The Long-Distance Anaphora Conspiracy: The Case of Korean

Kook-Hee Gill
University of Edinburgh

In the typological literature, e.g. Li & Thompson(1976) it is frequently observed that languages such as Japanese, Korean and Chinese present a certain cluster of common properties. Two particularly important of these properties are their so-called Topic-orientedness and their ability to have recursive specifiers of TP (or VP), i.e. the ³Multiple Nominative Construction². These languages also consistently exhibit long-distanced anaphoric patterns (ziji in Chinese, zibun in Japanese and caki in Korean). The aim of this paper is to show that the third property of these languages is best explained in terms of the interaction of the first two rather than in terms of the alleged high context and discourse dependency grammatically encoded in various forms (Perspective (Iida 1996), Logophoricity(Sells 1987) or Point of view (Kuroda 1973; Zribi-Hertz 1989)). We will base our argumentation on the behaviour of the Korean long-distance anaphor, caki, but our results naturally extend to the similar facts with zibun in Japanese and ziji in Chinese. Observe first the distribution of caki : it can take an argumental antecedent across a number of clause boundaries(1a) but can also be bound by a Topic in a A¹-position (1b).

(1)a. 	Johni-i   Billj-ekey Maryk-ka  cakii/j/k-lul cohahanta-ko malhayssta
	John-NOM  Bill-DAT   Mary-NOM  self-ACC      like-COMPL      told
	ŒJohni told Billj that Maryk likes selfi/j/k.¹
   b.  	Johni-un    ttal-i     cakii-pota   ki-ka       te        kuta
    	John-TOP  daughter-NOM self-than    height-NOM  more   {is tall} 
	ŒAs for Johni, (his) daughter is taller than selfi.¹

On the other hand, (2) illustrates an exchange between two speakers A and B. In B¹s utterance Caki occurs without an overt sentence internal antecedent.

     
(2) A:	Mary-ka  Ku  pati-e    kass-ni anim tarun salam-i    taysin   kass-ni?
	Mary-NOM the party-to  went-Q  or   other person-NOM instead  went-Q
	ŒIs it Mary who went to the party or somebody else instead?¹
    B:	Ani, caki -ka   kasse
	No,  self-NOM   went:  ŒNo, self went¹

In order to account for this distribution without making reference to the alleged ability of caki to refer directly to the discourse context we propose that caki -binding is very closely related to topicalised constructions which in turn, as we argue, are closely related to the null topic constructions and share the same underlying structure with double nominative constructions (3)

(3)a.	John-i/un     ton-i      issta
	John-NOM/TOP  money-NOM  exist:  ŒJohn has money¹

We argue that the binder for caki in (2) is not a topic in discourse, but actually an unrealised topic which originates in the outermost nominative position in an underlyingly multiple nominative structure, in other words the topic that binds caki in these instances is always in an A-position. The central evidence for this approach is twofold, syntactically it comes from the otherwise unexplained fact that caki can never be topic-marked itself which suggests that with the topic position already filled, had caki been topicalised, there would be no available position for its antecedent as in (4) (notice that Korean restricts the number of topics per sentence to one).

(4)	* Caki-nun kasse
	  Self-TOP went: ŒSelf went¹

Semantically on the other hand we observe that the topic antecedent of caki must be in a certain kind of relation with the nominative-marked element in the sentence in order for the binding of caki to be licensed. The relations in question are precisely the ones of possession, inalienable possession, part- whole, identity, and few others which are precisely (and only) the ones that must hold between the outer and inner nominative in a double nominative construction. It then follows that topic-binding of caki can only occur in a double nominative construction. Also, given that these relations must hold in double nominative constructions, the ungrammaticality of cases where a topic phrase inappropriately occurs, predicting wrong binding patterns as in (5), is immediately explained away with no further stipulation:

(5)	*John-un  Bill-i    ku  chayk-ul  caki-ekey  cwuessta
	John-TOP  Bill-NOM  the book-ACC  self-DAT   gave
	Œ*As for John, Bill gave the book to self.¹

In the last part of the paper we attempt to apply the same reasoning to languages such as Dutch and Icelandic which present in a sense minimal variation as they show long distance anaphora and allow transitive expletive constructions (a version of the Japanese/Korean double nominative construction). In conclusion, we have shown that a particular interaction of seemingly unrelated properties of the grammar of certain languages neatly explains the common patterns of anaphora, which in turn seem to need no extra stipulations in order to be explained away.

< Back to List of Abstracts | Back to PLC23 Home Page >

About the PLC23 Committee
Previously held Penn Linguistics Colloquium: PLC22 (1998), PLC21 (1997)

Penn Department of Linguistics
University of Pennsylvania