Date: Fri, 18 Oct 1996 15:32:57 -0500 (CDT)
From: Larisa Zlatic <zlatic@ccwf.cc.utexas.edu>
Subject: Reply to May/Seely from Zlatic
To: LINCONF@tamvm1.tamu.edu
Message-Id: <01IATZ2XB7IQ8WZI3F@EMUVAX.EMICH.EDU>
Below is my reply to the questions from R. May and Daniel Seely.
> Message 1:
>
> Date: Thu, 17 Oct 1996 18:06:37 -0700
> From: Robert May <RMAY@uci.edu>
> Subject: Comment on Zlatic
Robert May wrote:
> I have a couple of comments, of a very general nature.
>
> 1. I'm not quite sure I see what role thematic roles play in the
> account. Isn't all that is relevant are considerations as to what are
> arguments of (nominalized) predicates?
ANSWER: Based on the data presented in the paper, the answer is yes. Only
semantic arguments count as the antecedents for the reflexives.
> Does it ever matter what the
> thematic role of the argument is?
ANSWER: It seems that it does. The following example shows that it is the
argument highest in the thematic hierarchy that binds the reflexive.
(1) Jovanovo (i) vracanje dece (j) svojim(i, *j)/njihovim (j) roditeljima
John's returning children self's /their parents
'John's returning of the children to his/their parents'
In the above example, in which there are two syntactically expressed
arguments of the deverbal nominal, vracanje 'returning', only the agent
argument (i.e. Jovanovo) can be bound by the reflexive possessive
'svojim'. The theme argument, dece 'children', can only be bound by the
non-reflexive possessive pronoun 'njihovim'.
> 2. It seems that the author simply takes subject-orientation as
> primitive, and builds it into the Binding Theory in (11). Is there any
> account presumed which entails subject-orientation for these elements in
> Serbian?
ANSWER: Unfortunatelly, yes. My binding theory simply stipulates that
reflexives are subject-oriented (i.e. SUBJECT-bound) while pronouns are
antisubject-oriented (i.e. SUBJECT-free). I thus treat orientation
effects as a primitive of the theory.
In my other binding paper titled "Process Nominals and Anaphor Binding in
Serbian", which will appear in the Proceedings of the 4th Formal
Approaches to Slavic Linguistics, I have shown that neither LF-movement
theory (e.g. Pica 1987, Hestvik 1992) nor the relativized SUBJECT theory
of Progovac (1992, 1993) are able to account for the Serbian binding
facts, and consequently, for the orientation effects. Pica's theory of
LF-movement of reflexives, which assumes that monomorphemic reflexives
(being X-zero elements) move to INFL at LF (a position from which
reflexives can be c-commanded by a subject) is unable to account for the
NP-internal binding of Serbian reflexives (i.e. binding inside process
nominals), since a nominal domain has no INFL node for the reflexive to
move to. Similarly, Hestvik's (1992) theory of LF-movement of pronouns is
unable to account for the binding of regular pronouns by local objects
(cf. (2) below).
(2) Jovan je ispricao Mariji(j) sve o njoj (j)/sebi (*j)
John AUX told Mary everything about her / self
'John told Mary everything about her'.
In (2), the object 'Mariji' c-commands the pronoun in the minimal binding
domain, i.e. a clause, hence, violating a Condition B, which according to
Hestvik, applies at both levels of syntactic representation: S-structure
and LF.
Finally, Progovac's (1992, 1993) relativized SUBJECT theory (which does
not involve an LF-movement of reflexives) is unable to account for the
contrast between binding in process nominals and binding in non-process
nominals.
> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> Message 2:
> from Daniel Seely (dseely@emunix.emich.edu):
>
>
> 1. Some very interesting data in this paper! I wonder though to what
> extent it really supports the need for thematic structure in binding.
> In fact, on one view, you give evidence against thematic structure:
> you argue that coargumenthood is not a relevant factor, and you give
> no evidence for thematic hierarchy effects. It's only "semantic
> argument" that plays a role.
REPLY: See my reply to Robert May (to his question #1).
But couldn't this be stated in
> structural terms? (For instance, the subject of a process nominal is
> in an A-position, and the "subject" of a non-process nominal is A'; then,
> BT is sensitive to relativized minimality?)
ANSWER: This seems like an attractive suggestion. However, based on
morpho-syntactic evidence, these two types of 'subjects' behave
identically: the same morphology, the same word order within the noun
phrase, the same possibilities for extraction out of the noun phrase (see
Zlatic 1994 on this), the same behavior with respect to ellipsis, etc.
Thus, I conclude that these two types of subjects must have an identical
syntactic structure.
>
> 2. I wanted to clarify a number of (probably minor) data points:
>
> A. You state that the reflexive _sebe_ must be locally subject-bound
> within a tense clause, and that a pronoun must be locally subject-free
> in this domain (your example #(6)). But I don't recall seeing the
> confirming anti-subject orientation example. I guess it would be
> something like:
>
> Jovan (i) je ispricao Mariji (j) sve o nje(j) (*i, j).
> John AUX told Mary-DAT everything about pronoun
> 'John told Mary everything about her'
>
> [sorry if I've got the actual pronoun wrong, but you see
> the point]. I just wanted to be sure on this point.
ANSWER: Yes, in the above example, the pronoun must only refer to an
object, showing that pronouns must be subject-free in a local domain. (See
also my sentence (2) above).
> B. I also wanted to ask about the apparent instability of some of the
> data. It seems to be the case that non-local binding of the reflexive
> in a process nominal is strongly out (see your example #(3), which is
> relevantly starred), non-local binding of the reflexive in a
> non-process nominal is strongly in (that is the key distinction), but
> that local binding of a reflexive in a non-process nominal is
> unstable. I just wonder if you could comment on the ?* status
> of (4) on index _j_. (I find this potentially interesting
> in light of comments made in the Everaert-Anagnostopoulou
> paper.)
Due to lack of space, I was not able to comment on this issue in the
paper. The grammaticality judgment in the example (4), i.e. the label ?*
on the reflexive 'sebi' means that the majority of my informants (8 out of
11) did not accept binding of the subject of the non-process nominal by
the reflexive. (Similar judgments are also reported in Bennett's (1991)
paper "Anaphora in Serbo-Croatian", McGill Working Papers in Linguistics
7, 1-25.)
I have a processing/pragmatic explanation for this. Subjects of
non-process nominals are ambiguous, i.e., have many interpretations. If
the speaker uses a reflexive pronoun embedded in these nominals, the
hearer searches for the most prominent (i.e. unambiguous) binder in the
sentence, which is a clausal subject in (4). Hence, the 'long' distance
binding (i.e. binding across specified subjects of non-process nominals)
is preferred. Within process nominals, a reflexive can only refer to the
nominal's subject since the semantic role of that subject is unambigous.
A similar explanation can be offered for the 'local' binding of pronouns,
i.e. binding by the subject of the non-process nominal, as in example (9)
of the paper. In (9) the pronoun is used in order to avoid ambiguity.
More precisely, by using pronouns, which carry more morphological
information than the reflexives (which make no number, gender nor person
distinction), the speaker facilitates the processing time of the hearer,
despite the locality domain violation. With process nominals, where there
is no ambiguity in the semantic role of their subjects, the pronoun is
avoided because the reflexive can be used unambiguously.
I hope this helps.
--Larisa Zlatic