Re: ANNOUNCE: My new conlang S11

On Tue, Mar 01, 2005 at 11:10:08PM +0100, Henrik Theiling wrote:
[...]
> Motivation:
>    The main problem I was having with all conlangs so far
>    was the question of assignment of arguments to predicates:
>
>       - Which roles are direct arguments to verbs?
>         I.e., how to handle verbs that naturally have
>         three arguments like the prototype 'to give'?
>         Do we want two or more core cases (a core dative
>         case?)
Interesting. This is what motivated me to devise Ebisédian's (and
Tatari Faran's) case system (along with my general dissatisfaction
with the passive voice). During my early attempts to solve this
problem I decided that I needed at least 3 core cases, so that
trivalent verbs like 'to give' can be expressed without adjuncts.


[...]
>       - Which adjunct cases (or adpositions or whatever) do
>         we need?
>         How do be handle adjuncts/predicate modification
>         in general?
The Ebisédian case system tries to handle this by 2 additional core
cases, on top of the 3 prototypical for 'to give'. In retrospect, this
did not work very well, which is why Tatari Faran reverts to 3 core
cases, and uses postpositions for adjuncts.


[...]
>    I found the questions very hard an unsatisfactory to solve each
>    time and was searching for some grammar structure that ultimately
>    solved these questions without arbitrary borderlines e.g. between
>    arguments and adjuncts.
This is interesting. When I was working on Tatari Faran grammar and
decided that I needed postpositions, the first difficulty I had was
with how one would decide which core case would be governed by which
postposition. Some postpositions had easy mappings: 'out of' ->
originative, 'into' -> receptive, etc.. However, there were also many
postpositions which did not easily fit into this paradigm.

Eventually, I came upon a rather elegant solution (IMHO): since the 3
core cases were marked by postclitics, which were already treated as
separate words, why not open up the class and treat postpositions the
same way as well? And so, I decided that Tatari Faran postpositions
govern the unmarked NP (without case marker). I.e., they appeared in
the same position as where the case markers would appear, and
essentially behaved like the case markers. I even went so far as to
equate postpositional clauses with NP's at the syntactic level, in
that the indicative word order was subject-verb-arguments, and if one
fronts a postpositional clause, the order becomes
PC-verb-subject-arguments. I.e., the PC has the same status as a
'normal' NP.

Now of course, there is still a distinction between core case and
postpositional 'case'; so this isn't quite at your ideal yet. But I
quite like Tatari Faran's current system.


>    My current project S11 is my current attempt to get closer to a
>    solution.
>
> Grammar:
>
>    - two open word classes: nouns and verbs
>
>    - verbs have exactly (thus maximally(!)) *one* argument.
>      E.g. there are no transitive verbs.
>
>    - there are no adjuncts either, the whole structure is controlled
>      by using a sequence of noun-verb pairs.
Nice! This sounds almost like a fleeting idea I posted to the list
once, while thinking about Ebisédian grammar. It was sortof the
antithesis of my approach to the problem of adjunct/argument marking
then: I had been thinking along the lines of how to mark NP's so that
their relationship with the verb could be expressed; but it occurred
to me that perhaps the problem was with the fact that one needed
multiple arguments to a verb. What if, instead of a single verb to
express 'to give', one built the sentence from three verbs, 'to
offer', 'to transfer', and 'to receive'? Then "A gives B to C" could
be expressed as "A-offer B-transfer C-receive", which looks strikingly
similar to what you describe below. :-)

Pity I didn't pursue this idea any further after mistakenly trying to
shoehorn it into the existing Ebisédian system instead of *replacing*
the existing system.


[...]
>    For sentences that involve transitive concepts in other langs,
>    there are simply two (or more) verbs for each role.  I think they
>    will all be suppletives, and not derivationally related (or maybe
>    related by some irregular, chaotic, non-productive process in some
>    cases).
Very nice. I like the idea!


[...]
>    The word order is
>
>        noun verb noun verb ... noun verb.
Nice.


[...]
> Example:
>    There is no phonology yet, so the words are variables:
>
>        'LU'   verb: to ask something
>        'NI'   verb: for something to be asked
>        'MAT'  verb: for someone to be addressed
>        'JIT'  evidence: hearsay
>        'KHAN' noun: question
>
>    'John asks Mary a question.' =
>
>        John JIT LU   Mary MAT  KHAN NI.
>        noun ev. verb noun verb noun verb
>
> There is no need for voices, since all the verbs can be used on their
> own:
>
>    'John asks something.'     = John JIT LU.
>    'Mary is addressed.'       = Mary JIT MAT.
>    'A question is posed.'     = KHAN JIT NI.
Awesome! I think your solution is in many ways more elegant than mine.
:-)


[...]
> Modification: relative clause:
>    Use a normal sentence and suffix the relative particle
>    (let it be called 'GUP') to get a complex noun.  The
>    referent is in topic position in the sub-phrase:
>
>      'John, who asks a question, is addressed.' =
>      John JIT     LU   KHAN     NI    GUP  JIT     MAT.
>      noun ev.     verb noun     verb  rel. ev.     verb
>      John hearsay ask  question posed who  hearsay addressed
Nice.


[...]
> The whole thing will be polysynthetic again, I think, but that's more
> of an accident: I started with an isolating lang, like above, then
> made all the particles clitics that attach to the previous word.  This
> would yield the following words:
>
>      [John-JIT] [LU] [KAHN] [NI-GUP-JIT] [MAT].
>
> And then I decided to let the verbs also cliticise with the previous
> word.  I made some particles (e.g. the relative particle) free words
> again, to prevent multiple verbs per word.  So in the above sentence,
> you'd get the following words:
>
>      [John-JIT-LU] [question-NI] [GUP-JIT-MAT].
I like this. You could also have each component in the words blend
with each other through lenition/mutation/etc.. :-)


[...]
> So these are the basic ideas.  I'll post more later.
>
> What do you think?  Comments?
I really like the monovalent-only verb system. I think it's a much
better solution to the problem of argument/adjunct marking than the
Ebisédian / Tatari Faran case systems that I've come up with so far.
Makes me wish I'd pursued the similar idea I had back then instead of
tolerating the Ebisédian system for so long. :-)


T

--
Debugging is twice as hard as writing the code in the first place. Therefore,
if you write the code as cleverly as possible, you are, by definition, not
smart enough to debug it. -- Brian W. Kernighan

Reply