Effect of corrective feedback for learning verb second

Abstract

This paper presents a detailed study on the role of corrective feedback (CF) in the development of second language (L2) oral proficiency. Learners practiced speaking with a computer-assisted language learning (CALL) system that employs automatic speech recognition (ASR) technology to provide CF. The system tracks learner behaviour by logging the system-user interactions. Two language learning conditions are compared. In the CF condition learners received immediate, automatic CF on the grammaticality of their spoken output. In theOCF condition, learners practiced speaking with the option to self-correct. The target structure under investigation is Dutch verb second (V2) in the main clause. The results show that learner proficiency improved in both conditions. The CF condition shows an additional benefit for learning that is related to the learner’s initial knowledge of the target structure (which we call V2 proficiency). Learners at a lower V2 proficiency level benefitted more from practice with CF than learners in the NOCF condition. Learner evaluations are in line with these results: both the CF and the NOCF groups positively evaluated practice with the system, but the CF condition was preferred by learners starting at a lower V2 proficiency level. For more information on these outcome measures, we investigated the learners’ behaviour during practice. The two groups were found to receive equal amounts of input, but learners in the CF condition produced more (grammatically correct) output during treatment. We found that the CF group repaired their errors in fewer attempts as they progressed through practice. Learners in the NOCF condition generally did not (attempt to) repair their errors. However, the learners answered correctly more often as they progressed in the training. The log data, therefore, shows learning of the target structure in both conditions. We discuss these results and how learning outcome is related to learner behaviour.

Spoken grammar practice in CALL: The effect of corrective feedback and education level in adult L2 learning

“…While in the CF condition the higher number of spoken attempts and of grammatically correct utterances (OKs) had a beneficial effect in addition to the CF messages that alerted learners to errors, the absence of an additional effect of CF is remarkable. In a previous study with a similar setup the CF condition outperformed the NOCF condition (Penning de Vries et al, 2016). A crucial difference with that study is that those participants were high educated.…”

“…In meaning-based conversation, a V2 error will not result in communication problems. Our earlier study with high-educated participants working with GREET found that the CF condition was rated more positively (Penning de Vries et al, 2016). The current study showed a similar trend among HE learners.…”

“…Previous studies have shown the system to be effective: learners evaluated practice with the system positively, learner behaviour could be recorded to inform the proficiency test analyses, and learners improved their proficiency as a result of practice (Penning de Vries et al, 2015, 2016). These experiments were conducted with high-educated learners.…”

Intercultural competence development through a tele‐collaborative project supported by speech‐enabled corrective feedback technology

“…With the recent rapid advancements in machine learning and artificial intelligence, the field of technology‐assisted language learning has seen an increased focus on speech‐to‐text recognition and automated corrective feedback techniques (Aluthman, 2016; de Vries et al, 2016; Fichten et al, 2000; Konur, 2007; Kukulska‐Hulme et al, 2004). Speech‐to‐text recognition refers to a tool that can accurately transcribe speech input into text, which is then displayed on a whiteboard or computer screen for students (Ryba et al, 2006; Wald & Bain, 2008).…”