Re: THEORY: Morphosyntactic Alignment (again?), and Milewski

On Mon, 15 May 2006 11:22:43 -0400, Jim Henry
><jimhenry1973@...> wrote:
[er]
I think I was correct on the outline of Milewski's major ideas, but it is a
better-than-50-50 bet that I at least partially misunderstood at least one
important sub-idea.  If so I hope someone on-list can help.  Also, if so,
that may partly explain some of your objections to and/or confusions about
what I said.

[JH]
>On 5/14/06, Eldin Raigmore <eldin_raigmore@...>
>>wrote:
[er]
>>Milewski's extension, on the other hand, has been to
>>add what I will call "genitive phrases" to the mix;
>>that is, to align expressions in which one noun
>>modifies another noun (possessor-modifying-possessum
>>seems prototypical) with the prototypical bivalent
>>monotransitive clauses and the prototypical
>>monovalent intransitive clauses.
[er]
I'm pretty sure I got _that_ much right.

I wrote the original post because I thought for the first time that I
understood Milewski's alignment typology better than I had just from
reading the Wikipedia article, yet I still didn't feel that I understood it
completely.

I started feeling this way after reading all of one English article and the
beginning of one French article.

The English article is
Typological Similarities Between Caucasian and American Indian Languages,
Folia Orientalia IV [1962] pp. 221-230.

The French article is
La Structure de la Phras dans les Langues Indigenes de l'Amerique du Nord,
Lingua Posnaniensis II [1950], pp. 162-207.

He seems concerned with, among other roles ("functions"),
object of transitive clause
subject of transitive clause
subject of intransitive clause
predicate nominal
"substantif dans un emploi absolu" ("noun used in the absolute"?)

[er]
>>Some abbreviations will now be useful;
In the French article Milewski uses his own abbreviations;
S nominal subject
s pronominal subject
A nominal agent
a pronominal agent
O nominal direct object
o pronominal direct object
O' nominal indirect object
o' pronominal indirect object
P predicate
Pr possessor
Pe possessum
pos possessive pronoun
- ordering of independent words
+ ordering of grammatical words

[er]
>>A=Agent of prototypical bivalent monotransitive
>>clause
>>O=Patient of prototypical bivalent monotransitive
>>clause
>>S = Single argument of prototypical monovalent
>>intransitive clause
>>G=Possessor, or, noun which is being used to
>>modify another noun
>>C= Thing possessed, or, noun which is being modified
>>by another noun
[er]
Milewski referred to the thing I abbreviated by "G" as "determinANT" in
English, as "posseSSEUR" in French.
He referred to the thing I abbreviated by "C" as "determinATE" in English,
as "chose posseDEE" in French.

His prototype example was Latin "domus patris".  "Domus" was the
determinATE, "patris" was the determinANT, noun.

In the French article, however, he lets the "G" role be taken by what
appears to an English speaker to be an adjective, in his analysis of the
Maidu word "nenommaidum" ("old couple"); he says "nenom-" ("old") is
the "membre determinant du compose" ("the determining member of the
compound"?); leaving me to assume that "-maidum" ("man", subject) is the
determined member.

[er]
>>If a language has just two ways of marking these
>>roles, then,
>>* either S is marked like A or S is marked like O;
>>and independently,
>>* either S is marked like G or S is marked like C.
[er]
That comes from me, not from Milewski.  I'm still pretty sure what I
_meant_ to say is right, if not exactly earth-shaking.

[JH]
>In the languages I'm most familar with, G is marked
>and C can be marked like either A, P or S.  I.e., in a
>genitive language, why say that the possessum is
>prototypically S, or A, when it can just as easily be
>O? E.g.,
>
>"His dog chased her cat."
> G   A   V      G   P
>
>That is, it makes sense to say that in English and
>other genitive languages C is unmarked rather than
>that it is marked like S or A or P.
[er]
Well, I believe I failed to precisely and effectively communicate my
_exact_ meaning; because I used the words "marked" and "marking".

Actually it might be better to say one use of a noun is _treated_ like, or
differently from, a different use of the noun.

Milewski mentions, and consequently I am well aware, that there are
different ways of indicating the role a nominal has in a clause.

_My_ three-way classification (inspired by reading Joan Bybee) is as
follows; morphological (internal alternations to the stem or affixes),
lexical (using grammatical words in addition to the noun, for instance
adpositions or particles), or syntactic (for instance word-
order).  "Lexical" has a fuzzy boundary with "syntactic", since one must
decide _where_ to put the additional words; "lexical" also has a fuzzy
boundary with "morphological", because "clitic" has a fuzzy boundary
with "affix" and also has a fuzzy boundary with "independent word".  (For
that matter, is an imposition an adposition or an infix?)

_Milewski's_ four-way classification was;
positional (what I'd call "purely syntactic");
flexional (what I'd call "dependent-marking morphological");
incorporating* (what I'd call "head-marking morphological");
and flexional-incorporating.

I don't know how Milewski would have classified the ones I
called "lexical".  Perhaps he had such a classification, but didn't feel it
was needed in an article about mostly North American languages (also plenty
about South American languages, and some about "Palaeo-Asiatic" languages
and Caucasian languages.)

In any case, _all_ of the above techniques are covered by the concept I
called "marked" or "marking" in my original post.

I also meant to include -- without any non-theory-neutral commitment
concerning "markedness" a la Trubetzkoy 1939/1969 or Jakobson 1962, nor any
commitment to the existence of "zero morphemes" -- those treatments of
nouns which are characterized by nothing special happening to it -- no
alternations, no affixes, no adpositions; perhaps also no movements.

In this I was following in the spirit of Greenberg(1966)'s Universal
#38, "Where there is a case system, the only case which has only zero
morphemes is the one which includes among its meanings that of the subject
of the intransitive verb." (See
http://ling.uni-konstanz.de:591/Universals/FMPro?-DB=Universals&-Lay=Layout%
20%231&-Token=10&-Format=overview.html&-Max=10&-Findall

http://www.hku.hk/linguist/program/Typology5.html

http://ling.kgw.tu-berlin.de/Korean/Artikel02/morphology.html

http://ling.kgw.tu-berlin.de/Korean/Artikel02/Appendix3.html

http://www.cl.uni-heidelberg.de/kurs/ws03/sprachen/2.pdf

http://www.anukriti.net/pgdts/course421/ch1h.html

http://72.14.203.104/search?
q=cache:Meqv9CAXdnkJ:email.eva.mpg.de/~haspelmt/Parametric.pdf+Greenberg+uni
versal+zero+subject+intransitive&hl=en&gl=us&ct=clnk&amp;cd=7
which quotes it differently, as "If a language lacks overt coding for
transitive arguments, it will also lackovert coding for the intransitive
subject (Greenberg 1963, Universal 38)."

)

Note that neither Greenberg nor Milewski assumed every language has a case
system.  So Greenberg's universal 38 wouldn't apply to a non-case
language.  Therefore it is an implicational universal.  It turns out it is
also a statistical universal, since there seem to be a few known
counterexamples.

Milewski doesn't consider a language to have a case system unless at least
one morphological marking in that system is used both for a syntactic
purpose such as a grammatical relation, and for a "locative" or
directional.  Languages which use only suppletion, or only lexical means,
or only what he calls "incorporation", or only word-order or some other
purely syntactic means, to mark "case roles" or grammatical relations,
wouldn't have been considered "case languages" by Milewski -- if I have
understood him correctly.  Neither would a language which marked both
locatives and grammatical relations by means of inflections, if it so
happened that no such marking was used both for some grammatical relation
and for some locative.

*In the "incorporating" languages, clitic forms of pronouns representing
the core participants are incorporated into the verb; they may tell
something about the person, the gender, the number, and/or the "case role"
of the participants.  This is what is called "the cross-reference type of
agreement"; that is the verb agrees with its participants by cross-
referencing them.  Meanwhile, noun-phrases representing the participants
are appositive to these clitic pronouns.

>>This gives four alignments when there are only two
>>marking-types:
(er.i)
>>A=S=C (nominative), O=G (accusative=genitive);
(er.ii)
>>A=S=G (nominative),
>> O=C (accusative= construct state);
(er.iii)
>>O=S=C (absolutive), A=G (ergative=genitive);
(er.iv)
>>O=S=G (absolutive), A=C (ergative= construct state).
[er]
This was my own thinking, and I'm pretty sure it's correct; "whether or not
it's relevant" is part of what I was asking in the "do natlangs attest?"
questions, and "whether or not it's original" is part of what I was asking
in the "was this what Milewski meant?" questions.

My use of the term "construct state", like my use of the term "genitive
phrase", was based in part on my ignorance of any better term to use.
A "dependent-marking" language "marks" a phrase in which one noun modifies
another noun, by "marking" the modifying noun; this "marking" is frequently
called "genitive".  A "head-marking" language "marks" such phrases
by "marking" the modified noun; sometimes, some people call
such "markings", in some such languages, "construct state".

[JH]
>Esperanto has A=S=G, O -- unless I'm misunderstanding
>something here.
[er]
I don't know Eo; I know a little about it, but not enough not to simply
take your word for everything you're saying about it here.

According to what you say below, Esperanto has A=S, but does _not_ have A=G
nor S=G, because G is "marked" with the preposition "de".

[JH]
>G is marked with a preposition "de", and that
>preposition takes a nominative object.
[er]
Using the terminology the way I meant it, once the preposition "de" is
preposed to it, it isn't nominative anymore; now it's genitive.

More in line with the ideas I'm trying to talk about here, "A" and "S"
are "marked" by "zero"; "G" is "marked" by preposing "de".

[JH]
>C can be =A or =P.
[er]
The possibility that saying "in Esperanto C can be =A or =P" might be
correct, even using these terms the way Milewski meant, is one I can't be
sure I should eliminate nor be sure I should accept.

The problem is that in the two articles Milewski wrote about this that I
have read so far, he refers to marking the relationship between the G and
the C, but doesn't make clear which one he means to be the marked one in
each particular instance of which he speaks.

[JH]
>(Does "only two marking types" refer to cases/word
>order positions only or also to adpositions?)
[er]
See my earlier remarks.  It must have been the first time I read this
question that made me think of those remarks, but as I was adding my
thoughts to this reply I felt those particular remarks were relevant
earlier.

[JH]
>[I seem to vaguely recall someone saying that
>Esperanto's use of the nominative as the normal case
>for objects of prepositions violates some universal.
>If so, it's a universal of natural language
>evolution, not of the human language faculty, since
>people can learn to think fluently in Esperanto and
>automatically apply this rule.]
[er]
I don't know whether or not there is some (presumably statistical,
presumably implicational) universal that prepositions should get preposed
to the same form of the noun as the nominative -- supposedly this applies
only to prepositional (not to postpositional) languages with
accusative/nominative (not, e.g., ergative/absolutive) alignment.

But it isn't what I'm talking about here.  Here, what we would be saying is
that the bare stem is used as the Subject/Agent pivot, and the preposition
is preposed to that same bare stem.  The preposition isn't preposed to the
nominative.  If Esperanto were changed so that one or both of the S/A roles
were explicitly marked by an affix, instead of zero-marked, the
prepositonal phrases would _not_ be changed to match.

In any case, there are only two roles, among the ones involved in
Milewski's typology, which _are_ more-or-less frequently marked by means of
adpositions; the G role ("genitive") and the O role ("accusative").  In
English, the G is marked by "of", but the O is marked other-than-
adpositionally.

[er]
>>If, on the other hand, there are three
>>marking-types, fourteen other possibilities are
>>opened up:
(er.v)
>>A=S=C (nominative), O (accusative), G (genitive);
(er.vi)
>>A=S=G (nominative), O (accusative),
>> C (construct state);
(er.vii)
>>O=S=C (absolutive), A (ergative), G (genitive);
(er.viii)
>>O=S=G (absolutive), A (ergative),
>> C (construct state);
(er.ix)
>>S=A (nominative), O=C (accusative=construct state),
>> G (genitive);
(er.x)
>>S=A (nominative), O=G (accusative=genitive),
>> C (construct state);
(er.xi)
>>S=C (nominative/absolutive = construct state),
>> A=G (ergative=genitive), O (accusative);
(er.xii)
>>S=C (nominative/absolutive = construct state),
>> O=G (accusative=genitive), A (ergative);
(er.xiii)
>>S=G (nominative/absolutive = genitive),
>> A=C (ergative= construct state), O (accusative);
(er.xiv)
>>S=G (nominative/absolutive = genitive),
>> O=C (accusative= construct state), A (ergative);
(er.xv)
>>S=O (absolutive), A=C (ergative= construct state),
>> G (genitive);
(er.xvi)
>>S=O (absolutive), A=G (ergative=genitive),
>> C (construct state);
(er.xvii)
>>A=C (ergative= construct state),
>> O=G (accusative=genitive),
>> S (nominative/absolutive);
(er.xviii)
>>A=G (ergative=genitive),
>> O=C (accusative= construct state),
>> S (nominative/absolutive).
[er]
Once again, this is my own thinking; once again, it might be only about 33%
relevant; once again, I _hope_ it _isn't_ original.

[er]
>>As a matter of fact Milewski knew of only six of
>>these eighteen patterns being attested; he
>>hypothesized those six were the only ones that
>>actually occurred.
[JH]
>Which six?
Milewski listed six "syntactical types".

1 Treatment "a" covers both the subject-predicate relation and the agent-
predicate relation; treatment "b" covers the patient-predicate relation;
and treatment "c" covers the determinant-determinate relation.

{So Type 1 has A=S; has O!=S and O!=A; and has either G!=S and G!=A and G!
=O, or C!=S and C!=A and C!=O.
Thus it _could_ match any of (er.v), (er.vi), (er.ix), or (er.x).}


2 Treatment "a" covers both the subject-predicate relation and the patient-
predicate relation; treatment "b" covers the agent-predicate relation; and
treatment "c" covers the determinant-determinate relation.

{So Type 2 has O=S; has A!=S and A!=O; and has either G!=S and G!=A and G!
=O, or C!=S and C!=A and C!=O.
Thus it _could_ match any of (er.vii), (er.viii), (er.xv), or (er.xvi).}


3 Treatment "a" covers both the subject-predicate relation and the agent-
predicate relation; treatment "b" covers both the patient-predicate
relation and the determinant-determinate relation.

{So Type 3 has A=S; has O!=S and O!=A; has either G=O or C=O; and has
either G!=S and G!=A, or C!=S and C!=A.
Thus it _could_ match any of (er.i), (er.ii), (er.ix), or (er.x).}


4 Treatment "a" covers both the subject-predicate relation and the patient-
predicate relation; treatment "b" covers both the agent-predicate relation
and the determinant-determinate relation.

{So Type 4 has O=S; has A!=S and A!=O; has either G=A or C=A; and has
either G!=S and G!=O, or C!=S and C!=O.
Thus it _could_ match any of (er.iii), (er.iv), (er.xv), or (er.xvi).}


5 Treatment "a" covers the subject-predicate relation, the agent-predicate
relation, and the determinant-determinate relation; treatment "b" covers
the patient-predicate relation.

{So Type 5 has A=S; has either G=S and G=A, or C=S and C=A; and has either
O!=S and O!=A and O!=G, or O!=S and O!=A and O!=C.
Thus it _could_ match any of (er.i), (er.ii), (er.v), or (er.vi).}


6 Treatment "a" covers the subject-predicate relation, the patient-
predicate relation, and the determinant-determinate relation; treatment "b"
covers the agent-predicate relation.

{So Type 6 has O=S; has either G=S and G=O, or C=S and C=O; and has either
A!=S and A!=O and A!=G, or A!=S and A!=O and A!=C.
Thus it _could_ match any of (er.iii), (er.iv), (er.vii), or (er.viii).}


--

The four of "my" types that _couldn't_ be covered by any of Milewski's six
types, look to be (er.xi), (er.xii), (er.xiii), and (er.xiv).

I would, therefore, be very interested in any natlang which attests to one
of them.

ObConLang:  I would also be interested in any conlang which attests to one
of them.

--

Note that:
My (er.i) and (er.ii) could fall under
either Milewski's Type 3 or his Type 5.
Also my (er.iii) and (er.iv) could fall under
either his Type 4 or his Type 6.
Also my (er.v) and (er.vi) could fall under
either his Type 1 or his Type 5.
Also my (er.vii) and (er.viii) could fall under
either his Type 2 or his Type 6.
My (er.ix) and (er.x) could fall
either under Milewski's Type 1 or his Type 3.
Also my (er.xv) and (er.xvi) could fall under
either his Type 2 or his Type 4.

[er]
>>2) ObConLang: How do your conlangs fit into this
>>typology?
[JH]
>My gjâ-zym-byn is fluid-S active, with a variety of
>genitive postpositions for specific relationships
>(possession, ownership, entity-attribute, part-whole,
>authorship, kinship...), and no construct state.
>As a fluid-S language I don't think it fits into >Milewski's typology at
all.

[er]
That's right.  All of Milewski's types have either S=A or S=P; none of them
allow the S to be "split".

According to Klimov's 1972 article, which AFAIK was first published in
English in 1981, the difference between "active/stative" type
and "ergative" type wasn't published until Sapir did it sometime between
1917 and 1920, and wasn't "common" knowledge until Uhlenbeck published it
in 1930.  Klimov specifically mentions a Milewski article published, I
think, sometime in the 1960's, as one of several he mentions which
perpetuated the error.  AFAIK sufficient information about "languages of
active type" wasn't available to be satisfactorily worked into anybody's
typology until Klimov's 1972 article.  Milewski died before 1967, and the
latest article of Milewski's I have is from 1962.

[JH]
>There are at least three postpositions that can
>mark the subject of a sentence
[er]
_That_ is _very_ _interesting_!

[JH]
>(depending on animacy and volitionality)
>and at least six that can mark the object of a
>transitive verb,
[er]
That is interesting.

[JH]
>plus several others that can mark the predicate of a
>subject noun when there is no verb.
[er]
That sounds like the kind of phenomenon Milewski would have talked about;
unfortunately I can't figure out what he would have said about it.

At any rate, it's both interesting in its own right, and right on-topic for
this post.

[JH]
>One of my oldest conlangs, Pliv-Rektek, had both a
>genitive case and what I then called a
>contra-genitive, not having heard of the
>term "construct state".
[er]
Is "construct state" a "case", as it seems at the moment?  Or is it
like "definite" and "indefinite", whatever they are?

[JH]
>A simple statement of possession used a noun in each
>case with no verb; elsewhere one could use either
>case as needed to comment on a subject or object
>being a possessor or posessum of some other noun. So
>it would violate Milewski's tentative universal that
>no language marks A, P, G and C in four distinct ways.
[er]
Quite relevant.  Thanks.

I figure that was intended, or at least destined to become, a "statistical"
universal, anyway.

[JH]
>Jim Henry
>http://www.pobox.com/~jimhenry/gzb/gzb.htm
[er]
Thanks, Jim.

-----

Anyone else?

-----

eldin

Reply