Output, noticing, and learning: An investigation into the role of spontaneous attention to form in a four-stage writing task

Language Teaching Research 11 ,4 (2007); pp.  459–479

Output, noticing, and learning: An investigation into the role of spontaneous attention to form in a four-stage writing task

Osamu Hanaoka Tokyo International University, Japan

While the noticing function of output has been increasingly researched by a number of applied linguists, the nature of such noticing and its effect on subsequent learning in the context of EFLwriting have not been fully investigated. In a four-stage writing task consisting of output, comparison, and two revisions, this study examined what Japanese college students spontaneously noticed (1) as they wrote a story in response to a picture prompt (Stage 1), and (2) as they compared their original writing to two native-speaker models (Stage 2), and how such noticing affected their immediate and delayed revisions (Stages 3 and 4). The results suggest that the participants noticed overwhelmingly lexical features as they autonomously identified their respective problems, found solutions through models, and incorporated them in subsequent revisions. Regarding proficiency effects, more proficient learners noticed significantly more features than less proficient learners when they compared their original output with two models. Another finding was that, among the features of the models that the participants noticed, those that were related to the problems that they had noticed through output were incorporated at a higher rate and were also retained longer than unrelated features. Implications drawn from these findings are discussed.

I Introduction

There is now a general consensus that noticing is a prerequisite for learning to take place (e.g. Ellis, 1995; Robinson, 1995; Schmidt, 1990, 1995, 2001; Schmidt and Frota, 1986; Skehan, 1998). One important aspect of noticing research has concerned the role of output in promoting noticing. Swain (1995, 1998 , 2005) has proposed four functions of output, one of which is the noticing/ triggering function. It is claimed that, through output activities such as speaking and writing, L2 learners become aware that they cannot say what they want to say in the target language. However, how such spontaneous attention to form affects the subsequent learning process has not been adequately researched. Swain and Lapkin (1995) reported that their students consciously recognized linguistic problems through the act of writing and modified their output. However, Lapkin et al. (2002) argue that more L2 noticing studies are needed to provide

Address for correspondence: Osamu Hanaoka, 2-13-19-201, Kamisoshigaya, Setagaya-ku, Tokyo, 157-0065 , Japan; email: hanaoka@k8.dion.ne.jp

© 2007 SAGE Publications 10.1177/1362168807080963

direct empirical evidence that noticing leads to learning. Shehadeh (2002) also points out that while past research has focused on the occurrence of modified output, more research is needed to investigate how producing output can lead to acquisition. While some studies (e.g. Izumi, 2002; Morgan-Short and Bowden, 2006) have addressed this issue and shown the positive effects of output, relatively few studies have been conducted in the context of L2 writing.

In L2 writing contexts involving spontaneous focus on form, Qi and Lapkin (2001) conducted a case study in which two ESL learners at different levels of proficiency engaged in a three-stage writing task. The results indicated that noticing in the composing stage influenced noticing in the feedback processing stage, and that quality of noticing was an important factor in the improvement of the final written product. Moreover, the study suggested that quality of noticing may be related to the proficiency level of the learner. In this and other studies (e.g. Adams, 2003; Lapkin et al., 2002; Swain and Lapkin, 2002), a feedback technique known as reformulation was used and shown to be an effective feedback tool enabling learners to make cognitive comparisons and notice gaps between their own output and their reformulated versions. Qi and Lapkin suggested that ‘the positive modeling of native-like writing may be more helpful to the learner than error correction’ (p. 286). However, few studies have explored in this context the role of native speaker modeling which is not contingent on learner output. Building on the work of Qi and Lapkin, the present study investigated two broad issues: the role of spontaneous focus on form in L2 writing, and the potentially unique role of native speaker models as a feedback tool. This paper will discuss the data pertaining to the first issue (for the second issue, see Hanaoka, 2006).

1    Learner-initiated focus on form

Who should initiate focus on form is an important question to be answered in noticing research. In some studies, focus on form is planned in advance and learners are prompted, through task designs and input enhancement, to notice certain features of the input. However, it has been pointed out (e.g. Izumi et al., 1999; Kowal and Swain, 1994, 1997; Long and Robinson, 1998) that, in planned focus-on-form conditions, teachers’intended pedagogical focus does not always match the actual attentional focus of the students.

It should be noted that the original notion of focus on form proposed by Long (1991) assumed incidental focus on form in which linguistic problems are addressed ‘as they arise incidentally in lessons whose overriding focus is on meaning, or communication’ ( p. 46). Williams (2001) calls into question the effectiveness of planned focus on form and argues that ‘[I]f the effectiveness of FonF is ultimately determined by learner need, then it is essential to examine the episodes in which the learners themselves choose to focus on formal aspects of language’ (p. 304).

Ellis et al.’s (2001) study also points to the importance of respecting the learner’s internal syllabus. They analyzed 12 hours of meaning-focused instruction for focus-on-form episodes (FFEs). They found that there were many pre-emptive FFEs in which either the teacher or a student initiated attention to form when no actual problem had arisen. In their data, pre-emptive FFEs occurred as frequently as FFEs that occurred in reaction to students’errors (reactive FFEs). They also found that the majority of the pre-emptive FFEs were initiated by students rather than by the teacher and that the students were more likely to incorporate a form into an utterance of their own if the FFE was selfinitiated. It may be that ‘the most useful feedback comes from those areas of mismatch which students are themselves able to identify, because those areas will accord with the stage of their skill (or interlanguage) development’ (Johnson, 1988: 93). One of the important avenues of research, then, is to explore the nature of learner-initiated focus on form.

2    Factors influencing spontaneous attention to form

The findings of past research seem to indicate that, depending on the task conditions, learner attention is directed at different aspects of an L2. For instance, in grammar dictation tasks known as dictogloss (Wajnryb, 1990), where the purpose of the task is to reconstruct a passage as accurately as possible, learners tend to pay attention to grammatical as well as orthographic accuracy. For instance, Kowal and Swain (1994) reported that 70% of the language episodes produced during a collaborative dictogloss task dealt with accuracy. On the other hand, some studies in which learners engaged in producing original texts or in free interaction show that they tend to focus on lexical elements. In Swain and Lapkin’s (1995) study, 18 students were asked to write about an environmental issue. An analysis of the students’ think-aloud protocols revealed that, in the original editing phase, half of all language-related episodes involved lexical search. In oral contexts, Ellis et al.’s (2001) study showed that more than 66% of the student-initiated focus-on-form episodes addressed vocabulary. Williams (2001) confirmed that learners were capable of spontaneously attending to form and reported that an overwhelming 80% of all languagerelated episodes were lexical. Mackey et al. (2000) also reported in their study of learner perception of oral feedback that their learners of Italian as a foreign language were most often thinking about lexis.

Some studies indicate that the proficiency level of learners may also affect both the quantity and quality of their language-related noticing. Swain and Lapkin (1995) and Williams (2001) reported that more advanced learners initiated greater numbers of language-related episodes, while Qi and Lapkin (2001)  noted that writers with a relatively high level of L2 proficiency not only engaged more in language-related noticing but also better understood the nature of noticed gaps.

II The study

For the present study, a writing task was designed to provide the participants with the opportunity to (1) notice linguistic problems as they wrote a narrative in response to a picture prompt (Appendix A) (Stage 1), (2) notice gaps between their interlanguage forms and target forms based on a comparison of their original draft with two native-speaker models (Appendix B) (Stage 2), (3)  revise their original text based on what they noticed (Stage 3), and  (4) revise their original text again after an interval of over two months (Stage 4). There are two important differences between this study and that of Qi and Lapkin (2001). First, noticing in this study was measured by means of notetaking. Second, two native-speaker models of writing instead of individually reformulated texts were used for feedback to investigate the potentially unique role these independent models play in promoting learner noticing. Two models were used in this study to reduce the chance of the participants’mindless copying from a single model text, and to increase the chance of providing solutions to the problems that the participants incidentally noticed. The research questions posed in this study followed those investigated by Qi and Lapkin (2001: 284) and included the following:

1)      What aspects of language do L2 learners notice while composing a narrative on their own? (Nature of Stage 1 noticing)

2)      What do L2 learners notice as they compare their text to native-speaker models of writing? (Nature of Stage 2 noticing)

3)      What are the effects of Stage 1 and Stage 2 noticing on subsequent revisions?

4)      What are the learners’ proficiency effects on noticing and incorporation?

III Method

1    Participants

The participants were 37 Japanese students in two ability-based sophomore classes at a women’s university. One class was the most advanced class in one department, and the other class was an intermediate level class in another department. For convenience, the former class will be referred to as Class A and the latter class as Class B. The data to be analyzed for this study come from a total of 37 students (17 students from Class A and 20 students from Class B) who completed all the stages of the writing task.

2    Procedure

The students were asked to write a story in response to a picture prompt (see Appendix A). It was taken from the interview section of the second grade STEP Test in Practical English Proficiency conducted in 2000 by the Society for Testing English Proficiency, Inc. The picture prompt, consisting of two picture frames, helped to control the propositional content of the story that the students wrote. Throughout the task, instruction was given in Japanese and the students also took notes in Japanese. (The English quotations of their notes hereafter are my translations.)

In the Stage 1 writing task (pretest), the students were provided with Sheet 1, Sheet 2, and the pictures. On Sheet 1, they wrote a narrative and on Sheet 2, they took notes on whatever problems they noticed as they wrote on Sheet 1. The directions were written at the top of Sheet 2 with the following specific examples of note-taking in Japanese: ‘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’. This stage took 15 minutes in Class A, and 18 minutes in Class B. At the end of the Stage 1 task, the students were told that they would now receive native speaker models. They were then asked to indicate at the top of Sheet 2 how eager they were to read them on a scale of 1 to 5 (1 being ‘Not at all’ and 5 ‘Very much’). Then, Sheet 2 was collected. The students kept their original text (Sheet 1) and the pictures for the Stage 2 task.

In the Stage 2 task, which immediately followed the Stage 1 task, the students received Sheet 3 and two native-speaker models (see Appendix B). One of the models was written by an American teacher who taught English at the same university, and the other was written by a Canadian teacher who taught English at a different university. For ease of reference, the two models were titled (A) and (B) respectively. The students were asked to write on Sheet 3 whatever they noticed as they compared their original text with the models. Specific examples of note-taking were provided at the top of the sheet in Japanese. They were: ‘I couldn’t say X, but (A) puts it Y’, ‘(A) says X, but (B) says Y’, ‘I was impressed by (B)’s interpretation of one or the other picture’. This task took about 10 minutes in each class. At the end of the Stage 2 task, the native-speaker models and Sheet 3 were collected. The students kept their original text (Sheet 1) and the pictures to be used in the Stage 3 revision task.

In the Stage 3 task (post-test), the students were asked to rewrite their original text on Sheet 4. This task took 15 minutes in Class A, and 13 minutes in Class B. The Stage 4 task (delayed post-test) was conducted more than two months later after the summer break. The students had not been informed of the task in advance. For this task, they received their first draft (Sheet 1) and were asked to rewrite it on Sheet 5. This task took about 15 minutes in each class.

IV Analysis

For the sake of analysis, noticing was operationalized in this study as selfreports in the form of note-taking. Note-taking as a noticing measure entails some important merits and demerits (for discussions of strengths and weaknesses of various assessment measures, see Izumi, 2000, 2002; Jourdenais, 2001) . The weakest feature of note-taking may be the degree to which learners’ reports include what they noticed, because the act of writing is physically demanding and time-consuming. However, note-taking has some important advantages. First, it is likely to indicate the locus of learners’focused attention. In other words, learners’ notes are a good indication of where they allocated most of their attention. Second, descriptive notes may provide a clue as to the nature of learners’ awareness. Descriptive notes, for instance, may reveal whether a certain feature is new or already familiar to the learner. Third, as an online measure, it is less likely to be affected by memory loss than offline measures. It should be noted that reactivity of using note-taking was not included as a variable in this study. However, it is an important issue to be taken into account when interpreting the results of this study (cf. Adams, 2003; Bowles and Leow, 2005; Leow and Morgan-Short, 2004).

To investigate what aspects of language the students noticed, it was necessary to code into categories the problematic features noticed in Stage 1 (PFNs) and the features noticed in Stage 2 (FNs), and the features incorporated in Stage 3 and Stage 4. Qi and Lapkin (2001) coded language-related episodes ( LREs) broadly into lexical, form, and discourse types. This study classified PFNs and FNs into four categories: lexis, grammar, content, and other. The identification of lexical and grammar features followed Williams’s (2001: 330–31)  classification of LREs :

The lexical category … essentially includes anything that would fit into the categories “What does this mean?” “How do you say/spell this?” or “Which word should I use here?” In contrast, the grammar/morphology/syntax category includes LREs that revolve around tense choices, grammatical morphology, word order, and other features generally considered part of grammar.

The following examples illustrate the way in which the Stage 1 PFNs and Stage 2 FNs were coded into the four categories.

1    Lexis

‘I don’t know how to say “kotsu-jutai” in English.’ ( Stage 1 PFN )

‘“Angry” is better than “annoy,” I think.’ ( Stage 2 FN )

In the first example, the student wanted a lexical item for a Japanese expression. The second example involves the evaluation of two lexical choices.

2    Grammar

‘getobjectto? Usage’ ( Stage 1 PFN )

‘the man  a man I should say “a man” the first time.’ ( Stage 2 FN )

The student in the first example is not sure about the syntax of the verb get, while the second note suggests that the student noticed the grammatical usage involving the articles a and the.

3    Content

‘I wanted to express the man’s feelings while he was in the traffic jam.’ ( Stage 1 PFN )

‘40 minutes late. I didn’t look at the clock.’ ( Stage 2 FN )

The first note suggests that the student was unable to express the man’s feelings because of her inadequate linguistic resources. However, she stopped short of mentioning any specific vocabulary items or grammar features. Therefore, this note was classified into the ‘content’ category. In the second example, the student noticed in one of the native-speaker models a comment that she had not included in her narrative.

4    Other

The analysis of the data in the present study revealed that some notes were difficult to classify into any of these categories. In the first example shown below, the student noted the difficulty of the task without referring to any specific aspects of the difficulty. The second example represents cases where students simply noted their preference for one model over the other. Instances like these were grouped into the ‘other’ category:

‘It is difficult to describe the second picture.’ ( Stage 1 PFN )

‘I personally like B better. It’s easier to understand.’ ( Stage 2 FN )

A second researcher also coded the Stage 1 PFNs and the Stage 2 FNs. The two raters agreed on the classification of 130 (99.2%) out of the 131 Stage 1 PFNs and on all of the 162 Stage 2 FNs.

In this study, incorporation of features that the participants noticed from the native-speaker models was examined both quantitatively and qualitatively. It should be noted that, for quantitative analysis, a feature attempted in the revisions was counted as an instance of incorporation even if it contained minor errors. For instance, if a learner noticed the lexical item ‘traffic jam’ but misspelled it in a revision as ‘trafic jam’ or if a learner noticed the collocation ‘caught in a traffic jam’ but used the inappropriate form ‘catch in a traffic jam’, it was still counted as an instance of incorporation.

To examine proficiency effects on noticing and incorporation, a higherproficiency and a lower-proficiency group were set up based on the students’ scores on a cloze test. Those students who scored 10 or more out of 25 comprised Group A (N10) , and those students who scored 5 or less formed Group B (N14) . Thirteen other students who scored more than 5 but less than 10 were not included in either group. The mean scores of the two groups were statistically different (t11.00, df22, p.000) . It turned out that all the students in Group A were from Class A, and all the students in Group B were from Class B. For correlational analyses, data from one student was excluded because she did not take the cloze test.

IV Findings

1                    What aspects of language do L2 learners notice while composing a narrative on their own? (Nature of Stage 1 noticing)

The frequencies and proportions of problematic features that the participants noticed while writing their original narrative (PFNs) are shown in Table 1. They noted a total of 131 PFNs, or an average of 3.5 features per participant. These PFNs were overwhelmingly lexical. The percentage here is even higher than in Williams’(2001) study, which reported that about 80% of the LREs involving classroom interaction were lexically oriented.

2                    What do L2 learners notice as they compare their text to nativespeaker models of writing? (Nature of Stage 2 noticing)

At the end of the Stage 1 task, 35 students indicated how eager they were to read the native speaker models. On a scale of 1 to 5, the mean score was 4.3 (s.d..86) , while the median score was 5. This indicates that generally the participants were strongly motivated to study the models that were presented in the following stage of the task.

The frequencies and proportions of features noticed in the Stage 2 comparison task (FNs) are shown in Table 2. In this stage, the participants noted more features than in Stage 1 with each student generating an average of 4.4 FNs. Consistent with the Stage 1 noticing, by far the largest proportion of the FNs was lexical. One notable difference, however, was that nearly 30% of the FNs had to do with the content of the story. Some students noted how the models’ interpretations of the pictures differed from their own. For instance, one student commented that she had written that the meeting was to start at 10:30 while according to Model A it started at 10.

Table 3 displays the relationship of the 102 lexical FNs to the Stage 1 PFNs. It shows that 61 were related to the Stage 1 PFNs. These FNs involved noticing of words that the participants had wanted but had not been able to use or address by alternative forms, words for which they had made different

	All participants (N37)					Group A (N10)				Group B (N14)
	n	%	mean	s.d.	n	%	mean	s.d.	n	%	mean s.d.
Lexis	121	92.4	3.3	2.27	41	93.2	4.1	3.51	45	95.7	3.2	1.93
Grammar	3	2.3	0.1	0.28	0	0.0	0.0	0.00	1	2.1	0.1	0.27
Content	1	0.8	0.0	0.16	0	0.0	0.0	0.00	0	0.0	0.0	0.00
Other	6	4.6	0.2	0.44	3	6.8	0.3	0.67	1	2.1	0.1	0.27
Total	131	100	3.5	2.26	44	100	4.4	3.57	47	100	3.4	1.82

Table 1 Frequencies and proportions of problems noticed in the Stage 1 writing task

Note: meanmean number of features noted by participants

Table 2     Frequencies and proportions of features noticed in the Stage 2 comparison task

	n	%	mean	s.d.	n	%	mean	s.d.	n	%	mean	s.d.
Lexis	102	63.0	2.8	1.67	36	66.7	3.6	1.84	30	63.8	2.1	1.23
Grammar	7	4.3	0.2	0.46	2	3.7	0.2	0.42	2	4.3	0.1	0.36
Content	47	29.0	1.3	1.05	13	24.1	1.3	1.16	14	29.8	1.0	0.88
Other	6	3.7	0.2	0.37	3	5.6	0.3	0.48	1	2.1	0.1	0.27
Total	162	100	4.4	1.96	54 100      5.4			1.96	47	100	3.4	1.34

Note: meanmean number of features noted by participants

Table 3  Stage 2 lexical FNs by relation to Stage 1 PFNs

	n	mean	s.d.
Related to Stage 1 PFNs	61	1.7	1.23
Not related to Stage 1 PFNs	41	1.1	1.29
Total	102	2.8	1.67

Note: meanmean number of features noted by participants

lexical choices, and the same words used differently. The 41 FNs that were not related to the Stage 1 PFNs represented noticing of words which had not been consciously searched for during the original writing stage. It should be noted that the noticing of 12 of these 41 features involved delayed noticing of problems with their original output. For instance, one student, who originally wrote ‘a woman who worked the same office with him’noted that she should have used the word ‘co-worker’. She had not focused on this issue as problematic during the Stage 1 noticing task.

3 What are the effects of Stage 1 and Stage 2 noticing on subsequent revisions?

First, the problematic features that the students noticed while composing their original text (PFNs) were classified as either ‘solvable’or ‘unsolvable’from the two native speaker models presented to them later. A ‘solvable’PFN refers to a problematic feature for which the models (either Model Aor Model B, or both) provide at least one solution. To illustrate, in the present study, many participants noted that they wanted an English lexical item for the Japanese ‘doryo’. Model Aprovided one solution, ‘co-worker’, whereas Model B offered another solution, ‘colleagues’. Therefore, PFNs involving the Japanese ‘doryo’ were deemed ‘solvable’. An ‘unsolvable’ PFN, on the other hand, refers to a problematic feature for which the models do not provide a solution. For instance, one student in this study noted that she wanted an English expression for the Japanese ‘yoko wo muku’ (to look to the side). However, neither Model A nor Model B provided any solutions to this problem because neither of them made any reference to this point. Therefore this PFN was regarded as ‘unsolvable’. The six PFNs in the ‘other’ category did not include any specific focal points, and therefore they were excluded from this analysis.

The participants in this study produced a total of 125 PFNs in the ‘lexis’, ‘grammar’, and ‘content’ categories, of which 89 PFNs (71%) were solvable, and 36 (29%) were unsolvable. For the classification, the two raters reached an agreement on 122 (97.6%) out of the 125 PFNs. All but one of the 89 solvable PFNs were lexical features. During the Stage 1 writing task, participants had solved on their own 10 of these lexical PFNs (N88)  by using exactly the same features that were included in the models to be presented later. These features were therefore excluded from the analysis. The remaining 78 lexical PFNs (23 for Group A, 34 for Group B) were then analyzed to examine the frequencies at which the students noticed solutions to these PFNs in the Stage 2 comparison task, and the frequencies at which they actually incorporated those solutions in their subsequent revisions.

Table 4 shows that the participants noticed approximately two-thirds of the lexical solutions available from the models, and incorporated 92% of them in the revision task that immediately followed. It should be noted here that an analysis of the data revealed that four students (two of them being Group B students, with the other two belonging to neither group) incorporated an additional six lexical features (three of them attributed to Group B) in their revisions without noting them in the comparison task, indicating incomplete note-taking by the participants. With regard to the Stage 4 delayed revision task that took place two months later, the participants retained 40% of the solutions that they had incorporated in their first revision.

In order to examine the effects of noticing that occurred during the output stage, a further analysis was conducted regarding the incorporation of the 102 lexical FNs in terms of their relevance to the Stage 1 PFNs. As shown in Table 3, these lexical FNs were divided into those that were related to the Stage 1 PFNs

Table 4     Solutions to lexical PFNs noticed and incorporated

All participants (N37)						Group A (N10)					Group B (N14)
	n	%	mean	s.d.	n		%	mean	s.d.	n	%	mean	s.d.
Solvable PFNs	78	100	2.1	1.43	23		100	2.3	1.49	34	100	2.4	1.65
Stage 2 noticing	51	65.4	1.4	1.04	17		73.9	1.7	0.95	21	61.8	1.5	1.16
Stage 3 Inc.	47	60.3	1.3	0.96	17		73.9	1.7	0.95	18	52.9	1.3	0.99
Stage 4 Inc.	19	24.4	0.5	0.77	8		34.8	0.8	1.14	6	17.7	0.4	0.51

Note: Inc.Incorporation; meanmean number of features noted by participants

(N61) and those that were not (N41) . Table 5 shows the incorporation of each of these two types of features in the two revisions. It was found that those features that were related to what the students had noticed during the original composing stage were incorporated significantly more in the immediate revision (²5.403, df1, p.02)  as well as in the delayed revision that took place two months later (²6.511, df1, p.01) . Also, with respect to the retention of the features actually incorporated in the first revision, the related features were incorporated at a higher rate in the second revision two months later than the unrelated features. This difference approached siginificance (²3.462, df1 , p.06) . These statistical numbers, however, should be interpreted cautiously because some participants contributed more than others to the frequency counts.

4 What are the learners’proficiency effects on noticing and incorporation?

As shown in Table 1, during the original composing stage, the students in Group A, the more proficient group, noted more PFNs (M4.4) than the students in Group B (M3.4) . However, a t-test revealed that this difference was not significant (t.94, df22, p.357). In addition, no significant correlation was observed between the scores of the 36 participants and the number of PFNs that they produced (r.206, df34).

As shown in Table 2, during the Stage 2 comparison task, the students in Group A noted more FNs (M5.4) than the students in Group B (M3.4). A t-test revealed that this difference was significant (t3.05, df22 , p.006). A significant correlation was also observed between the cloze test scores of the 36 students and the total number of FNs that they produced (r.500, df34, p.01).

With regard to the solvable lexical PFNs, the students in Group A noticed nearly three-quarters of the solutions and incorporated all of them in their revisions (see Table 4). Group B students, on the other hand, noticed about 60% of the solutions and incorporated 86% of them. However, the difference between the two groups in the ratio of noticing solutions was not statistically significant (².911, df1, p.34). Neither was the difference in the ratio of incorporating these solutions in the immediate revision task (²2.637, df1, p.10).

Table 5     Incorporation of Stage 2 lexical FNs by relation to Stage 1 PFNs

	n	%	mean       s.d.	n	%	mean s.d.
Stage 2 lexical FNs	61	100	1.7           1.23	41	100	1.1 1.29
Stage 3 incorporation	51	83.6	1.4           1.16	26	63.4	0.7 0.97
Stage 4 incorporation	25	41.0	0.7           0.92	7	17.1	0.2 0.46

Note: meanmean number of features noted by participants

In the Stage 4 revision, the students in Group A and Group B respectively incorporated 35% and 18% of the solutions available from the models. Although the percentage gap may seem large, it is based on small numbers of instances and the difference was found to be statistically nonsignificant (²2.174, df1 , p.14) . With regard to the retention of the features incorporated in Stage 3, the students in Group A retained 47%, while the percentage for the students in Group B was 33% (excluding three features not noted in Stage 2). However, the difference was again not significant (².686, df1, p.41).

V Discussion

1    Learner noticing during output and comparison with feedback

The first research question posed in this study was: What aspects of language do L2 learners notice while composing a narrative on their own? The answer to this question was that the students noticed overwhelmingly lexical problems. The second research question was: What do L2 learners notice as they compare their text to native-speaker models of writing? The answer to this question was also that a large majority of student noticing was lexical. It should be pointed out as well that in this stage the participants noticed more features than in Stage 1 and that content features accounted for nearly 30% of the participants’ noticing. This indicates a useful role of native-speaker models in promoting learner noticing and specifically in drawing the learners’ attention to the content of what they wrote.

The findings regarding the Stage 1 and Stage 2 noticing suggest that the participants noticed lexical ‘holes’ ( Swain, 1998) in their interlanguages through output and that this perceived need for vocabulary led to a lexically oriented search for solutions in the two models presented later. Overall, these results indicate that ‘learners focus, above all things, on words’ (Williams, 2001: 338) . The following possibilities also need to be recognized, however. First, it is possible that lexical features were simply easier to express and report than grammatical ones. This may have been the case especially under the physically demanding note-taking condition. Second, the directions given to the participants when they took notes may also have inflated the proportion of lexical PFNs. These methodological problems need to be addressed in future studies.

With respect to the noticing function of output (Swain, 1995; Swain and Lapkin, 1995), it should be noted that during the Stage 2 task, the participants noticed some new problems with their original output while studying the TL models. Therefore, it may be argued that output not only results in the immediate recognition of linguistic problems (holes) but also facilitates further noticing of problems (gaps) in subsequent processing of TL input (Swain, 1998) . These two types of problem recognition may be distinguished from each other. In the case of the former, the learner typically wonders, ‘How can I write (say) this?’while in the latter case, the learner may say, ‘I should have written (said) it this way.’ In this sense, those problems that the participants notice during output, or ‘holes’(Swain, 1998) may represent proactive recognition of problems, whereas those problems that they notice for the first time during the comparison stage without being preceded by noticing of ‘holes’ may be characterized by reactive recognition of problems.

2    From noticing of problems to incorporation of solutions

The third research question was: What are the effects of Stage 1 and Stage 2 noticing on subsequent revisions? First of all, the participants noticed lexical problems while composing a story and later noticed about two-thirds of the solutions available from the models, indicating a strong relationship between Stage 1 and Stage 2 noticing. Qi and Lapkin (2001) also noted that their two participants’recent experience of output greatly influenced what they noticed in the comparison task. They provided the following psycholinguistic explanation (Qi and Lapkin, 2001: 289):

the failure to reach a satisfactory solution to a problem with existing linguistic knowledge may result in a sense of uncertainty or lack of fulfillment on the part of a learner… It is perhaps this sense of lack of fulfillment that may push a learner to look out for any future relevant information available that he/she believes might help solve the problems in a better way.

What is described here may be known as the Zeigarnik effect (Zeigarnik, 1999) , which states that unfinished tasks create psychological tension and thus tend to be remembered better than finished ones. Noticing of holes through output may trigger this psychological effect and accelerate the subsequent learning of the relevant form. This may be recognized as the motivating function of output (see also Izumi, 2003).

In terms of completing unfinished tasks, the Zeigarnik effect may also predict incorporation of solutions. In the present study, participants felt a strong need for lexical elements through output and generally indicated a strong desire to study the models. After noticing solutions to about two-thirds of the problems whose solutions were available from the models, they immediately incorporated 92% of them into their revisions. This high incorporation rate was achieved despite the fact that the models and the noticing sheet were taken away when the participants engaged in the revision task. This seems to indicate that output motivated the participants not only to seek solutions to the problems that they identified but also to use these solutions upon noticing them.

Another important finding was that among the lexical features noticed from the models, those that were related to the questions that the participants had posed during the original output stage were incorporated significantly more in the immediate revision as well as in the second revision two months later than those that were not related. One explanation for this finding may be that those features related to the problems that the participants had noticed during original output were probably the features that they needed most to revise their stories, whereas those features that were not related were optional elements for the revision tasks. It may also be argued that the linguistic agendas that participants had identified for themselves through output increased their motivation not only to seek relevant features in the models but also to use them, leading to a significantly better retention of those features than other features not related to their original agendas. This finding also seems to support Brown’s (1993) contention that it is pedagogically useful to create a gap by having L2 learners ‘[experience] a need for a word form before they encounter it’ ( p. 266). Lexical items thus learned may be better retained in their long-term memory. In this regard, it is noteworthy that, among the features actually utilized in the first revision, those features related to the participants’original agendas were more likely to be incorporated again two months later than the other unrelated features. Is it possible that these two types of features were not equally affected by time? If delayed effects of output were at play here, it might be hypothesized that noticing of a linguistic problem during output, or what Swain (1998) termed ‘noticing the hole’, not only serves as ‘an important stimulus for noticing the gap’ ( Swain, 1998: 66) but also facilitates retention of the solution in short-term as well as long-term memory. The findings in this study indicate a need for further research to explore this possibility.

The participants’ performance in their Stage 4 revision may allow for various interpretations. In this delayed revision task that took place two months later, the participants were able to incorporate only about one fourth of the solutions available from the models. However, they retained 40% of the features that they had incorporated in their first revisions. In this sense, it may be argued that the task consisting of output, comparison, and immediate revision was effective. On the other hand, the fact that the participants failed to retain much of what they noticed and incorporated over the course of two months suggests that learners need frequent and extended opportunities for rehearsing those features (Izumi, 2002; Qi and Lapkin, 2001; Robinson, 1995).

Finally, potential reactivity of note-taking used as a noticing measure in this study needs to be considered. It is possible that the act of note-taking allowed the participants to engage in metalinguistic reflection and thereby enhanced the entire learning process investigated in this study. Specifically, note-taking may have amplified the positive effects of output by providing participants with an opportunity to further process what they noticed through output, and this in turn may have improved retention of the noticed features. This study, however, was not designed to tease apart this potential reactivity factor. This issue and concomitant pedagogical implications of taking notes need to be addressed in future studies.

3    Proficiency effects on noticing and incorporation

The fourth research question in this study was: What are the learners’ proficiency effects on noticing and incorporation? Although, during the original composition stage, the more proficient group noticed more problems than the less proficient group, the difference was not significant. The only statistically significant difference emerged during the Stage 2 comparison task, in which the higher-proficiency learners noticed more features from the two nativespeaker models than the less proficient group. These findings lend partial support to previous studies (Qi and Lapkin, 2001; Swain and Lapkin, 1995 ; Williams, 2001), which found that more advanced learners initiated greater numbers of language-related episodes. One question that arises in the context of L2 writing is whether the proficiency level affects the number of features noticed during feedback processing (i.e. noticing of gaps) more than the number of features noticed during output attempts (i.e. noticing of holes).

With regard to the problems noticed during Stage 1 that were solvable from the models, the higher-proficiency group noticed a higher percentage of solutions and incorporated a higher percentage of them in the immediate revision. However, the differences were not significant. This was probably due to the fact that both groups were able to find solutions to a high percentage of their selfidentified problems and that they almost always incorporated solutions once they were noticed. Overall, the significant difference observed in the total number of features noticed from the models and the general trend found in favor of the higher-proficiency group suggest that more proficient learners may potentially benefit more from spontaneous focus-on-form tasks employing models.

4    Quality of noticing

In this study, the participants were overwhelmingly concerned with lexical features and the difference in the extent to which they noticed collocation of a word directly affected the quality of their revision. For instance, some of the participants were often satisfied to find the key word or phrase for a Japanese expression and stopped short of noticing its collocation. One student, who wrote in her original text, ‘The man late for …’ noted in Stage 1 that she did not know the expression for ‘chikoku suru’ ( to be late). Then in the Stage  2 comparison task, she wrote that she was right in using the word ‘late’ and failed to notice that the word needs to be preceded by a BE verb. As a result, she again failed to use a BE verb in both her Stage 3 and Stage 4 revisions.

The following three instances involving the noticing and incorporation of the noun phrase ‘traffic jam’ illustrate how the level of noticing a collocation directly affected the subsequent revisions. Since the noun phrase constituted one of the key expressions in the description of the pictures, most of the participants noticed it being used in the two models. However, some of them only noticed the phrase ‘traffic jam’ and failed to notice its collocation. Student A, who noted in Stage 1 that she did not know the word(s) for ‘juutai’ ( traffic jam), wrote during the comparison stage :

‘For the first time, I learned that “juutai” was “a traffic jam”.’

In the revision task which followed, she incorporated the expression ‘traffic jam’ ( omitting the article ‘a’) and wrote :

‘Because it is traffic jam in the morning.’

On the other hand, some students who already knew the expression ‘traffic jam’ went on to notice its collocational property. Student B noted in Stage 1 that she knew the expression ‘traffic jam’, but did not know how to say ‘juutai ni makikomareru’ ( to be caught in a traffic jam). While studying the models, she noted:

‘“Juutai ni au” is “catch in a traffic jam”.’

Consequently, in her subsequent revision, she incorporated, albeit incorrectly, the longer chunk and wrote:

‘She didn’t catch in the traffic jam.’

The third instance further indicates that learner noticing is strongly influenced by the learner’s prior knowledge. Student C, who successfully used the pattern ‘be in a traffic jam’ in her first draft, noticed an alternative expression in the Stage 2 comparison task. She noted:

‘I learned that both “catch in a traffic jam” and “be in a traffic jam” are possible.’

In her Stage 3 revision, she immediately tried out the alternative expression and wrote,

‘He was caught in a traffic jam.’

It may be argued that the models, by providing an alternative expression, helped her expand her collocational schema involving the expression ‘traffic jam’. This indicates a potential advantage of independent modeling over individually provided reformulation, which may not address what is already correct (see Hanaoka, 2006, for further discussion). It should also be pointed out that the fact that she tried out and rehearsed the new expression, although the task did not require it, shows that she actively engaged in language learning.

5 Type of noticing required for learning collocations

The three examples cited above may be described in terms of scope of noticing. While the concept of scope of noticing may be usefully applied to various aspects of lexical knowledge such as meaning, form, and use (Nation, 2001), here the discussion focuses on the scope of noticing collocation, that is, the extent to which the collocates of a word are noticed. The above examples demonstrate that the scope of a lexical chunk that learners notice is the part of the chunk that they may learn. It is also helpful to apply the concept of integrative processing here. Izumi (2002) argues that integrative processing, which involves noticing of elements that are related to each other, is essential in the acquisition of grammatical structures. Since collocation dictates the use of words in combination, the learner needs to notice the relationship that holds among them and identify them as an integrated whole. Therefore, it may be argued that the scope of noticing involving collocations is determined by the level of integrated processing that accompanies such noticing.

VI Limits and suggestions for future research

The present study has some important limitations, which in turn suggest some new avenues for future research. First, this study found that the learners’main concern was with vocabulary. However, it is possible that the act of notetaking oriented the participants to report lexical items more frequently than other features. Also, the fact that participants incorporated more features than they reported noticing in Stage 2 suggests that the participants’ notes were incomplete. Therefore, it is necessary to conduct further studies employing different noticing measures such as think-alouds to compare the results. Second, the high incorporation rate of PFN-related FNs raises an interesting question: Is ‘noticing the gap’preceded by ‘noticing the hole’more conducive to learning than ‘noticing the gap’ alone? In other words, does the subjective experience of a need for a word increase the impact of noticing a solution and thereby the chance of its retention in memory? Third, some questions regarding proficiency effects remain to be answered. One question that arises from the present study is whether, in L2 writing contexts, the proficiency level affects to a greater extent the number of gaps noticed when exposed to feedback than the number of holes noticed during the original output stage. It is also important for future studies to investigate how the proficiency level is related to the retention of noticed features by using individually tailored posttests. Finally, this study suggests that, as a feedback tool, models may play a useful but different role from individually prepared reformulations in promoting learner noticing and subsequent learning. This is also an important issue to be explored in L2 writing research.

VII Conclusion

The findings of this study indicate an important noticing function of output. While composing a story, the participants, regardless of their proficiency level, noticed their respective linguistic agendas (holes), autonomously found solutions in the models, and incorporated them in subsequent revisions. In particular, lexical features of the models that were related to the problems that the participants had noticed through output were incorporated at a higher rate and were also retained longer than unrelated features. In other words, it may be stated that output had a positive domino effect on learning. This suggests that output plays a useful role in both helping learners identify the linguistic features they need and facilitating subsequent learning of these features. Spontaneous focus-on-form activity in L2 writing, which allows learners to decide target forms based on their own respective needs, may, thus, have a strong motivational advantage in driving forward the subsequent learning process. In pedagogical terms, two implications may be stated. First, when designing a task, we need to consider the possibility that more proficient learners are capable of noticing more and also learning more from such noticing than less proficient learners. In this regard, the language level of the model needs to be carefully considered so that the learners can maximally benefit from it. Second, the participants’spontaneous noticing of collocations was often inadequate and this resulted in inadequate incorporation. This suggests that follow-up activities are necessary to enhance the effectiveness of the original spontaneous focus-on-form task.

Acknowledgements

I would like to thank Kensaku Yoshida and Shinichi Izumi for their helpful comments, suggestions, and guidance. I am also grateful to the two LTR reviewers for their valuable comments on an earlier draft of this paper. Thanks are also due to Yoko Kanazawa and Richard Sutton for their consistent support.

VIII References

Adams, R. 2003: L2 output, reformulation and noticing: implications for IL development. Language Teaching Research 7: 347–76.

Bowles, M. and Leow, R. 2005: Reactivity and type of verbal report in SLA research methodology: Expanding the scope of investigation. Studies in Second Language Acquisition 27: 415–40.

Brown, C. 1993: Factors affecting the acquisition of vocabulary: Frequency and saliency of words. In Huckin, T., Haynes, M. and Coady, J., editors, Second language reading and vocabulary learning, Norwood, NJ: Ablex, 263–86.

Ellis, R. 1995: Interpretation tasks for grammar teaching. TESOL Quarterly 29: 87–105.

Ellis, R., Basturkmen, H. and Loewen, S. 2001: Preemptive Focus on Form in the ESL Classroom. TESOL Quarterly 35: 407–32.

Hanaoka, O. 2006: Exploring the role of models in promoting noticing in L2 writing. JACET Bulletin 42: 1–13.

Izumi, S. 2000: Promoting noticing and SLA: an empirical study of the effects of output and input enhancement on ESL relativization. Unpublished doctoral dissertation, Georgetown University, Washington, DC.

—— 2002:  Output, input enhancement, and the Noticing Hypothesis: an experimental study on ESLrelativization. Studies in Second Language Acquisition 24: 541–77.

—— 2003:  Comprehension and production processes in second language learning: in search of the psycholinguistic rationale of the output hypothesis. Applied Linguistics 24: 168–96.

Izumi, S., Bigelow, M., Fujiwara, M. and Fearnow, S. 1999: Testing the output hypothesis: effects of output on noticing and second language acquisition.

Studies in Second Language Acquisition 21: 421–52.

Johnson, K. 1988: Mistake correction. ELT Journal 42: 89–96.

Jourdenais, R. 2001: Cognition, instruction and protocol analysis. In Robinson, P., editor, Cognition and second language instruction, Cambridge: Cambridge University Press, 354–75.

Kowal, M. and Swain, M. 1994: Using collaborative language production tasks to promote students’ language awareness. Language Awareness 3:   73–93.

—— 1997:  From semantic to syntactic processing: how can we promote it in the immersion classroom? In Johnson, R.K. and Swain, M, editors, Immersion education: international perspectives, Cambridge, UK: Cambridge University Press, 284–309.

Lapkin, S., Swain, M. and Smith, M. 2002: Reformulation and the learning of French pronominal verbs in a Canadian French immersion context. Modern Language Journal 86: 485–507.

Leow, R. and Morgan-Short, K. 2004: To think aloud or not to think aloud: the issue of reactivity in SLA research methodology. Studies in Second Language Acquisition 24 ,  35–58.

Long, M. 1991: Focus on form: a design feature in language teaching methodology. In de Bot, K., Ginsberg, R. and Kramsch, C., editors, Foreign language research in cross-cultural perspective, Amsterdam: John Benjamins, 39–52.

Long, M. and Robinson, P. 1998: Focus on form: theory, research and practice. In Doughty, C. and Williams, J., editors, Focus on form in classroom second language acquisition, Cambridge: Cambridge University Press, 15–41.

Mackey, A., Gass, S. and McDonough, K. 2000: How do learners perceive interactional feedback? Studies in Second Language Acquisition 22: 471–97.

Morgan-Short, K. and Bowden, H.W. 2006:  Processing instruction and meaningful output-based instruction: effects on second language development. Studies in Second Language Acquisition 28: 31–65.

Nation, I.S.P. 2001: Learning vocabulary in another language. Cambridge: Cambridge University Press.

Qi, D.S. and Lapkin, S. 2001: Exploring the role of noticing in a three-stage second language writing task. Journal of Second Language Writing 10: 277–303.

Robinson, P. 1995: Review article: Attention, memory, and the “noticing” hypothesis. Language Learning 45: 283–331.

Schmidt, R. 1990: The role of consciousness in second language learning. Applied Linguistics 11: 129–58.

Schmidt, R. 1995: Consciousness and foreign language learning: a tutorial on the role of attention and awareness in learning. In Schmidt. R., editor, Attention and awareness in foreign language learning ( Technical Report No. 9), Honolulu : University of Hawai’i, 1–63.

—— 2001: Attention. In Robinson, P., editor, Cognition and second language instruction, Cambridge: Cambridge University Press, 3–32.

Schmidt, R. and Frota, S. 1986: Developing basic conversational ability in a second language: a case study of an adult learner of Portuguese. In Day, R., editor, Talking to learn: conversation in second language acquisition, Rowley, MA: Newbury House, 237–326.

Shehadeh, A. 2002: Comprehensible output, from occurrence to acquisition: an agenda for acquisitional research. Language Learning 52: 597–647.

Skehan, P. 1998: Acognitive approach to language learning. Oxford: Oxford University Press.

Swain, M. 1995: Three functions of output in second language learning. In Cook, G.

and Seidlhofer, B., editors, Principle and practice in applied linguistics: studies in honour of H. G. Widdowson, Oxford: Oxford University Press, 125–44.

—— 1998:  Focus on form through conscious reflection. In Doughty, C. and Williams, J., editors, Focus on form in classroom second language acquisition, Cambridge: Cambridge University Press, 64–81.

—— 2005: The output hypothesis: theory and research. In Hinkel, E., editor, Handbook of research in second language teaching and learning, Mahwah, NJ: Lawrence Erlbaum Associates, 471–83.

Swain, M. and Lapkin, S. 1995: Problems in output and the cognitive processes they generate: a step towards second language learning. Applied Linguistics 16: 371–91.

—— 2002: Talking it through: two French immersion learners’ response to reformulation. International Journal of Educational Research 37: 285–304.

Wajnryb, R. 1990: Grammar dictation. Oxford: Oxford University Press.

Williams, J. 2001: Learner-generated attention to form. In Ellis, R., editor, Formfocused instruction and second language learning, Malden, MA: Blackwell, 303–46.

Zeigarnik, B. 1999: On finished and unfinished tasks. In Ellis, W.D., editor, A source book of Gestalt psychology, London: Routledge, 300–14.

Appendix A: The picture prompt

From the second grade STEP Test in Practical English Proficiency, 2000. Used with permission from the Society for Testing English Proficiency, Inc.

Appendix B: The two native speaker models

( A )

While riding her bike to work one morning, a woman passed a co-worker who was driving his car. He was caught in a traffic jam. One hour later the woman was at work when the co-worker she passed finally arrived at work. His boss and co-workers were angry because a meeting had already started and he was 40  minutes late.

( B )

A man and woman are on their way to work. The woman is riding a bicycle and looks very happy and refreshed. The man is driving a car and looks frustrated. He is in a traffic jam. An hour later the man storms into his office all sweaty and embarrassed. He is late and a few of his colleagues are scowling at him. The woman, meanwhile, is sitting at her desk very relaxed and amused.