free word-order conlangs
|From:||Eldin Raigmore <eldin_raigmore@...>|
|Date:||Monday, July 17, 2006, 17:14|
On Sun, 16 Jul 2006 19:41:33 +0100, And Rosta <and.rosta@...> wrote:
>The discussion of Warlpiri prompts me to solicit information about
>conlangs in which word-order is in some sense very free but without
>ambiguity resulting from the freedom.
This is about Adpihi and its descendant Reptigan.
I have just begun committing Adpihi to the web, and have not yet begun to
commit Reptigan to the web, so these answers are provisional.
>1. How free is free?
>Is freedom limited to within some subsentential domain such as the clause?
Well, yes and no.
In Adpihi (and probably also in Reptigan),
Words may be re-ordered within phrases; phrases may be re-ordered within
clauses; and clauses may be re-ordered within sentences.
But complete and total "scrambling" is much dis-preferred.
I'm not sure yet I won't allow words to "leave their home phrases" and move
around the clause.
I can't even imagine why I would want to allow phrases to "leave their home
clauses" and move around the sentence. Does that occur in any natlang? If
so, does anyone on-list have an example?
If it occurs in a conlang, can anyone give me/us an example?
>Within the domain of freedom are all orders permissible, or just very
Probably "most". Maybe all.
Here are the "rules" (actually, I suppose, "violable constraints");
1. If the following rules don't change it, the language is thorough-goingly
2. In keeping with that, the pragmatically neutral "word"-order is Verb-
3. The "presentative movement" is highly influential; the part of the
utterance which will be most needed in the subsequent discourse moves as
close to the end of the utterance as it's allowed to do.
4. The opposite idea to the "presentative movement" is also highly
influential (though less so); the part of the utterance which was most
discussed in the previous discourse moves as close to the beginning of the
utterance as it's allowed to do, without screwing up rule 3.
5. Other things being equal "given" usually comes first and "new" usually
6. Other things being equal "topic" usually comes first and "comment"
usually comes last.
7. Other things being equal if "topic" doesn't come first then "topic"
usually comes last.
8. Other things being equal "focus" usually comes last and "ground" usually
9. Other things being equal if "focus" doesn't come last then "focus"
usually comes first.
10. Other things being equal "agent" usually comes first, non-agent
participants usually being not-first (unless there _is_ no agent).
>2. What mechanism allows the freedom (without ambiguity)?
Haven't quite decided.
I'm leaning towards this, and also towards case-marking.
I expect the verb to cross-reference its participants -- or, at least, its
core terms -- by pronouns cliticized to the verb.
I expect to require the verb to agree with _all_ human participants,
regardless of pragmatic status; with _all_ definite participants,
regardless of gender; and with _all_ specific, animate participants.
I expect the clitic pronouns used for these cross-referencing purposes will
include case-marks or some other way tell which case-role the participant
to which they refer has in the clause.
That's the way I'll allow the noun-phrases to move around within the clause.
As for how I'll allow the words to move around within the noun-phrase, or,
even, to break off from the noun-phrase and still be joinable to it; I
haven't even really decided to do that at all, yet.
I imagine I'll do it by requiring adjectives to concord with the person,
number, gender, case, and pragmatic status of their head noun; requiring
nouns to concord with the degree-of-comparison of their dependent
adjectives; requiring adpositions to concord with the
person/number/gender/etc. of their dependent nouns; and, obviously,
requiring nouns to agree with the "case" of their head adpositions.
In any case I see no reason why Adpihi shouldn't be both head-marking and
>Or something else?
>3. Is the freedom structural or just 'informational'?
>By 'structural freedom' I mean that linear precedence is of little
>importance to syntax.
I expect then Adpihi will have a lot of what you call "structural
freedom". The order of the parts of a word within a word may be very
important; that may take the place of the importance of the order of words
within a sentence.
Adjacency, however, I expect to remain quite important. That is, if a
phrase contains the words A and B, I expect the syntax will make little
but I expect the syntax may "care" a _lot_ about the difference between
those two and either of;
"A N O P Q R S T U V W X Y Z B"
"B N O P Q R S T U V W X Y Z A".
However it may allow the latter to still be grammatical; or may not.
>By 'informational freedom', I mean that even if syntax is highly sensitive
>to linear precedence, the grammar nevertheless has resources such that for
>any combination of a meaning and an order of content words, some syntactic
>structure is available to express that combination. (An example
>of 'informational freedom' would be "The farmer killed the duckling"
>vs "The duckling was killed by the farmer", allowing both F-K-D and D-K-F
>orders, but with structural changes.)
I expect Adpihi to have a lot of "informational freedom", then. I haven't
gotten far enough to say that it _does_; but I hope it will.
I imagine it will have more "informational freedom" than "structural
freedom", though I expect it to have a lot of "structural freedom" as well.
>The Latin & Warlpiri natlang examples of freedom strike me as
>comparatively uninteresting, because they can be analysed in terms of flat
>clause structures without internal ordering -- nothing that looks like
>thoroughgoing scrambling. But conlangs very possibly have more of interest
>to offer here...
As I say I'm not sure Adpihi will have thoroughgoing scrambling. So far I
haven't planned that in, at least, not at all levels.
So far it has rather complete "informational freedom" of "order of
meaningful constituents" of clauses; the kind of "word-order" Greenberg
meant. The immediate constituents of a clause can be re-ordered within a
clause in nearly any order -- or, at least, any element can be made first
or last -- and the different orders will have different pragmatics (and, to
the degree that pragmatics influences semantics, may therefore have
somewhat different meanings).
So far this will be accomplished by means of "cross-reference"-style
agreement marking on the verb.
>(To start the ball rolling: my Livagian has
>no structural freedom
>but lots of informational freedom, using a
>mechanism other than rampant concord,
Tell me/us about this mechanism.
>and no limitation to certain subsentential domains.)
Can you explain this in a little more detail?
(If you were to ask me to explain for Adpihi, I'd have to say I can't yet;
so I'll understand if you can't go into more detail yet.)