(7) Incremental anaphora resolution in Korean: Evidence from early gender mismatch

Poster session 1
Monday, September 5, 17:30
Jieun Kiaer
King's College London
jieunjoekiaer2003atyahoodotcodotuk
This paper argues that (i) anaphoric resolution is incremental in pro-drop languages like Korean and (ii) the implicitness of arguments doesn't cause delay or confusion in parsing, contrary to general assumptions. I adopt Dynamic Syntax (DS, Kempson et al. 2001) to capture incrementality found in both structure-building and anaphora resolution during the course of parsing. DS assumes that verbs in head-final languages project full template structure with semantically underspecified arguments, which will be updated during the structural update process. So, whether there are explicit argument expressions or not, DS allows parsing to take place incrementally without any delay. Such expressions, when explicitly introduced, are presumed to find the anaphoric values from the construction process. However, when they are not explicitly introduced, the values of the arguments are established from context.

To demonstrate incremental semantic update in Korean parsing, I devised a self-paced reading experiment (number of subjects: 25) using gender mismatch effects as a diagnostic for early (or incremental) anaphoric resolution of implicit arguments. I assume that if all such arguments are resolved non-incrementally, any gender mismatch between possible host NP and the implicit argument would not be detected at the early stage of parsing. However, if all such underspecification is resolved incrementally, delay caused by gender mismatch between the host NP and implicit argument will appear at the early stage of parsing. In Korean, a certain set of family relational words bear implicit arguments (i.e., her/his in her/his brother is often dropped) which are furthermore gender-sensitive (e.g., oppa 'brother' from a girl's perspective. hyung 'brother' from a boy's perspective) As in (1), if a girl's name is used (Maria) without further context, oppa should be used to refer to her brother - not hyung. If such resolution is delayed, parsing (2) shouldn't cause early delay, as in (1). The result shows severe delay, however, right after parsing hyung 'brother (M)'.

This implies: (i) parsing implicit arguments is not delayed; (ii) anaphora resolution is, like structure-building, incremental in Korean. This is fully compatible with DS claims that structural properties of languages such as Korean are displayed through the on-line incremental parsing process.

Reference

Kempson, R., W. Meyer-Viol, and D. Gabbay (2001) Dynamic Syntax. Blackwell.

Examples
(1) No Delay
  Maria-ka oppa-ka linguistics-lul kongpwuhanta-ko haysse.
  Maria(F)-NOM brother(F)-NOM -ACC study-COMP said
   
(2) Early Delay after Gender Mismatch
  Maria-ka hyung-i linguistics-lul kongpwuhanta-ko haysse.
  Maria(F)-NOM brother(M)-NOM -ACC study-COMP said
  'Maria said that her brother studies Linguistics.'