNM+f at .../ --> /ÆN:+f at .../ (deletion of nasal segment, comepnsatory
lengthening
                                of nasalized vowel)
/ÆN:+f at .../ --> /ÆNW+f at .../ (transitional nasal insertion; by /W/ I mean the
                                labio-dental nasal; in IPA it is a lower-case
                                "m" with the right side shaped like an angma.)

AHA! The transitional nasal insertion rule is obviously sensitive to the
following sound, and, if in effect, can "tell" us if the speaker has
underlying /f/ or /p/. UNLESS, story one is correct, in which nasalization
of the vowel works but deletion of the nasal segment does not, In this
case, the identity of the underlying form is "hidden."

I believe both are possible (and the third with /p/).

So, a person with an underlying /f/ and "delayed" nasal segment deletion
after vowel nasalization can have a different surface form. All other
roads, it seems to me, lead to the same product, and what underlies them
may have to be determined speaker-by-speakedr (and, as we all know, worse,
speaker-by-speaker's variety).

dInIs









Dennis R. Preston
Professor of Linguistics
Department of Linguistics and Languages
Michigan State University
East Lansing MI 48824-1027 USA
preston at pilot.msu.edu
Office: (517)353-0740
Fax: (517)432-2736