1. Introduction

Errors unavoidably take place in language classes and they illustrate the point that learning is proceeding Furthermore, corrective feedback stands for “constitutes a reaction to learners' incorrect linguistic form in order to help them notice their incorrect utterance and correct it” (Zhang & Rahimi, 2014, p. 429).There are diverse approaches to error correction. Teachers have wide-ranging standpoints toward error correction, some consider that learners' errors ought to be overlooked and some others believe that they ought to be corrected instantly whereas others think that errors should be corrected implicitly and indirectly. Teachers utilize diverse ways of error correction to correct the learners' errors. They may give corrective feedback in explicitly or implicitly to rectify the learners’ erroneous utterances.  Based on Yoshida (2010), a teacher’s selection of corrective feedback type may well be subjective to their recognition of specific learners and the type of error. From the examination and analysis of the learners' errors, instructors can gather information about the nature of learners' knowledge at that particular point in their learning and apprehend what they still ought to learn (Abbasi & Karimnia, 2011). In the last decade, there always has been a developing awareness regarding the role played by corrective feedback in SLA .Some studies grounded on data obtained from classrooms (e.g., Lyster, 2004; Panova & Lyster, 2002) and on information gathered in a laboratory -type setting (e.g., Iwashita, 2003; Mackey et al., 2003; Philp, 2003)have inspected the type of corrective feedback given to learners and the degree to which the feedback is taken note by the learners, or uptaken, or both. Some experimental studies endeavored to look at the commitment that corrective feedback brings into acquisition (e.g., Leeman, 2003; Ayoun, 2004; Lyster, 2004; Han, 2002).

Corrective feedback varies in relation to the amount of explicitness or implicitness. When it comes to implicit feedback, no plain marker exists to show us that an error has been occurred, while in explicit feedback, there is. A point can also be mentioned for the commitment of corrective feedback types that are evidently corrective to learning by themselves. The autonomous inductive theory of Carroll (2001), sets that feedback can contribute to acquisition if and only if the corrective aims of the feedback are known to the learner. Moreover, learners should be capable of locating the error; Carroll mentions that “most of the indirect forms of feedback do not locate the error” (p. 355). Recasts for example don't explicitly indicate that the learner has made an error and should help in finding and localizing the error according to the following conditions: when the recast is full (i.e., there is a reformulation of the complete incorrect expression) or partial (i.e., just the incorrect portion of the expression is reformulated), as Sheen (2006) demonstrates. On the other hand, explicit feedback forms make the remedial drive or corrective power clear for the learner and at the same time provide hints as to the precise place of the error. In this way, they may well be more likely to persuade learners to perform the cognitive assessment and comparison between the target form and their error (Ellis, 1994) that is supposed to cultivate acquisition (Schmidt, 1994). Connectionist models moreover provide evidence and support for explicit error correction. Ellis (2005) makes a distinction between unconscious and conscious learning mechanisms, accentuating the role of awareness, consciousness, and attention in the former and of connectionist learning in the latter. He then suggested the sequence of learning (1):

Researchers in the field of L2 writing contend that CF can only hit the learners’ explicit knowledge. For example, Truscott (1998) notes that although grammar error correction can contribute to the explicit knowledge that is required for checking, monitoring, and revising the texts, it essentially ignores the “genuine knowledge of language” (Truscott, 1998, p.120) which can facilitate L2 advancement (i.e., implicit knowledge). In the same way, Bitchener (2012) indicated that written CF can lead to the advancement of explicit knowledge. Moreover, Polio (2012) anticipated that written CF will improve explicit knowledge.

As mentioned earlier, the role of written CF is fervently discussed. This discussion has brought forth some empirical and observational studies about written CF utilizing quasi-experimental designs to explore this question: is CF effective at all and in case it is, what kinds of CF are more successful and operative? These explorations drop into 3 main categories: (a) inquiries that have inspected the impact of CF on the reexamined writings of learners; (b) inquiries that have compared diverse types of CF—feedback on content Vs. feedback on form, indirect Vs. direct correction, underlining Vs. error codes—and (c) studies that have examined the impact of CF on new parts of writing in the course of time. As Truscott (2007) mentioned, the reality that learners are capable of reexamining and revising their papers does not guarantee that they would be able to exchange this ability for different piece of writing.

Undoubtedly, written CF inquiries have endured some methodological restrictions (e.g., the need for a control group in Lalande, 1982). Because of that, the findings of earlier research studies did not supply strong support that written CF assists learners improve linguistic accuracy as time passed by. Nevertheless, more new studies (e.g., Sheen et al., 2009; Ellis et al., 2008) have proven that written CF can uphold interlanguage improvement, which disputes the claim of  Truscott ( 1996, 1999 ) that the written forms of grammar error correction are destructive and ineffective. Researchers who examined written CF have made a comparison between indirect versus direct CF beside different methods of indirect correction provision and regularly, have given attention to a varied range of categories of linguistic errors. Some more new studies (e.g., Ellis et al., 2008; Bitchener & Knoch, 2008), though, have been arranged to explore focused CF—which is, feedback aimed at correcting and modifying a particular linguistic error. These inquiries have displayed that focused error correction results in a picks up in linguistic correctness and that the more focused and explicit the correction is, the more prominent the advantage for the learners.

Habitually, the factors of CF mode (inductive exemplar-based vs. Deductive rule-based) and indirectness of negative evidence (explicit vs. implicit provision of an error evidence) are conflated in studies assessing the viability of CF. For instance,  sheen (2006) contended that rule-based CF provides negative evidence explicitly whereas exemplar-based CF gives negative evidence of diverse degrees of implicitness in research studies comparing different types of CF. This conflation makes it bothersome to decide if it is the explicitness of negative evidence or the mode of CF which helps learning (see, e.g., the talk in Li, 2010). The present study endeavored to address the issue of conflated variables and evaluates CF mode alone. It does so by making comparison between 3 CF types all of which provide negative evidence explicitly and additionally prompt the revision of learners: rule-based metalinguistic CF, exemplar-based explicit corrections, and exemplar-based analogy-based CF.

These three  CF types are compared to each other on the acquisition of the third person singular ‘s’ in the writing samples of Iranian intermediate EFL learners; the motive for choosing this grammatical point is that Iranian learners don’t have this structure in their L1, so they sometimes face difficulties in applying third person singular ‘s’. Additionally, an attempt is made to see which one of these three types of corrective feedback is more influential in the acquisition of that grammatical rule. We hope the results of this study provide EFL teachers with useful outlooks regarding types of corrective feedback in teaching, and make them attentive to their practices in seeking professional advance in the ELT context. All in all to meet the objectives of the current study, the following research question was addressed:

2. Empirical studies

The role played by corrective feedback in the ESL/ EFL setting has been explored extensively (Swain 2006; Ellis et al., 2006).  Ellis et al. explored the impact of implicit VS explicit feedback types on learners’ capacity to advance the past tense in English. The outcomes of the delayed posttests showed that metalinguistic feedback was a better mediational tool than implicit feedback or recasts since the explicit feedback containing metalinguistic explanations may help learners to improve implicit and explicit L2 knowledge (Ellis et al., 2006).

More recently, a study was conducted by Thomas named “Comparing Explicit Exemplar-Based and Rule-Based Corrective Feedback: Introducing Analogy-Based Corrective Feedback” (Thomas, 2018, 1). She presented a framework to provide feedback to L2 learners which is named analogy-based feedback that's begun from parallel learning speculations and grammatical arrangement in conversation. Students were provided a comparative identical form of their products in which errors produced by them were rectified, and they must decipher and translate the analogy-based corrective feedback to get the rectification. A quasi-experimental classroom-based investigation was done to explore the adequacy of corrective inputs that were diverse in type (deductive rule-based or inductive exemplar-based) on agreement between subject and verb in English. Analogy-based and metalinguistic remedial feedback besides explicit correction was evaluated by untimed and timed grammaticality judgment tasks for ungrammatical things. The distinction wasn't seen between the study groups in terms of rule-based or exemplar-based corrective feedback. Descriptive statistics showed varied changes through numerous testing periods, wherever analogy-based feedback usually resulted in the lowest doing on the immediate posttest but probed progress on the delayed posttest, contrasting other forms of corrective feedback.

Masjedi and Tabatabaei (2019) worked on the study which is named, “Analogy-Based Corrective Feedback or Metalinguistic Feedback: Which One Is Better in EFL Classrooms?” They examined the adequacy of recast followed by an explicit correction that may be referred to as analogy-based corrective feedback and metalinguistic corrective feedback amongst Iranian students of English at the elementary level. Results disclosed that the analogy-based group did better in learning grammar as compared to the group that had been given metalinguistic feedback and also the control group (Masjedi & Tabatabaei, 2019, 513).

3. Methodology

3.1Participants

 Participants of the present study comprised three intact classes of female EFL learners (N=75, intermediate level) in one of the private English institutes in Khorramabad, Iran who were randomly selected. All the participants were considered to be at the intermediate level of language proficiency. Their age ranged from 18-22. No student was an English speaker, and none of them had the experience of living in an English-speaking country and Persian was their L1. Although the classes were considered homogeneous according to the result of a placement test administered by the institute, to check and make sure about the homogeneity of the participating students, the Oxford Placement Test (OPT) was run. Levene’s test was checked to see whether the variance in scores of the oxford placement test was the same for each of the three groups. Sig = .94 for Levene’s test, the Sig value is greater than .05; consequently, it can be determined that three groups were homogeneous in terms of their language proficiency based on their OPT scores. Moreover, there was not a statistically significant difference at the level p< 0.05 level in the OPT scores for the three groups.

     The participants were randomly assigned to three experimental groups, each containing 25 students. Each experimental group was exposed to one kind of corrective feedback e.g. explicit exemplar-based, rule-based, and analogy-based written corrective feedback. At the outset of the study, the consent of the institute and the participating teachers were obtained. 

3.2 Instrumentation

To meet the objectives of the current study, four instruments were used for collecting the required data.

3.2.1 Oxford Placement Test

For homogenizing the participants concerning their general English proficiency an Oxford Placement Test (OPT) was run before conducting the treatment. The paper and pen version of this test was used in the present study. It’s worth noting that, this test was designed by Oxford University Press and Cambridge ESOL and it can be used for all levels of language proficiency. This test contained 60 multiple choice items which lasted for about 30 minutes and it included grammar, vocabulary, and cloze tests. 

3.2.2 Multiple Choice Subject-Verb Agreement Pre-test

For evaluating the participants’ knowledge of subject-verb agreement before the treatment sessions, a test that focuses mainly on the third person singular ‘s’ was designed by the researchers. The items of the test were collected from authentic books available in the market as Touchstones, and other online sources for preparing learners for the IELTS exam as Interactive JavaScript Quizzes for ESL students by Charles I. Kelly. Moreover, they were checked by two TEFL teachers. The pre-test and post-tests all included 66 items on the target structure in the SVA.

3.2.3 Writing Task

The treatment in this study was in the form of pieces of writing that participants were demanded to deliver each session. Ten related topics that focused mainly on 3^rd person singular s were chosen meticulously by TEFL teachers. In every session the same topic was given to the participants in all experimental groups and they received the appropriate corrective feedback accordingly.

3.2.4 Multiple Choice Subject-Verb Agreement Immediate and Delayed Post-test

The post-tests (immediate and deferred) also contained 66 items targeting on SVA. As we ran both immediate and delayed post-tests in this investigation, the test re-test technique was taken into account. It means that the same test in the immediate post-test was administered as the delayed post-test two weeks after the treatment.

3.3 Procedure 

In addition to the placement test held by the institute, an Oxford Placement Test (OPT) was also run for homogenizing the students. Then, the learners were randomly divided into three experimental groups. The classes for all of the groups were held three sessions per week, each lasting for two hours.  Before the treatment, a 66-item achievement test was administered as the pre-test to assess the students’ prior knowledge of the grammatical structure under study which was third person singular s. The treatment lasted for ten sessions. The treatment sessions were managed in this way: in each session the teacher asked the learners to write a piece of writing on the specific topics that were meticulously selected by the researchers in textbooks and they were focused mainly on simple present sentences that elicit third person singular ‘s’. For any one of the experimental groups, the learners got a special kind of written corrective feedback on their compositions. Experimental group one was exposed to analogy-based CF (for instance, when a student writes or says, “My sister assist my mom in the kitchen,” the corrective feedback based on “analogy” probably respond: “More or less. You might say: your sister helps your mother. Now May you modify your statement with your original words?”), experimental group two received exemplar-based CF (An explicit correction probably react to the previous error: “More or less. Your sister assists your mom) and experimental group three took rule-based CF (Consequently, for the former error, Rule- based CF probably respond: “More or less. Your subject is singular now therefore you have to use a singular form of verb.”). After admitting different corrective feedback, the papers were given to the learners and they were asked to correct their grammatical mistakes by analyzing their structure and the feedback that was written for them. Finally, the 66-item immediate post-test was run right after the final treatment to test the learners on the target structure they had covered during the treatment sessions which lasted 30 minutes. Additionally, the delayed post-test which was almost like the immediate post-test was run to measure any possible changes in the learners’ achievement and performance. 

4. Results

This study attempted to investigate the efficacy of exemplar-based, rule-based, and analogy-based corrective feedback on Iranian intermediate learners’ writing based on analogical learning theories and syntactic alignment in dialogue. In order to answer the research question which investigates the differences among the immediate and delayed post-tests of three groups, first of all, a one-way between-groups analysis of variance (ANOVA) was conducted to study the effect of three different types of corrective feedback on the acquisition of third person singular ‘s’ in the immediate post-tests of the three groups. Table1 displays descriptive statistics to evaluate the mean difference among the post-tests of the three groups. As can be seen in table1, the results of descriptive statistics show the mean score for three groups (group 1, analogy-based corrective feedback, (M = 46.04, SD = 9), group 2, exemplar-based corrective feedback, (M = 48.68, SD = 6.47), and group 3, rule-based corrective feedback, (M = 50.56, SD = 5.18). Based on the results, the mean scores for the rule-based corrective feedback group were shown to be greater than the other two groups. 

Table 1

Descriptive Statistcs of Immediate Posttests of three Groups

To discern the statistical significance of the observed difference, an ANOVA was run and as can be discerned in table 2, the results of inferential statistics showed that the difference cannot be interpreted as significant (p=.71>.05) therefore we can claim that there is not any significant difference in the immediate posttest of EFL participants who were exposed to analogy based, rule-based and exemplar-based corrective feedback in terms of their acquisition of 3^rd person singular “s”.

Table 2

Analysis of Variances of Immediate Posttests of three Groups

The 2^nd  part of the research question was an effort to inspect the differences in the delayed posttest of the participants. To answer this, another one-way between-groups analysis of variance was conducted to examine the impact of three types of corrective feedback on the delayed post-tests. As can be discerned from table 3, the results of descriptive statistics showed differences in the mean scores of the 3 groups: the mean score of analogy-based corrective feedback (M = 52.68, SD = 3.70) was different from exemplar-based corrective feedback, (M = 46.44, SD = 9.43) and rule-based corrective feedback, (M = 46.76, SD = 6.29). 

Table 3

 Descriptive Statistics of Delayed Posttests of three Groups

To find out the statistical significance, an ANOVA was conducted. Results of table 4 show that there exists a statistically significant difference between the three experimental groups in the delayed posttest: F = 47.12, p = .00. Utilizing the Tukey HSD test for Post-hoc comparison showed us that the mean score of analogy-based corrective feedback was significantly different from exemplar-based corrective feedback (tables 5&6), to put it in another way, the analogy-based corrective feedback group outpaced the other groups; however, there was not any statistically significant difference between exemplar-based corrective feedback and rule-based corrective feedback groups. 

Table4

 Analysis of Variances of Delayed Posttests of three Groups 

 Table 5

 Multiple Comparisons of Delayed Posttests of three Groups

Table 6  

Tukey Test of Delayed Posttests

5. Discussion

As mentioned before, results of the immediate post-test comparison of the groups showed that in general, the analogy-based group did comparatively poorly on the immediate post-test, while the other two experimental groups inclined a rise on the immediate post-test, falling somehow on the delayed post-test. Also, the rule-based group scored marginally higher than the explicit exemplar-based group but the difference was not found to be significant, so the first null hypothesis which asserts that there is not any significant difference among the immediate post-tests of the three groups is accepted. Furthermore, in the 2^nd part of the research question, the mean score of the groups in the delayed posttest was compared and the analysis showed us that the analogy-based group outperformed the other groups, while there was a significant difference in the achievement of the learners of the rule-based and explicit exemplar-based group compared with the analogy-based group. This might also additionally mirror a preferred tendency for the analogy-based corrective feedback group to carry out at a suitable level relative to the other two groups on delayed post-test conditions. These orientations to more profound comprehension, construction and absorption may be connected to levels of processing proposed by Craik and Lockhart’s model (1972). This framework contended for a progressively structured cognitive framework where initial sensory breakdown and analyses are shallow and bolstered into future semantic breakdown and analyses that as a rule necessitate more thoughtfulness and attentional assets. Remembering For Lockhart and Craik (1990) is attained through recovery that includes a memory footprint of conceptual or perceptual analysis (cf. p. 89). They claimed that analyses including superior meaning abstraction (depth) influence more on retention and memorizing. In analogy-based CF Shared, analogical alignment can be assumed as an elaboration form, where about the learner’s authentic grasp of syntax is compared to the innovative equivalent form.

A surprising finding of this study was that the analogy-based group was inclined to drop below rule-based and exemplar-based groups on the immediate post-test and after that rise once more on the delayed post-test. This can be clarified by the claim made by Loaiza et al.’s (2011) that the consequences of more profound processing might seem later instead of instant, probably illuminating the decline in post-test 1. By way of explanation, this drop-rise pattern may show better rearrangement of morphosyntactic knowledge for analogy-based corrective feedback. The reality that the drop-rise design and structure are unambiguously observed with the analogy-based group and not with the other two groups appears to show that something different has happened here. This arouses the question of whether or not an extended interim between the 2 post-tests or a lot of corrective feedback together with distractors throughout instruction would have influenced the outcomes. Another feasible rationalization is that analogy-based corrective feedback was excessively puzzling and confusing and improvements on the delayed post-test might mirror the results of confusion step-by-step dispelling. Be that as it may, the very fact that analogy-based corrective feedback considerably assists acquisition for a few conditions opposes this justification. Fascinatingly, directed induction studies by Hwu et al. (2014) and Cerezo et al. (2016) failed to mirror the drop-rise pattern observed here. It’s going to be the fact that differing kinds of inductive learning procedures end in totally different depths of understanding and processing.

Analogy-based remedial input is more challenging than other forms of corrective feedback as learners must compare the two closely resembling expressions to find similitudes, discover the distinction and after that employ the abstracted design within the corrective feedback to the first original expression, manufacturing reformed output instantaneously to react to the modification provoke. 

Finally, this finding can be related to the role of corrective feedback in general and different feedback types in particular. In other words, this study endeavored to investigate whether analogy-based, rule-based, and explicit exemplar-based corrective feedback could assist L2 learners to acquire the grammatical point under study. The findings obtained from the study substantiated that analogy-based corrective feedback had a positive impact on the improvement of the learners’ grammatical knowledge in the delayed post-test. Further, it came to light that analogy-based, rule-based, and explicit exemplar-based corrective feedbacks are potent enough to make significant changes in the learners’ L2 grammatical knowledge. 

6. Conclusion

The position of Corrective Feedback within the second language acquisition (SLA) process may be an issue that's receiving heaps of analysis consideration. Sheen (2010a) attributed this continued research attentiveness in corrective feedback to, “the significance it carries for both SLA theory building and language pedagogy” (p. 177). It is claimed that written corrective feedback that is comprehensive can supply a learning outcome. Till now, the sole proof of the long-run influences of written corrective feedback came from research investigating the effectiveness of focused or targeted rectification, that's corrective feedback aiming at one particular kind of error (e.g. Bitchener, 2008; Bitchener & Knoch, 2009; Bitchener & Knoch, 2010a, 2010b, Ellis et al., 2008; Sheen 2007; Sheen, 2010b).

Nonetheless these discoveries may be for the theory of SLA; a few researchers have addressed their commonsense pertinence (e.g. Ferris, 2010; Storch, 2010). They claimed that once giving corrective feedback, instructors typically choose to enhance the general accuracy of their students’ writing, not simply the employment of one particular linguistic feature.

This may well be significantly right in settings in which the focus of educational philosophy and pedagogy is mainly on “context communicating” and not on “language as an object” (e.g. Anderson, 2010 ). It has in this manner been claimed that inclusive remedial input could be a more bona fide input technique (e.g. Anderson, 2010; Ferris, 2010; Hartshorn et al., 2010; Storch, 2010). The conclusion that a dependable remedial technique such as comprehensive mistake rectification leads to L2 acquisition might in this way be considered of extraordinary pertinence to instructors.

The suggestion that can in this way be drawn from investigations just like the present study is that instructors ought to permit learners the chance to reexamine their writings based on the given criticism and feedback. Asking students to perform modification exercises might be anticipated to cultivate SLA since, “providing the correct form may assist learners to automatize their L2 Production” (Loewen, 2004, p.157).  In addition, manufacturing and generating correct revisions may well be thought of as an indicator of pushed output, and may so be anticipated to push L2 acquisition by activating noticing the gap and hypothesis testing (Swain, 1985; 2005).

Taking all these issues into consideration, it can be regarded as vital to make students mindful and responsive to the objective and worth of corrective feedback revision and provision exercises, and to provoke learners’ motivation and inspiration to interact with the provided feedback. In light of all these the findings have meaningful suggestions for instructional practices in L2 classrooms. If EFL teachers want this to be operative and practical in their classes, they should attempt to reconcile corrective feedback with the educational activities of their classrooms and use corrective feedback for internalizing knowledge. At a broader level, it is up to curriculum designers to incorporate corrective feedback within the language curriculum of the specialists.