Re: Indicating verbs valence? (Was: The disappeared conlang)
|From:||Stephen Mulraney <ataltanie@...>|
|Date:||Friday, January 25, 2002, 1:56|
On Thu, 24 Jan 2002 10:53:53 -0500
Jeff Jones <jeffsjones@...> wrote:
> I already have verb prefixes to indicate which argument position maps to
> which core case role. I call this grammatical voice, although it only
> shuffles arguments without deleting any. Currently, the 2nd object can be
> omitted with no other indication. For a different language this would be a
> good solution.
Interesting. Maybe in this kind of 'stuck' situation you should ask yourself -
"what would a natlang do?". Natlangs are absurdly resilient to misdesign
and damage. I imagine that none has ever gotten stuck ;). In other words,
from observing the way natlangs act, one can surmise that there exists
a possible, and relatively simple modification to your lang that will
introduce the missing functionality AND retain 'backwards compatability'.
I suppose that natlangs in such situations usually resort to round-about
'lexical' solutions. For example, when the loss of the Latin future tense
was felt in Romance (or whatever ancestor of modern romance langs it was)
something like it - a 'future of obligation' was constructed using the verb
'to have' (like the English idiom 'to have to do something'), so e.g.
(in proto-French) 'laver ai' "I have to wash" > Mod Fr 'je laverai' "I'll wash".
>> If this sounds like the method of semitic languages, you're rightish.
>> PL had a triconsonental root for each word, with different vowel values
>> and affixes to form different kinds of words, including different valences
>> of verbs. E.g. the root S`-g-(r) [(r)=voiceless 'r', S`=retroflex s] is for
>> "cut" and we get "caer" < S`Ego(r), while "coirin" < S`egu(r)eX\i.
>> In ML the 3-root system is no longer productive, and the ML forms are the
>> reflexes (descendants) of these two PL forms... in theory. All I need now
>> is some convincing sound-laws connecting them... In principle, you should
>> not need to know PL grammar to know ML grammar, since by analogy there
>> should be only a few conjugations of verbs, and a few simple rules for
>> changing intransitive->transitive etc.
> Coming up with sound-laws is especially tricky when _both_ endpoints have
> already been determined.
Yes ;). In practice this won't be a problem - the current method is
just to 'bootstrap' the creation of enough words for me to invent a
grammar (it's especially necessary to see what the words look like since
there are lots of phonetical tricks used to implement grammar). I could
just pick the words out of the air, but then knowing my imagination,
they would all be CVCCV and have 'eth' in them. So I can get some
more varied form in this way, and the restricted number of shapes
in the Primitive Lang will give rise to some inflectional classes -
conjugations, declensions etc.
One the grammar has been stabilised, all etymologies and primitive roots
will be thrown out, and I'll be in a position where I can employ the
simplest sound laws that suit the data. Chomskian recontruction of the
rule-system, I suppose (and of the underlying forms - is that allowed,
or does Noam say No?). [Damn Chomskians - what couldn't they be just
>> Oh, I meandered... how unusual ;)
Anyway, I'll post a sample of this lang soon.
> I wanted to first review the possibilities, before seeing how they might
> apply. But most solutions won't work without a complete redesign).
See above. ;)
It seems to work. Looks very interesting - I'll give it a perusal.