The effect of oral corrective feedback on the article errors in l3 English: Prompts vs. recasts

This study examines the effect of prompts and recasts in providing CF for the article errors by Kurdish-Arabic bilingual adolescents who learn English as a third language. For this purpose, 39 lower-intermediate Kurdish-Arabic bilingual adolescent learners of L3 English were tested on three tests: pre-, post-, and delayed post-tests involving articles in four contexts; [+def, +spec], [+def, -spec], [-def, +spec], [-def, -spec], in L3 English. The participants were randomly put into three groups: (1) prompt group (n =15), (2) recast group (n = 14), and (3) no feedback group (n = 10). Each group completed 28 dialogues, which included articles in a forced choice elicitation task as a pre-test. The same test was given to the three groups as post- and delayed posttests. Between the pre-test and the post-test, the prompt group and the recast group took a treatment. The treatment involved an interactional activity (dialogue build) which aimed the FCET, eliciting the students to make dialogues in which the former took CF in the form of prompts, and the latter took it as recasts for their article errors in L3 English during the activity. Results of the FCET showed that both prompt and recast groups developed on the posttest. Whilst, prompt group significantly outperformed recast group on the long term, recast and no feedback performed equally at delayed posttest. A two-way ANOVA of the accuracy data showed that there were no significant differences among the three groups in the pretest and posttest. However, there were very significant differences between the prompt and the other two groups in the long term in which the prompt group outperformed the other two groups. Analysis of the error data revealed that the errors mostly fell in substitution, which indicated that the students were fluctuating between choosing the correct article to use in all the semantic contexts especially in [-def, +spec] context. Given the Fluctuation Hypothesis, L1 speakers of non-article languages will either fluctuate between definiteness and specificity when learning a language that encodes the features [+definite] or [+specific] or will select the appropriate value for the target language (Ionin, 2003, p. 23). However, L1 Kurdish and Arabic both are [+article] languages based on definiteness. Our results suggest that a specificity-based choice of articles can occur even when the L1 is an article-based language which is based on definiteness. Prompt group significantly made fewer errors than the recast and the no feedback groups in the long term.