|Date: Tue, 30 Mar 2004 13:13:51 -0600
From: Eric McCready
Subject: Ellipsis and Reference Tracking in Japanese
AUTHOR: Nariyama, Shigeko
TITLE: Ellipsis and Reference Tracking in Japanese
PUBLISHER: John Benjamins
Eric McCready, Department of Linguistics, University of Texas at Austin
This book considers how the reference of ellipsed nominal arguments are
recovered in Japanese. Since ellipsis is ubiquitous in Japanese, an
answer to this question is important not only theoretically but in
practical applications such as machine translation systems. Nariyama
provides a detailed overview of various mechanisms at work in reference
recovery, which are finally integrated into an algorithm for
The book consists of seven chapters, footnotes, references and an
index. The first two chapters describe general issues related to
anaphora resolution, focusing on Japanese, and provide an overview of
the book. Chapters 3-6 describe a range of facts that influence the
resolution of ellipsed arguments. Chapter 7 integrates these facts into
a general algorithm. In the following each chapter is described in
turn. I defer most critical discussion until after the summary.
Chapter 1. 'Introduction'
As its title suggests, this chapter introduces the phenomena which the
book is concerned with, primarily instances of zero anaphora within
narrative and expository texts. Some general issues relating to
anaphora and ellipsis are also discussed: the different types of
anaphora (intra- vs. inter-sentential) and some possible functional
motivations for the use of ellipsis.
Chapter 2. 'Various approaches to anaphora'
This chapter is a critical review of some previous analyses of the
anaphora facts at issue. Nariyama begins with a general consideration
of syntactic approaches to the distribution of anaphors, and then
discusses pragmatic (Gricean) approaches to anaphora resolution in
somewhat greater detail. The last part of the chapter focuses on
analyses of Japanese anaphora resolution: in particular, the
centering-based theory of Kameyama (1985) and approaches designed
specifically for machine translation. It was a bit surprising to see
that more recent centering-based work on Japanese was omitted from the
discussion (e.g. Walker et al. 1993, Iida 1998). Some evaluation of
this more current research would have made the review more useful.
Chapter 3. 'Predicate devices: Argument-inferring morphemes'
This chapter and the next introduce constructions that allow an
interpreter to narrow the space of possible referents for an ellipsed
nominal. Chapter 3 considers morphemes and predicates which select for
arguments of a particular person, social status, or discourse status.
Nariyama also presents the interesting thesis that Japanese is an
inverse-marking language, although not such an extreme case as the
Algonquian languages, based on speaker judgements about the
naturalness of a wide range of examples.
Chapter 4. 'Sentence devices I: The principle of direct alignment'
This chapter is an elaboration of the inverse-marking idea introduced
in the previous chapter. Nariyama shows that, in general, sentences in
which an argument of higher animacy or discourse salience follows an
argument lower on these scales is infelicitous (compared to similar
sentences in which linear order respects animacy/salience hierarchies).
She argues that such sentences are ordinarily "restructured" (by
passivization, intransitivization, etc.) so that linear order of
arguments in the sentence becomes isomorphic to their position on the
hierarchies. Finally, she discusses some cases in which the
requirement for restructuring is overruled by other factors: for
example, when a nominal receives contrastive focus.
Chapter 5. 'Sentence devices II: The principle of argument ellipsis'
This chapter brings together the observations in chapters 3 and 4 and
begins to gather them into a concrete theory of resolution, which is
further elaborated in chapter 7. Nariyama also introduces a principle
which is meant to describe which arguments are prone to ellipsis:
essentially, those high on the hierarchies described above. To this
extent, the focus of this part of the book is different from the rest,
considering production rather than interpretation.
Chapter 6. 'Discourse devices: Ellipsis as the unmarked representation
While previous chapters focused on the resolution of intrasentential
null arguments, this chapter discusses the behavior of such arguments
in discourse. In line with theories of null argument resolution
proposed within centering theory, Nariyama states that null arguments
are interpreted as referring to the discourse topic, which corresponds
to the highest-ranked object on a totally-ordered list of salient
arguments. In cases where there is more than one null argument in a
clause, an ellipsed subject is coreferential with the topic, while a
second null argument is coreferential with the next highest argument on
the list. Nariyama does not consider instances of more than two null
arguments. At the end of the chapter, the analysis is extended to
discourse uses of the reflexive pronoun 'zibun' and null genitives.
Chapter 7. 'Algorithm'
This chapter refines the proposal in chapter 5 into a concrete
algorithm for reference resolution. The basic idea is that null
arguments are coreferential with the discourse topic (the most recent
wa-marked argument), unless information from the sentential and
predicate devices discussed in previous chapters provides conflicting
information. After proposing the algorithm, Nariyama works through
several examples in detail and mentions some possible problems and
areas where more work is needed.
The contribution of this book, as I see it, is primarily that it makes
available a broad range of data relating to structures that contribute
to the disambiguation of Japanese null pronouns. The discussion of how
certain predicates force, for instance, orientation of a sentence to
the viewpoint of a speaker, hearer, or third party is convincing and
interesting not only for anaphora resolution but for more general
problems in theoretical pragmatics. More specifically for the study of
resolution, the algorithm presented seems to perform well in capturing
the data Nariyama considers. However, it's not clear that it
generalizes to all examples, as she herself notes. There are also some
broader questions that arise as one reads through the book.
The status of the constraints she assumes is not always entirely clear.
The discussion in chapter 4 seems to indicate that Nariyama takes
constructions in which her direct alignment principle is violated to be
infelicitous (or even ungrammatical), suggesting that there is a
prohibition of such constructions in the grammar. However, she also
states (p. 228) that inverse alignment is possible when all arguments
are overt---that is, when the sentence includes no null pronouns. In
addition, sentences violating direct alignment are possible in answers
to wh-questions and focused constructions, as she notes. It also seems
fairly straightforward to construct examples in which lingustic
context, in the form of preceding discourse, allows the constraints to
be overriden. These facts suggest that the constraints she uses must be
defeasible. The discussion at times appears as if Nariyama herself
would agree with this position. However, in the actual algorithm, no
provision is made for overriding the constraints induced by direct
alignment or predicates (except for manual revision). Ultimately, it
seems plausible that a more flexible method than the strict algorithm
she proposes might do a more satisfying job.
Some claims made in the text could, I felt, have benefited from
additional exposition. To take one example, Nariyama argues that aspect
plays a role in determining reference: perfective aspect signals a
first person subject, while use of the imperfective signals that the
subject is not first person (p. 142). This constraint seems rather
implausible as stated. Again, while there may be something to it for
individual sentences (though I am not convinced), as soon as a sentence
is embedded in a discourse any preference here can be immediately
overriden. Perhaps more discussion would have made the idea clearer.
Another example is the claim that certain complex sentences are in fact
monoclausal (p. 164). It wasn't clear to me how such a situation can
hold. Again, more explanation would have been welcome here. A more
trivial complaint is the relatively large number of typos, especially
in the second half of the book.
In sum, although not every bit of the book was completely convincing,
it is a useful resource for scholars with interests in anaphora
resolution and Japanese syntax.
Iida, Masayo. 1998. Discourse Coherence and Shifting Centers in
Japanese Texts. Centering Theory in Discourse, Marilyn Walker,
Aravind Joshi, and Ellen Prince, eds. Oxford University Press.
Walker, Marilyn, Masayo Iida and Sharon Cote. 1994. Japanese Discourse
and the Process of Centering. Computational Linguistics 20-2.