"Make these work together" game.

A forum for all topics related to constructed languages
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Fri 26 Jul 2013, 16:39

Click wrote:Ok, here's the challenge.
  1. The language is mostly head-final, especially with regards to adverbial and noun phrases.
  2. The language orders words in a VSO word-order.
  3. There is a quite plausible diachronic explanation how the language came to have these two rather conflicting features.

The Most Plausible Universal Diachronic Explanation Ever: nothing essential has changed WRT the features in question in the last 5 KY.

(And I was almost serious writing that.)

However, the above does not appear to be what Click expects. Let's go further.

I have several problems with understanding the challenge.
The language is mostly head-final, especially with regards to adverbial and noun phrases.
*Mostly* is the difficult bit. If these constructions are only statistically head-final in the descendant, and (suppose) they used to be only statistically head-initial in the ancestor, then we have just a shift of markedness in the set of constructions which were already available in the ancestor. Generally, such changes seem to be diachronically cheap; finding a plausible motivation for the shift may be more interesting, but I don't think it makes for a real challenge.

Perhaps, this condition must be strengthened somehow: either the ancestor's NP's were *strictly* head-initial, or direction of branching has not changed in NP's.

Also, WRT "adverbial phrases": do these include adpositional phrases as well?
There is a quite plausible diachronic explanation how the language came to have these two rather conflicting features.
Here, the difficulty is about "conflicting".

It is true that V1 + head-final NP's seems to be an exceptionally rare combination. But that's not enough to call it problematic, since the sample can be severely biased.

One can notice that the symmetrical situation, V-last plus head-initial NP's, is not particularly rare. Moreover, it seems to have been produced by two opposite types of diachronic scenarios: mostly head-final language developing head-initial NP's (Farsi) and mostly head-initial language becoming Verb-last (part of Ethiosemitic langs). (In both cases, the ancestral WO was in fact much freer, but that's irrelevant for assessing the quirkiness of the result.) Therefore, if V1 + otherwise head-final is indeed a problematic configuration, the problem is about those aspects where right-branching and left-branching are not truly symmetrical options. That is, it's probably not about parsing difficulties with deep embedding and the like.

Frankly, I know close to nothing about such asymmetries.

So, Click: can you elaborate on what type of conflict(s) you meant, of either synchronic or diachronic nature?

(Without such elaboration, I can propose a lazy solution which, I suspect, will look rather boring...)

And BTW, I don't think Creyeditor's scenario is plausible.

Agreement markers reinterpreted as free pronouns?

NP placement changed on the analogy of pronoun placement?

In both cases, the opposite direction of change would look more natural, and not without a reason.
User avatar
Click
darkness
darkness
Posts: 3328
Joined: Sat 21 Jan 2012, 12:17

Re: "Make these work together" game.

Post by Click » Fri 26 Jul 2013, 18:27

[O.O] I don't know what to write now.
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Fri 26 Jul 2013, 19:22

Comment. Share your thoughts. Speak out :)

Also, answer the questions.

I was just thinking aloud there, not what you thought
User avatar
eldin raigmore
fire
fire
Posts: 5684
Joined: Sat 14 Aug 2010, 18:38
Location: SouthEast Michigan

Re: "Make these work together" game.

Post by eldin raigmore » Fri 26 Jul 2013, 21:52

Click wrote: [O.O] I don't know what to write now.
Lucien Tesnière, the same guy that first defined "valency" IIANM, was the first guy IIANM to classify languages according to head-directionality.
Basically he proposed five classes; namely:
  1. very head-initial
  2. very head-final
  3. "ameliorated" head-inital
  4. "ameliorated" head-final
  5. none of the above
I think Click was intending to say that the current language fits in class d (if not in class b).

@Click: Do you think that is in fact what you meant?

@Click, and also maybe @basilius:
Can you look up Tesnière's definitions, or some later linguists' modifications of them, to find out what the difference is between very-head-final, sort-of-head-final, and not-really-head-directional? If you can do so, I recommend doing so.

It has been proposed that most languages mostly don't like to insert stuff between the Verb and the Object (or Patient?).

In keeping with that, SVO and VOS languages, this proposal says, mostly have adverbs before verbs ("head-final" "verb phrases", for a given sense of "head" and a given sense of "verb phrase", not necessarily the ones usually meant); and mostly have adjectives etc. after the noun ("head-initial" noun-phrases, again for a certain sense of "head").

OTOH this proposal says that SOV and OVS languages mostly have adverbs after verbs ("head-initial" "verb phrases", for a given sense of "head" and a given sense of "verb phrase", not necessarily the ones usually meant); and mostly have adjectives etc. beforer the noun ("head-final" noun-phrases, again for a certain sense of "head").

It has also been proposed that SVO and VOS languages mostly have prefixing verb-morphology and suffixing noun-morphology, while SOV and OVS languages mostly have suffixing verb-morphology and prefixing noun-morphology.

Off the top of my head I can't right now recall which parts of those proposals are in fact statistically "true", and which are not even statistically "true", at least not to a significant degree.

Note that the sense of "head" used above is not one that could apply to every phrase; it can only apply to phrases of the forms
X --> X + m
or
X --> m + X
for some part-of-speech X. That is, phrases consisting of a modifier (or "dependent") and a "head" that's the same part-of-speech as the entire phrase.

In my own conlanging I find more useful the definition of "head" as follows.
An immediate constituent of a phrase is the "distributional head" of the phrase if it determines the distributional word-class (aka "part-of-speech") of both the other constituent(s) and of the entire phrase.

That would mean that when we have phrases like
NP --> AdjP + NP
NP --> NP + AdjP
VP --> AdvP + VP
VP --> VP + AdvP
it's the AdjP that's the "distributional head" of the NP, and it's the AdvP that's the "distributional head" of the VP.

However, the usual definition of "head-word" of a phrase is, the word (or, rather, constituent, since it's possible the constituent is a smaller phrase), that has the same part-of-speech as the entire phrase, and that has a meaning closest to the meaning of the entire phrase. The head-word of the phrase could be substituted for the entire phrase and that would still make a grammatical clause; and it's meaning would be related to the meaning of the clause with the whole phrase instead of just the head-word.
Or, if that's not the more usual meaning, at least (I think) it's the one usually meant when we talk about "head-initial vs head-final" and "head-marking vs dependent-marking".

You would need to look up Tesnière's definitions, or some later linguists' modifications of them, to find out what the difference is between very-head-final, sort-of-head-final, and not-really-head-directional.

WALS.info doesn't discuss head-directionality, nor branching-direction (they may be related but they are certainly not synonymous).
OTOH it does discuss "word"-order.
"Word"-order is not synonymous with head-directionality; nor with branching-direction.

See
http://www-01.sil.org/linguistics/Gloss ... sAHead.htm
and
http://www-01.sil.org/linguistics/Gloss ... Phrase.htm
and
http://www-01.sil.org/linguistics/Gloss ... adword.htm.
But think about them, don't just adopt them wholesale without comparing with some other source.

Unrelatedly you could look up http://www-01.sil.org/linguistics/Gloss ... alency.htm.
Not only verbs have valency. For instance some adpositions, such as "between", have valency higher than 1. It might be argued that comparative adjectives and comparative adverbs (like "bigger" and "sooner") have valency 2. But AFAIK only verbs have valency higher than 2.
User avatar
Linguifex
roman
roman
Posts: 920
Joined: Fri 03 Aug 2012, 07:07
Location: Ohio

Re: "Make these work together" game.

Post by Linguifex » Sat 27 Jul 2013, 02:54

eldin raigmore wrote:Unrelatedly you could look up http://www-01.sil.org/linguistics/Gloss ... alency.htm.
Not only verbs have valency. For instance some adpositions, such as "between", have valency higher than 1. It might be argued that comparative adjectives and comparative adverbs (like "bigger" and "sooner") have valency 2. But AFAIK only verbs have valency higher than 2.
I'd think you could get valency higher than 2 with adpositions ("It was among the rocks, the trees, and the soil"; "I had to choose between the gun, the aspirin, and the hammer").
(Avatar via Happy Wheels Wiki)
Index Diachronica PDF v.10.0
Conworld megathread

AVDIO · VIDEO · DISCO
User avatar
Creyeditor
mongolian
mongolian
Posts: 3948
Joined: Tue 14 Aug 2012, 18:32

Re: "Make these work together" game.

Post by Creyeditor » Sat 27 Jul 2013, 19:42

eldin raigmore wrote:@Creyeditor, I like what you posted in response to Click's make-these-work-together challenge; in fact I like it a lot.
But am I missing something?
I don't see how (or where, or whether, whatever the right word is) you've shown that it satisfies all three requirements.
In particular, how does it satisfy the requirement:
"1.The language is mostly head-final, especially with regards to adverbial and noun phrases."
?

In did not mention that explicitly, but in my first sentence I called the first version of the language entirely head-final. What I wanted to say was, that it had SOV word order and noun phrases followed their adjectives, nouns followed their genitives, nouns followed their numerals, etc.

eldin raigmore wrote: I find it interesting that your diachronic process, though Step 1 starts with APV ("SOV") "word"-order, seems to pass through a PVA (aka "OVS") word-order in step 4 in case the "Patient" is a full noun-phrase.
Wijay tay vi.
wijay tay vi
sheriff see 1.NOM
I see the sheriff.


Kivi tay wijay.
kivi tay wijay
towel see sheriff
The sheriff sees the towel.


(Step 4 appears to be VAP ("VSO") if both participants are pronouns):
Tay vi hi.
tay vi hi
see 1.SG.NOM 3.SG.ACC
I see him.

Yep, I know that some of the intermediate word orders are rather rare. This is one of the shortcomings of my solution.
Creyeditor
"Thoughts are free."
Produce, Analyze, Manipulate
1 :deu: 2 :eng: 3 :fra: 4 :esp: 4 :ind:
:con: Ook & Omlűt & Nautli languages & Sperenjas
[<3] Papuan languages, Morphophonology, Lexical Semantics [<3]
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Wed 31 Jul 2013, 18:56

eldin raigmore wrote:Lucien Tesnière, the same guy that first defined "valency" IIANM, was the first guy IIANM to classify languages according to head-directionality.
Basically he proposed five classes; namely:
  1. very head-initial
  2. very head-final
  3. "ameliorated" head-inital
  4. "ameliorated" head-final
  5. none of the above
I think Click was intending to say that the current language fits in class d (if not in class b).
<...>
@Click, and also maybe @basilius:
Can you look up Tesnière's definitions, or some later linguists' modifications of them, to find out what the difference is between very-head-final, sort-of-head-final, and not-really-head-directional? If you can do so, I recommend doing so.
No, I don't have anything handy. Anyway... from the little I know about WO variations, I suspect "none of the above" will collect most instances of neither strictly-head-final nor strictly-head-initial orderings.
It has been proposed that most languages mostly don't like to insert stuff between the Verb and the Object (or Patient?).
English being one of the prototypical languages illustrating this constraint...

There may be some dissent on which object is synchronically "more direct" in The boy gives his father a book, but diachronically there is none: his father used to be an indirect object in dative, and a book was the direct object in accusative.

Another type of English constructions that illustrates former ordering constraints is subject inversion (questions, conditionals, no sooner..., etc.): nothing can be put between the verb and the inverted *subject*, and in particular not a DO. Mandatory use of an auxiliary (Does the boy give his father a book?) obscures the mechanics in today's English, but e. g. Shakespeare's texts offer a lot of examples showing the original constraint ranking. More conservative Germanic languages, e. g. Danish, illustrate this point very well, too:

I går slog drengen ikke manden.
'Yesterday the boy didn't beat the man.'

I går gav drengen ikke manden et æble.
'Yesterday the boy didn't give an apple to the man.'

(Underlined is material that *must* be placed between the verb and its DO, if neither is placed before the verb; ikke 'not' fills the slot that can be also occupied by some other "sentential adverbs". The examples have been constructed by myself, and properly need an approval by a native speaker.)

In older Germanic languages, the ordering rules for full NP's were much freer, but there was a strong tendency to put any cliticized material in contact with the verb (thus, if the DO was not a clitic, it could be separated from the verb by a chain of enclitics). More interestingly, there was a tendency towards a fixed WO for the postverbal clitics, and the commonest ordering seems to be the following:

finite verb | subject pronoun | dative pronoun | accusative pronoun

(The above is a huge simplification: I don't want to even touch upon quirky subjects, quirky objects, antisubjects and the like; yet it seems that by default there were at least two slots between the verb and its DO, within the maximum clitic chain template.)

Speaking of clitics, in Romance languages there are examples like the following French ones:

Je l'y mets.
'I am putting it there.'

Je le lui donne.
'I give it/him to him/her/it.'
(This last ordering rule works only for 3rd person clitics in French.)

Also, French often allows for placing an adverb (not clitic) after the verb, thus before the DO, but normally not between the subject and the verb.

The above examples aren't about particularly exotic languages.

They show that there are categories of dependents that are quite regularly put closer to the verb than DO is; namely, subjects, "dative" objects, certain types of adverbs, and all types of clitics when the DO isn't a clitic.
It has also been proposed that SVO and VOS languages mostly have prefixing verb-morphology and suffixing noun-morphology, while SOV and OVS languages mostly have suffixing verb-morphology and prefixing noun-morphology.
I don't know about the rest (although some counterexamples do immediately come to my mind), but SOV languages certainly tend to be suffixing, including WRT noun morphology.
In my own conlanging I find more useful the definition of "head" as follows.
An immediate constituent of a phrase is the "distributional head" of the phrase if it determines the distributional word-class (aka "part-of-speech") of both the other constituent(s) and of the entire phrase.

That would mean that when we have phrases like
NP --> AdjP + NP
NP --> NP + AdjP
VP --> AdvP + VP
VP --> VP + AdvP
it's the AdjP that's the "distributional head" of the NP, and it's the AdvP that's the "distributional head" of the VP.
This is an interesting idea. But I think "head" isn't a very convenient word in this context, just because it mostly means something quite different.
WALS.info doesn't discuss head-directionality, nor branching-direction (they may be related but they are certainly not synonymous).
Wikipedia (first paragraphs here and here), on the other hand, says they're basically the same, and that's also how I've always have understood it.

However, I'd be curious to know if some theories try to draw a subtle distinction between the two.
Not only verbs have valency.
Sure. Obligatorily possessed nouns, too, can be viewed as having a strong valency for possessor.

Also, a word can have a valency not only for a dependent, but also for its head. Thus, in the following examples:

John gave Mary a book.

- and:

The book given by John to Mary.

- all the lexical valencies of give are also present with its participle given, only with one of them being filled by the participle's syntactic head (the book).

* * *

Thinking again about Click's challenge, I recalled something curious (and potentially relevant).

Sanskrit is mostly head-final, but the default position of *predicate nominal* is clause-initial (thus, the formula Tat tvam asi 'You are that', lit. 'That thou art', illustrates the default ordering rather than some marked emphatic option).

I don't think anyone has ever explained why.
User avatar
eldin raigmore
fire
fire
Posts: 5684
Joined: Sat 14 Aug 2010, 18:38
Location: SouthEast Michigan

Re: "Make these work together" game.

Post by eldin raigmore » Thu 08 Aug 2013, 16:47

basilius wrote: ....[fascinating, well-reasoned stuff clipped out]....
WALS.info doesn't discuss head-directionality, nor branching-direction (they may be related but they are certainly not synonymous).
Wikipedia (first paragraphs here and here), on the other hand, says they're basically the same, and that's also how I've always have understood it.
However, I'd be curious to know if some theories try to draw a subtle distinction between the two.

....[yet more well-reasoned fascinating stuff clipped out]....
Prepare yourself for a shocker [;)] :
Spoiler:
Wikipedia is sometimes wrong.
:roll:

"Branching direction" has nothing to do with heads.
As you are aware in context-free grammars there are many productions ("rules") of one of the forms:
X --> Y + Z
X --> Y + z
X --> y + Z
X --> y + z

where the uppercase letters represent "non-terminal symbols" (that is, things that can go on the Left-Hand Side (LHS) of a production rule with more than one symbol on the RHS, or with at least one non-terminal symbol included in the RHS) and the lowercase letters represent "terminal symbols" (symbols each of which can be replaced by at most one word, coming from some fixed finite list of words, depending on the symbol).

For a rule of one of the forms
X --> Y + z
or
X --> y + Z ,
the terminal symbol on the RHS is called the "non-branching" constituent. The non-terminal symbol is the "branchable" constituent.

The strongest form of "Language L is a left-branching language" is that any grammar of Language L has no production rules at all of the form
X --> y + Z .
In other words when any phrase's two immediate constituents consist of one which can branch and one which cannot, the one which can branch always comes first ("on the left").

The strongest form of "Language L is a right-branching language" is that any grammar of Language L has no production rules at all of the form
X --> Y + z .
In other words when any phrase's two immediate constituents consist of one which can branch and one which cannot, the one which can branch always comes last ("on the right").

Note that "left-branching" or "right-branching" say nothing at all about production-rules of the forms
X --> Y + Z
or
X --> y + z .
Possibly if both constituents of an X-phrase "branch", or neither does, then the language's grammar might allow them to come in either order.

One typical weaker form of "left-branching" requires that the branchable constituent come first only in case it actually does branch (that is, some production rule is applied to it resulting in its eventual replacement by a phrase of two or more words).
One typical weaker form of "right-branching" requires that the branchable constituent come last only in case it actually does branch (that is, some production rule is applied to it resulting in its eventual replacement by a phrase of two or more words).

Another much-weakened form of "left-branching" says that the language has no production rules of the form:
X --> y + X .
That is it applies only to rules where the branchable constituent of the RHS is "of the same kind or type" as the LHS.

Another much-weakened form of "right-branching" says that the language has no production rules of the form:
X --> X + z .
That is it applies only to rules where the branchable constituent of the RHS is "of the same kind or type" as the LHS.

Head-direction, OTOH, requires knowing which constituent of a phrase is its head.
It could talk about not only rules of the forms
X --> Y + z
or
X --> y + Z ,
but also about rules of the forms
X --> Y + Z
or
X --> y + z .
In a head-initial language's grammar all four forms of production rules might exist. We would just know that the Y or y was always the "head" of the X-phrase.
In a head-final language's grammar all four forms of production rules might exist. We would just know that the Z or z was always the "head" of the X-phrase.

The very weakest form of "Left-Branching", namely that "in any production rule (X --> X + z or X --> y + X) in which the LHS is a constituent of the RHS and the other constituent is a terminal symbol, then unless the X on the RHS happens not to branch in this particular case it must come first", might be considered equivalent to a special case of "head-initial", if we define "head" in such a way that whenever an LHS of a production-rule appears on the RHS of the same rule then the appearance on the RHS is the "head" of the phrase being produced. But that's a lot weaker than the actual definition of "head-initial".

The very weakest form of "Right-Branching", namely that "in any production rule (X --> X + z or X --> y + X) in which the LHS is a constituent of the RHS and the other constituent is a terminal symbol, then unless the X on the RHS happens not to branch in this particular case it must come last", might be considered equivalent to a special case of "head-final", if we define "head" in such a way that whenever an LHS of a production-rule appears on the RHS of the same rule then the appearance on the RHS is the "head" of the phrase being produced. But that's a lot weaker than the actual definition of "head-final".

We might relate all this to X' ("X-bar") theory.

In its original form X'-theory noted that context-free grammars of natural languages often had several sets of rules of the forms:
  • XP --> X' + XSpec or XP --> XSpec + X' or XP --> X' + xspec or XP --> xspec + X'
  • X' --> X' + XAdj or X' --> XAdj + X' or X' --> X' + xadj or X' --> xadj + X'
  • X' --> X + XComp or X' --> XComp + X or X' --> x + XComp or X' --> XComp + x or X' --> X + xcomp or X' --> xcomp + X or X' --> x + xcomp or X' --> xcomp + x

In all of these the "head" is the x or X or X'.

The XP and X' are always "branching"; or at least "branchable".
Ordinarily the X in the last row is non-branching, or at least it is in the examples usually given to explain the concepts (which may be cherry-picked for simplicity).
Ordinarily the XSpec, XAdj, and XComp can be branching.

For the grammar to be "head-initial" we'd need the rules to be like the following:
  • XP --> X' + XSpec or XP --> X' + xspec
  • X' --> X' + XAdj or X' --> X' + xadj
  • X' --> X + XComp or X' --> x + XComp or X' --> X + xcomp or X' --> x + xcomp
For the grammar to be "head-final" we'd need the rules to be like the following:
  • XP --> XSpec + X' or XP --> xspec + X'
  • X' --> XAdj + X' or X' --> xadj + X'
  • X' --> XComp + X or X' --> XComp + x or X' --> xcomp + X or X' --> xcomp + x
For the grammar to be "left-branching" we'd need the rules to be like the following:
  • XP --> X' + XSpec or XP --> XSpec + X' or XP --> X' + xspec
  • X' --> X' + XAdj or X' --> XAdj + X' or X' --> X' + xadj
  • X' --> X + XComp or X' --> XComp + X or X' --> XComp + x or X' --> X + xcomp or X' --> x + xcomp or X' --> xcomp + x
For the grammar to be "right-branching" we'd need the rules to be like the following:
  • XP --> X' + XSpec or XP --> XSpec + X' or XP --> xspec + X'
  • X' --> X' + XAdj or X' --> XAdj + X' or X' --> xadj + X'
  • X' --> X + XComp or X' --> XComp + X or X' --> x + XComp or X' --> xcomp + X or X' --> x + xcomp or X' --> xcomp + x
A grammar could be both head-initial and left-branching:
  • XP --> X' + XSpec or XP --> X' + xspec
  • X' --> X' + XAdj or X' --> X' + xadj
  • X' --> X + XComp or X' --> X + xcomp or X' --> x + xcomp
A grammar could be both head-initial and right-branching:
  • XP --> X' + XSpec
  • X' --> X' + XAdj
  • X' --> X + XComp or X' --> x + XComp or X' --> x + xcomp
A grammar could be both head-final and left-branching:
  • XP --> XSpec + X'
  • X' --> XAdj + X'
  • X' --> XComp + X or X' --> XComp + x or X' --> xcomp + x
And so on.
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Thu 08 Aug 2013, 20:25

Cool and very illuminating! I need some time to digest it :)
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Thu 22 Aug 2013, 15:23

I think I don't understand a couple bits.

For example.
eldin raigmore wrote:A grammar could be both head-initial and left-branching:
  • XP --> X' + XSpec or XP --> X' + xspec
  • X' --> X' + XAdj or X' --> X' + xadj
  • X' --> X + XComp or X' --> X + xcomp or X' --> x + xcomp
Why is this called specifically left-branching rather than both left- and right-branching (if it also allows for capitalized XAdj and XComp on the right side)?

Also. If we are speaking of not just abstract entities that happen to be called "grammars", but of grammars as formal models of human languages' structures, it seems important to me which "branch" (or branches) allows for unrestricted multiple embedding (for a branch which lacks this quality is kinda flat even if it's capitalized in X-bar notation). In this sense, it seems to me that only the position of XComp is relevant for determining the direction of branching. Is this sheer nonsense?

The main source of my questions is ignorance, sorry.
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Thu 22 Aug 2013, 17:04

Also...
basilius wrote:I går gav drengen ikke manden et æble.
'Yesterday the boy didn't give an apple to the man.'

(Underlined is material that *must* be placed between the verb and its DO, if neither is placed before the verb; ikke 'not' fills the slot that can be also occupied by some other "sentential adverbs". The examples have been constructed by myself, and properly need an approval by a native speaker.)
Upon reading a bit more on the topic, I discovered that my example above is probably problematic in colloquial Danish (because of the non-pronominal prepositionless "dative" object - colloquially, it would be construed with the preposition til, and the prepositional phrase put in the end); it may be still valid for a more conservative written language.

However, a rather literal match to my artificial Danish example exists in Swedish.

Curiously, there is some variation among Swedish speakers in that point. Namely, for Skógvur and Aszev the most neutral WO is where the negation (inte) follows the subject (like in my constructed Danish example):

Igår gav pojken inte mannen ett äpple.

For DrGeoffStandish, more neutral is negation placed immediately after the finite verb:

Igår gav inte pojken mannen ett äpple.

But in either case there are at least three slots between V (gav 'gave') and DO (ett äpple 'an apple'); it is interesting to watch how two elements fight for the position closest to V, while DO is just idly gazing at them from the far end of the clause...
User avatar
eldin raigmore
fire
fire
Posts: 5684
Joined: Sat 14 Aug 2010, 18:38
Location: SouthEast Michigan

Re: "Make these work together" game.

Post by eldin raigmore » Thu 22 Aug 2013, 19:32

Who will post the next challenge?

basilius wrote:
It has also been proposed that SVO and VOS languages mostly have prefixing verb-morphology and suffixing noun-morphology, while SOV and OVS languages mostly have suffixing verb-morphology and prefixing noun-morphology.
I don't know about the rest (although some counterexamples do immediately come to my mind), but SOV languages certainly tend to be suffixing, including WRT noun morphology.
Like I said, it's been proposed. I don't think it's true, though. See http://wals.info/feature/combined/26A/83A and http://wals.info/feature/combined/51A/83A and http://wals.info/feature/combined/69A/83A and http://wals.info/feature/combined/51A/69A.
Also see http://wals.info/feature/combined/33A/83A.
It looks like both nominal morphology and verbal morphology are predominantly suffixing regardless of whether the language is OV or VO, but the predominance is more pronounced in OV languages than in VO languages both for nouns and for verbs.
http://wals.info/feature/combined/33A/69A also contradicts the proposal; plural prefixes on nouns go with tense-aspect prefixes on verbs, and plural suffixes on nouns go with tense-aspect suffixes on verbs.

http://wals.info/feature/combined/57A/83A upholds the proposal, though.

See also http://wals.info/feature/combined/51A/57A and http://wals.info/feature/combined/57A/69A.
Also see http://wals.info/feature/combined/33A/51A.

basilius wrote:I think I don't understand a couple bits.
For example.
eldin raigmore wrote:A grammar could be both head-initial and left-branching:
  • XP --> X' + XSpec or XP --> X' + xspec
  • X' --> X' + XAdj or X' --> X' + xadj
  • X' --> X + XComp or X' --> X + xcomp or X' --> x + xcomp
Why is this called specifically left-branching rather than both left- and right-branching (if it also allows for capitalized XAdj and XComp on the right side)?

"Branching direction" applies only to phrases of whose immediate constituents just one is branching (or branchable) and at least one is non-branching.
XP --> X' + XSpec, X' --> X' + XAdj, and X' --> X + XComp, both have two branching (or branchable) constituents on the RHS, and no non-branchable constituents.
So they are not relevant to the "branching side" requirement.
If an RHS has no nonbranching constituents but has more than one branchable constituents, the "branching side" allows any order among the branching constituents.
(Similarly If an RHS has no branching constituents but has more than one nonbranching constituents, the "branching side" allows any order among the nonbranching constituents.)

basilius wrote:Also. If we are speaking of not just abstract entities that happen to be called "grammars", but of grammars as formal models of human languages' structures, it seems important to me which "branch" (or branches) allows for unrestricted multiple embedding (for a branch which lacks this quality is kinda flat even if it's capitalized in X-bar notation). In this sense, it seems to me that only the position of XComp is relevant for determining the direction of branching. Is this sheer nonsense?
No, not nonsense, not if you meant XAdj instead of XComp (and I'm sure you did).

X' --> X + XComp or X' --> X + xcomp or X' --> x + xcomp can only be applied once per X-Phrase (it's the first production-rule applied, if it's applied at all), so it can't lead to unboundedly deep multiple embeddings.

XP --> X' + XSpec or X' --> X' + xspec also can only be applied once per X-Phrase (it's the last production-rule applied, if it's applied at all), so it also can't lead to unboundedly deep multiple embeddings.

It's X' --> X' + XAdj or X' --> X' + xadj which can be applied unboundedly often per X-Phrase, so can lead to unboundedly deep multiple embeddings.

(Even in later revisions of X-bar theory, there's still only one XComp and only one XSpec but there can still be multiple XAdjs.)

"Left-branching" or "right-branching" has the effect of requiring any unlimited-depth multiple-embedding to "pile up" "only on the left" or "only on the right" respectively.
It still has that effect even if the language's grammar does not conform to X-bar theory.

But though it wasn't "sheer nonsense", it was incorrect.

None of X' --> X + XComp or X' --> x + xcomp or XP --> X' + XSpec or X' --> X' + XAdj are relevant for branching-directionality; for three of them both RHS constituents are branching, and for one of them both RHS constituents are non-branching.
Branching direction is relevant only for those production rules whose RHS contains a branching constituent and a non-branching constituent.
For Left-Branching languages the branching constituent(s) of any RHS must precede the non-branching constituent(s) of the same RHS;
for Right-Branching languages the branching constituent(s) of any RHS must follow the non-branching constituent(s) of the same RHS.
Branching-direction says nothing about what happens in those production-rules whose RHS contains either no branching constituents or no non-branching constituents.

basilius wrote:The main source of my questions is ignorance, sorry.
Perhaps, also or instead, either not reading carefully what I wrote before, or my writing being too hard to follow.
If that last is what it was, it's my fault and I apologize and will try to do better.
Maybe someone else here on the CBBoard can make it more accessible to you?
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Tue 27 Aug 2013, 21:04

eldin raigmore wrote:Perhaps, also or instead, either not reading carefully what I wrote before, or my writing being too hard to follow.

The former. Absolutely.
Maybe someone else here on the CBBoard can make it more accessible to you?
In fact, I know very few people, not to say just one person, who'd undertake writing an intro to a theory like X-bar in a forum comment format... and I certainly know exactly one person who performs such feats regularly :)

I mean, thank you. It's worth the effort, even if some of your readers are lazy and not focused enough. Like, some places become really worth visiting because it happens.

I have the feeling that I owe Click a more constructive comment on his challenge.
Click wrote:Ok, here's the challenge.
  1. The language is mostly head-final, especially with regards to adverbial and noun phrases.
  2. The language orders words in a VSO word-order.
  3. There is a quite plausible diachronic explanation how the language came to have these two rather conflicting features.
Spoiler:
The ancestral language is strictly head-final.

At some point, split constructions with dislocated dependents become common, of the type that can be imitated in English as follows:

He looked at her, John, at Mary.

Or, to match the head-final syntax of the language:

He at-her looked, John, at-Mary.

Or even more faithfully:

He her-at looked, John, Mary-at.

(I use look at instead of a canonical transitive verb in this illustration for at nicely imitates an accusative marker.)

In about 300 years, such dislocation-based constructions are reanalyzed as a legitimate clause subtype (i. e. not perceived as dislocations anymore).

After some 300 years more, this construction becomes prevailing in main clauses. Pronominal elements before the verb become agreement markers. (It could work a bit differently if originally the language was more pro-drop, but this is irrelevant for solving the challenge.)

After some extra 500 years, the former regular verb-last construction becomes marginalized, i. e. mostly disallowed in independent declarative clauses. (Dependent clauses may preserve some remnants of the former word order.) NP's and other dependent phrases mostly remain head-final.

The last stage has a clear functional motivation: parsing becomes easier if one can rely on clause-initial V being always the main verb. The motivation behind the changes in the two earlier stages isn't clear (although, as usual, one can always resort to contact influences).
Does this look like a solution you wanted, Click?
User avatar
Click
darkness
darkness
Posts: 3328
Joined: Sat 21 Jan 2012, 12:17

Re: "Make these work together" game.

Post by Click » Tue 27 Aug 2013, 21:15

Yes. [:)]
User avatar
eldin raigmore
fire
fire
Posts: 5684
Joined: Sat 14 Aug 2010, 18:38
Location: SouthEast Michigan

Re: "Make these work together" game.

Post by eldin raigmore » Wed 28 Aug 2013, 01:15

Click wrote:Yes. [:)]
Great!

(By the way, basilius, I, also, liked your proposed solution to Click's challenge.)

-------------

Does that make it basilius's turn?
Or maybe Creyeditor's?
What do you think, Click?
basilius
MVP
MVP
Posts: 244
Joined: Thu 27 Jan 2011, 18:22

Re: "Make these work together" game.

Post by basilius » Wed 28 Aug 2013, 01:56

Thanks :)
eldin raigmore wrote:Who will post the next challenge?
OK, I think I can try to propose one more.

It's about finding a plausible diachronic scenario; and I don't know a good solution myself. Also, it'll require a lengthy explanation, so I won't be surprised or upset by a "tl;dr" type of reaction =)

I imagined a system for agreement marking on verbs which appears logical, functional and learnable (if you think it isn't, please comment), but I don't see a natural-looking diachronic scenario leading to it.

The template for a transitive verb looks like this:

Verb stem | Role-1 | Some other marker(s) | Role-2 | Some other marker(s)

Some preliminary notes.

I don't want to precipitately decide on whether the system is accusative, ergative or something else; I'll discuss only transitive verbs, and I'll speak of agent-like arguments (A) and patient-like arguments (P). The agreement works solely for A and P (thus, there's no agreement with obliques).

"Verb stem" is already marked for some categories (it's irrelevant by which morphological techniques). The categories whose markers are part of the "Verb stem", or are placed in the two slots I labelled "Some other marker(s)", include voice, plurality of participants, and some additional categories (tense-aspect-mood and perhaps others, not discussed below since they're irrelevant for the challenge).

Voices probably include active, reflexive and reciprocal (and perhaps something else); besides, a marker from the same set is used to differentiate transitive verbs from intransitives. The important thing here is that the agreement system does not permit marking like "1>1" or "2>2" (see below), and "3>3" cannot have a reflexive or reciprocal reading. It looks natural to convey the latter meanings by voices having their own markers on the verb (although I can imagine some alternatives).

Plurality of participants distinguishes situations where both A and P are singular from situations where A, or P, or both are plural (thus, the system will not differentiate "I saw them" from "We saw it" or "We saw them"); but it may be a bit trickier than that; for example, dual inclusive may be still treated as singular, since it allows for analysis like "1S + 2S"; or perhaps "plurality" should better be made more differentiated, allowing for things like "Pl>Sg", "Pl>Pl" etc.; but this isn't very relevant either.

Now, the real thing.

"Role-1" is the slot for an element marking the role of first person (the speaker).

"Role-2" is the slot for an element marking the role of second person (the listener).

There is no such thing as "first person plural" or "second person plural"; these meanings are treated as "the speaker plus someone else" and "the listener plus someone else".

The "Role-1" slot, or the "Role-2" slot, or both, or neither can be filled with one of the following two markers: "A" (for agent-like role) and "P" (for patient-like role). The "A" and "P" markers used in the "Role-1" slot are the same morphemes as the "A" and "P" markers used in the "Role-2" slot.

If neither slot contains the "A" marker, the agent-like participant is 3rd person; and if neither slot contains the "P" marker, the patient-like participant is 3rd person. In other words, the role of 3rd person participants is not marked as such, but can be deduced from lack of a marker coding the respective role for either 1st or 2nd person.

The "A" marker in the "Role-1" slot (Role-1=A for short) means that the speaker, alone or with someone else, is the agent-like participant in the event; translated "I" or "we".

Role-1=P means that the speaker, alone or with someone else, is the patient-like participant in the event; translated "me" or "us".

Similarly, Role-2=A is translated "thou" or "you (Pl)", and Role-2=P "thee" or "you (Pl)".

Role-1=A + empty Role-2 (Role-2=0 for short) means "1.excl>3", e. g. "I saw him" (or "we saw them" etc., depending on which readings the plurality marker permits).

Role-1=A + Role-2=P means "1.excl>2", e. g. "I saw you" or "we saw you" (the latter, where plurality marker permits such reading).

Role-1=P + Role-2=0 means "3>1.excl", e. g. "he saw me" (etc., depending on plurality).

Role-1=P + Role-2=A means "2>1.excl", e. g. "you saw me" (etc.).

Similarly, Role-1=0 + Role-2=A is "2>3", and Role-1=0 + Role-2=P is "3>2".

"Role-1" and "Role-2" filled with identical role markers refer to first person inclusive; that is, Role-1=A + Role-2=A means "1.incl>3" ("You and me saw him" etc.), and Role-1=P + Role-2=P is "3>1.incl" ("He saw you and me" etc.).

Role-1=0 + Role-2=0 is "3>3" ("He saw them" etc.).

As stated above, the system does not permit "1>1" or "2>2", so meanings like "I imagined both of us old and respectable" are conveyed in some periphrastic ways (and "I imagined myself..." is rendered by the reflexive voice).

In short.

There is no such thing as "first person marker" or "second person marker"; instead, there are "position for the marker of first person's role" and "position for the marker of second person's role".

In other words, two things are swapped in the above system.

Normally, persons are coded by markers, and roles, (mostly) by those markers' positions (and, often, specific shapes/allomorphs).

With the alternative in question, roles are coded by markers, and persons, solely by positions of those role markers.


The challenge is to find a diachronic scenario where the above unusual agreement marking strategy develops from something reliably attested (or else, to find substantial arguments for there being no plausible scenarios of that sort).

(Thus, the two things that must "work together" is the above strategy and any less bizarre ancestral state.)

Like I said, I don't know a good solution myself; this means that solutions to a modified form of the challenge (e. g. with a different order of elements within verb template), partial solutions (e. g. modelling the "positional" coding strategy for patients only, and the like), and just considerations that may seem on-topic - will be very, very welcome and helpful.
User avatar
eldin raigmore
fire
fire
Posts: 5684
Joined: Sat 14 Aug 2010, 18:38
Location: SouthEast Michigan

Re: "Make these work together" game.

Post by eldin raigmore » Wed 28 Aug 2013, 21:41

Well, I love the challenge (but I don't have an answer for it yet).
Thanks.
User avatar
eldin raigmore
fire
fire
Posts: 5684
Joined: Sat 14 Aug 2010, 18:38
Location: SouthEast Michigan

Re: "Make these work together" game.

Post by eldin raigmore » Thu 08 Jun 2017, 12:45

Deliberate thread necromancy:
I wish someone would try to answer basilius's challenge.
When I have more free time I will try to answer it myself.
It might not be today, though I hope it will be within 24 hours;
but I really, really hope (and also, surmise) it will be within 7 days if I can't get to it today!
User avatar
Creyeditor
mongolian
mongolian
Posts: 3948
Joined: Tue 14 Aug 2012, 18:32

Re: "Make these work together" game.

Post by Creyeditor » Fri 09 Jun 2017, 13:19

basilius wrote:
Spoiler:
Thanks :)
eldin raigmore wrote:Who will post the next challenge?
OK, I think I can try to propose one more.

It's about finding a plausible diachronic scenario; and I don't know a good solution myself. Also, it'll require a lengthy explanation, so I won't be surprised or upset by a "tl;dr" type of reaction =)

I imagined a system for agreement marking on verbs which appears logical, functional and learnable (if you think it isn't, please comment), but I don't see a natural-looking diachronic scenario leading to it.

The template for a transitive verb looks like this:

Verb stem | Role-1 | Some other marker(s) | Role-2 | Some other marker(s)

Some preliminary notes.

I don't want to precipitately decide on whether the system is accusative, ergative or something else; I'll discuss only transitive verbs, and I'll speak of agent-like arguments (A) and patient-like arguments (P). The agreement works solely for A and P (thus, there's no agreement with obliques).

"Verb stem" is already marked for some categories (it's irrelevant by which morphological techniques). The categories whose markers are part of the "Verb stem", or are placed in the two slots I labelled "Some other marker(s)", include voice, plurality of participants, and some additional categories (tense-aspect-mood and perhaps others, not discussed below since they're irrelevant for the challenge).

Voices probably include active, reflexive and reciprocal (and perhaps something else); besides, a marker from the same set is used to differentiate transitive verbs from intransitives. The important thing here is that the agreement system does not permit marking like "1>1" or "2>2" (see below), and "3>3" cannot have a reflexive or reciprocal reading. It looks natural to convey the latter meanings by voices having their own markers on the verb (although I can imagine some alternatives).

Plurality of participants distinguishes situations where both A and P are singular from situations where A, or P, or both are plural (thus, the system will not differentiate "I saw them" from "We saw it" or "We saw them"); but it may be a bit trickier than that; for example, dual inclusive may be still treated as singular, since it allows for analysis like "1S + 2S"; or perhaps "plurality" should better be made more differentiated, allowing for things like "Pl>Sg", "Pl>Pl" etc.; but this isn't very relevant either.

Now, the real thing.

"Role-1" is the slot for an element marking the role of first person (the speaker).

"Role-2" is the slot for an element marking the role of second person (the listener).

There is no such thing as "first person plural" or "second person plural"; these meanings are treated as "the speaker plus someone else" and "the listener plus someone else".

The "Role-1" slot, or the "Role-2" slot, or both, or neither can be filled with one of the following two markers: "A" (for agent-like role) and "P" (for patient-like role). The "A" and "P" markers used in the "Role-1" slot are the same morphemes as the "A" and "P" markers used in the "Role-2" slot.

If neither slot contains the "A" marker, the agent-like participant is 3rd person; and if neither slot contains the "P" marker, the patient-like participant is 3rd person. In other words, the role of 3rd person participants is not marked as such, but can be deduced from lack of a marker coding the respective role for either 1st or 2nd person.

The "A" marker in the "Role-1" slot (Role-1=A for short) means that the speaker, alone or with someone else, is the agent-like participant in the event; translated "I" or "we".

Role-1=P means that the speaker, alone or with someone else, is the patient-like participant in the event; translated "me" or "us".

Similarly, Role-2=A is translated "thou" or "you (Pl)", and Role-2=P "thee" or "you (Pl)".

Role-1=A + empty Role-2 (Role-2=0 for short) means "1.excl>3", e. g. "I saw him" (or "we saw them" etc., depending on which readings the plurality marker permits).

Role-1=A + Role-2=P means "1.excl>2", e. g. "I saw you" or "we saw you" (the latter, where plurality marker permits such reading).

Role-1=P + Role-2=0 means "3>1.excl", e. g. "he saw me" (etc., depending on plurality).

Role-1=P + Role-2=A means "2>1.excl", e. g. "you saw me" (etc.).

Similarly, Role-1=0 + Role-2=A is "2>3", and Role-1=0 + Role-2=P is "3>2".

"Role-1" and "Role-2" filled with identical role markers refer to first person inclusive; that is, Role-1=A + Role-2=A means "1.incl>3" ("You and me saw him" etc.), and Role-1=P + Role-2=P is "3>1.incl" ("He saw you and me" etc.).

Role-1=0 + Role-2=0 is "3>3" ("He saw them" etc.).

As stated above, the system does not permit "1>1" or "2>2", so meanings like "I imagined both of us old and respectable" are conveyed in some periphrastic ways (and "I imagined myself..." is rendered by the reflexive voice).

In short.

There is no such thing as "first person marker" or "second person marker"; instead, there are "position for the marker of first person's role" and "position for the marker of second person's role".

In other words, two things are swapped in the above system.

Normally, persons are coded by markers, and roles, (mostly) by those markers' positions (and, often, specific shapes/allomorphs).

With the alternative in question, roles are coded by markers, and persons, solely by positions of those role markers.


The challenge is to find a diachronic scenario where the above unusual agreement marking strategy develops from something reliably attested (or else, to find substantial arguments for there being no plausible scenarios of that sort).

(Thus, the two things that must "work together" is the above strategy and any less bizarre ancestral state.)

Like I said, I don't know a good solution myself; this means that solutions to a modified form of the challenge (e. g. with a different order of elements within verb template), partial solutions (e. g. modelling the "positional" coding strategy for patients only, and the like), and just considerations that may seem on-topic - will be very, very welcome and helpful.
So I think there might actually be a partial solution to this, based on syncretism and grammaticalization of auxillaries. I can give you a slot for any first person agreement and one for any second person agreement, but I wasn't able to get differences for different roles. I also set aside the first person plural inclusive.

We start with an SOV language with suffixal agreemnt for person on verbs, no auxillaries and a ban on having more than one agreement suffix (all of this attested). Which agreement suffix surfaces is based on the following hierarchy 1 > 2 > 3. There might be a tense morpheme intervening.
Thus
Spoiler:
1 2 greet-1
`I greet you.'

1 3 greet-1
`I greet her.'

2 1 greet-1
`You greet me.'

2 3 greet-2
`You greet her.'

3 1 greet-1
`She greets me.'

3 2 greet-2
`She greets you.'

3 3 greet-3
`She greets him.'
In the next step, phonological changes render the third person and second person agreement homophonous. We have a first vs. non-first pattern. This change is definitely attested. This yields
Spoiler:
1 2 greet-1
`I greet you.'

1 3 greet-1
`I greet her.'

2 1 greet-1
`You greet me.'

2 3 greet-N1
`You greet her.'

3 1 greet-1
`She greets me.'

3 2 greet-N1
`She greets you.'

3 3 greet-N1
`She greets him.'
In the next step, two things happen. The non-first marker get's worn off by phonological changes and an auxillary is introduced to show aspectual distinctions. This auxillaries agrees in the same way the main verb used to agree (except that the hierarchy is 2>1>3), except that it's based on independent pronouns and thus not yet phonologically worn off. This gives us:

Spoiler:
1 2 greet-1 PROG.AUX-2
`I am greeting you.'

1 3 greet-1 PROG.AUX-2
`I am greeting you.'

2 1 greet-1 PROG.AUX-2
`You are greeting me.'

2 3 greet PROG.AUX-2
`You are greeting her.'

3 1 greet-1 PROG.AUX-1
`She is greeting me.'

3 2 greet PROG.AUX-2
`She greets you.'

3 3 greet PROG.AUX-3
`She greets him.'
In the next step, the first and the third person markers on the auxiallary become homophonous and we get a second vs. non-second system on the auxillary.

Spoiler:
1 2 greet-1 PROG.AUX-2
`I am greeting you.'

1 3 greet-1 PROG.AUX-2
`I am greeting you.'

2 1 greet-1 PROG.AUX-2
`You are greeting me.'

2 3 greet PROG.AUX-2
`You are greeting her.'

3 1 greet-1 PROG.AUX-N2
`She is greeting me.'

3 2 greet PROG.AUX-2
`She greets you.'

3 3 greet PROG.AUX-N2
`She greets him.'
In the last step, the non-second person marker on the auxillary is phonologically worn off. In addition the auxillary becomes incorporated into the main verb phonologically.
Spoiler:
1 2 greet-1-PROG.AUX-2
`I am greeting you.'

1 3 greet-1-PROG.AUX-2
`I am greeting you.'

2 1 greet-1-PROG.AUX-2
`You are greeting me.'

2 3 greet-PROG.AUX-2
`You are greeting her.'

3 1 greet-1-PROG.AUX
`She is greeting me.'

3 2 greet-PROG.AUX-2
`She greets you.'

3 3 greet-PROG.AUX
`She greets him.'
Feel free to expand this solution. And I hope there are not too many typos.
Creyeditor
"Thoughts are free."
Produce, Analyze, Manipulate
1 :deu: 2 :eng: 3 :fra: 4 :esp: 4 :ind:
:con: Ook & Omlűt & Nautli languages & Sperenjas
[<3] Papuan languages, Morphophonology, Lexical Semantics [<3]
User avatar
eldin raigmore
fire
fire
Posts: 5684
Joined: Sat 14 Aug 2010, 18:38
Location: SouthEast Michigan

Re: "Make these work together" game.

Post by eldin raigmore » Fri 09 Jun 2017, 18:33

Thank you, @Creyeditor.
I like that!
BTW I don't know whether or not basilius is still with us;
but look at her/his remark Posted: Fri 26 Jul 2013, 10:39 about your solution to Click's challenge.
There s/he objected to agreement-markers morphing into independent pronouns.
In your solution to the current challenge posted by basilius, you have the opposite direction happening; independent pronouns morphing into agreement-markers.
I don't think basilius would object to that direction; what do you think basilius would think?
Post Reply