| While most researchers acknowledge that error correction (EC) is most effective in meaningful contexts, few studies have addressed collaborative EC or longitudinal language development during oral conversations---especially conversations where new knowledge is continually integrated. By observing how the tutor helped two college-age beginning students of Chinese learn three inter-related and chronologically-offset target grammatical structures (TG) during nine weeks of hourly one-on-one tutorial sessions, the study investigated: (a) the types of assistance the tutor provided in spoken conversation; (b) changes in this assistance within and across sessions; and (c) how errors towards TG were eliminated.; Analysis of protocols (transcripts marked up with visual cues), learners' questionnaires, and graphs revealed that: (a) the tutor provided two types of contingent assistance: regulation in participation (RinP), and EC on emergent errors; (b) EC was effective and its explicitness depended only on the learner's Zone of Proximal Development---same finding for RinP; (c) during the goals-oriented activity, language, serving both social (active and accurate meaning-exchange) and cognitive (tutor's EC and RinP, and learners' meta-comments) functions, was responsible for learners' transformation from other-regulation to self-regulation---language serving a cognitive function on an inter-personal level gradually became intra-personal; (d) RinP was instrumental in transferring not only the responsibility for participation (elaboration, initiation, and elicitation of TG) but also, through EC consequent to elicitation of TG, the responsibility for grammar-accuracy; and, (e) TG lacking an English counterpart required not only learners' cognitive understanding of the TG form but also where (which contexts) to use it---here, RinP efficiently co-constructed contexts for elicitation of TG and its differentiation, through EC.; In line with Vygostkian principles, the tutor's collaborative RinP improved learners' participation while the collaborative EC improved the learners' grammar accuracy within that improving participation. Implications include: (a) grammar accuracy is not an end-product but depended on not only task-difficulty and subject-matter but also degree to which similar TG were differentiated; and, (b) all errors, salient and not, must be corrected from the beginning---ignoring errors deemed "unimportant" was myopic. |
