classic coherence

Bill Mann bill_mann at SIL.ORG
Tue Jan 11 19:13:36 UTC 2000


Dear RST list folk:

Before we set aside the subject of coherence completely,  I want to expose and
comment on some thoughts that come from David Weber, also of our list.

Here is a message he sent, which I have edited:

  You had written:
(1) a. I love to collect classic automobiles.
        My favorite car is my 1899 Duryea.
     b. I love to collect classic automobiles.
        My favorite car is my 1977 Toyota.

(David:)... for me, 1b is just as coherent as 1a.

You (Bill Mann) wrote:
> On coherence:  if coherence is an intuitive judgment, then yes, without
> explanation, 1b is incoherent.  If I take an RST approach to explaining that,
it
> is that A. only the elaboration relation is anywhere close, and B. elaboration
> does not fit; specifically, elaborating "classic cars" , I cannot believe that
> the writer thought (in reporting year 1983) that a  1977 Toyota was in a
> set:member relation to "classic cars."
>
..

> What relation would you use?

(David:)  The RST relation I would posit is elaboration. I can imagine many
contexts in which it works fine. For example:
(0) The speaker is in love with old Toyotas, and says 1b as a way to
convey his elevated opinion of them.
(1) The speaker is plagued with problems with his old car and says 1b
ironically (up to his ears in grease from under the hood).
(2) The speaker has various cars but the are all junkers, and says 1b as
a humorous way to characterize his collection.
(3) The hearer collects cars which he regards as classics and the
speaker says 1b poking fun of the hearer's collection.
(and so forth).
..

Note: I don't think this undercuts RST: presumably  we would still
impose some relation on the two clauses in the process of interpreting
them.
=====================
 First of all, I want to cheer for David's treatment of "context" as something
far more extensive than the nearby parts of the text in question.  Clearly this
sort of  broad approach is needed to make consistent sense of these cases.  The
word "situation" could also be used.

Part of the problem comes from saying "Here is a text fragment (or two). How are
they coherent?"  The fragment of course comes from Mann and Thompson, and they
must take full blame.  (In another place they say that some of the illustrations
were made up, but the theory rests on natural texts.  Given space limitations in
journals, that is often required.)

The methodology of offering a text fragment and saying "Is it coherent?" is
flawed, just because it lacks the situation or context needed to produce an
impression of coherence, or to fail to do so.

For example (1),   the situation yields irony.  But if we apply that situation
to the 1899 example, it does not.  Situation (2) does not permit 1a at all.
Situation (3) produces ridicule or teasing with 1b, but not with 1a.

The illustrative purpose of this example was to compare 1a, spoken in a
situation in which according to the situation it was clearly sincere and
informative, with 1b EMBEDDED IN THE SAME SITUATION.  In 1a, the set:member
relationship is plausible; in 1b in that situation and reading, it is not.  So
the example works, as an illustration, when specified far enough.

We should avoid building our results solely, or even largely, on constructed
examples.  That practice produces even more difficulties in discourse studies
than it does in syntactic studies.   We should use  fragments of natural texts
with caution.

Yes, context and situation are necessary.  Notice that Weber, for all of his
examples, gave facts about speaker or hearer, not wording.

=====================================================

Having said all of that, let me use this case to produce an amusing (to me)
example, violating some of my advice.

Take the two examples 1a and 1b above, and change the brand names to Ford.
(Fords have been produced from classic automobile times up through the present.)
 Use the situation described above, where the speaker is known to be sincere.

We then have:

     a. I love to collect classic automobiles.
        My favorite car is my 1899 Ford.
     b. I love to collect classic automobiles.
        My favorite car is my 1977 Ford.

Imagine that coherence is a well bounded category of this language E, with a
limit that makes year 1899 work and 1977 fail.  The general case is:

    c. I love to collect classic automobiles.
        My favorite car is my <year N> Ford.

Now all of the mathematicians can tell us that there exists an N such that
coherence arises but for M=N+1 it does not.  The empirical issue is: What is N?
How do we find out?  From the speakers of E? What do we ask them?

(Or perhaps is it the case that there are degrees of coherence even for a single
reader?)

Enough.  Or more.  Apologies for the length of this.

Bill Mann












______________________________

The RST website is at http://www.sil.org/linguistics/RST   .



More information about the Rstlist mailing list