Re: S7 grammar in a nutshell (long)

Hi!

And Rosta <a.rosta@...> writes:
> > Comments appreciated!
>
> Case: How many cases?
First of all: thanks for your thorough reading and commenting!

I havn't completed the case system yet.  But I can give you some
hints:

   - There will be at least five basic spatial cases: locative (at),
     allative (to), ablative (from), perlative (through),
     spatiorative (around).  For the latter, I might adjust the
     name.  Any suggestions?

     Spatial cases will be optionally combinable with other categories
     proximity modifiers (e.g.  touching vs. on-touching), precision
     of location (exactly the position vs. vague direction), viewpoint
     ('in front' from my perspective vs. the perspective of the
     object), and, of course, the point of relation: in, at, over,
     under, on.

   - Some cases will be available in different shades and/or will
     be modifiable, e.g.
        dative as the general case, benefactive and malefactive
        as a case that has a conotation.

   - There will be at least causative, genitive (possibly more than
     one), instrumental, sociative/comitative, essive.

   - I have to think about a system similar to spatial cases for
     time.

> What is the max number of arguments a predicate can have?
Two: agent and patient.

Further, S7 will add adjuncts and use serial-verb constructions.

> You could, like Livagian, use alternate case frames as a kind of
> polysemy-generating device.
Is there a webpage describing what Livagian does?

I think S7 valence already works like that: it distinguishes
semantical and syntactical valence, both of which is marked.  When you
change the semantical valence, the semantics are probably changed.  I
havn't decided what will be possible, but at least, you can do
something like middle constructions with this by eliminating a
semantical agent.  Also, 'the colour red' and 'to be red' are surely
the same stem and only distinguished by the semantical valence (first:
valence 0, latter: valence 1: patient).  Maybe some inherently passive
words can add a causative by this means.  Then, you'd also regularly
get 'to make red' (valence 2: agent+patient).

Is this what you meant?

> The rest is all very much to my tastes, except:
Ah, thanks! :-)

> Head-first, VOS.
Well, *I* like it! :-)

> I prefer to have all orders possible. This allows information
> sequencing to be in the domain of pragmatics, and it allows light
> phrases to be ordered before heavy, thus making sentences more
> processable.
Yes.  S7 is only *primarily* VOS.  There are already valence markers
that allow for heavy-NP shift, which will definitely be kept in the
grammar for non-incorporated arguments (but probably removed for
incorporated ones).

As to right-dislocating the V-part / fronting A or P, I havn't yet
decided anything.

I'm aware of needs to shift: Tyl-Sjok suffers from this problem.  It
cannot easily shift anything, it is embedding (internally headed
relative clauses) and it has no morphology, so you get more levels in
the syntax tree.  Further, it is very ambiguous since you can drop
almost all function words.  I thought it was a simple structure, but
unfortunately, it is very hard to process.  I suspect I introduced
some confusion in relay five with this...

> Clitics. I'm not sure what functional virtue these have.
It's probably a misnomer.  I wanted a distinction between affixes,
which are non directly derived from stems, but a distinct, closed
lexical class, and stripped stems -- stems without their class prefix.
I called the latter 'clitics', because there cannot be used in
isolation, but since that is not the only criterion for clitics, the
name is probably misleading.

Do you have a good name for it?  'Stripped stem that cannot be used in
isolation' in one word.

> Why should evidence be mandatory?
Because I like it. :-)

I should have noted that the primary design goal is personal taste.
This includes violation of all other design goals...

> Case and valence are necessary for an unambiguous parse, but
> evidence isn't. I'd have thought it should be unspecified by
> default.
Yes, very right.  I just wanted a language that has evidence as its
only 'superfluous' mandatory category.  Maily for fun and experiments.
In fact, I tried to construct the language in such a way that the
evidence is hard to drop even if you try.  In S7, the phonetactics
must be violated to drop the evidence.  I wanted to prevent early loss
of mandatory evidence when a society starts to speak S7.  I did it on
purpose!

> I don't see much virtue in derivational morphology; it merely
> reduplicates syntax, so needlessly complicates the grammar.
That's certainly right.  Although I'd doubt that it is 'needless'
in all cases.

I wasn't content in some respects with Tyl-Sjok, where I decided
against having morphology at all.  However, I found it very hard to
get a language that is very underspecified and isolating and not
having morphology and where idiomatic vs. generic differences are
easily expressible:

Ger: 'Rotwein'  vs. 'roter Wein'
     RED wine       red WINE
     idiomatic      generic

Mand: 'hongputaojiu' vs. 'hong de putaojiu'
      RED wine            red WINE
      idiomatic           generic

Eng: 'hot-dog' vs. 'hot dot'

I wanted a related but yet different level of compositional abilities
that -- on purpose -- resemble parts of the syntax, but which can be
used in a more idiomatic way.  Morphology is still lightweight: S7
drops some markers (class and case) and then agglutinates.  This
allows for idiomatic expressions to combine 'closer' that with syntax.

This way, I have something powerful between stem (atomic) and
clause-level (syntax).

You will probably have to put some compounds into the lexicon, since
the morphological composition is more ambiguous than the syntactical
one, but that's ok, since it's for more idiomatic things anyways.

Does this convince you or would you suggest other means to do this?

> >     This is to achieve feasibility for both poetry and law.
>
> Fine, except that while law requires precision, poetry does
> not inherently require imprecision. However, *conversation*
> tends to require imprecision.
Yes, that's right.  It was not a listing of extremes, but a listing of
examples where it might be useful.

> >   - minimal grammar-only affix inventory, maximal degree of
> >     lexicon ~ grammar coincedence (e.g. tenses are also
> >     free-standing words as are aspects etc.)
>
> Why have affixes at all?
For the same reason as with resembled syntax in morphology: I was not
totally convinced by the totally analytic structure of Tyl-Sjok.  Or
at least wanted something different.  I wanted a clear affix that is
used for 'past tense' (for example), that is *not* merely the verb 'to
happen in the past', but which, although related to the lexicon, has
new function of precifying tense.

> My preferences are either (i) all affixes are functional, and there
> are no function words, or (ii) there are no affixes.
Yes, I understand that.  Tyl-Sjok went way (ii), Fukhian went (i).

> Why are degree infixes not open class words?
Hmm.  I don't really know.  I have to think about that.

> >        a) You want a predicate: add a valence vowel
> >        b) You don't: add a case vowel
>
> It sounds like they're all predicates, but not all predications;
> that is, valence vowel = predication = has truthvalue; case
> vowel = argument of a predicate (expressed by the stem) = lacks
> truthvalue.
Errm, yes.  I understand I used wrong terminology?

The truth value difference is indicted in the following way, too: if
you add a valence infix, you *must* add an evidence affix.  If you
don't, you *cannot*.

**Henrik

Reply