Selective vs. comprehensive grammar correction in writing pedagogy: Counter evidence to Truscott’s view

Introduction

Writing teachers have traditionally viewed written error correction as playing an integral role in improving Second Language (L2) writing accuracy (Brown, 2007; Ferris & Roberts, 2001; Lee, 2004), but debate regarding its effectiveness has emerged in the past decades (Chandler, 2003; Truscott, 1996). Results of some studies (e.g. Kepner, 1991; Krashen, 1985, 1999; Semke, 1984; Truscott & Hsu, 2008) indicate that error correction is not only ineffective, but also potentially detrimental to L2 writing development. Truscott (1996) following a non-interventionist point of view and considering naturalistic second language acquisition (SLA) and the Natural Order Hypothesis (Krashen, 1981, 1982, 1985) raises his objection against grammar correction. He argues that classroom time should not be devoted to such matters as grammar instruction and error correction, thus, grammar correction in L2 writing classes should be abandoned. He provides evidence for this contention through an extensive review of the past studies that demonstrate grammar correction to be ineffective and unhelpful. Truscott (2001) believes that grammar errors are not good targets for error correction. In other words, since grammatical errors stem from problems in the syntactic system, they are the least correctable features.

Contrary to what some critics have stated, advocates of corrective feedback (e.g. Bitchener, 2008; Bitchener & Storch, 2016, Bitchener & Ferris, 2012; Bitchener, Young & Cameron, 2005; Ellis, 2009) challenge Truscott's claims against the effectiveness of grammar correction and have produced research evidence that supports the potential benefits of providing written error correction as far as development and improvement in learners' L2 writing accuracy are concerned.

Accordingly, interventionists (Lalande, 1982; Lightbown, 1998; Long & Robinson, 1998; Lyster, Lightbown & Spada, 1999) argue against reacting to learners’ errors. These arguments have led to two general approaches to providing written error correction (i.e., comprehensive vs. selective) in the currently available literature (Ellis, 2009; Van Beuningen, 2010). The comprehensive/unfocused approach involves the teachers correcting all errors in a learner's text, irrespective of the error category. On the other hand, the selective/focused approach targets specific linguistic features only, leaving all other errors outside of the focus domain uncorrected.

Different predictions have been made regarding the effectiveness of either approach. The comprehensive approach in written error correction may lead the attention of the student not just towards errors in writing but also to new features of the target language, thereby, promoting more effective language learning (Corpuz, 2011). Some researchers have found evidence that systematic correction of all student errors leads to lower error rates (e.g., Lalande, 1982) and accuracy development in the revision of a particular text (Van Beuningen, De Jong, & Kuiken, 2008, 2012); others, on the other hand, have called for selective/focused correction of specific error types (e.g., Ellis, Loewen, and Erlam, 2006; Ferris, 2006; Van Beuningen, 2010).  Ellis et al. (2006) argue that a comprehensive approach in providing written error correction may not be the most effective approach because L2 learners have a limited processing capacity. They claim that asking L2 learners to cope with written error correction that covers a wide range of linguistic features at the same time may lead to a cognitive overload that might prohibit the students from processing the feedback they receive. Furthermore, Ellis (2009) claims that a selective approach in written error correction may prove more effective as L2 learners are able to examine multiple corrections of a single error. Because of this, L2 learners might obtain not only a richer understanding as to why what they wrote was erroneous, but also opportunities to acquire the correct form.

Controversy among the scholars regarding effectiveness of the general written corrective feedback (WCF) and the relative advantages of different WCF options remains to a large extent unresolved. Accordingly, the purpose of this study was to assess student uptake of corrections received through various forms of intervention, that is, comprehensive feedback, selective feedback, and no feedback as far as their immediate effect on writing accuracy was involved. In addition, it aimed at investigating such an effect in the long run as it is assumed that the delayed effect provides evidence as to whether the input has been not only comprehended but acquired as part of the learners’ developing competence in the L2. Unfortunately, the longitudinal piece, that is, the assessment of the delayed effects of corrective feedback, is lacking in many studies of error feedback in L2 writing and merely shows up in the literature. Hence, it can be argued that further research is needed on error correction in its various forms unless its ineffectiveness and harmfulness have been conclusively proven (Corpuz, 2011) as a better understanding of WCF has important pedagogical implications for language instruction in various contexts (Liu & Brown, 2015).

Literature Review

Truscott’s (1999) strong opposition to WCF has faced numerous challenges and received critiques from researchers, who, through empirical research or other scholarly synthesis, have basically argued that grammar feedback is essential for second language acquisition (SLA) and should remain an important component of L2 instruction  (Liu & Brown, 2015). What follows is brief account of some of these empirical research studies.

Sheen, Wright, and Moldawa (2009) examined the differential effects of three treatments, namely, direct focused CF, direct unfocused CF, and writing practice alone, on the use of the English articles by 80 English as a Second Language (ESL) learners. Students in the focused CF group only had their article errors corrected, whereas the unfocused CF group had their errors in five categories (articles, copula be, regular past tense, irregular past tense, and preposition) corrected. These two CF groups completed two written narrative tasks. The writing-practice groups only did the narrative tasks receiving CF; the control group received no CF. All four groups completed the pre-test, immediate post-test, and delayed post-test; and acquisition of the articles was measured by three versions of a narrative writing test, which asked the students to write a story based on a series of pictures. The results showed that all three treatment groups outperformed the control group. For the acquisition of the articles, the focused group showed an advantage over the unfocused group at both the immediate and the delayed post-tests, suggesting that focused CF is more effective than unfocused CF. Surprisingly, the focused group, who did not receive corrections on features beyond the articles, also outperformed the unfocused and the other two groups in terms of the overall accuracy in the five targeted features. Sheen attributed this to the possibility that when CF addresses a range of errors, learners might be less able to process the feedback effectively. Another reason, according to Sheen, might have to do with the manner in which the CF was provided, that is, the feedback which the focused group received was systematic, but that which the unfocused group received was much less so.

Ellis et al. (2006), adopting some methodological features from Sheen (2007), also compared the effects of focused and unfocused written CF on Japanese EFL learners’ use of the two English articles. On three written narratives, students in the focused group received corrections of article errors, the unfocused group received corrections of all errors, and the control group received no CF. Accuracy-gain on the article uses was measured using narrative writing tests and error correction tests. The study found that both CF groups performed significantly better than the control group at the delayed post-test. The researchers contended that contrary to Truscott’s (1996, 1999) claim, CF can indeed facilitate acquisition.

Araghi and Sahebkheir (2014), Farrokhi and Sattarpour (2012), and Pashazade and Marefat (2010) evidence the same pattern of findings in an Iranian context. They investigated whether focused CF and unfocused CF can cause any differential effects on the accurate use of the simple past tense, and of definite and indefinite English articles, respectively. The results suggest that focused CF promotes learners' grammatical accuracy in L2 writing more effectively than unfocused CF.

Similarly, Van Beuningen et al.’s (2008) study challenges Truscott’s (1996) suggestion that having students to do additional writing practice might be more worthwhile than giving them CF. Their study provided evidence to support the beneficial role of unfocused CF. The study was carried out on 62 secondary-school first language (L1) and L2 learners of Dutch, who were randomly assigned to one of four groups: direct unfocused CF, indirect unfocused CF, practicing writing (practice), and self-correction. The participants in a three-session experiment completed two writing tasks of different topics, accompanied by a series of pictures, for Session 1 and another two tasks for Session 3. Students in the CF groups received either direct corrections or coded CF on their writing. The practice group and the self-correction group received no CF. During Session 2, the CF groups revised their texts based on the corrections they received; the Practice group did not revise their uncorrected texts but instead, completed two additional writing tasks; and the Self-correction group was required to revise their texts without the benefit of CF. The results revealed that the two feedback groups significantly outperformed the Self-correction group on the revised texts. With regard to performance on the new tasks given at Session 3, only Direct CF resulted in improved accuracy. Of the four groups, the Practice group performed the worst, displaying no improvement from Session 1 to Session 3.

Following up on their (2008) study, Van Beuningen et al. (2012) investigated the effect of direct and indirect unfocused CF on 268 Dutch learners. The study, also, sought to test Truscott’s (2007) hypotheses that CF might only have value for non-grammatical errors and that CF compels students to simplify their writing. The students were randomly assigned to one of four conditions: direct CF, indirect CF (via error codes), self-correction, and practice. The experiment consisted of four sessions: a pre-test session (Session 1), a treatment/control session (Session 2), a post-test session (Session 3), and a delayed post-test session (Session 4). At Session 1, all four groups completed the first writing task. One week later at Session 2, the two CF groups revised their first written texts based on the CF they received; the control group self-corrected their original texts without the help of CF; and the practice group completed a new writing task. During session 3 and session 4, all four groups produced a new text based on a new topic. The results showed that unfocused CF led to improved accuracy in both the revised texts and the new texts. The positive effect of unfocused CF observed at the revision stage was retained four weeks later. In terms of the differential efficacy of direct and indirect CF, the study found that only direct CF facilitated “durable grammatical accuracy improvements of a medium size” (p. 32), and that indirect CF had a greater effect on non-grammatical errors. The findings rebutted Truscott’s (1996) hypotheses that CF harms the complexity of students’ writing and that additional writing practice may be more beneficial than the provision of CF. In short, this study demonstrated that comprehensive treatment of errors can help students improve their grammatical accuracy over time, not supporting Truscott’s claim that grammar correction may bring about harmful effects.

All the studies discussed so far reported positive evidence in support of CF. A few research studies (e.g. Hartshorn et al, 2010; Liu, 2008, Truscott & Hsu, 2008), however, have concluded that written CF did not have any benefit.

Although the effectiveness of oral CF is well-established (Li, 2010; Lyster & Saito, 2010; Mackey & Goo, 2007), and a number of theoretical SLA insights predict that written CF can enhance L2 development, yet, the usefulness and efficacy of written error correction are still topics of considerable debate (Liu & Brown, 2015). This seeming stagnation is attributed to methodological challenges (Liu & Brown, 2015). Otherwise stated, earlier studies that compared the effects of CF types across separate error categories have been too heterogeneous (with respect to the learner variables, types of errors targeted, CF type, research design and context, etc.) to result in any definitive conclusions. Clearly, then, further studies seem warranted so that the points of contention, to use Bitchener and Ferris’ (2012) terms, over the complex and multivariate subject of WCF is resolved.

Research Questions

Based on the above observations, the present study was guided by the following questions:

1. Is there a significant difference in writing accuracy of Iranian elementary EFL learners provided with various patterns of corrective feedback (comprehensive, selective and none)?

2. Does the possible effect of various patterns of corrective feedback last in the long run?

Method

Participants

By means of an institutional placement test, from a total of 100 EFL learners, 66 female elementary EFL learners were selected and assigned into three treatment groups (n = 22 in each group) receiving comprehensive corrective feedback, selective corrective feedback and  no feedback at all, that is,  following Truscott’s ( 1996) view, no correction of grammar errors. The participants who were within the age range of 14 to 17 had at least 1 year experience of learning English at secondary school and institute(s) and had been involved in writing tasks ( e.g., writing questions, summaries, etc.).

Instruments

To practically set the ground for the sake of putting into practice the already mentioned theoretical aspects of the current study, at the macro level, a reliable in-house proficiency test (α=.90), institutionalized by Iran Language Institute, was put to use. It served the purpose of homogenizing the participants in terms of language proficiency at the outset of the study. Furthermore, the participants were assigned a writing task that assessed learners’ writing accuracy at the beginning of the study prior to the treatment. Then, the researchers selected some topics covered in the students' books for which the learners were required to write compositions. At the end of the treatment, two other topics were used in immediate and delayed post-tests. What follows is a brief account of the procedures followed in data collection

Procedure

Prior to the treatment, the researchers made  sure that the groups were not different from each other at the outset of the study, that is, through a set of  initial ANOVAs (F=.97, ρ=.38 > .05) and (F=.97, ρ=.42>.05), homogeneity of the participants’ was ensured in terms of their general English proficiency and writing accuracy, respectively. During the study, the participants in all groups wrote one composition per week for six weeks on general topics (e.g., write about a bad memory you had in the past; write about the last trip you took; write about your best holiday) covered in their students' books. They were asked to write compositions within a word limit of at least 150 in each composition in 40 minutes. During writing time, the teacher monitored and observed learners and provided hints (e.g., the equivalent words without any reference to their past forms in English) whenever needed. Learners doing their first composition (write about a good memory you had in the past), that served the purpose of pre- and post-tests, were not allowed to have access to any resources and assistance.

Participants’ written pieces were reacted by the teacher following three methods. In line with Lalande (1982), the teacher underlined all the errors in the written performances of the first treatment group. As for the second group, in accordance with Ferris (2006), Hendrickson (1981), and Pashazadeh and Marefat (2010), she focused on errors such as simple past tense, countable/uncountable, and comparative adjectives which are considered global errors in this study since they inhibit communication (Burt, 1975; Hendrickson, 1980).  However, the third group did not receive any correction from the teacher and was just commented on the content through such terms as great, good, ok, etc.

Learners in the first group (i.e., comprehensive feedback), due to the larger number of errors corrected, were given 25 minutes to check and reckon the errors corrected and underlined by teacher; and those in the second group (i.e. selective feedback) were given 15 minutes to check and observe the frequently happened errors corrected by the teacher. Due to the lack of correction, learners in the third group were advised to reflect about their written pieces and revise their texts without the help of CF. The rationale behind no correction was that the writing process alone would lead to the development of accuracy (Truscott, 1996). This procedure was followed for six weeks and at the end of the sixth composition, immediate post-test (i.e., the 6th composition) and delayed post-test, within a month time interval (i.e., week eleven) were conducted. A high inter-rater reliability of .77 was established through double coding 25% of the written data by a research assistant.

Data Analysis

By means of the SPSS (Statistical Package for Social Sciences) software, a series of one-way ANOVA tests were run: (1) in the pretest to establish homogeneity across the participants, (2) at the end of the study, that is, the immediate post-test to figure out the possible effects of various treatment patterns, and (3) in the delayed-post-test to find out whether such an effect, if there was any, lasted through time. Moreover, post-hoc comparisons using the Tukey HSD test were conducted in post test stages to find out where exactly the difference among the pairs existed.

As regards writing accuracy, following the scoring procedures applied by Chastain, (1988), Kroll (2001), and Zhang, (1995), we applied the following scoring procedures for the selective and comprehensive groups, respectively.

Selective group

× 100

Comprehensive group

× 100

Results

Table 1 shows descriptive statistics as far as the participants’ performance is concerned in terms of writing in the immediate post-test stage.

Table 1. Descriptive statistics: Participants’ Performance in the Immediate Post Test

S: Selective       C: Comprehensive     NC: No correction

As the mean scores indicate, the number of errors in the group receiving selective CF is fewer in comparison to the other two groups. Further significant results were found through one-way ANOVA presented in Table 2.

Table 2. ANOVA Results: Immediate Post-test Results across the Three Treatment Groups

A one-way between-groups analysis of variance was conducted to explore the differences among the effect of feedback types on writing accuracy measured through immediate post tests. Statistically significant difference at the p < .05 was observed across the three groups: F (2, 63) = 22.81, p = .00. Further post-hoc tests summarized in Table 3 were also run to clarify where exactly the difference across the paired groups existed.

Table 3. Post-Hoc Multiple Comparisons of Means: Immediate Post Test

*. The mean difference is significant at the 0.05 level.

Post-hoc comparisons indicated that the mean score of the selective feedback group
(M = 62.88, SD= 19.16) was significantly different from that of the comprehensive feedback group (M = 86.73, SD = 8.50) and no-correction group (M = 84.95, SD = 8.43).

As mentioned in the forgoing sections, another focus of the current study was to find out whether the effect of the types of corrective feedback lasted in the long run. Table 4 shows descriptive statistics as far as participants’ performance in delayed post test was concerned.

Table 4. Descriptive statistics: Participants’ Performance in the Delayed Post test

As the summary statistics indicate, the group receiving selective feedback, compared to the other treatment groups, obtained lower mean scores, that is, fewer number of errors in writing. This difference turned out to be significant as the one-way analysis of variance (ANOVA) results show in Table 5.

Table 5. ANOVA Results: Delayed Post-test Results across the Three Treatment Groups

One-way between-groups analysis of variance indicates a similar pattern at this stage, that is, the delayed post test. There was a statistically significant difference at the p < .05 level among the three groups: F (2, 63) = 20.51, p = .00. The results are further confirmed through the follow up pair-wise comparisons presented in Table 6.

Table 6. Post-Hoc Multiple Comparisons of Means: Delayed Post Test

*. The mean difference is significant at the 0.05 level.

As the results of post-hoc comparisons, that is, the Tukey HSD test, indicate the mean score obtained by the selective feedback group (M= 63.76, SD= 19.28) was significantly different from that of the comprehensive) and no-correction (M = 86.18, SD = 8.31), (M = 84.87, SD = 8.25) groups, respectively.

Discussion

Given a limited range of studies on the effect of feedback type (comprehensive, selective, no correction) upon writing accuracy in an EFL setting and the existing controversy among them, the present study investigated the effect of different types of feedback on written accuracy of elementary EFL learners. It was found that selective type of feedback compared to the other types had a significant effect on learners' accuracy in using simple past tense, countable/uncountable nouns, and comparative adjectives both in the short term and the long run. This finding indicates that teachers' decision on written corrective feedback should depend on the acting time and the educational focus on the type of grammatical features.

The results of the study are in line with Ellis et al.’s (2006) and Pashazade and Marefat’s (2010) research that focused on the effectiveness of selective approach, which targeted specific linguistic features and left errors outside the focus domain uncorrected compared to the unfocused approach. They indicated robust positive effects of focused CF and durable accuracy gains. This can be due to the fact that learners are more likely to notice and understand corrections when they target a specific error type. This suggests that CF needs to be aligned to the learner’s current level of L2 development. Theoretically, this suggestion can be explained through the limited processing capacity model of L2 acquisition (Schmidt, 2001; Van Patten, 1996, 2004). They claim that asking L2 learners to cope with written error correction that covers a wide range of linguistic features at the same time may lead to a cognitive overload that might prohibit the students from processing the feedback they receive.

Nonetheless, the findings contradict Lalande's (1982) and Van Beuningen, et al.'s (2012) study in which systematic correction of all student-errors led to lower error rates; and others (e.g. Krashen, 1984; Semke, 1984; Truscott, 1996, 2007) who argue that every type of error correction should be eliminated because it is ineffective in the long run.

What the findings may suggest is that despite the process of providing corrective feedback being frustrating, difficult, and time consuming, teachers still are recommended to provide written error correction because it allows for individualized teacher-to-student communication that is rarely possible in the day-to-day operations of an L2 writing class (Ferris, Pezone, Tade, & Tinti, 1997). Selective error correction, in the meantime, could function as a noticing facilitator, to use Schmidt’s (1994) notion of Noticing Hypothesis that directs the attention of the L2 student not only towards error, but also towards new features of the target language. As Schmidt maintains, the more L2 learners notice, the more they learn the L2.

It is thus proposed that grammar correction should be based on learners' current stage of development and interlanguage since it is thought that comprehensive feedback may demotivate and discourage learners from taking risks and trying more sophisticated language forms. It seems that, as Pienemann (1984) maintains, some of the errors corrected may be related to those cognitively demanding ones that the learners are not ready to absorb; therefore, corrective feedback, especially comprehensive type, may not lead to improvement in their accuracy. Accordingly, in line with what Schmidt (1994) suggests, for learners to improve their writing, they have to be provided with appropriate feedback, at the right time and in the proper context. Learners have to notice the feedback and be given ample opportunities to apply the corrections. However, when everything is said and done comprehensively, unfortunately, if the learners are not committed to improving their writing skills, they will not improve, no matter what type of corrective feedback is provided. In order to help learners refine their output in these areas, selective feedback can help learners close the gap between their current and desired state of interlanguage. It is thought possible that selective feedback gives assurance to writers that although they may have problems in writing, parts of their writing is error-free which adds to their motivation to solve minor writing problems they have. The long term effect of WCF found in the current study can be partly explained in terms of the dynamic processes involved in learning. Noticing and learning that occur represent dynamic learning processes which can be an indication of self-initiated focus on form, that is, learners come to pay attention to forms they need for communication in the L2 independently (Ellis, Basturkmen, & Loewen, 2001; Williams, 1999, 2012). This learner initiation can promote learner autonomy in the long run, which is another key factor for long-term L2 development (Dickinson, 1995; Holec, 1981). This study, therefore, may elucidate some key issues: noticing during output, provision of feedback that meets the learners’ needs, promotion of self-initiated focus on form, and learner autonomy. Furthermore, considering the nature of fossilization, it can be argued that providing effective written error correction, more particularly through focusing on certain problematic ones (i.e. global errors), has a vital role in language instruction in order to preclude its occurrence. Provision of error correction may attract the attention of the L2 student and aid him not only to discover his/her errors in his/her output, but also the feature of the target language (Corpuz, 2011).

Conclusion and Implications

Whereas the value of written CF for L2 acquisition has been heavily contested (e.g. Truscott, 1996; 2007), this study provides evidence on the efficacy of error correction in L2 writing. The fact that the accuracy improvement brought about by written CF was shown to be durable, rebuts Truscott’s (1996) claim that correction can only lead to a superficial and transient type of L2 knowledge. It is concluded that by providing learners with the opportunities to notice the gaps in their L2 interlanguage system, to test the hypotheses, and to get involved in metalinguistic reflection, written CF has the ability to improve SLA and to lead to accuracy development. The present empirical work, thus, advances the theoretical understanding of the language learning potential of written CF, and shows that selective CF is a useful pedagogical tool for elementary learners.

Furthermore, the outcomes of the present study imply that EFL instructors should reflect on when and how to use WCF. Otherwise stated, it is important to know what types of errors require more attention from the teacher, it warrants further investigation though.  This means that, to avoid learners' frustration and discouragement due to comprehensive feedback, learners' current stage of development and interlanguage should be paid due attention if a teacher intends to help learners through grammar correction. This pattern of CF helps learners refine their output, and close the gap between their current and desired state of interlanguage.

One can conclude, then, that the thought-provoking recommendations of such well-known figures as Truscott (1996, 1999, 2010), who predicted CF to have no potential value for the development of grammatical competence and suggested that CF could only be beneficial for errors that are relatively simple such as spelling errors and lexical errors must be reexamined.

As any human production, this study has some limitations, thus the findings of the study need to be interpreted after the due considerations of this drawback. Catering for these limitations may also provide some applicable insights for conducting further studies. The results of this study may be unique to this particular population under investigation, and may not be universal in nature. In order to gain more reliable information and findings about the study variables, other studies should be carried out with more participants in different contexts with bigger sample sizes. In addition, given the time span of the study over a semester, there may have been other intervening variables such as participants’ individual study efforts, variability in classroom instruction and teaching style, and motivation, which may have influenced how students responded to written CF. Moreover, this study focused on the effect of direct CF on learners’ writing accuracy. However, more studies need to be carried out with regard to other CF types (e.g., indirect) so that more comprehensive conclusions and findings can be drawn. The last limitation of the current study relates the errors focused in the research. Conducting the study with only focusing on some grammatical structures leaves the results with other error types in an aura of ambiguity.

We are still facing many unanswered questions. As SLA researchers who have encountered the challenges involved in designing and executing classroom studies that address pedagogically relevant questions, one thing has become very clear to us, that is, research cannot provide language classroom teachers with clear-cut answers regarding what kind of CF to provide or how it should be provided. There are simply too many variables involved (Guenette, 2007). It is always easy to critique what researchers investigating written CF should and should not have done. The way forward is to try to investigate systematically the variables that are pedagogically relevant and to conduct replication studies. Thus, with respect to the aforementioned limitations of the study, certain points deserve further exploration in the future. Researchers should triangulate the findings both qualitatively and quantitatively in order to get reliable data to generalize. The subjects of this study were Iranian EFL learners. Other research can be done in the ESL setting.If possible, even a comparative study that compares the learners’ attitudes in EFL vs. ESL settings can be conducted. Further studies on this special issue with more samples are needed to provide more insights over just how far the findings can be generalized beyond the immediate and relatively small-scale world of the experimental studies. The learning potential of comprehensive or unfocused CF deserves more attention. Whereas a few studies recently provided evidence on the effectiveness of comprehensive correction, the focus of the current CF studies has mainly been on establishing the value of selective or focused CF. The CF responsiveness of different types of errors to CF types should be studied more. Therefore, due to the pedagogical and contextual restrictions, further research is called upon to push the frontier of knowledge so as to provide a fruitful English teaching and learning conditions especially regarding writing skill in English classes.