The present study explored a reading-based speaking task for its potential to develop EFL students' fluency. Based on an L1 speaking model by Levelt (1989, 1993), the task was developed in a way that students can maximize their fluency in a scaffolding-rich condition: one in the stage of message generation and the other in linguistic encoding. The use of reading texts for the development of speaking fluency was also hypothesized to induce authentic interaction among interlocutors. How the reading-based speaking task functions as hypothesized was tested with five Korean college students who participated in a 4-week reading-based speaking program. The analysis of the group activity transcripts and interviews revealed that scaffolding in message generation took place in three types of language episode: a personal experience episode, a vocabulary-related episode, and a contentclarification/ elaboration episode. Linguistic encoding in L2 speaking was scaffolded via shared context, which enabled the participants to develop and utilize a paraphrasing skill. Finally, the participants’ engagement in speaking activities increased, affecting their willingness to communicate and their motivation to improve speaking.

This study investigated the relative contributions of linguistic knowledge and strategy use to L2 listening success, especially in bottom-up and top-down dominant listening tasks. Participants (n = 130) were Korean college students in a required listening course. The tested variables for linguistic knowledge were sentence processing speed, grammar, receptive vocabulary, and productive vocabulary. Listening strategy use was measured with a metacognitive awareness listening questionnaire. We hypothesized that linguistic knowledge will make greater contributions to Bottom-Up-Listening-Comprehension (BULC) than to Top-Down-Listening Comprehension (TDLC), and different aspects of strategies will be accessed in each comprehension type due to different psycholinguistic features of the tasks. A series of stepwise multiple regressions were conducted and confirmed our prediction. The unique variance explained by linguistic knowledge was 27.8% in BULC, but 22.4% in TDLC. Strategy items that address problem solving and mental translation were significantly related to BULC, while items dealing with directed attention and person knowledge had significant explanatory power for TDLC.