Re: Impersonal Passives and Quirky Case in Subject-Prominent Languages (was: Copula)

---In conlang@yahoogroups.com, R A Brown <ray@...> wrote:
>Eldin Raigmore wrote:
>>On Tue, 20 Mar 2007 20:15:03 +0000, R A Brown
>><ray@...> wrote:
>[snip]
>>>I do not see how this is possible if we are talking in purely
>>>_grammatical_ terms, i.e. we are not dealing with semantic roles.
>>>what I mean is that the nominative case is defined as "in a
>>>morpholologically *accusative*, the case used for both subjects of
>>>intransitive verbs and subjects of transitive verbs" [Trask]. That
>>>is, the nominative case is defined by being the grammatical
>>>subject.
>>>Similarly the accusative case (in nominative-accusative languages)
>>>is the case that marks the direct objects.
>>Well, if we follow those two definitions too slavishly, there can
>>be no such thing as "quirky case" subjects and "quirky case"
>>objects in Icelandic and German and Malayalam and Hindi;
>
>How else do you define the nominative and accusative case?
In a morphologically nominative/accusative language, the nominative case is
used for the subject of nearly all basic monovalent clauses, and for the agent
of prototypical transitive clauses, and for the donor of nearly all basic
ditransitive clauses.

In a morphologically nominative/accusative language, the accusative case is
used for the patient of prototypical monotransitive clauses, and if the
language is also morphologically directive in its dependent-marking, for the
theme of nearly all basic ditransitive clauses.

>I really do not understand,
That could be because I did not overcome the difficulty of saying what I
wanted to say.

>and I'm sorry you feel it is being 'slavish' if one understands a
>definition in a certain way.
"Slavish" may have been a poor choice of word on my part.
Forgetting that, let me try to explain what I meant.

Consider
>>>the nominative case is defined as "in a morpholologically
>>>*accusative*, the case used for both subjects of intransitive
>>>verbs and subjects of transitive verbs" [Trask]. That is, the
>>>nominative case is defined by being the grammatical subject.
If this is accepted as being literally true and word-for-word true, it would
mean that any clause without a nominative participant would have no subject;
and any clause with more than one nominative participant would have more
than one subject; and there could be no such thing as a clause with a
nominative object.

But there are languages that do have (a minority of their) basic clauses with
non-nominative subjects and/or with nominative non-subjects; or at least that
is the accepted analysis.

Examples of non-nominative subjects include those of Icelandic's equivalents
of verbs (some transitive, some intransitive) meaning "long to", "dream
about", "melt", "drift", "lack", "freeze", "speak", "capsize", "slope", "seem", "rese
mble", "recover from", "be noticeable".

Other examples of non-nominative subjects include Tamil's equivalent of the
verbs "want", "understand", "know", "like", "hunger",
Malayalam's "hunger", "hurt", "sorrow", "hate", Hindi's "saw", "become
happy", "fear", Tamil's "have", Kannada's "have", Malayalam's "have",
Hindi's "find" and "receive".

More examples include the non-volitional (but not the volitional) subjects of
Malayalam's "understand" and "weep", Kannada's "learn" and "catch fever",
Hindi's "see" (vs. "look at"); and also the subjects of moods involving
permission or obligation or desire in Malayalam and Hindi.

Examples of nominative non-subjects include the inanimate direct objects in
Malayalam and Hindi.

Now also consider
>>>Similarly the accusative case (in nominative-accusative languages)
>>>is the case that marks the direct objects.
If this is accepted as being literally true and word-for-word true, it would
mean that any clause with no accusative participant had no direct object, and
any clause with two or accusative participants had two or more direct
objects.  Also, there could not be any such thing as a clause with an
accusative subject.

But there are languages that do have (a minority of their) basic clauses with
non-accusative direct-or-primary objects and/or with accusative subjects; or
at least that is the accepted analysis.

Examples of non-accusative objects include German antworten, begegnen,
danken, fluchen, folgen, gefallen, helfen, trauen, gedenken; and
Icelandic "rescue", "finish", "await", "visit".

Examples of accusative subjects include Icelandic "long to", "dream
about", "melt", "drift", "lack", and "freeze".

>either words do have meanings or they behave as Humpty Dumpty made
>them behave. In the latter case, rational discussion becomes
>meaningless.
Well, (a) if we don't agree on the meanings of words during our conversations,
our conversations may be confusing when they need those words; (b) but
even if words have fixed and universal meanings, Trask's statements of them
may not have been absolutely correct, or may have been slightly mutated
during quotation.

Anyway if we accept that;
(1) Every subject is nominative.
and
(2) Everything nominative is a subject.
Then there can not be any such thing as "quirky-case subjects", and all of the
books and papers written about them were just a waste of trees' lives.

And if we accept that
(3) Every direct object is accusative.
and
(4) Everything accusative is a direct object.
Then there cannot be any such thing as "quirky-case objects", and all of the
books and papers written about them were wastes of time.

>>contradiction in terms.  So, Eldin, please shut up and quit wasting
>>bandwidth."
>I did not say that, and I'm sorry you feel that way. I said that I >did not
understand what you meant and explained why, that's all.

OK.

>[snip]
>>18a Icelandic Active
>>Ég hjálpaði þeim.
>>I(NOM) helped them(DAT).
>>
>>19a Icelandic Passive
>>Þeim var hjálpað.
>>Them(DAT) was helped.
>>
>>49a German Active
>>Ich habe ihm geholfen.
>>I(NOM) have-1sg him(DAT) helped.
>>
>>49b German Passive
>>Ihm wurde geholfen.
>>Him(DAT) was helped.
>
>Sorry, but for the life of me I do not see how these examples differ
>from Latin:
>Active
>eis succurri
>them(DAT) helped(1st.S)
>
>Passive
>eis succursum est.
>them(DAT) was helped
>
>Neither in Latin has a direct object - one wouldn't expect that of a
>passive anyway. Both have an indirect object (dative) which, from
>the English point of view, is quirky.
But do either of them have a subject?  Apparently the Active one is usually
analyzed has having a 1st person subject, and the Passive one is analyzed as
being impersonal.  But why?  I just took your word for it.  About Latin and
Welsh, and some other things, it seems to me that you are "a usually reliable
source", as the saying goes; I am prone to question you, if at all, not because
I doubt you, but because I want to know something you haven't said, such as
the reasoning or causes behind what you do say.

>>He quotes Zaenen, Maling, and Thráinsson (1985)* as proving Þeim
>>is the grammatical subject of 19a, by the tests of reflexive
>>binding, subject ellipsis, subject-verb inversion, and raising.
>
>Does he? Surely a prominent mark of the subject in Indo-European
>languages is that they control verb agreement;
Not in all I-E languages, if I remember correctly.  I believe someone on this list
mentioned a Scandinavian language which he or she speaks, in which the
verbs don't agree in any way with any of their participants.  So, triggering
agreement is not a necessary condition of subject-hood in every I-E language.

If there are any I-E languages with polypersonal agreement (any verb-initial I-
E languages would be the likeliest place to look for examples), then triggering
agreement would not be a sufficient condition for subject-hood in every I-E
language either.  I don't know of an I-E language with polypersonal agreement,
but there are several languages in which bivalent-or-higher-valency verbs
agree with two participants; several in which trivalent-or-higher-valency verbs
agree with three participants; a few in which a sufficiently-high-valency-verb
agrees with four participants; a couple in which a verb may agree with five
participants; and apparently at least one has been reported that sometimes
requires verbs to agree with more than five.  So, triggering agreement is not a
sufficient condition of subjecthood cross-linguistically, whether or not it is in I-
E languages.

But actually there is an I-E language, Hindi, which shows triggering agreement
is not sufficient.  In Hindi, perfective-aspect (or is it past-tense?) clauses
have ergative agents and nominative patients.  There is good reason to
believe that the agent is still the subject of these clauses.  But in Hindi the
verb agrees with the highest-ranking nominative participant.  If that happens
to be the subject (and it usually is), then the verb agrees with the subject;
but in most perfective-aspect (or past-tense?) transitive clauses, the subject
(agent) is in ergative case instead of nominative case, and the object
(patient) is in nominative case, and the verb agrees with the object (patient),
not with the subject (agent).

>this is so even if, as in Welsh, a language has abandoned all case
>markings, even for pronouns.
But clearly not in those that have abandoned all agreement marking on the
verbs.

>Does _Þeim_ control verb agreement in Icelandic?
I don't know.  Does anyone else on-list know?  I'd like someone to say.

>If it does not then it would seem that Paul R. Kroeger
>defines 'subject' differently to what I understand by the term
That's beginning to seem probable.

I assumed you knew ("knew" is the appropriate verb only if I'm correct here)
that there was, not long ago, a lot of disagreement about this.

Many linguists think all languages "have" subjects"; many linguists think some
languages don't.  But it's generally accepted that _most_ languages "have"
subjects; and it's generally accepted that some languages are _not_ "subject-
prominent".

Most linguists seem to think that most languages have at least two
grammatical relations; the first is "subject" and the second is either "primary
object" or "direct object".  But it seems to be accepted that some languages
(maybe only a few) don't have more than one; for instance, most of the
languages which have been controversially advanced as candidates for not
having _any_ grammatical relations, _uncontroversially_ don't have _more_
than one.

Most linguists seem to accept that some languages have a third grammatical
relation -- "indirect object" or "secondary object" -- and some don't, and for
some, even though all the evidence which will ever be collected probably has
already been examined, it's hard to decide whether or not they have a third
grammatical relation.

More confused/confusing, and longer-lasting, than the lack of clarity about
subjects, has been the lack of clarity about the difference between an adjunct
and an argument, and between an oblique argument and a direct argument
(a.k.a "core argument" or "term").

In 1976 the Academic Press published a book edited by Charles N. Li, written
by "the Symposimm on Subject and Topic, University of California at Santa
Barbara, 1975"  ("Subject and Topic" for short).  The tenth paper in the book
(starting on page 303) is "Towards a Universal Definition of 'Subject'", by
Edward L. Keenan.

Keenan says, in that paper, that the definition of subject varies from language
to language.  He also says that in any language, the question of whether or
not a sentence has a subject, and which participant is that subject, is much
clearer in "basic" sentences than in others.  (A "basic" sentence is a sentence
than which no other sentence is "more basic". In other words, it is maximally
basic.  A sentence X is more basic than a sentence Y, if X can be understood
without reference to Y, but understanding of Y implies, or pre-requires,
understanding of X.)

Since Keenan's paper, and possibly before it, there had been published
statements by professional linguists that "subjecthood" didn't have any cross-
linguistically necessary or cross-linguistically sufficient conditions.[*see
below]  Rather there was a "basket" of subject-like properties, such that in
each language, some collection of them was considered "relevant" by that
language, and most subjects in that language had most of those relevant
properties, and most things that had most of those relevant properties were
subjects, and the relevant properties tended to be correlated with each other
in that language.

(Thus, for example, even in a language in which every verb has to agree with
one and only one participant, it may not be the case that that participant is
necessarily the subject.  One would have to check that out.)

Keenan came up with a list, the "subject properties list".  He suggested that
any NP in any basic clause which had most of these properties, and also had
more of them than any other NP in that clause, was the "subject" of that
clause.

(For example, triggering agreement is on that list.)

See
http://www.chrisdb.me.uk/wiki/doku.php?
id=grammatical_relations&DokuWiki=7e05cdd3ebf481af1bf2f999167792c5#subje
ct
or
http://tinyurl.com/33lbq4
for Keenan's list, as near as I could make it out.

See
http://pluto.huji.ac.il/~msyfalk/Subjecthood/Typology.pdf
for a more modern take on the problem.

>(of course it may be that _Þeim_ does control the verb form unlike
>_Ihm_ in your German example or _eis_ in my Latin  example - but you
>do not say so or give examples to show this).
I'm sorry, I don't know.

>If we are defining 'subject' differently then quite clearly our
>definitions of the 'nominative case' must be different.
Well, my definition above:

|In a morphologically nominative/accusative language, the nominative
|case is used for the subject of nearly all basic monovalent clauses,
|and for the agent of prototypical transitive clauses, and for the
|donor of nearly all basic ditransitive clauses.

differs from Trask's mainly in saying "nearly all", "basic", and "prototypical", at
various points.

>>But he shows Ihm is not the subject of 49b, because it fails the
>tests of
>
>I agree 100% about Ihm    :)
>
>>To me, the only fly in the ointment, is that I cannot find where he
>>proves (or refers to someone else’s proof) that þeim and ihm are
>>the direct objects of 18a and 49a respectively. That's just one of
>>the reasons I want an "Object Properties List" similar to
>>Keenan's "Subject Properties List".
>
>But then we shall indeed need to define 'object' very carefully.
>There are different sorts of objects.
Yes.  I was particularly interested in a list of properties that would allow one
to conclude, if a given NP was not the subject, but had more of the properties
on the list than it lacked, and also had more of them than any other NP, then
it must be either the "Primary Object" or the "Direct Object", depending on
which the language has.

>For example, as Jeffrey Jones pointed out in an email yesterday
>Direct Object and Primary Object are not the same things.
Right.

>If one understands by 'object' "a generic term for any noun phrase
>occupying an argument other than the subject" [Trask], the object is
>in effect defined simply as being 'an argument which is not the
>subject.'
I wasn't quite doing that; Trask's definition as quoted above would allow
oblique arguments to still be considered "objects".

My definition was
"'object': a generic term for any noun phrase occupying a grammatical relation
or a _direct_ argument or a _core_ argument other than the subject; the
object is in effect defined simply as being 'a _term_ which is not the subject.'"

>But clearly different types of object need defining.
Yes, and I'd also like to define "indirect object" and/or "secondary object" in a
cross-linguistically useful way.

But having read Patrick Farrell's "Grammatical Relations", as well as some other
literature, I've come to the impression that many linguists feel that for
_direct_ objects and/or _primary_ objects, this task is imminently doable, and
perhaps has already been done.

>Ray
[*this is the "below" that I asked you to see]
There is a school of linguists, and I think it is the Lexical-Functional
Grammarians to which Paul R. Kroeger says he belongs, which would say that
the question of which of these properties is/are relevant, and which are more
or less important than which others, varies not only from language to
language, but even from one type of construction to another type of
construction within a language.  In his book that I have I believe Kroeger tried
not to be biased in favor of LFG; but in his other work I think he'd admit he is.

Thanks again, Ray.  I apologize for the misunderstanding(s) I hope I have (at
least partially) cleared (at least some of) them up.

Replies