This study investigates how task complexity, learners’ L1 backgrounds and proficiency levels influence the ways in which learners introduce and maintain referential topics in L2 Japanese oral narratives. Narrative discourse produced by two groups of learners of Japanese, one whose L1 is typologically parallel to and one whose L1 is distant from Japanese (Korean and English, respectively) was investigated at two different proficiency levels (intermediate and advanced levels, based on the ACTFL guidelines). Native speakers (NS, hereafter) and non-native speakers (NNS, hereafter) of Japanese constructed two narratives looking at a series of pictures, and reconstructing a story after seeing a silent film. The former task was supposed to elicit narratives in a ‘here and now’ context (H/N, hereafter), while the latter, in a ‘there and then’ context (T/T, hereafter). The latter task is considered more complex than the former, as in the T/T task, the speakers would need to code, store and search the content of the story in viewing the film as well as re-telling the story, and thus it is cognitively more challenging than the H/N setting.
The results showed that different narrative tasks brought out differing topic management patterns in both NS and NNS narratives. Especially notable was that the task complexity affected the L2 Japanese narrative discourse produced by lower proficiency level English speakers. The T/T narrative task revealed higher grammatical accuracy in English speakers’ narratives, such as the use of NP+ga’ and non-use of particle omissions. Yet, English speakers under-produced zero anaphora in their T/T narratives, whilst the production of zero anaphora reached target-level in their narratives in the H/N context. In contrast, Korean speakers showed consistent grammatical accuracy irrespective of task types or proficiency. It was argued that positive L1 influence was observed in Korean speakers’ Japanese L2 narratives in referential topic management, and negative L1 influence was observed in English speakers’ corresponding narratives, as Korean and Japanese share comparable linguistic features with respect to code topic continuity and discontinuity. While Korean speakers showed similar patterns of use of particles and zero anaphora to NS of Japanese throughout the narratives, their performance in the production of passive structures differed from NSs, and resembled the performance of English speakers. As proficiency increased, both learner groups produced more passive structures. The advanced level learners produced nearly twice as many passive structures in H/N than T/T, whereas task complexity did not have an influence on the number of passive structures produced in NS narrative discourse. Moreover, even advanced level learners in the H/N context produced fewer instances of passive structures than their NS counterparts. It is argued that this resulted from the different ways in which Japanese and English/Korean speakers place their focus in developing a story in their L1. Specifically, Japanese tend to place their focus on the main characters and tell a story from their viewpoints, whereas English and Korean speakers put their focus on action or fact. Japanese speakers in the current study kept main characters in the topic position; sometimes as agent in active structure, other times as patient in the passive structure. English and Korean speakers tended to switch the topic of the sentence between protagonists and antagonists, instead of keeping the protagonists in the topic position, and as a result, produced fewer instances of passive morphemes.