In this study, I analyze how Japanese native and non-native speakers use referring expressions in Japanese narratives. The data were elicited by having informants (20 native, 24 non-native) tell the story of an animated film to a friend in Japanese (Chafe 1980, Fuller and Gundel 1987). I demonstrate that native speakers deviated from the general tendencies of referential choice proposed by Givon (1983), Gundel et al. (1993), etc. in order to create hierarchical structure in their narratives.
The native narrator in (1) describes the episode of asking the mother for help and her response in a-c, and the child's reaction in d-e. The topic sita no ko 'younger child' introduced in (1) a is ellipted in (1) b and c, but is reintroduced with a full NP form at the transition to a new episode in (1) d, even though ellipsis would not cause any ambiguity.
(1) a. Sositara sita no ko ga outi no naka ni itte,
then the younger child went into the house, and
b. sinbun yonde ru okaasan ni nanka totte mitai ni itta
kedo,
(she/he) said to the mother who was reading a newspaper
something like "get (the kite) down," but
c. torenai wa: mitai ni iwarete,
(she/he) was told like "(I) can't get (it) you know", and
d. koo sita no ko akiramete
in this way the younger child gave up, and
e. dete itta no.
it's that (she/he) went out.
I conclude that native speakers deviated in their referential choice with full NPs to create major transitions and ellipsis for minor transitions in their narratives. In contrast, many non-native speakers used NPs at every topic shift and did not distinguish between main and sub characters. Results of this study have direct applications for teaching JSL students how to tell stories in Japanese.