2.6 Corrective Feedback
2.6.3 Research on Corrective Feedback in Language Learning
2.6.3.3 CALL systems
Here, I will discuss three research studies on CF provided in CALL systems. Heift (2004)’s is the first one, followed by Nagata (1997), and the last one is Ferreira (2006). Heift (2004) carried out an experiment to investigate which one of three CF types is the most effective on learners’ uptake in a parser-based CALL system. The CF types are metalinguistic, metalinguistic + highlighting, and repetition + highlighting. The highlighting technique here is referred to elicitation feedback as given in a normal classroom. The experiment uses a CALL system known as E-Tutor which is a system to practise various exercises on German vocabulary and grammar. There are a total of 177 participants for this study. A pre-test is given to all participants, then all participants use the CALL system for 15 weeks. After that they take a post-test. During the E- Tutor usage, at least four chapters were taught. Each participant randomly receives
one type of CF utterance, that is either metalinguistic, metalinguistic + highlighting, or repetition + highlighting for each chapter. This means, all participants receive each CF type at least once. The CF is provided as responses to the participants’ incorrect responses.
Results show that the group which received metalinguistic + highlighting feedback produced the biggest number of correct responses. The group which receives metalin- guistic feedback made fewer correct responses than the metalinguistic + highlighting group. A repetition + highlighting group produces the least number of correct re- sponses amongst the other two groups. The result also shows that those that receive metalinguistic + highlighting feedback are most likely to correct their errors (87.4%) compared to those who receive metalinguistic (86.9%) and repetition + highlighting (81.7%) feedback. Pair-wise comparison tests that are performed to determine an inter-group variation show there is a significant different between metalinguistic + highlighting and repetition + highlighting, and between metalinguistic and repetition + highlighting. Even though a pre-test and a post-test are given to all participants, no test results are reported in Heift’s paper. Overall, Heift concludes that the metalin- guistic + highlighting feedback is the most effective feedback at leading to learners’ uptake when learning a language using a CALL system.
Nagata (1997) conducted an experiment whose the objective was to study the effec- tiveness of deductive feedback and inductive feedback provided in response to Japanese particle errors, in a parser-based CALL system called BANZAI. Both types of feedback are of metalinguistic type. Deductive feedback utterances give detailed linguistic in- formation on particle errors and besides, the correct particle is also given. Inductive feedback is also the deductive feedback, with an additional of two or three examples of correct sentences that use the particles.
A testing procedure consists of six sessions over a duration of 15 days. The testing starts with a pre-test, followed by four computer sessions, and ends with two tests (a post test and a comprehension test).10 During the computer sessions, 30 participants are divided evenly into 2 groups. One group receives deductive feedback and the other
one receives inductive feedback in response to erroneous utterances.
A result of the post-test shows that the deductive-based group performs significantly better than the inductive-based group. Again here no results are yielded on statistical significance tests for a difference between the pre-test and the post-test score for each group. Overall, the deductive-based feedback provision is more effective than inductive- based feedback for language learning.
The third research is Ferreira’s study. This is also an experiment to investigate which CF is more effective between two CF strategies, that is provided when using a Web-based CALL system. Ferreira classifies the CF into two types: Giving-Answer Strategies (GAS) and Prompting-Answer Strategies (PAS). GAS consists of repetition and explicit correction and PAS includes metalinguistic and elicitation feedback re- spectively. The target structure learnt in the CALL system is Spanish subjunctive mood. The research question is Are PAS or GAS feedback strategies more effective for teaching the Spanish subjunctive mood for foreign language learners?
In this study there is a total of 24 participants, who are divided: two experimental groups and one control group. The experiment consists of a pre-test, treatment sessions, and a post-test. The pre-test is given to all group participants, then all groups use the CALL system during the treatment sessions. In response to incorrect answers, the first experimental group receives PAS feedback, and the second group receives GAS feedback. The control group receives only positive and negative acknowledgment in response to their answers. The sessions end with a post-test which is conducted at the end of the treatment activities.
Statistical significance tests are performed on the differences between pre- and post- test scores in the different groups, to see whether there are more gains in some groups. The gain of the PAS group was found to be significantly greater than the gain of the control group, and also (marginally) significantly greater than the gain of the GAS group (which was itself marginally significantly greater than the gain of the control group).
Table 2.3: The tractability of different kinds of CF.
Notice Locate Perceive Uptake Repair
1 Elicitation directly directly easy yes easy
2 Metalinguistic directly directly easy yes easy
3 Explicit Correction directly directly easy least easy
4 Repetition directly directly difficult yes easy
5 Clarification Re-
quest
directly indirectly difficult yes difficult
6 Recast indirectly indirectly difficult least easy