On Thu, 22 Nov 2007 09:16:27 -0800, Gary Shannon <[log in to unmask]> wrote:

>--- Geoff Horswood <[log in to unmask]> wrote:
>> --- Gary Shannon <[log in to unmask]> wrote:
>> <snip>
>> > Perhaps instead of each word adding more meaning to
>> > the sentence, each wourd
>> > could subtract meaning from the sentence. Use words
>> > so broad as to be
>> > essentially meaningless, but when combined with
>> > other words take the meaning(s)
>> > _shared_ by the two words.
>> >
>> > For example, "glundi" means "large, blue, glowing,
>> > animal", and "finja" means
>> > "red, glowing, automobile, animal", and "wagnus"
>> > means "tree, horse, broken,
>> > cloud". So it's clear that "glundi finja wagnus" can
>> > only mean "horse".
>>  Unless it means "large broken automobile" or "glowing
>> cloud", both of which I can think of a context for.
>> Geoff
>Perhaps I didn't explain it very well. The meaning is the _intersection_ of the
>definitions. A phrase means ONLY what ALL the words have in COMMON:
>"glundi" and "finja" have only "glowing" and "animal" in common so all other
>meanings are disarded.
>Then (glundi+finja) and "wagnus" have "animal" vs "horse" as their only point
>of intersection. All other meanings are not possible.

Well, you can _almost_ say that each word  w  has a set  M(w)  of meanings,
and the meaning of a phrase  w_1 ... w_n  is an (the unique?) element of the
intersection of the sets  M(w_1), ..., M(w_n).

But you can't quite say that: 'animal' and 'horse' aren't the same meaning.
 You perhaps want to say instead that the meaning of the phrase  w_1 ... w_n
 is an element  m  of one of the sets  M(w_1), ..., M(w_n),  such that all
of these sets contain a ... hyperseme? of  m.  By "hyperseme of" I mean
'meaning the same as or broader than': a word  w  in a vanilla language is a
hypernym of  x  just if the meaning of  w  is a hyperseme of the meaning of  x. 

The semantics I've just laid down above I find somewhat inelegant, in that
it's essential that you say that  m  is an element of one of the  w_i. 
Omitting this condition, you would get exactly the behaviour Geoff sketches.
 'Large broken automobile' is a meaning, just as much as 'horse' is.  And,
unless you make some dodge about lexical categories ("nouns and adjectives
don't mix") that feels out of place in a semantics, 'large' and 'automobile'
and 'broken' are all hypersemes of 'large broken automobile'.  So, in my
conception, the thing ruling out 'large broken automobile' as the meaning of
_glundi finja wagnus_ is its not occurring as the meaning of any one of the
words alone.  

There's another way out, I suppose, which is to deny that 'large broken
automobile' is a meaning.  But that's only tenable, I think, if you can
define what an allowed meaning is, and you'll likely have to resort to
either listing them all or stealing them from some other language.  
Can you actually state, in a clear-cut and well-defined fashion, what
property the meaning 'automobile' has that 'broken automobile' doesn't? 
"Well-defined" means in particular you can't resort to "well, the former is
just a noun and the latter is an adjective plus noun"; these aren't
properties of the meanings themselves but of their glosses, and glosses
aren't unique.  You can say "the former _has_ a one-word English gloss and
the latter doesn't", but then that's the steal-from-English solution.