Joe:
> And Rosta wrote:
> >
> > Which constructions would you say are the most nontrivial?
>
> Paul Kay once put it this way to me: every construction is a
> construction from Hell.
>
> Offhand, my *favorite* constructions in English are:

not all seem clearly nontrivial, though ...

> - Mad Magazine-sentences
>
>   I like these because of the interaction between prosody and syntax.

prosody aside (as, alas, it always is, until we know it better),
the syntax seems pretty ordinary: an infinitive verb with an
overt unraised subject.

> - Just because X doesn't mean Y and the Two-BE construction
>
>   I like these constructions because they illustrate the notion of
>   unusual configurations really well.

OK.

> - Paul Kay Structures such "I didn't get rich by throwing my money
>   away."
>
>   This is funny because the interpretation is that you did get rich
>   precisely because you didn't throw your money away (which is the
>   exact opposite of what is actually said).

? The literal meaning is "NOT: I got rich by throwing my money away",
which is not the exact opposite of "I got rich by not throwing my
money away".

A better example would be "I miss not going to WG seminars", meaning
"I miss going to WG seminars". I don't know if Americans say that
too.

> - What's X doing Y?; the "way" construction; nominal extraposition;
>   ditransitives; the X-er, the Y-er construction.

What do you mean by "nominal extraposition", exactly?

>   These are all interesting (= nontrivial) for their own reasons.
>   The list could go on quite a ways.

OK. Based upon this list, I would say that constructions do not
present a great challenge to Syntax As We Know It... That is, most
can I think be straightforwardly accommodated using existing analytical
methods, while the few challenges (e.g. the more the more) remain just
the well-known ones.

> I have some cool Spanish examples too, but they would take some
> explaining.
>
> > I observe also:
> >
> >  - He's (as) stubborn as the Rock of Gibraltar.
> >    This is (as) clear as daylight.
>
> Good point. I'll have to keep it in mind.
>
> >  - He's *(as) stubborn as she is. [comparative AS]
>
> This is fine for me, so I'm not sure I see what you're getting
> at here. (This is probably due to my ignorance.)

I mean:

     He's as stubborn as she is.
    *He's stubborn as she is.

> > I don't see here much of a challenge to compositional semantics.
> > It's not easy for me to say exactly what's going on, because I
> > have no idea about how the syntax of comparatives works, so have
> > no basis to work from. But impressionistically it seems that
> > the as+NP construction means something like "like" == "he is
> > stubborn in such a way that he is like the R of G" or, better,
> > "he has a stubbornness that is like the R of G" (which is not
> > odd, because likening stubbornness to a big rock is not a
> > weird simile), while the AS + gap-containing clause means "He is
> > as stubborn as the Rock of Gibraltar is stubborn", which is odd
> > because the R of G is not stubborn. To summarize, then, the
> > simile-like semantics of one construction but not the other
> > makes it pragmatically more compatible with more metaphorical
> > meanings. Nowhere here do I perceive noncompositionality.
>
> Okay (I think). BTW, my view of compositionality is that it means
> that the whole is the mere sum of the individual parts. That is
> your view too, right?

Yes. One has to remember that of course adding extra elements of
meaning *restricts* the meaning of the whole.

> But you don't seem to accept that compositionality is a matter of
> degree.

That's right. I don't.

But you must remember that the individual elements of meaning come
not just from lexical entries but also from constructional meaning.

> It seems to me that that's where we disagree.
>
> What I would like to say is that the meaning of the individual
> parts *motivates* the meaning of the whole: zero motivation =
> utter noncompositionality and full motivation = completely
> compositional. So, what the individual words of an expression
> (plus their dependencies, of course) do is provide cues for
> interpretion.

Given that I accept that within noncompositional structures there
is a scale of translucency, there is no more than a terminological
difference between us: what I'd call noncompositional and n%
translucent, you'd call n% (non)compositional.

However, I'd insist that there is an important distinction between
100% compositional, where the structures are simply built up bit
by bit on the basis of lexical and constructional meanings, which
follow automatically from syntactic structure, and less than 100%
compositional, which requires some sort of idiomaticon which
lists complex syntactic structures and states their meanings
which involve discarding/overriding the meaning of some of their
parts. The idiomaticon shades into our 'usage bank', our memory
of quotations and common expressions, and hence the grammar/usage
boundary is blurry here. The real boundary is between Syntax,
which is made up of word-sized elements that combine together
compositionally, and Usage, the patterns of which can be
conventionalized to a greater or lesser degree and translucent
to a greater or lesser degree. (The more conventionalized and
less translucent patterns, we call 'idioms'.)

A relatively empirical prediction of this view is that idioms
or any other 'construction' with noncompositional meaning do
not have formal/syntactic properties that differ from ordinary
syntax. That is, there are idioms with ordinary syntax, and
probably idioms with no syntax, but no idioms with idiomatic
syntax.

To summarize the last two paragraphs, I am accepting that there
is no boundary between knowledge of usage and knowledge of
grammar, if 'grammar' means conventional sound--interpretation
correspondences. But whether or not Syntax (in the broad
sense covering all compositional properties of the sentence) is
(therefore) subsumed by knowledge of usage, it remains a
crisply definable autonomous system.

> To be sure, at the end of the day the semantic interpretation of
> a structure will be "compositional": the extra (or just "different")
> meaning will emerge with the help of semantic principles such as:
>
> - semantic-network knowledge (= frame-semantic knowledge),
> - organizing principles such as protoypes and basic-level concepts,
> - nonliteral processes such as metaphor metonymy,

The defining essence of noncompositionality is (as I see it) that
elements of compositional meaning are discarded. So I don't think
it's possible to always argue that at the end of the day the semantic
interpretation of a structure will be "compositional".

> - constraints having to do with conventional implicatures
>   associated with grammatical constructions

Do you necessarily mean "conventional implicatures"? I understand
these to be elements of encoded meaning that are outside the logical
scope of the illocutionary operator at the root of every sentence;
i.e. they're not part of truth-conditional meaning. Constructional
meaning seems to me to be no different from word meaning -- sometimes
conventional implicature and sometimes not (i.e. "conventional
explicature").

--And.