We covered four studies this week: Oliver (1995), Lyster & Ranta (1997), Ortega & Long (1997), and Doughty & Varela (1998).
The comparability is always a thorny issue, but we could still be better informed about the role of feedback by trying to compare each study’s design and main findings from a single line. The following summary would be insightful from that sense:
-----------------------------------------------------------------
Oliver (1995)
Laboratory study
Descriptive study (No particular focused targets; No casual explanation)
* Ss: 8-13-year-old ESL learners (16 dyads)
* Two feedback types in focus: Negotiation vs. recasts
* Main findings: Negative feedback was given to 61% of the students’ entire error moves (that is, 39% of error moves were ignored). Negotiation tended to be used when the meaning is opaque to NS interlocutors, whereas recasts were common when the meaning was transparent but the form was problematic. “Negotiation seems to serve to make the picture clear, whereas recasts are like straightening the picture on the wall” (p. 473). NS responses tended to be affected by several factors such as the type and complexity of learner errors. Learners seemed to successfully incorporate negative feedback (35% of recasts were incorporated).
Lyster & Ranta (1997)
Classroom study
Descriptive study (No particular focused targets; No casual explanation)
* Ss: 10-year-old content-based French immersion learners
* 20 hrs classroom observation, 4 classrooms, 4 teachers
* 6 types of feedback and subsequent uptake moves (4 types of repair & 6 types of needs-repair) were in focus
* Main findings: Recasts were used most often (55% of error turns induced teacher recasts). 69% of recasts were unnoticed, only resulting in topic continuation. Of the rest 31% of recasts which led to student turns with uptake, 18% and 13% resulted in repair and needs-repair turns, respectively. Compared to recasts (which only led to simple repetition of previous feedback turns), elicitation and metalinguistic feedback were more effective in that they might cause student-generated repair. More important is the observation that elicitation and metalinguistic feedback did not interfere with the flow of communication.
Ortega & Long (1997)
Laboratory study
Quasi-experimental study (Particular focused targets)
* Other related studies: Long et al. (1998), Inagaki & Long (1998)
* Ss: 3rd semester Spanish (low-intermediate level adult learners) (30 dyads)
* Two feedback types in focus: Recasts (negative feedback) vs. models (positive feedback)
* Targets: Object topicalization & Adverb placement (Both were previously unknown structures)
* Pre-posttest control group design
* Main findings: Both types of feedback did not bring about significant learning of object topicalization. Recasts were more effective than models in learning of the adverb placement rule.
Doughty & Varela (1998)
Classroom study
Quasi-experimental study (Particular focused targets)
* Ss: 11-14-year-old content-based ESL learners
* Targets: past tense –ed & conditional would
* One feedback type in focus: Corrective recasts
* Pre-post-delayed posttest control group design
* Main findings: The focused recasts led to substantial gains on oral mode tests and the beneficial effects were maintained 2 months later. Gains on written mode tests were less robust. FonF is feasible, but should be brief, immediate, focused, and not to be overused.
-----------------------------------------------------------------
After outlining the contrasting features of the four studies, we focused more on Ortega and Long’s (1997) quasi-experimental laboratory type study. We could be aware of the detailed experiment procedures by listening to actual task samples. (It was of particular interest to see that the GJT used by the researchers was not a conventional, decontextualized one).
In light of the study findings, even though the two targets were presumed to be at the same developmental stages, only the adverb placement rule was learned when recasts were provided to Spanish learners. Object topicalization rule might have been too difficult to observe the expected learning outcomes. By contrast, the adverb placement rule, which is more related to lexical items, could have been more learnable. What was interesting from the follow-up interview data was the fact that some learners, although they could not reply to the question as to what they actually had learned overall, tended to state the adverb placement rule accurately, indicating noticing of the rule truly occurred during the task performance.
An improvement of the study design could have been made with delayed posttest measures included. Also, rather than the repeated measures design, a combinatorial design with two feedback types and two target structures would have enabled us to have a clearer understanding of the roles of the two feedback types.
On Thursday, we focused on two of the four studies. Oliver (1995) and Lyster & Ranter (1997) that we covered are all descriptive recast studies. In pairs, we tried to find answers to the following five questions. I am including the answers and ideas we shared as a whole-class discussion format:
1) How frequent was negative feedback in each study?
* Oliver (1995; hereafter O): 38.82%
* Lyster & Ranta (1997; hereafter LR): 62%
* Different task conditions as well as idiosyncratic participant characteristics of the two studies may have resulted in this huge difference in the amount of feedback. For example, in the case of Oliver (1995), it was kids who gave feedback to their peers and the study was laboratory-based, which might have affected the reported smaller amount of negative feedback.
* LR: There seems a difference across teachers in the amount of negative feedback.
* O: Providing raw frequency data (instead of percentage values) would have been more desirable.
2) How was negative feedback provided (how many different ways, what range of explicitness) in each study?
* O: 2 types; negotiation and recasts
* LR: 6 types; recasts, elicitation, clarification request, metalinguistic feedback, explicit correction, repetition; these 6 types of feedback may exist on the implicit-explicit continuum.
* It came to be acknowledged that recasts may take an explicit form.
3) What evidence does each of the two studies consider in order to talk about "effectiveness” of negative feedback? (Do they talk about “effectiveness,” and if so what arguments do they consider?)
* LR: Effectiveness is discussed in terms of the extent of uptake and repair. Particularly, student-generated repair, which is different from uptake and simple repair, is important in telling the effectiveness of negative feedback.
* O: Feedback is available and usable. Negative feedback seems effective from the point that learners tended to incorporate NS’s feedback in subsequent turns (e.g., 35% of recasts were incorporated).
4) How did participants respond to negative feedback in each study? (what did they do with it, if anything?) -- Skipped
5) What did each study have to say about type of error?
* O: Detailed categories of grammatical errors were identified. Recasts were significantly more common than negotiations for errors in singularity, plurality, and subject-verb agreement. For other cases (e.g., Aux/copula, pronoun, word order/omission, word choice, no subject), negotiations were the preferred feedback option.
* LR: Grammatical errors (50%) are the most common error type, followed by lexical (18%), phonological (16%), and L1-related errors (16%). Lexical and phonological errors (80% and 70%) induced teacher feedback more than grammatical and L1-related errors (56% and 43%). This reminded us the findings from Mackey et al.’s (2000) study, suggesting the saliency and correctability issue in relation to each type of error. Recasts were the more preferred option across different error types, whereas negotiation of form was more common than recasts for lexical-related errors.
