Investigating Noticing in Narrative Writing Tasks and its Effect on EFL Learners’ Writing Performance

Introduction

The attainment of L2 writing skill in relation to the noticing of language features is a
well-researched area of ELT. The proposed noticing conditions for improving L2 writing in pedagogical contexts have ranged from providing various types of negative evidence as corrective feedback to reformulation and modeling as well as learners’ self-initiated noticing through output-provoking tasks. In this regard, the crucial role of learners’ noticing of linguistic features in the overall development of writing skill has been widely discussed within theoretical predictions of Output Hypothesis (Swain, 1995; 2005). According to Swain, learners’ own production of language imparts at least four important types of effects to the process of language learning, one of them being the “noticing or triggering function” (Swain, 1995, p. 135).

Relying on the fruitfulness of noticing function of output, language production activities draw learners’ attention towards problematic linguistic features, making learners aware of what language resources they actually need in communicating their intended meaning. Engagement in written output leaves the learners in an optimal position to notice problems in their turnout efforts as a result of which they formulate hypotheses and try it out in their writing. Swain and Lapkin (1995), for instance, have reported that writing consciously offers the students the opportunity to make out linguistic deficiencies in their productive act of communication and try to repair them. However, the explanation on how such a noticing influences the handling of future input or any upcoming positive or negative feedback requires further scrutiny (Izumi, 2002; Shehadeh, 2002).

As Lapkin, Swain, and Smith (2002) argue, more noticing studies are required before we are left in the position of making a sound judgment on how L2 noticing leads to learning. In addition to exploring the nature of noticing through writing and through exposure to native models in Iranian EFL context, the present study aimed at inquiring into short-term and
long-term pedagogical effects of these two types of instructional noticing strategies in terms of grammatical accuracy in narrative writing tasks.

Literature Review

Schmidt’s Noticing Hypothesis (Schmidt, 1990), which hinges on the premise that the process of noticing empowers  language learners to turn input into intake, has boasted for enormous developmental effects and is now regarded as a mainstream SLA construct (Bergsleithner, Frota & Yoshioka, 2013). Once the constructive role of attention to language forms was substantiated through the extended research findings of the focus-on-form movement (Ferris & Roberts, 2001; Leow, 2000; Robinson, 1995, Schmidt, 1990; Tomlin & Villa, 1994), the focus partially shifted to the nature of noticing as a dynamic learning process and the way it might be operationalized and measured in the process of learning L2 (Philip, 2012).

The type of the linguistic form that learners regularly focus on depends on the nature and the goal of the task at hand. While learners’ attention is channeled towards grammatical or orthographic features in a reconstruction task such as dictogloss (Wajnyrb, 1990), in a free production task the learner-writers focus dominantly on lexical features (Garcia Mayo & Labandibar, 2017; Hanaoka, 2007). Learners’ attempted output at progressing stages of their interlanguage may include tentative use of lexical or grammatical items or even their resorting to communication strategies (Hanaoka & Izumi, 2012). When learners’ production deviates from target-like forms, they need the exposure to fine-tuned input in order to be able to incorporate accurate forms into their language.

Reformulations and models as two types of instructional strategies to provide learners with some evidence of positive feedback have been advocated in literature against more traditional types of feedback mainly involving negative evidence (Hanaoka, 2007; Qi & Lapkin, 2001). The argument is that exposing learners to some alternative ways of expressing meaning, which is beyond their present state of interlanguage, will engage learners in a more profound level of language processing than that allowed by learners’ existing linguistic resources. Some studies have indicated the utility of reformulation strategy in enhancing learners’ noticing (e.g., Qi & Lapkin, 2001; Lapkin, Swain, & Smith, 2002; Yang & Zhang, 2010).While Qi and Lapkin (2001) concluded that both reformulations and models “may be a viable way of helping non-native writers move nearer to native-writer norms” (p. 7), other researchers including Hanaoka (2007) have emphasized the superiority of models as an optimal form of noticing-provoking feedback. According to Hanaoka, models play a constructive role in the enrichment of learners' linguistic resources by offering them more viable alternatives to their trial output efforts as learners notice and incorporate them into their writings.

Hanaoka (2007) investigated the noticing function of output and the possibility of this noticing in affecting subsequent writing ability of Japanese EFL students. The key finding in his study was the priority given by Japanese learner-writers to the problems they had recognized during their output over the revisions suggested by native speakers throughout their reformulated models. By accommodating the research scheme offered by Hanaoka (2007) into Iranian EFL context, the current study aimed to investigate the nature of self-initiated noticing of linguistic forms as learners are involved in actual writing accompanied by input-enhancement opportunity provided by native model compositions. The main concern here was to see what types of linguistic features are dominantly noticed by learners. The other concern of the research was to inspect the future influences of noticing grammatical features on the quality of learners’ written performance in terms of grammatical accuracy.

Based on the research aims and the existing gap, the following research questions were formulated:

1. What type of language problems do Iranian EFL learners encounter during the composition in narrative tasks?

2. What do Iranian EFL learners notice when they compare their written output to native speaker models?

3. What is the Iranian EFL learners' attitude toward receiving models as feedback in writing tasks?

4. Does Iranian EFL learners' noticing of grammatical forms during writing tasks have any short-term effect on the written accuracy of their later performance?

5. Does Iranian EFL learners' noticing of grammatical forms during writing tasks have any long-term effect on the written accuracy of their later performance?

The first two research questions were scrutinized through a qualitative, descriptive procedure. The other three questions were approached quantitatively. Meanwhile, bearing upon the theoretical underpinnings in operation, we anticipated meliorated short-term and long-term pay offs in terms of the accuracy of EFL narrative composition writing as a result of noticing structural forms by being exposed to native-speaker modeling feedback.

Methods

Design

The study resorted to both qualitative and quantitative methods of data analysis in a complementary fashion in order to delve into the nature and pedagogical outcomes of the noticing process in the short and long run. The type of linguistic features noticed by learners during their own production and following the exposure to native writing models was studied through qualitative analysis. To scrutinize the actual effect of noticing grammatical forms on L2 learners’ performance in terms of grammatical accuracy, their writing samples of narrative writing tasks were collected in three planned occasions, i.e., pre-test, post-test and delayed post-test. Hence, the design on the quantitative side of the study was
quasi-experimental based on repeated measurement.

Participants

The sample consisted of 43 learners of English as a Foreign Language (19 males and 24 females) from Zabansara Language Institute in Malekan, East Azerbaijan, Iran. According to the institute’s placement standards, the English proficiency level of the participants was assumed to be intermediate. Their age ranged between 17 and 34, ninety-seven percent of them being native speakers of Turkish. The participants were selected according to the principle of convenient sampling since they were recruited from intact classes. Forty-three respondents out of total 65, having willingly consented to participate, were included in the study based on the results of a Preliminary English Test (PET) administered prior to study proper. The mean scores of the learners on the PET were calculated along with their standard deviation. The candidates who scored within the range of one standard deviation above and below the mean were selected, leaving out22 outliers.

Instrumentation and Tasks

The writing task involved a story writing activity based on a picture prompt (Appendix). Each participant received the same six descrambled pictures. To complete the task, the participants had to rearrange the pictures in the right order for presenting a written narration of the story implicated by the pictures. Thestory-completion task was adapted from the Wet Paint Sign task in the SPEAK Practice Test ofIU School of Liberal Arts website (available at https://liberal arts. iupui.edu/english/files/documents/pdf/SPEAK_Sample_2011.pdf), and it was used to give the participants the chance of noticing linguistic problems as participants were engaged in a multi-stage writing task including four stages following Hanaoka (2007). First, the learners wrote a narrative based on the picture prompt (Stage1). Second, they were given an opportunity to notice gaps between their current knowledge of English forms and native speaker forms by comparing their own writing with two native-speaker models (Stage 2). Following the rationale raised by Hanaoka (2007), two model writings were used in this study to obviate the risk of mechanical copying from one given model, and to encourage spontaneous problem-solving attempts. Third, the participants were asked to rewrite their early draft inspired by what they had noticed in models (Stage 3). Finally, the learners rewrote their original text once again after a time laps of two months (Stage 4).

A single-item, Likert-type question on a scale of 1 to 5 (1 being ‘Not at all’ and 5 ‘Very much’) was included at the top of Sheet 2 in order to evaluate learner's attitude toward reading model texts. Furthermore, the researchers made use of self-report in the form of note-taking in order to have an appropriate measure for the noticing process (Garcia Mayo & Labandibar, 2017).

Procedure

The four-stage procedure for carrying out the experiment was a replication of the procedure adopted by Hanaoka (2007). The instructions were rendered in Persian to the learners throughout task implementation to ascertain students' acquaintance with the procedure and the purposes entailed in performing the task. In addition, the students were allowed to take notes.

In the Stage 1 writing task (pretest), the learners were availed with Sheet 1, Sheet 2, and the pictures. While they wrote their narratives on Sheet 1, they jotted down their notes on Sheet 2 about the linguistic points they felt short of producing while composing narratives (PFNs). Following Hanaoka (2007), the directions appeared at the top of Sheet 2, illustrated by some sample notes in Persian corresponding to: ‘I don’t know how to say X in English’, ‘I wrote X, but I’m not sure if this is correct’, ‘what is the past tense of X?’ and ‘I’m not sure whether the picture is describing X’. At the end of Stage 2 the participants were informed of being given two native model writings related to the same picture narration task. To implement a single-item attitudinal enquiry, the students were asked to demonstrate their level of interest in reading the model texts at the top of Sheet 2 ranging from 1 (Not at all) to 5 (Very much). The sheet 1 and the pictures were left with the students for later activities.

In the Stage 2 task, the Sheet 3 and two native-speaker models already collected from two Native Americans teaching English at a Turkish school were handed out. The students wrote on Sheet 3 their noticed linguistic points (FNs) as they compared their own writings with the model texts. At the end of Stage 2 task, the model text and Sheet 3 were taken away.

As a post-test, in the Stage 3 task, the students put the revised version of their first draft on Sheet 4. Two months later, we implemented the delayed posttest through which the participants were once again given their original text and were asked to rewrite it on Sheet 5.

As mentioned earlier, students’ notes on the problems they encountered while composing a narrative acted as the indication of the cognitive process of noticing. This study also adopted Hanaoka's four categories of lexis, grammar, content, and other (e.g., discoursal features or prefabricated expressions) in classifying the problematic features noticed (PFNs) and features noticed (FNs) (Garcia Mayo & Labandibar, 2017; Hanaoka, 2007; Hanaoka & Izumi, 2012).

Data Analysis

The descriptive and inferential statistics were employed to capture the nature of data according to the quality-based and quantity-based research questions. Research questions 1 and 2 which concerned the type of linguistic problems learners noticed during the writing task and after the exposure to native models, respectively, were examined through qualitative analysis of the descriptive data. Research question 3 was inspected based on the statistics from the single-item self-report attitudinal query. On the quantitative side, paired-samples
T-tests were run to answer the research questions 4 and 5.

Language Problems Encountered by Learners

In examination of Research Question 1, the descriptive statistics of PFNs, i.e., the problematic features noticed by participants during the composition of their narrative texts (Hanaoka, 2007) have been displayed in Table 1.

Table 1. Descriptive Statistics of PFNs in Stage 1 Writing Task

The participants noticed a total of 443PFNs which makes an average of10.3 PFNs per participant. As far as the nature of the PFNs is concerned, the noticed features were mainly lexical (46.72%). The averages of PFNs per participant were 4.81, 2.27, 1.88, and 1.32 respectively for lexical, other, grammatical, and content type. The summary of the percentages is illustrated in Figure 1.

Figure 1. The Percentages of Types of PFNs

Language Features Noticed by Learners due to Native Speaker Models

The descriptive data concerning the features that were noticed in the Stage 2 composition task after being exposed to native models (FNs), as queried in the Research Question 2, are demonstrated in Table 2. Overall, the participants noted a total of 486 FNs which makes an average of 11.3 features per participant. When compared with stage 1, a greater number of features were noted by students. Concerning the type of features noticed, the largest number of cases of FNs concerned lexical features (51%). This finding is in congruence with the Stage 1 noticing pattern. In contrast to the first stage, the second most frequent FN was content category with 24 percent. The third most common type of problem faced by Iranian EFL learners was grammar (17 %). And finally the least attended item by learners was other comprising six percent of all FNs due to native model writings. In line with the first stage, the range of lexical problems noticed by the students was 9 which is remarkable when compared with grammar, content, and other which were 5, 4, and 5, respectively.

Table 2. Descriptive Statistics of FNs in Stage 2 Writing Task

Table 3 displays the similarity level of the 250 lexical PFNs at Stages 1 and 2.According to the data, 130 features from stage 2 had something to do with PFNs in Stage 1. Arguably, these FNs targeted participants' lexical needs in language production for which they lacked the appropriate lexical knowledge or they had come up with some less appropriate substitutions. The remaining 120FNswhich never appeared at the Stage 1 PFNs involved those lexical needs the participants had never noticed during their own written production.

Table 3. The Identical Nature ofStage2 Lexical FNs and Stage1 PFNs

Learners’ Attitude toward Reading Model Texts

Learners’ eagerness in using native speaker narratives as models for revising their original texts was assessed with a single-item query on a scale of 1 to 5 (Likert scale). The mean score was 4.1 (SD= .89) which settles down a high level of interest in participants to consider the native speaker texts as a means of improving their writings. Even while conversing with the researchers, they were very zealous for being given the native models and they claimed that they felt much more confident for the subsequent writings. This finding is in line with Hanaoka (2007) where learners were highly motivated to receive models ( M=4.3; SD=.86) and in contrast with Garcia Mayo and Labandibar (2017) who came up with a relatively low level of learners’ willingness to receive models (M=2.7; SD=.97).

Short Term Effects of Noticing on Accuracy of Narrative Composition

An important aspect of the current study was to see whether students' noticing of grammatical features during their independent writing task had any significant effect on later writing performance of learners. The objective here was to statistically compare the portions of PFNs of Stage 1 and FNs of Stage 3 that had been incorporated into learners' revisions.  Therefore, a paired samples T-test was run to examine the difference between incorporated features in the two writing tasks at stages 1 and 3. The frequency count of grammatical mistakes in the original writing of the students were rendered as scores and compared with their counterparts in learners' second writing at stage 3. The results are presented in Tables 4 and 5.

Table 4. Descriptive Statistics for the Number of Grammatical Mistakes at Stages 1 and 3

Having checked for normality of distribution of scores, a paired samples t-test was run to compare the accuracy rate of learners’ writings in stages 1 and 3 (Table 5).

Table 5. Paired Samples T-test for Grammatical Accuracy at Stages 1 and 3

Table 5 indicates a significant difference between the grammatical accuracy of students prior to and after the noticing. The number of grammatical mistakes decreased statistically from stage 1 to stage 3 (t =11.58, p<.001). In other words, the grammatical accuracy in the narrative writing task improved overwhelmingly. The mean decrease in the number of grammatical mistakes was 1.51 with a 95% of confidence interval ranging from 1.24 to 1.77.

Long Term Effects of Noticing on Accuracy of Narrative Composition

To answer the fifth research question, the number of grammatical mistakes on stage 3 and stage 4 writings of the learners was compared. A paired-sample t-test was conducted to compare students' accuracy performance in the two occasions (Table 6).

Table 6. Descriptive Statistics for the Number of Grammatical Mistakes in Stages 3 and 4

Having checked for normality of distribution of scores, a paired samples t-test was run to compare the accuracy rate of learners’ writings in stages 3 and 4 (Table 7).

Table 7. Paired Samples T-test for Grammatical Accuracy in Stages 3 and 4

As shown in Table 7, despite the slight increase in the number of mistakes in the performance on delayed post-test, the p-value of comparing the grammatical mistakes between the post-test and delayed post-test is much higher than .05 which means that there is no significant difference between the grammatical accuracy of students from pre-test to the delayed post-test. In other words, there was no statistically significant change in the number of grammatical mistakes from stage 3 (M=.37, SD=.57) to stage 4 (M=.57, SD=.59), t (42) =11.58, p>.001. The mean decrease in the number of grammatical mistakes was -.13 with a 95% of confidence interval ranging from -.24 to -.3. Hence, it is established that the grammatical accuracy of students in the narrative writing task remained fairly stable from post-test to the delayed post-test. The durability of noticing after an interval of over two months is an indication of the effectiveness of self-initiated noticing through writing task and exposure to native model compositions in language achievement in general and grammatical accuracy in particular.

Discussion

The initial concern of the study was to investigate what aspects of language Iranian EFL learners notice when they get involved in free narrative writing task. In line with the previous studies in this regard (Garcia Mayo & Labandibar, 2017; Kang, 2010; Hanaoka, 2007; Hanaoka & Izumi, 2012; Qi & Lapkin; Williams, 2001; Yang & Zhang, 2010), EFL learners in Iranian context indicated a prominent tendency toward noticing lexical problems throughout the unfocused output task. This finding provides further evidence in support of the alleged “lexically-biased” (Garcia Mayo & Labandibar, 2017) and “semantically-oriented” (Hanaoka, 2006) noticing during output activities which is mainly explicable by the limited cognitive capacity of EFL users. Language production requires primarily decision makings about conceptual meanings that are crystallized as lexemes in the process of output (Levelt, 1989; Manchon, 2011). The fact that learners concentrate their attentional resources chiefly on lexical problems rather than grammatical problems implies pedagogically that instructional tasks and procedures for writing should not be devised in a way that they overoccupy learners' mind with grammatical concerns.

The extent to which Iranian EFL learners were inspired by being exposed to native narrative models of writing was another concern of the current study. In strict compliance with previous research (Yang & Zhang, 2010; Kim, 2015; Martinez-Esteban & Roca de Larios, 2010; Saeidi & Sahebkheir, 2011), a remarkable portion of students’ noticing of problematic features was lexical (51 %), content features accounting for 24 % of their noticing. This evinces that native-speaker models play a useful role in increasing the rate of learner noticing and transmitting the focus from grammatical accuracy to the content of writing. The finding is in line with Hanaoka and Izumi (2012) where the model text turned out to be equally beneficial in promoting the noticing of both overt and covert
(pre-articulated) problems. The findings here provide further evidence to related studies (Garcia-Mayo & Labandibar, 2017; Hanaoka, 2007; Hanaoka & Izumi, 2012) according to which output production through independent writing helps learners to recognize problematic features they have not already been conscious about. This level of deficiency in using language is what Hanaoka (2007) calls “holes”. These shortcomings are compensated for by later noticing of the gaps between learners’ present knowledge and native-like performance (Swain, 1995).

The results concerning learners’ attitude toward using model texts indicated a rather positive attitude of participants toward modeling in writing as the perception enquiry used to assess the extent to which native models appealed to the learners indicated a high level of interest. This finding concerning Iranian EFL learners is in line with the high eagerness of participants reported by Hanaoka (2017) and Yang and Zhang (2010), but in contrast with low eagerness reported by Garcia Mayo and Labandibar (2017). This contrasting attitudinal pattern has been attributed by Garcia Mayo and Labandibar to the differences in learners’ ages in the two studies (university students in Hanaoka (2007) and adolescents in theirs). The participants in the present study were adult learners aging between 17 and 34 which covers late teenage to adulthood. Therefore, the soundness of ascribing learners’ motivational tendencies toward using models to learners’ age is not corroborated here. Further studies are required to account for these contradictory findings.

A main concern of the present study was to examine short-term and the long term effects of noticing grammatical forms through self-initiation and modeling on the accuracy of narrative composition written by Iranian EFL learners. Research questions4 and 5concerned the effects of noticing grammatical forms on the accuracy of narrative composition. In line with previous studies (Saeidi & Sahebkheir, 2011), the findings verified the
input-enhancement role of modeling as it has been claimed in the literature. A statistically significant difference was obtained between grammatical accuracy of the learners’ writing between the first and the third stage. The difference revealed that the learners did see the gaps and attempted to incorporate the noticed grammatical features in their later revisions. The deduction is that noticing is not a one shot momentous phenomenon taking place at a critical stage of education; rather, noticing linguistic deficiencies by the language learner can be piecemeal and progressively formed as the learners’ input and output processing capacities interact with each other. This argument is specifically ratified here by the fact that learners learned to use those grammatical features from the models that their attention had already been drawn to during their own output. This was true for both their immediate and delayed attempts in rewriting their original texts.

Another proof for the progressive nature of noticing linguistic gaps comes from the examination of the fifth research question. In the delayed rewriting task with a two-month time laps, a comparably smaller number of solutions picked up from native models were recaptured in revising the original text. We witnessed about 37 % increase in the number of the grammatical features that students had noticed earlier but were unable to incorporate into their revised text in delayed post-test. However, the retention of the grammatically sound forms did not vary significantly from that of Stage 3 revision regardless of the fact that the mean score of their overall performance dropped slightly. This finding provides further evidence to establish the long-term effects of the instructional sequence of initial noticing through output production combined with the provision of enhanced input. The slight fall in learners' performance in using PFNs accurately in their writings two months later underlines the cyclical and progressive nature of grammatical development and the urgency for complementary activities in a course of time subsequent to learner’s initial noticing of the “holes” and later noting of the “gaps” and further establishment of their “gains”.

Conclusion

The present study provides a more comprehensive grasp of the nature of noticing during written output by Iranian EFL learners by shedding light on the type of problems noticed as well as the short-term and long-term consequences of noticing learners’ later written performance. During written production, students noticed particular gaps and/or holes in their L2 knowledge, were inspired by utilizable grammatical solutions in the model input, and fused them into a follow-up written output task. It is argued that the participants’ engagement in output activity has increased their awareness of the required linguistic features, facilitating their noticing of those features through exposure to positive feedback they received in the form of models. The findings provide further evidence for the constructive role of output in language learning.

This study relays two conceptual implications concerning the essential qualities attributed to noticing in language learning. First, it attests to the isochronal nature of grammatical development where noticing plays an important role. In the meantime, noticing less seems to be a “once and for all” event than a cyclical phenomenon. The current study indicated that providing positive evidence promoted students’ initial noticing of linguistic features from their output. However, this noticing requires recurrent consolidation involving interval output-based and input-based learning tasks to ensure a lasting effect. Further studies involving such interval tasks with a more longitudinal focus are suggested. Second, although learners adopted a predominantly lexico-semantic orientation in noticing through both their own output and exposure to models, their limited expenditure of attentional resources on grammatical problems has been sufficiently effective in promoting the accuracy performance of learners on local tasks. To complement our understanding of how noticing certain grammatical items end up with improvements in written accuracy, we need focused studies where particular structural forms become the subject of noticing and are traced well into learners’ accuracy performance on those specific items. Researchers following such a design need to commit themselves to more longitudinal instructional patterns in order to inhibit overgeneralizations.

Teachers, teacher educators, and EFL learners can benefit from the findings of this research by receiving clues on the types of the most frequently-noticed gaps cited by the learners and tuning the teaching or learning practice to learners’ orientations in noticing gaps and even orienting learners’ attention to incorporate certain types of noticing in the process of language production. Over all, the process of written output is overwhelmingly lexical and improvements in grammatical accuracy are the by-product of subsidiary noticing of grammatical forms.