Direct and Indirect Feedback in the L2 English Development of Writing Skills

The present study examines the role that feedback plays on the development of second language (L2) English learners‟ writing accuracy over time. Earlier formal accounts and empirical works have focused on the relevance of corrective feedback (CF) in L2 writing learning (Ellis et al., 2008; Sheen, 2007), and what kind of CF (i.e. direct or indirect) has proved to be the most effective one, especially at low L2 levels (García Mayo and Labandibar, 2017; Ismail et al., 2008). We have analyzed 3 pieces of writing produced by 8 L2 English participants (aged 11 to 12). The participants were randomly divided into two groups, one of them received direct CF on their written tasks and the other group was exposed to indirect CF. Results revealed that both groups seemed to improve their mean scores from the pre-task to the post-task, regardless of the type of CF implemented. However, the direct CF group has proven to benefit more from teacher‟s written CF, when compared to the indirect CF group. This is especially the case in the development of grammar accuracy.


Introduction
The present work aims to analyze the role that feedback plays on the development of second language (L2) English participants" writing skills (Ellis, 2008;Ferris and Roberts, 2001;Sánchez Calderón, 2014;Sheen, 2007). Feedback is defined as the information provided by an agent (e.g. the teacher) regarding the learner"s performance (Hattie and Timperley, 2007;Sadler, 1989). Therefore, it helps to fill the students" gap between what is understood and what is aimed to be understood.
In the present study, two types of corrective feedback (CF) are examined, namely, direct and indirect. As shown in (1), direct CF implies that the teacher identifies an error and provides the correct form (Ellis, 2009).
(1) a. He was waking up and his father frying an egg [students" error] b. When He he was waking up woke up, and his father was frying an egg [direct CF] With regards to indirect CF, the teacher identifies errors without providing explicit correction. As illustrated in (2), a cross is provided to indicate a missing word or group of words. Also, underlined words are included to show errors.
(2) a. He was waking up and his father frying an egg [students" error] b. He was waking up and his father frying an egg [indirect CF] Learning how to write in an L2 is a challenging process, especially for low level L2 English students (Hyland and Hyland, 2006). Thus, L2 English teachers are responsible for encouraging learning, helping students improve their writing skills and increase their motivation when accomplishing their writing tasks via written CF (Bitchener et al., 2005;Ellis et al., 2008;Hyland and Hyland, 2006).
Using written CF as a teaching tool has been discussed extensively over the last few years (Ellis, 2009;Ferris and Roberts, 2001;Sheen, 2007). However, there is a dichotomy in previous studies with regards to the effects of CF on L2 students" L2 performance. For instance, while Truscott (1996) has claimed that CF is non-effective and even harmful, Ferris (2012) has argued that CF improves student"s writing skills.
In the present study, we have compared the effectiveness of direct and indirect CF via a pre-test and a post-test. In order to do so, we have analyzed 3 pieces of writing produced by 8 L2 English participants (aged 11 to 12) whose first language (L1) is Spanish. The participants were randomly divided into two groups, one of them received direct CF on their written tasks, and the other group was exposed to indirect CF. This paper is organized as follows. Section 2 presents the defining properties and the taxonomy of feedback. Section 3 reviews empirical works on the role played by CF in L2 learning. Section 4 formulates the research questions (RQs) that will guide the data analysis. Section 5 presents the study, namely the participants, the data extraction and the procedure. Section 6 displays the data analysis and the discussion of results, respectively. Conclusions and suggestions for future research are drawn in section 7.

Defining Properties
Feedback can be defined as the reaction to a process or an activity, along with the information obtained from such a reaction (Abedi et al., 2010;Liu, 2008). Therefore, an interactive ex-change is established between readers and writers.
Two issues could play a role in the use of feedback in education, namely, the teacher"s reaction to the student"s performance and the student"s reception of effective information from the teacher"s response (Hattie and Timperley, 2007). In other words, feedback is said to fill in the gap between what is understood and what is aimed to be understood (Sadler, 1989).
Students are expected to benefit from the teacher"s valuable information, regardless of the written or the oral feedback provided (Clarke, 1998;Clarke et al., 2001;Clarke, 2004;Ellis, 2009;Hyland and Hyland, 2006). Thus, feedback helps to reinforce language learning in identifying students" errors (Bitchener et al., 2005;Ellis, 2009;Jamieson, 2011). In order for students to benefit from this practice, teacher"s feedback should be brief, clear and easy to understand (Sánchez Calderón, 2014).
Praising has also been considered to be a motivating factor when feedback is offered (Ferris, 1995;Hyland, 1998;McGrath et al., 2011). In this way, teachers can build students" confidence and encourage constructive learning along the L2 learning process.

Feedback Taxonomy
As shown in table 1, feedback can be classified according to the following parameters: (a) source (peer feedback vs. teacher feedback); (b) aim (summative feedback vs. formative feedback); and (c) teacher"s response to students" performance (positive feedback vs. negative feedback) (Ellis, 2009;Hyland and Hyland, 2006). Learner"s incorrect performance With regards to the source, feedback can be provided by the teacher or by another student (Hyland and Hyland, 2006). In the case of the latter, collaborative learning is enhanced, allowing students to learn by interacting with other peers. In other words, peers" response helps learners to understand how others see their L2 performance.
When feedback focuses on learning as a final product, that is to say, by achieving a specific goal, we refer to it as summative feedback. On the other hand, formative feedback guides the learner"s progress while providing instruction. Hyland and Hyland (2006), argue that, contrary to summative feedback, formative feedback points forward to the student"s future writing and the development of his or her writing processes.
Feedback can also be classified according to the teacher"s response to the students" performances (Ellis, 2009). Positive feedback focuses on the correctness of the learner"s response, regardless of whether it refers to content or to linguistic responses. As discussed earlier, positive feedback provides learners with affective support and fosters motivation to keep learning. By contrast, negative feedback is concerned with the learner"s errors.
One of the types of negative feedback is the so-called CF. This means that learners" errors present a response or a reaction. More specifically, CF consists of the following elements: the distinction of the error made, the provision of the correct target language form, the metalinguistic information about the nature of the error or the combination of the first two elements .
Based on the teacher"s response, two types of feedback can be addressed, namely, direct or explicit CF and indirect or implicit CF Ellis, 2009). Regarding the former, the teacher identifies the error and provides the correct form along with explicit guidance about how to correct the students" errors as well as grammatical information about the error nature (also known as metalinguistic feedback). As for indirect CF, the teacher identifies errors without providing correction. This implies students should be capable of diagnosing and self-correcting their own errors. Therefore, the main factor that distinguishes these two types of CF is the learner"s degree of involvement in the correction process.
Out of these types of feedback, the present study will investigate the role played by direct or explicit CF and indirect or implicit CF in the development of L2 English writing skills.

Corrective Feedback in the Development of L2 English Writing Skills
Teacher"s feedback has been argued to have a significant impact on the L2 writing process (Emig, 1977). Indeed, L2 teachers feel they should write substantial comments in order to help students" development of their writing skills (Hyland, 2003;McGrath et al., 2011). However, feedback requires time and effort.
There is a debate as to whether feedback influences efficiency on student"s writing skills. For example, while Truscott (1996) argues that error correction does not significantly benefit L2 writing skills, Ferris (2012) claims that error correction is widely considered as an important factor in enhancing students" performance.
Despite this controversy, there are other studies that lend support to the importance of CF in L2 writing learning Ellis, 2009;Ferris and Roberts, 2001;Sheen, 2007). Nevertheless, they do not agree on which type of feedback is most effective, or whether feedback helps improve the quality in student"s writing overall and the linguistic development of specific grammatical features.
There are other studies that argue for the benefits of using indirect feedback. Such is the case of Ferris (2012) who suggests that teachers should ask students to reflect on their work based on the teachers" feedback. This type of feedback will encourage students to develop cognitive skills in problem-solving tasks.
Considering the previous taxonomy analyzed, the present study will focus on two types of CF, direct and indirect. The subsequent section will revise empirical works on L2 written CF.

Empirical Studies on the Role Played by Direct and Indirect Corrective Written Feedback
Two main issues have been investigated with regards to the use of CF in the development L2 writing skills, namely, (a) what kind of CF has proved to be the most effective to help L2 students" written output (Coyle and Roca De Larios, 2014;García Mayo and Labandibar, 2017;Ismail et al., 2008;Sarvestani and Pishkar, 2015); and (b) what kind of feedback is most effective at low L2 levels (Coyle and Roca De Larios, 2014;García Mayo and Labandibar, 2017). Coyle and Roca De Larios (2014), investigated the role played by two forms of CF, error correction and model texts, on L2 English children aged 10 to 12 (26 boys and 20 girls) from a Spanish primary school. The children had been learning English for 4 or 5 years and were exposed to English 3 hours per week. They were not familiar to writing freely and had never produced a piece of collaborative writing before the study took place. The children were placed in proficiency-matched pairs based on their performance obtained in English tests. Data for the study were collected over a period of 4 weeks. The study was carried out in a three-stage collaborative writing task. In Stage 1, a four-frame picture story prompt was given to each pair and the participants were asked to compose their task together and note down any linguistic difficulties they experienced while performing the task. Stage 2 took place one week later and, while half of the pairs had their stories returned with the teacher"s explicit error correction, the other pairs were given two model texts of the story. Stage 3 occurred one week later, and the pairs were given the picture prompt again and were asked to rewrite the story. Coyle and Roca De Larios (2014), results revealed that L2 English children significantly showed linguistic acceptability and comprehension of the revised texts after receiving teacher"s feedback. As shown in table 2, there was a significant increase in the production of clauses when the initial non-revised versions (where feedback was not provided) are compared to the revised versions of the texts. Error correction (n=9 pairs) 6.5 2.1 6.5 1.6 2 1.6 1.1 1.8 3.5 2.6 1.5 1.6 1 1.4 3.6 2.8 Thus, the error correction condition, a form of direct feedback, proved to be more effective when compared to the use of models. In fact, models may be considered to be an indirect way to provide feedback as participants have to compare their drafts to their teacher"s models instead of getting the correction directly.
García Mayo and Labandibar (2017), also examined the role played by model texts as an implicit way of increasing participants" error awareness. The use of models has been claimed to push learners to actively identify and understand mistakes. They analyzed 60 Basque-Spanish participants, 30 of whom came from a secondary school (13 years old) with a low English level, and 30 of whom came from a high school (16 years old) with a lowerintermediate English level. Three groups were distinguished at each proficiency level: (a) a control group, which completed the composing and the rewriting stages but did not receive the teacher"s models; (b) a guided noticing group, which completed the three stages and used a guided form of noticing their errors; and (c) a group which completed the three stages but used a non-guided form of noticing their errors. Based on the results of this study, models proved to be useful in increasing the participants" lexicon array. In other words, indirect feedback modelling appeared to be an effective device in improving L2 English writing skills. Sarvestani and Pishkar (2015), examined the role of written CF in a group of 60 intermediate L2 English learners (age range: 19 to 32) at Navid English language institute, Shiraz (Iran). Participants were divided in three groups and were given eight topics to write about along eight weeks.
While, in the first experimental group, participants received direct written CF on their grammatical errors, in the second experimental group, participants received indirect written CF. There was a control group who did not receive any sort of feedback in their writing. Furthermore, Sarvestani and Pishkar (2015) administered a post-test to compare it to the participants" performances in the three groups in the pre-test and examine what kind of feedback is more influential in the development of the participants" L2 English writing skills. As shown in table 3, L2 English participants who received direct written feedback showed higher scores in the post-test (M = 69.00) when compared to those ones who received indirect written CF (M = 60.30). Regarding the control group, feedback was not provided in their writings (M = 52.45). Thus, this study shows that direct written CF reflects that CF helps participants to improve their writing accuracy.
Further studies on the role that CF plays in students" development of writing skills have also been analyzed in adults (Abedi et al., 2010;Ahmadi et al., 2012;Liu, 2008;Shintani and Ellis, 2013). For example, Liu (2008) investigated the extent to which two types of CF feedback (direct and indirect) could help L2 English participants to improve writing accuracy. The study was conducted in a southwestern university in the United States where 12 first year L2 English students enrolled a 3-unit composition class. The participants were randomly divided into 2 groups (group A and group B) with six students in each group. As for the first draft, the participants in group A received direct CF, where the errors were both underlined and corrected, and the participants in group B received indirect CF, where the errors were only underlined. Both groups were required to submit a second draft after revising the errors in the first one. As illustrated in figures 1 and 2, indirect CF significantly decreased morphological and semantic errors throughout the three drafts.  (Liu, 2008) Therefore, Liu (2008) study suggests that indirect CF improves self-correction when writing. That is to say, it helps to find out the error and self-correct it, as reflected in the lack of morphological errors that were previously included in the first draft. Shintani and Ellis (2013), investigated whether written error feedback has an effect on L2 English adult learners" implicit and explicit knowledge, and the effect of direct CF and of another type of error feedback -the provision of metalinguistic explanation. The effect of these two types of error feedback was measured by an Error Correction Test (ECT) and by examining the accuracy in the of use of indefinite articles in English in both a revised text and in new pieces of writing. They examined 49 low intermediate L2 English students in the United States (31 boys and 18 girls), aged 18 to 48. These participants were divided into three groups when compared to a control group that was not offered feedback. The direct CF involved pointing out errors in the text and providing the correct form. They were given five minutes to look over the writing corrections and they were told to rewrite the story. They completed a third writing task two weeks later. The metalinguistic feedback group did not receive any individual CF; instead, it was given a metalinguistic explanation handout. The researcher read out the explanation to the class and then gave the participants five minutes to check the errors from the first draft and rewrite the story. The third written narrative was completed two weeks later.
The results indicated that L2 English students did not benefit from direct CF (F = 1.79 (2, 41), p = .180) in the error correction test after completing the first writing task and in the accuracy test when writing new pieces. Conversely, the metalinguistic explanation group showed significantly higher scores in the second piece of writing after receiving metalinguistic explanation of errors (F = 3.20 (4, 82), p = .017); post-hoc effect p = .002, r = 0.47).

Research Questions
Considering earlier formal accounts and empirical acquisition works on L2 English feedback, we formulate the following RQ: RQ 1. Are there differences between direct CF and indirect CF in L2 English learners" development of writing skills?
We predict that direct CF has a greater impact when compared to indirect CF in the L2 English learners" awareness of error correction (Coyle and Roca De Larios, 2014;Sarvestani and Pishkar, 2015); among others). These findings will result in the L2 English learners" accuracy of written productions.

The Participants
A total of 8 L1 Spanish children participated in this study, namely, 3 boys and 5 girls. Their ages range from 11 to 12 and they are studying sixth grade of Primary Education. They have been learning English for 6 years, and they have been exposed to English for 3 hours, which equals 3 English lessons per week. The participants" individual characteristics are very diverse in terms of their learning needs, their abilities and their pace while completing tasks.
All the students were informed about the study and they agreed to cooperate with the teacher. Their parents and their tutor were also informed. They were told that participation was optional, and it would not affect their grades in anyway.
Although there are 24 students in this class, some students had to be excluded from the analyses because they failed to produce any piece of writing. This means that their basic learning standards were not met for the present study given that they could not write an English text without the teacher"s help. Other students were excluded since they did not finish the writing tasks or did not hand in all the tasks required. The 8 target participants were randomly divided into two groups (4 participants per group). While one of the groups received direct CF on their writing tasks, the second group received indirect CF.

Data Extraction: Types of Feedback
Two types of feedback were used as the variables under analysis, namely, direct CF and indirect CF. The direct CF group received their tasks with the ill-formed target structures explicitly corrected. As illustrated in figures 3, all the errors were either underlined or indicated with symbols along with the correct forms provided nearby.

Figure-3. Example of Direct CF
Corrections can take a number of different forms, namely, by "crossing out an incorrect word, phrase, or morpheme, adding a missing word or morpheme, and writing the correct form above or near to the incorrect word/phrase/morpheme provided" (Ellis, 2008).
The indirect CF group had their writing tasks returned without any corrections provided. As displayed in figure  4, information about the correct way of writing was not shown; rather, errors were underlined, crossed or drawn in a circle.

Figure-4. Example of Indirect CF
When a grammatical target structure is provided in the indirect CF group, two symbols were used in the pieces of writing, namely, a + or a .

Procedure
Three written tasks were implemented. New pieces of writings were required in the test and in the post-test stages rather than asking participants to revise and rewrite their first drafts produced in the pre-test. These drafts are in line with earlier empirical works since there is an CF via creative pieces of writing has proven to be an effective methodological approach (Bitchener et al., 2005), also known as delayed-post-tests (Sheen, 2007). Teachers should ensure both the design of effective learning situations which help students" achievements and the teaching syllabus timing.
As shown in table 4, each task included a different topic, namely, Christmas in the pre-test, stories in the test, and animals in the post-test. The three tasks were designed for the students" development of the knowledge and the competences regarding the production of written texts. Three writing tasks in the pre-test, the test and the post-test were designed for the L2 English participants" development of the following linguistic features: (a) a specific grammatical structure (namely, present simple tense, "be going to", past continuous vs. past simple and comparative adjectives); (b) vocabulary related to the writing task topic (that is, Christmas, outdoors activities and animals); (c) textual coherence (i.e. the appropriate use of paragraphs, logical organization of ideas and punctuation).
As shown in table 5, the L2 English participants carried out the three task stages along 6 sessions that took place during the regular 1-hour lessons. Two sessions were conducted in the pre-test, two sessions were carried out in the test and two sessions were implemented in the post-test. Step 1: Introduction of the task and objectives Step 2: Teacher"s detailed explanation of the task Step 3: Participants" written production of texts and teacher"s collection of writings 2 Step 1: Teacher "s return of participants"writings Step 2: Participants" revision of corrections and doubt solving Step 3: Teacher"s collection of writings Each of the 3 stages described in table 5 plays a fundamental role in our study given that they allow to measure the development that will occur as a result of the CF provided. This is seen in the implementation of the pre-test and the post-test that will assess participants before and after the target test, respectively, as well as in the CF effectiveness.
The following lines will discuss the methodological procedure that was followed in each phase of the study. The pre-test included prompts which helped participants to develop the production of paragraphs. As illustrated in figure  5, prompts are given in a worksheet that required participants to write a Christmas letter addressed to an American pen pal.

Figure-5. Pre-test worksheet
Participants were asked to write a personal letter by taking into account the following linguistic and textual items: (a) the use of present simple tense when describing Spanish Christmas traditions; (b) the use of "be going to" when discussing Christmas holiday plans; (c) the use of Christmas-related vocabulary and (c) the use of paragraphs and punctuation.
Although participants were given oral instructions at the beginning of the task, they were also written in the digital whiteboard so that notes could be taken in the students" workbooks. In order to complete the written task, a 20-word limit per paragraph was given in the worksheet. All the participants were given 30-40 minutes to finish the task. All the worksheets were collected to be marked by the teacher based on the evaluation rubric designed for the task. As depicted in table 6, the participants" written production was measured by assigning a score from 1 to 10 that corresponds to the assessable items represented in the rubric (namely, grammatical accuracy, lexicon and textual coherence). In the test writing task, the participants were shown a visual prompt. They were asked to write their own texts in a white sheet. As illustrated in figure 6, the visual stimuli consisted of 3 pictures related to outdoors free-time activities and did not include any explanatory text. Although the learners were given oral instructions in English at the beginning of the experiment, directions were also written at the top of each sheet in the same language. The task procedure was similar to the one conducted in the pre-test. The slight difference when compared to the pre-test procedure is that participants were expected to write a story based on the visual prompt, the target grammatical structures under investigation was past simple in contrast with past continuous, and the target vocabulary was related to outdoor activities. The writings were again marked according the evaluation rubric and CF was provided.
The final task was designed as a post-test. It focused on writing a report based on an animal. Participants were given oral instructions at the beginning of the task and a presentation was also shown on the digital whiteboard to help participants understand what they were expected to do (Appendix 1). Grammar (that is, the use of comparative adjectives, the verb "have got", the modal "can" and the present simple tense), lexicon related to animals, and textual coherence based on the writing of 3 paragraphs were also assessed according to the evaluation rubric.

Data Analysis and Discussion of Results
As discussed in section 5.3, the three tasks under investigation required two sessions to be implemented, namely, one session was aimed for participants to write the texts and the second session was used when the corrected texts were returned to the participants with the direct or indirect CF provided by the teacher.
A general oral discussion of the most frequent errors found in the texts was shared in class. The participants were encouraged to ask questions about the error corrections and the CF they had in their own writings. Finally, the teacher collected the worksheets again to compare the participants" performances in the pre-test and in the post-test.
The aim of the present study is two-fold: (a) to shed light on whether CF is effective on L2 English students" improvement of written errors, and (b) to explore what type of CF (direct or indirect) proves to be more effective in terms of grammar, lexis and text organization when L2 English students develop creative writing.
Regarding the first objective, the data collected from the pre-test and the post-test were analyzed. An average score was given to each task. Scores ranged from 0 to 10 (0 being the lowest score and 10 being the highest score). As illustrated in table 7, the participants improved their performance from the pre-test (mean score = 7.33) to that in the post-test (mean score = 8.66), regardless of the CF they received (that is, direct or indirect). Moreover, the participants" differential mean performance in the direct and in the indirect CF groups was, on average, 1.33 out of 10. Thus, the means scores were higher in the post-test when compared to those ones in the pretest. Considering these data, we can infer that either direct CF or indirect CF helped L2 English participants to improve their writing accuracy. These results are in line with earlier empirical works on the influence of CF on L2 participants" writing performance (Ahmadi et al., 2012;Coyle and Roca De Larios, 2014).
The second objective of the present study refers to grammar, lexicon and text organization as linguistic factors to be considered when L2 English students benefit from written CF. As displayed in figure 6, data showed that there was an overall significant progress in grammar accuracy, text coherence and lexicon when both direct CF and indirect CF were applied. Nevertheless, when comparing across the three linguistic factors, participants obtained the highest scores in the post-test when grammar accuracy was assessed. Average grammar scores appeared to be higher when compared to textual coherence scores and lexicon assessment, respectively. These data confirm Sarvestani and Pishkar (2015) and Ismail et al. (2008) findings since students in these two studies also benefit from teachers" CF on grammar accuracy.
As reflected in table 8, direct CF group showed a slightly better performance (i.e. +0.25) when compared to the indirect CF. This is equally seen in the three categories, namely, grammatical accuracy, lexicon and textual coherence. Therefore, the participants in both groups seemed to improve their mean scores in each of the linguistic categories. However, the direct CF group benefitted more from teacher"s written CF over time. These results go hand in hand with earlier works on the higher positive effect of direct CF when compared to indirect CF on written accuracy García Mayo and Labandibar, 2017;Sarvestani and Pishkar, 2015).

Conclusion
In this study, we have investigated the influence of feedback on the development of L2 English students" writing skills. According to our results, L2 English students (age 11 to 12) seemed to improve their writing accuracy over time after receiving CF on their writings, as also reported in previous studies (Bitchener et al., 2005;Ellis, 2009;Ferris and Roberts, 2001;Sheen, 2007). In addition, providing CF feedback resulted in an upward pattern of improvement over time, regardless of the type of CF offered, that is, direct or indirect. Therefore, these results underline the importance of providing CF, particularly in a beginner L2 English level context where teachers" instruction and feedback are the most important ways through which learners can improve their written skills.
Furthermore, direct CF appeared to be slightly more influential on grammar accuracy when compared to lexicon and textual coherence. These data are in line with earlier works on measuring whether direct or indirect CF is most effective in L2 students writing accuracy (Sarvestani and Pishkar, 2015). Their data showed that the two types of CF are equally effective in L2 English learning despite the fact that low L2 English levels were argued to receive more direct CF to be able to use grammar more effectively.
The results of the present study are also consistent with Ferris and Roberts (2001) study since direct CF pointed to a better written performance when compared to indirect CF at L2 English learning stages. Ferris (2012), also concluded that, although most teachers should provide indirect CF to engage students in cognitive problem-solving tasks, L2 English learners with a low language proficiency should not be provided with this type of feedback since they may not possess the linguistic competence to self-correct and/or benefit from the teachers" written CF assessment.
The findings of the study, however, are limited by the small sample size. Only preliminary conclusions can be drawn when comparing the 8 participants" performance in the present work. Thus, we leave this study open to further work to observe the effects of direct CF on a larger number of students with different language level proficiencies. Future research can also look into different direct CF strategies to find out the most appropriate one depending on the learners" language proficiency.
Thus, provided that CF on Primary Education L2 English students" writing skills is not a sufficient way by itself to improve their written accuracy, oral discussions on specific grammar structures, lexical topics or text coherence techniques should also be taken into account. The use of other CF techniques such as the use of model texts is also a very interesting topic for further analysis (Coyle et al., 2018;García Mayo and Labandibar, 2017).