another Hocank/Helmbrecht article question

Thu Jul 18 14:37:07 UTC 2002

This gets a bit long and messy... perhaps a key will help.  SO -- Johannes'
words have >, John's replies have nothing, and my (Catherine's) replies
have ##.

> Now, I would like to apply the criteria, Catherine mentioned in her
> contribution to the observations in Hocank. Catherine said that if the
> head noun were always the first constituent in the RC the head noun
> would be external. I do not understand why this is a criteria for
> external headedness, but if this is so, well, as I showed above, then
> this is the case in Hocank.

I think Catherine waffled on this to some extent herself.  I suppose the
issue is that if the head is at the margin and you're trying to justify
putting it inside the margin, it helps if you can show that something
clearly within the margin is "outside" of it (outer-more than it??).

##Yes, exactly.  Pam's question about "what happens when there is a noun
subject on an object-headed RC in Hocank (e.g. 'the bear (that)the man
saw....')" is a good example of this.  If it can be "man-bear-saw-def",
corresponding to a normal sentence "man-bear-saw" meaning 'the man saw the
bear', it strongly indicates that the head "bear" internal to the relative
clause.  (I won't say it absolutely PROVES it -- there's the remote
possibility of some sort of topicalization of "man" (as for that man, the
bear he saw was...) etc. -- but it's a very very strong indication,
especially if it's a semantically/pragmatically unmarked order.) On the
other hand, if it turns out that only the order "bear-man-saw-def" is
possible, it again doesn't prove anything, but it does make an
external-head analysis more likely.  The head "bear" could be obligatorily
first because it is actually outside the clause, or it could be that, for
information-structure reasons (the head being the topic of the relative
clause, presumably)it has to come first but is still within the clause.
      There -- that's completely muddied the waters, right?  John said it
more clearly:  the issue is whether something clearly belonging to the
clause ("within the margin") can precede the head or not.   Looking at the
examples people have brought up:

ku'=niNk=(g)a, [hiaN'c^=ha=ra  ware=hu'= iNgigi'           ]=ra
o grandmother   father  my DEF work come he made me his own  DEF

tuuxu'ruk=    s^aNnaN.
I finished it DEC
Grandmother, I have finished the [work for which my father sent me here].

##This one looks to me like a clear example of something preceding the head
-- ware 'work' certainly seems to be the head, as John said, unless the
gloss is somehow misleading.  (By the way, John -- I don't see why this
would be "strictly speaking a noun clause" ... what do you mean?)  So this
example, assuming head = ware, seems to argue for a head-internal
structure.

>..word order is pretty fixed in Hocank RCs. The order is always  Noun Head
- Predicate - Determiner and this >order exactly replicates the order in
the ordinary noun phrase in Hocank except that the determiner is not
>obligatory. Permutations in this order are not accepted by Hocank
consultants. In (1)a there is an example of a >transitive clause including
a RC modifying the subject noun huNuNc^ 'bear'. In (1)b-c, it is shown that
the >predicate of the RC cannot be moved before the head noun, and it seems
to be the case that the adverbial particle >gojá 'over there' needs to
appear betwen head noun and predicate. What is possible is the permutation
of the >whole RC behind the predicate of the main clause, cf. (1)d.

>(1)a    huNuNc^-zí-   ra       gojá         hac^a-rá
hiN-nuNxé-jiree-naN
>           bear-   brown-DEF  over there I.saw.it-DEF
me-chased-started-DECL
>           The brown bear I saw over there started to chase me.
>(1)b    *gojá hac^a-rá huNuNc^-zí-ra ...
>(1)c    *hac^a-rá huNuNc^-zí-ra  gojá
>(1)d    hiN-nuNxé-jiree-naN        huNuNc^-zí-   ra    gojá
hac^a-rá
>           me-chased-started-DECL bear-brown-DEF   over there I.saw.it-DEF

>           He started to chase me, the bear I saw over there.

##Now, this set, on the other hand, seems to lean toward a head-initial
structure (i.e. EITHER external head or else "topicalization" (or
something) of the head within the clause)  The interesting example is (1b).
I assume that Goja' hac^a-ra' huNuNc^zi-ra hac^a-X is a perfectly fine
sentence? (With "X" being some main-clause verb ending instead of -ra';
meaning "I saw the brown bear over there.")  If not, then the example is
irrelevant -- there's some independant reason why the adverbial can't
precede huNuNc^-zi-ra, having nothing to do with its status as head of a
relative clause.

> The second criteria, Catherine mentioned is the status of the head
> noun with respect to definiteness. As can be seen from the examples,
> the head noun almost always carries a definite article.

Which makes it useful to know for sure if we know that -ra marks
definitenes.  Suppose it marked something like referentiality/specificity?

##Yes, this is absolutely crucial.  Anyone have any ideas how to find
out???

> ... but I have the impression, that the head noun is still definite,
> even if the -ra is missing. So, if it is correct that internal heads
> of RC are indefinite, than the Hocank head noun of RC are clearly
> external. But I have to admit, that this criteria is not clear to me
> either.

I think maybe what Catherine is getting at here is not that the head is
indefinite as such, but that the definiteness of the head is marked on the
clause as a whole, providing an argument for internal status of the head.

##Yes.  In languages with internal-headed relative clauses, the head is
marked with an indefinite determiner (or none) EVEN IF IT IS SEMANTICALLY
DEFINITE.  There's usually a clause-external determiner which marks the
semantic definiteness or indefiniteness of the entire construction (and
thereby, of the head).  John Boyle's recent paper (from the Spearfish
meeting) shows this very nicely in Hidatsa, and it's well known in Lakhota
and a bunch of other languages, from various families ... Some syntactic
analyses of internal-headed RCs involve raising the head to the position of
the external determiner or coindexing it with that determiner to account
for its semantics.  This is actually very similar to what Johannes shows in
the next example:

>Now, if they [positional demonstratives including je'ga 'standing' in 2
--CR]  are used as subordinating forms as >suffixes to the embedded verb,
they always classify the head noun, no matter which semantic/ syntactic
role this >constituent may have in the RC, cf. the example in (2)

>(2)    huNuNc^- rá       gojá         hac^a-jéga
>        bear-      DEF     over there I.seeing.it-DEM('distal'; standing)
>        hiN-nuNxé-jiree-naN
>        me-chase-  start-DECL
>        The bear I am seeing over there (standing) starts to chase me.

>In exampl (2) the bear is the direct object of the verb 'to see' in the RC
and the attr. demonstrative classifies the >bear as a standing one. If the
bear were head noun and subject of the RC this demonstrative had the same
effect. >This demonstrates the close syntactic bond of the elements within
the RC and the head noun, and it is a further >similiarity to the ordinary
NP where these attributive demonstratives have the same function (to
classify the referent >of the NP with regard to position and to mark the NP
as definite)

##Exactly!  But I wouldn't just say it demonstrates a "bond" between "the
elements within the RC and the head noun", I'd say it (may?) show that the
head noun IS ONE OF "the elements within the RC".  This example in fact
provides the strongest argument so far that the head is actually internal.
An attribute of the head is marked on the clause as a whole:  [[bear.
-ra-there-I.see.it]standing]  Which makes me wonder all the more about the
status of the -ra element.

And, if I am correctly anticipating the next step, we can argue that the
determiner itself is not the head if we can show that it is optional.

##Huh?  Sorry, John.  You've lost me here.  My fault, probabably!
Actually, I might well claim that the determiner (the clause-final one) IS
the head in the sense that the whole construction is a DP.  But that's not
the one Johannes is saying is optional, I think (???) And this is a
different "head" than the nominal head of a relative ...

 >As can be seen from the examples, the head noun almost always carries a
definite article. I browsed through my >notes to find examples with no
definite article on the head noun and I could find examples for this only
if the >embedded predicate has a attributive demonstartive of the type
shown in (2). In this case, the def article on the >head noun seems to be
optional

##What about cases where the whole construction is indefinite?  (A bear
that I saw...)

##Enough for one message!  Thanks for your reply, Johannes, and I hope
you'll write again when you are back home.  Great examples.  Once again, I
do realize that this issue wasn't relevant to your paper -- I was just
excited to see someone mentioning RCs at all!

Catherine