coherence

[James Cornish]

Please forgive the length of this, but it'll give you guys more to shoot
at:

In reply to my friend and mentor, Bill Mann's assertion that

>It would be very difficult to divide these two sentences into a
beginning, middle and end,
>and it does not seem insightful to do so.

 I think I can show through mixing the sequence of the text up and
asking speakers of English to reassemble the order that this text does
have a beginning, middle, and end.  And hope that the value of this
drive humans have to make these connects will arise for you readers on
its own.

THE TEXT

1.As a child, you had a pet;
        2.the pet was named "Sparky;"
        3.you threw vegetables under the table (presumably to avoid
eating them);
        4.Sparky ate them;
        5.Eating the vegetables affected Sparky's health.


A first shot at an analysis would take the trajectory of:

The beginning cannot be #5 because we do not as yet know who/what Sparky
is

 '5.Eating the vegetables affected Sparky's health.'  although is a
candidate for a 'beginning' because of the fullness of the lexemes (no
anaphors/cataphors nor direct extra-textual references beyond who/what
Sparky is (a dog or small male child are the most likely references in
my dialect of English).

#4 coming after #5 might work because of the anaphor, but to be
coherent,  4 would need an overt connector like 'so'.  For the sake of
this experiment, we don't add or subtract anything.

"so '4.Sparky ate them;' "  but #4 could not function as a beginning
within the setup of the available lexical items in the text because of
the pronoun reference problems.

#3. might function as a beginning, but there are much better choices
available given the set of lexemes in the text and how they are
presented.

'3.you threw vegetables under the table (presumably to avoid eating
them);'

And so on.  The most obvious choice for a 'beginning' is #1

'1.As a child, you had a pet;'  It seems to be a canonical setup phrase
for a story.

And #2 sets up the information flow for 'Sparky'

'2.the pet was named "Sparky;" '


I have used this type of mixed-up text activity in composition (writing)
classes with longer texts to make the pre theoretical point that we can
find the beginning, middle, and end intuitively through information flow
features (given/new etc.) and the intuitions that RST seems to get at so
nicely relationally.

I argue this point because I believe that competent speakers of a given
language have learned cues for deciding the possible sequences of items
in a text which makes one system's 'coherence' another system's
'incoherence'.  The extension of this would be in RST.  Do we learn this
type of cueing system differently in different languages for what is
labeled by RST relations?  I say "yes."   The insight available here
(which I do not claim for myself) is that we expect to have beginnings,
middles, and ends in texts.  Incoherence could come in some part, from
this effort being thwarted.  And there is always the notion that some
people put texts together better than others because of experience,
training and sheer tenacity.

When looking for RST relations, I find this sense of
beginning-middle-end helpful for predicting (or trying to) what the
writer/speaker is trying to do with the spans ALTHOUGH that is only part
of the process for deciding on the relation labels.  I need all the help
I can get.

And the other point Mann makes I agree with:

> I think of coherence as a pretheoretical notion, something to be
> explained by linguistic theories, and as spanning nearly all taxonomies
> of texts.  (I want to qualify "texts" here as really already restricted
> to communicative texts.  That is tricky to do, even pretheoretically, but
> explaining the independence of text type and finding coherence does not
> require doing that.)
>

But I think the value of the word 'genre' here adds the dimension of
"text type to whom for what purpose."   There lies one of the values of
RST that other ways of viewing coherence do not address very well.
These qualities must be considered (to whom and why) to some extent to
get the most useful analysis.   I really don't think one can get away
from the human originators nor the human interpretors of texts without
falling into the same set of problems that American Formalists
do--exceptions don't count.