Abstract
In the past 20 years, research in the field of ESL writing has made great progress. However, research on EFL writing strategies is still a latecomer, especially in China. A review of the previous studies at home shows that most of the research subjects are adults; research on local techniques outnumber those on global strategies and the researchers employed either qualitative or quantitative instrument in the research. Motivated by the limitation of the previous research, the present study focuses on the writing strategies of Chinese college EFL writers. Conducted in Dalian University of Technology, the study aims at analyzing the relations among writing proficiency, writing strategy and writing scores. A combination of quantitative instrument, questionnaire and qualitative instrument known as Stimulated Recall Protocol (SRP) is employed in this study and computer software SPSS 10.0 is used to compile statistics. It is found that there exist significant differences in strategy use between English major and non-English major writers; English major students hold a more positive attitude towards English writing; both English major students and non-English major students cannot handle the strategy of “coherence” well and English major students pay more attention to “global” strategies while non-English major students pay more attention to “local” strategies. Not only are the findings of the present study pedagogically helpful for EFL teachers, but also they reveal the writing habits of Chinese EFL writers with different writing proficiency, which can help them improve their own writing.

Key words: EFL, writing strategy, Stimulated Recall Protocol, writing proficiency

1. Introduction

As ESL teaching develops more and more rapidly, an increasing number of linguists pay attention to research in the field of ESL writing strategies. They seek to develop an inventory of writing strategies and pedagogically use it as a guide in ESL writing to help students improve their writing abilities.
In the past 20 years, relevant research in this field has been carried out. However, research on EFL writing is still a latecomer in this field. Especially in China, most of the research in English writing focuses on the detailed techniques rather than global strategies owing to the test-oriented education system.

The purpose of the present study is twofold. On the one hand, it compares the writing strategies used by Chinese college EFL learners which will have pedagogical implications in EFL teaching and learning. On the other hand, the reliability and validity of Stimulated Recall Protocol is validated which provides suggestions for further studies.
To achieve these goals, the present study tries to answer the following research questions:

1. Are there differences in writing strategy use among three groups of subjects with different writing proficiency?
2. What are the key strategies differentiating good writers and weak writers?
3. What are the relations between composition score and writing strategies?
4. Is Stimulated Recall Protocol reliable and valid for a Chinese college EFL writing strategy study?

Both qualitative and quantitative instruments are employed to improve the reliability and validity of the study.

2. Methodology
The present study focuses on the writing strategies used by Chinese EFL writers and was conducted at Dalian University of Technology, Peoples’ Republic of China. The subjects are 60 EFL learners who are all undergraduates. As an alternative to think-aloud protocol, the present study employs a less intrusive method, Stimulated Recall Protocol, to examine the writing process of the three groups of writers. Besides Stimulated Recall Protocol, the present study also uses a questionnaire as a quantitative instrument. Combining a qualitative instrument with quantitative instrument, the present study elicits more reliable data.

2.1 Questionnaire
The present study employs the revised questionnaire used in Yang Shuxian’s (2002) study. The language in the questionnaire is Chinese, which guarantees that the subjects can comprehend the items well and makes sure language comprehension will not be an obstacle for the subjects. The questionnaire contains two sections: basic information and a five-point Likert scale. As for the basic information, except for the grade and gender, a general question on the personal attitude towards English writing is also included.  The second section of the questionnaire is divided into 5 parts: pre-drafting strategies (Part A), drafting strategies (Part B), post-drafting strategies (Part C), audience awareness (Part D) and mother tongue avoidance (Part E). There are 5 items in Part A (A1 to A5), 9 items in Part B (B1 to B9), 7 items in Part C (C1 to C7), 2 items in Part D (D1 to D2) and 3 items in Part E (E1 to E3). So in total, there are 26 items in the second section of the questionnaire and the five-point Likert scale is from 1 to 5, in which 1 represents “totally or almost totally unsuitable for me”; 2 represents “suitable for me less than half of the time”; 3 represents “suitable for me about half of the time”; 4 represents “suitable for me more than half of the time”; 5 represents “totally or almost totally suitable for me”.
For example:
Standard: 1= totally or almost totally unsuitable for me
2= suitable for me less than half of the time
           3= suitable for me about half of the time
           4= suitable for me more than half of the time
           5= totally or almost totally suitable for me

Pre-drafting Strategies	1	2	3	4	5
Make an outline first	1	2	3	4	5

For each item, the subjects have to choose from number 1 to 5 according to the degree of suitability of their own situation. 

2.2 Stimulated Recall Protocol (SRP)
As a qualitative instrument, Stimulated Recall Protocol (SRP) has frequently been used in research in recent years. SRP focuses on the writing process of the subjects and uses “pauses” during the writing process as a “window” to get data of the strategies used by the subjects while they are writing. Compared with the previously and frequently used Think-aloud Protocol, SRP is less intrusive and helps elicit more reliable data from the subjects. According to Jiménez (2005), SRP has the following advantages:
1. That which is hidden, becomes visible.
2. It provides information on the cognitive processes of language formation.
3. It helps in the fine-tuning of the analysis.
4. Speakers can make corrections to their own narratives.
5. It helps in the interpretation of pauses.
6. It helps interpret message abandonment.
7. It provides unexpected comments.
8. It provides post-recall impressions.

Considering the above reasons, SRP is employed in the present study. Since SRP uses video images instead of audio materials to elicit data, a digital camera, a computer and a tape recorder are needed in the study process. The subjects are asked to come to a room individually and write English articles in a quiet atmosphere. The whole writing process is video-taped and monitored. We focus on “pauses” during the writing process and record the time point and length of the “pauses”.

Immediately after the subjects finish writing, they are asked to watch the video of their own writing process with us, during which the subjects are asked questions on their “pauses” during the writing process. The questions focus on what the subjects were thinking during the “pauses” and from the answers of the subjects, we can learn what strategies the subjects use during their writing. The process of watching video, asking and answering questions is also audio-recorded to make sure that we can get more exact data afterwards.
The writing topic is an argumentative one. The reason for this choice is that they had been used in previous studies (Sasaki, 2000; Bosher, 1998) and according to Grabe and Kaplan (1996), argumentative topics demanded more complex processing and tended to elicit more meaningful information.

The time span from 30 to 60 minutes is chosen for the subjects to write for the following reasons. Firstly, the topic is about the current situation of the job market in China, which is familiar to the college students. So it is a relatively easy topic to write on. Secondly, for most of the Chinese university students, the time spent on writing in most of the English tests should be less then 30 minutes or they cannot finish the tests. So compared to the time limit in their English test, 30 to 60 minutes in the present study is enough for them to finish their writing. Finally, this time span of writing makes the following video-watching and interview session relatively short. Based on the result of the pilot study, the overall SRP process for one subject is about 2 hours. If the subjects spend more than 60 minutes writing, it will make them too tired to recall their thoughts during the writing.

The tips also made it easier for the subjects to understand the procedure of the whole SRP experiment. The writing instruction and the tips were written in both English and Chinese, which ensured that all the subjects could comprehend the writing topic correctly and fully. All the subjects were given pieces of paper to write on. On the paper, they also have to give their personal information, such as grade and gender. The scripts of the subjects were collected after the SRP process.

Similar to Sasaki’s research (2000), before the subjects started to write the title of their essays, they were asked some questions about their writing plan, such as “What are you going to write?”, “Do you have an overall writing plan?”, “What do you plan to write at the beginning, in the middle and in the end?”. When they answered the questions, they were told that the time of this question session would not be included in the 30 to 60 minutes for writing. All the questions were asked in Chinese so that each subject could understand correctly. Then, the subjects began to write their articles. The camera focused on the movement of their writing hands and the content of their writings. We monitored the whole writing process and took down on the SRP Time & Strategy Table the beginning and ending time points of pauses of 3 seconds or longer. For each pause, we designed several strategies according to the items in the questionnaire for the interview part, such as spelling, grammar, correction, planning (for what to write next), reviewing, expressing in a different way (expression), translation, and so on. While the subjects were writing, they were not allowed to use dictionaries. After they finished writing and before watching the video tape, they were also asked several questions, such as “Does the video camera affect your writing?”, “Did you write the same content at the beginning, in the middle and in the end according to your original writing plan?” These questions were also asked in Chinese.

In the following interview session, the subjects watched their writing process with us. The subjects watched the tape with their own writings as reference. The reason is that in the pilot study, we learnt that it was easier for the subjects to recall their thoughts with the help of the texts. While watching the videotape, every time the subjects stopped for 3 seconds or longer, we asked them questions “What were you thinking about then?” in Chinese. The subjects answered the questions in either Chinese or English. If their answers of a certain pause belonged to the strategies in the SRP Time & Strategy Table, we marked a “√” in that pause. If not, we wrote down the strategies they used in the column “others”. This process continued until they finished the writing process of the subjects on the videotape. Through the questions on the pauses during the writing process, we got the strategies used by the subjects respectively. The whole process of interview session was audio-taped by MP3 and the audio and video materials were all transferred to the computer for future use.

After the video-watching process, the subjects were asked to fill in the questionnaire. For evaluation, all the writing texts collected were sent to national English examination graders, two EFL professors with more than 15 years’ teaching experience.
From the entire SRP session, the following materials or data were obtained:

1. Personal information of the subjects
2. Written texts and drafts (if the subjects made one)
3. Time points and length of the “pauses”
4. Time spent on writing of the subjects
5. Strategies used during the “pauses”
6. Subjects’ answers and comments on the whole writing process
7. Data of the questionnaire for SRP session
8. Audio and video materials of SRP process

2.3 Research subjects
We chose the undergraduate students in Dalian University of Technology as subjects. Three groups of subjects were asked to take part in the study. They were junior English majors (Group 1), freshman English majors (Group 2) and sophomore non-English majors (Group3). The reason to choose these three groups of subjects is that the differences of English writing proficiency among the three groups are distinctive. Until the present study, junior English majors and freshman English majors have 2.5 years and 0.5 years of professional English writing training respectively. For the non-English majors, their English proficiency is presumed highest in the sophomore year since they have to pass College English Test Band Fouri (CET-4) his year. But compared with English majors, their English writing proficiency is still relatively low. So these three groups of subjects form a clear hierarchy of English writing proficiency. It is hypothesized that from top to the bottom, the order of writing proficiency of the three groups is Group1, Group 2 and Group 3.

The present study includes two sessions: questionnaire and SRP experiment.  In the questionnaire session, the software SPSS is employed to analyze the data.  For each group, 20 subjects are needed for the scientific and reliable data analysis by SPSS. The total number is 60.  (See Table 1)

Table 1: Group Information of Questionnaire Session

	Group 1	Group 2	Group 3	Total
Major	English (junior)	English (freshman)	Architecture
Male	4	7	15	26
Female	16	13	5	34
Total	20	20	20	60

From Table 1, it is clear that the different classes of majors have different gender distributions. In the first two groups, there are more female subjects than male ones, but in the last group there are more male subjects than female ones. The reason is that for English major classes, there are always more female students than male ones. On the contrary, in non-English major classes, especially science major classes, there are always more male than female students. Since the subjects of the questionnaire session are chosen randomly, the distribution of the gender is in accordance with the major difference.

For the SRP experiment session, the same groups were used. SRP is a way to get qualitative data and it is relatively time-consuming. After referring to the previous relative studies (Sasaki, 2000; Bosher, 1998; Yanga & Shib, 2003), 4 subjects were chosen randomly from each group to take part in SRP experiment. The total number of subjects taking part in the SRP experiment is 12. (See Table 2)

Table 2: Group Information of SRP Experiment

	Group 1	Group 2	Group 3	Total
Major	English (junior)	English (freshman)	Architecture
Male	2	1	3	6
Female	2	3	1	6
Total	4	4	4	12

In Table 2, it is coincidental that the number of male subjects equals the female ones. It in turn makes it easier to analyze and compare the qualitative data that was obtained from the SRP session.

3. Data analysis and results
The data and SRP experiment were analyzed in different ways. Since a questionnaire is a quantitative instrument, the data were firstly recorded and then analyzed by the software SPSS 10.0. As for the SRP experiment, the materials were observed and analyzed. However, the results from only subjective observation were far from enough, so some of the data, such as frequency of pauses, pausing length and so on were recorded by Microsoft Office Excel since only simple statistical processing was needed. For strategies used and other relative data, SPSS 10.0 was used for more complex statistical analysis.

3.1 Questionnaire data analysis
60 valid questionnaires were retrieved after the class visit. The subjects taking part in the questionnaires were chosen randomly so the data was reliable and valid. To analyze the data, we at first numbered the questionnaires by groups: majors Group1 (from number 1 to 20), majors Group 2 (from number 21 to 40), and majors Group 3 (from number 41 to 60).
After grouping and numbering all the questionnaires, we used SPSS 10.0 to analyze whether there were significant differences of item choices among the three groups and between English majors and non-English majors. The results would reveal the different strategies used by the subjects in the three groups.

Since the number of subjects in each group was 20, less than 30. Nonparametric Tests in SPSS 10.0 was used to analyze the significance of differences. We used Crosstab Test to compare the three groups of data. Pearson Chi-Square Correlation Coefficient was used to indicate the significance of item choices difference among the three groups.

For the comparison between English majors and non-English majors, another test was used to analyze the data. The total number of subjects in Group 1 and Group 2 is 40, which is larger than the number of the subjects in Group 3. So we first used SPSS 10.0 to select randomly 20 subjects from Group 1 and Group 2. These 20 subjects of English majors form a new group of English majors, Group 4. The 20 subjects in Group 4 were then put together with the 20 subjects in Group 3 for further Crosstab analysis.

3.2 SRP Experiment data analysis
Four sets of data were obtained from each subject in the SRP experiment, including an SRP Question Sheet, an SRP Time & Strategy Table, a composition, and a questionnaire. The data of each subject were put in a portfolio together and then grouped and numbered. Therefore, the data of the 12 subjects in SRP experiment were categorized respectively to Group 1 (number 1 to 4), Group 2 (number 1 to 4), and Group 3 (number 1 to 4). Then, the four sets of data were analyzed one by one.

The questions on SRP Question Sheet are about the writing plan and the validity of the video-taping method. The aim of the SRP questions is twofold. On the one hand, the questions can check whether the students write consistently and how much planning they do before writing. On the other hand, the questions also investigate whether the video-taping method is intrusive during the writing process of the subjects and is helpful for them to recall their thoughts after the writing. Besides, the main problems in the subjects’ writing are also revealed by the questions.

Some of the questions and answers were recorded by SPSS 10.0 and some by Microsoft Excel. The reason is that SPSS is good for statistical analysis but Microsoft Excel can obtain more subjective information. The “yes-or-no” questions are recorded by SPSS:

a. Do you have an overall writing plan?
b. Did the questions have negative influence on your writing?
c. Did the video-taping affect negatively your writing?
d. Do you write according to your writing plan?
e. Is video-watching helpful to recall your thoughts during
the writing process?

The “open-ended” question is recorded by Microsoft Excel:
f. What are the main problems in your writing?
For the “yes-or-no” questions, SPSS was used to analyze the frequency of the answers of “yes” and “no”. And for the “open-ended” question, the answers were recorded for further analysis.
SRP Time & Strategy Table is used to record the thoughts or strategies used by the subjects during their writing process. For each subject, there is an SRP Time & Strategy Table. For example, Table 3 is the SRP Time & Strategy Table of subject Number 2, Group 1.

Table 3: SRP Time & Strategy Table of subject Number 2, Group 1
TP=Time Point
TS=Time Span         

		Strategy
TS (sec)	TP	S	G	C	P	R	E	T	O
6	01:48-01:54				√
6	03:55-04:01		√				√
8	04:24-0432				√
9	05:00-05:09						√	√
4	06:10-06:14						√
10	07:30-07:40				√
12	08:40-08:52				√
12	09:50-10:02						√
13	11:24-11:37				√
15	11:40-11:55				√
9	12:02-12:11						√
15	13:40-13:55				√
8	14:00-14:08				√
11	14:36-14:47						√
11	16:08-16:19				√
9	16:56-17:05				√
3	17:17-17:20						√
9	18:35-18:44				√
5	19:28-19:33						√
4	21:20-21:24				√
5	21:40-21:45				√
5	22:40-22:45	√

S=spelling
G=grammar
C=correction
P=planning
R=reviewing
E=expression
T=translation
O=others

TP is the time point in the tape. We identified the TP while monitoring the video camera. TS is the length of each pause in the writing process, and it is calculated by seconds. The strategies are what subject Number 2, Group 1 used during her writing process. She told us the strategies she used in each pause while watching the video tape of her own writing. From the SRP Time & Strategy Table, we got the strategies used by the subjects, pausing length, and frequency of pauses. During the data recording, it was found that more strategies were used by the subjects than those in the SRP Time & Strategy Table, such as transition, borrowing (from the instruction), using a dictionary, and overall proofreading (after finishing writing). So as a whole, 11 strategies were recorded.
Then, the tape of each subject was studied again to record the time for writing.  The time span for writing was then calculated in seconds. After that, for each of the subjects, the following data were recorded by Microsoft Excel for basic statistical analysis: frequency of different strategies used, frequency of pauses, pausing length, and the time span of writing.

After that, the data were analyzed and compared respectively by groups (Group1, Group 2, and Group 3) and gender (male vs. female). In the group analysis, for each group, the following data were first calculated by Microsoft Excel: group frequency of each strategy respectively, group frequency of all the strategies, group frequency of pauses, total pausing length, total writing time span, and proportion of total pausing length to total writing time span.

In gender analysis, similar data were calculated for male and female subjects respectively: frequency of each strategy, total frequency of all the strategies, frequency of pauses, total pausing length, total writing time span, and proportion of total pausing length to total writing time span.

After the basic analysis by Microsoft Excel, the following data were recorded by SPSS 10.0 and each subject was treated as one case: frequency of each writing strategy, frequency of all the writing strategies, frequency of pauses, mean pausing length, writing time span, and the proportion of total pausing length to writing time span. All the data were used in further correlation analysis.

In the SRP experiment, questionnaires were used as a complementary tool to get more data. Similar to the data collection process in the questionnaire section, the data of the 12 copies here were also recorded by SPSS 10.0. Since the size of the sample is too small to do quantitative analysis, the data from the questionnaire here were used as backup information for the subjects in SRP experiment. The research focused on the feelings of the subjects when they filled in the questionnaires immediately after finishing one piece of writing.

3. 3 Composition scores analysis
After all the texts of the writing of the subjects were collected, they were sent to be evaluated by two EFL experts of the Foreign Languages School of Dalian University of Technology who have more than 10 years’ experience in the field of EFL teaching. (see Appendix A) was used for the evaluation.
The ESL Composition Profile was developed by Hughey (1981). There are 5 components in the profile, and different values are given to the different aspects of ESL writing: content (30), organization (20), vocabulary (20), language use (25) and mechanics (5).
For each aspect, there are 4 levels according to the different writing levels of the ESL students. The criteria of each level are also listed. In the present study, the total scores are needed.

4. Findings and Discussion
4.1 Questionnaire findings and discussion
SPSS Crosstab Test is used to analyze the significance of differences among the choice of the groups. The items with value of P smaller than 0.05 are discussed.

Table 4: Do you like English writing?

		Group
		1	2	3	4
	Like it	6	7		5
	No special feelings	10	11	12	11
	Dislike it	4	2	7	4
	Dislike it very much			1
Total		20	20	20	20

Table 4 is the result of the general question “Do you like English writing?” From the table it is clear that a quarter of the subjects in Group 4 like English writing and only 4 subjects do not like it. On the contrary, none of the subjects in Group 3 like English writing and 8 of them dislike it or dislike it very much. The result of the general question shows that most of the English major students like English writing. But for the sophomore non-English major students, the most important thing they are facing is the CET-4. They have to learn how to write in English if they want to pass or get a good result in the test and they drill a lot in exam writing. As a result, English writing has become a hard task for them and they get bored in the learning process. This is the reason why none of the sophomore non-English major students like English writing.

Table 5: Items with P<0.05

A	totally or almost totally unsuitable
B	suitable in less than half of the time
C	suitable in about half of the time
D	suitable in more than half of the time
E	totally or almost totally suitable

Item	Group	A	B	C	D	E
Planning before writing	1				16	4
	2				16	4
	3		1	6	11	2
	4				16	4
Write a detailed outline	1	1	15	4
	2	4	11	3		2
	3	5	5	8	1	1
	4	4	16
Write soon after reading the topic	1	5	7	7	1
	2	7	10	2	1
	3		10	6	3	1
	4	6	8	5	1
Pay attention to coherence	1			7	12	1
	2			5	13	2
	3		4	2	11	3
	4			8	11	1
Translation	1	1	5	10	4
	2		6	7	7
	3		7	2	7	4
	4	1	4	9	6
Delete or add things while writing	1		4	2	12	2
	2		4	6	8	2
	3	2	2	10	6
	4		5	5	8	2

Table 5 presents the items with P smaller than 0.05. Items involved are “Planning before writing”, “Write a detailed outline”, “Pay attention to coherence”, “Translation”, “Delete or add things while writing”, and “Write soon after reading the topic”. The first three strategies are about planning.
As for the strategy of “planning before writing”, the value of P is 1.000 between Group 1 and Group 2. It means that all the subjects in Group1 and Group 2 made the same choice in item “planning before writing”. It indicates that planning plays an important role in the writing of English major students, but the value of P between Group 3 and Group 4 is smaller than 0.05. It means that for the non-English major students, they still have not formed the habit of planning every time before writing. Accordingly, planning is a unique strategy that differentiates EFL writers with higher level from those with lower level.

For the strategy of “outlining”, the value of P between Group 3 and Group 4 is 0.03, indicating the difference is extremely significant. It is clear from the table that the choices of the subjects in Group 4 center on “unsuitable”. On the other hand, about half of the subjects in Group 3 write a detailed outlining at first. This shows that English major students do not always write a detailed outline before writing, while most of the non-English major students use the strategy of detailed outline frequently. The reason is that although English major students have learnt the importance of outlining in their professional English writing training, they do not have to write a detailed outline every time because they are more frequently exposed to an English environment, so they are used to thinking in English. As a result, most of the English major students only brainstorm relative ideas before writing and then select the useful ones quickly.

On the contrary, the writing ability of non-English major students is still relatively weak, so they have to write a detailed outline to help them keep in mind what to write and how to organize the ideas. This is the reason why non-English major students use the strategy of detailed outlining more frequently than English major students.

Concerning the strategy of “write soon after reading the topic”, the choices of the subjects in Group 2 center on “unsuitable” while those of the subjects in Group 3 center on “suitable”. Although the subjects in Group 2 are only freshman English major students, they have already learnt the importance of planning before writing. For those students, planning has become a necessity every time before they want to write English articles. Nonetheless, the subjects in Group 3, sophomore non-English major students, have not realized the importance of planning even though they have been taught by their English teachers. Most of them still cannot get rid of the habit of starting to write immediately after reading the topic without a primary writing plan. From the data, it is concluded that “planning” is one of the key strategies that differentiate professional writers from unprofessional ones.

As for the strategy of “coherence”, it is obvious that about a quarter of the subjects in Group 3 chose “suitable in less than half of the time” while more than half of the subjects in Group 4 chose “suitable in more than half of the time” or “totally or almost totally suitable”. Although the choices of the subjects in Group 4 center on “suitable”, there are still 8 subjects in this group that chose “suitable in about half of the time”. The data indicate that most of the English major students know the importance of coherence in English writing, but some of them are still not sure that they pay attention to coherence in their writing. On the other hand, some of the non-English major students have not yet realized the importance of coherence in English writing. So it is clear that for the EFL learners as a whole, they do not have a strong awareness of coherence strategy.

As for the strategy of “translation”, the choices of the subjects in Group 4 center on “unsuitable” while the choices of the subjects in Group 3 center on “suitable”. This means that most of the English major students do not use translation in their writing while a large proportion of non-English major students tend to think in Chinese first and then translate their ideas into English during their writing. The difference lies in the degree of translation in writing. For English major students who are more proficient writers, the degree of translating ideas consciously in writing is lower than the less proficient writers, the non-English major students.

For the strategy of “delete or add things while writing”, the data show clearly that about three quarters of the subjects in Group 1 use deleting or adding during their writing, while only about a quarter of the subjects in Group 3 chose “suitable in more than half of the time”. Half of the subjects in Group 3 chose “suitable in about half of the time,” meaning they are not sure whether or not they delete or add things during writing. Another significant difference is that 2 subjects in Group 1 chose “totally or almost totally suitable” while 2 subjects in Group 3 chose “totally or almost totally unsuitable”. In the table of Chi-Square Tests, the value of P here is 0.017. It means that the difference between the two groups here is very significant. All the data mentioned above indicate that junior English major students delete or add things frequently during writing, but sophomore non-English major students do not use it frequently.

The reason lies in the fact that the writing proficiency of junior English major students is higher. In order to make their writing better, English major students always correct what they have written during the writing process by adding or deleting things. But for the non-English major students, whose writing proficiency is relatively lower than the English major students, expressing their ideas in English correctly is already challenging, let alone making their expressions better. Consequently, more junior English major students use the strategy of “deleting or adding things during writing” than sophomore non-English major students.

Table 6: Items with P=1.00

Item	Group	A	B	C	D	E
Pay attention to spelling &grammar	1		5	2	11	2
	2		5	2	11	2
	3	1	4	5	9	1
	4		3	3	13	1
Fluent and authentic expression	1		3	9	7	1
	2		3	9	7	1
	3	3	1	10	5	1
	4		3	9	7	1

Table 6 presents the items with the value of P equal to 1.00, which indicates that the choice of the items between Group 1 and Group 2 is exactly the same. Items involved are “pay attention to spelling and grammar” and “fluent and authentic expression” in writing. The data show that although the writing proficiency of English major students is relatively high, they are still EFL learners, and they have problems in spelling and grammar as well. The strategy of paying attention to spelling and grammar is used frequently in their writing. 

As for the strategy of fluent and authentic expression, the data indicate that in both Group 1 and Group 2, fewer subjects pay attention to the fluency and authenticity of expression in English writing. It also shows that even for junior English major students who have had 2.5 years of professional English writing training, it is hard for them to maintain fluency and authenticity of expression.

4.2 SRP experiment findings and discussion
12 subjects took part in the SRP experiment: 4 subjects in Group 1 (junior English major students), 4 subjects in Group 2 (freshman English major students), and 4 subjects in Group 3 (sophomore non-English major students). There are three parts of data in the SRP experiment: the SRP Question Sheet, SRP Time & Strategy Table, and composition evaluation. Qualitative data were obtained from the SRP Question Sheet and quantitative data were acquired from the primary analysis of SRP Time & Strategy Table.

4.2.1 SRP Questions findings and discussion
For the data from the SRP Question Sheet and SRP Time & Strategy Table, frequency of different variables (such as strategies and pauses) was analyzed because it is the most obvious and easiest indicator of a descriptive analysis.
The design of the SRP questions followed two rules. On the one hand, the questions can check the consistency of the students’ writing and whether and how much planning they do before writing. On the other hand, the questions investigate the validity of the method of video-taping during writing and video-watching after writing. What’s more, the questions about the main problems in the subjects’ writing are also included in the sheet.

According to the answers of the subjects and the observations of the researcher concerning the consistency of writing, subjects with a clear writing plan always wrote fluently and smoothly, while those with no or an unclear writing plan had to stop many times during the writing, and it was difficult for them to continue and finish the writing as well. From the video-tapes that were observed and analyzed, it is clear that planning does play an important role in EFL English writing.

What is more, almost all subjects in three groups stuck to their writing plan, but from the interview, the researcher was told by some subjects that they changed their ideas more or less during their writing process, especially the subjects in Group 1 and Group 2. The reason lies in the fact that subjects in Group 1 and Group 2 are English major students whose English writing proficiency is relatively higher than the non-English major students. And for these English majors, organizing new ideas during writing will not be so big a problem as it may be for the non-English major students.

As for the validity of the method of video-taping and video-watching, the answers of the subjects show that Stimulated Recall Protocol is much less intrusive compared with Think-Aloud Protocol. However, it still bothers the subjects if they are video-taped while writing, especially when the camera is visible to them. Video-taping the subjects from another room may be a better way.

As for the effectiveness of the video-watching process, all subjects agreed that the video-watching process did help them recall their thoughts during the “pauses”. It further proves the validity of SRP. Additionally, all of the subjects recognized that the writing texts of their own and the video-watching process could help them recall more accurate information.

4.2.2 Pauses & Strategies findings and discussion
Strategies and pauses are the center of the entire Stimulated Recall Protocol. The aim of the SRP experiment is to know what the subjects were thinking while they were writing, and what strategies they were using during the “pauses”. From the SRP Time & Strategy Table, the researcher got the raw information of each subject and then did a primary analysis using Microsoft Excel. Personal data of “pauses” and strategies used during the pauses were analyzed. There are more strategies in Personal data of Strategy (Appendix B) than in SRP Time & Strategy Table, such as transition, borrowing (from the written instruction, using a dictionary and overall proofreading. The reason is that some of the strategies used by the subjects during their writing are not listed in the SRP Time & Strategy Table designed by the researcher. Take using a dictionary as an example. The writing proficiency of two subjects in Group3 was so low that they could not finish their writing without the help of dictionary. As a result, the researcher had to allow them to look some of the words up in dictionary. But for other subjects who do not have this kind of problem, using a dictionary is not allowed in order to prevent them from copying sentences directly from the dictionary, which will lower the reliability of the data obtained from the SRP process.

Group data of “pauses” and “strategies” are presented respectively as follows.

Table 7 Group data of “pauses”


Table 7 is the group information of “pauses”. It shows that in the three groups, subjects in Group 3 paused most frequently during writing and the subjects in Group 2 paused least frequently. The total pausing length follows the same order. It indicates that sophomore non-English major students and junior English major students pause more often and longer than the freshman English major students. For the non-English major students, pausing is inevitable because their writing proficiency is relatively lower than other two groups. They have to stop from time to time to organize their ideas or consider the expression of the language. But for the junior English major students, pausing is a period of time when they try to use what they have learnt in the years of professional writing training and write in a better way. So they tend to pause longer than the subjects in other two groups. That explains why the mean pausing length and writing time span of Group 1 are both the longest among the three groups.

As for the proportion of pausing length to writing time span, again Group 3 is the highest and Group 2 is the lowest. It means that although the subjects in Group 1 spent more time in writing, and their mean pausing length is longer than other two groups, their total pausing length is shorter than that of Group 3. On the other hand, the subjects in Group 3 spent less time in writing, but their total pausing length is the longest among the three groups. As a result, the proportion of pausing length to writing time span of Group 3 is higher than the other two groups. The order of the proportion indicates that the subjects in Group 1 wrote more fluently than those of Group 3.
It is also obvious from the table that all the data of Group 2 are the smallest among the three groups. Compared with Group 1 and Group 3, the subjects in Group 2 are freshman English major students. They are EFL writing beginners with half a year of professional writing training, which helps them write more proficiently than sophomore non-English major students in Group 3. As a result, they wrote more naturally, affected less by the “rules” or “skills” in writing, paused less, and often wrote quickly.

Graph 1 presents the frequency of each strategy and the total number of strategies in each group. From the graph it is clear that for each strategy, the frequency order of the three groups is different. The subjects in Group 1 used most frequently correction, planning and reviewing. The subjects in Group 3 used most frequently grammar, expression, spelling, translation, borrowing(from the instruction), transition, using a dictionary, overall proofreading and the total frequency of strategies is the largest among the three. Group 2 had no preference concerning strategy use.

According to the hypothesis, the subjects in Group 1 are the most proficient writers among the three groups, so they pay more attention to the quality of their writing and try their best to write flawlessly. Therefore, they use more planning, correction and reviewing than the subjects in the other two groups.

As for the subjects in Group 3, their writing proficiency is the lowest among the three groups. As a result, they have to make the effort to make sure that their grammar and expressions are right. Further, it is recognized that spelling, translation and borrowing (from the writing instruction) are the basic strategies in English writing. And to some degree, the frequency of these basic strategies used in EFL writing negatively correlates with the level of writing proficiency according to the result. During the writing process, 2 subjects in Group 3 could not finish writing without using a dictionary. In addition, only the subjects in Group 3 used overall proofreading. It is an unexpected result, because all the English major students have learnt the importance of proofreading. The reason may be that they did not take the writing in the present study as seriously as did the subjects in Group 3.

The frequency of pauses of Group 3 is the highest among the three groups. Since “pauses” are used in the present study as a “window” to see what strategies the subjects use while writing, and usually subjects use more than one strategy in one pause, the frequency of all the strategies used more or less positively correlates with the frequency of the pauses. This is the reason why the order tendency of the frequency of all strategies used is the same as that of the pauses.

4.2.3 Composition scores findings and discussion
A special program was written in Microsoft Visual Basic (VB) to calculate the Z-score of the 12 subjects taking part in SRP experiment. Two sets of T-scores and Z-scores for each subject are presented in Appendix C. Since 2 experts were involved in the composition evaluation session, there are two sets of T-scores and Z-scores for 12 subjects, so the mean T-score and Z-score of each subject are used in the present study (see Appendix B). Standard Deviation (SD) and mean of each set of T-score are analyzed in the following part.

Table 8: SD and mean of two sets of T-scores
1=scores given by expert NO. 1
2=scores given by expert NO. 2

Standard Deviation (SD)		Mean (M)
SD 1	SD 2	T-score 1 M	T-score 2 M
12.75	11.92	68.25	69.25

Table 8 presents the standard deviation (SD) and the mean of the two sets of T-scores. It is clear that for the same 12 subjects, the SD of T-score given by expert NO. 1 is 12.75, not largely different from 11.92, the SD of T-score given by expert NO.2. In the mean time, the mean score of the two sets of the T-score is 68.25 and 69.25, only 1 point different from each other. The data shows that there is not a large difference between the two sets of T-scores. The close SD and mean of these two sets of T-scores confirm the reliability and validity of the ESL Composition Profile. The reason is that different EFL experts used the same ESL Composition Profile to evaluate the compositions of the subjects and the result is almost the same. In addition, the reliability and validity of ESL Composition Profile in turn guarantee the reliability of the scores.

Table 9: Mean T-score and mean Z-score of each group

Group	Group mean T-score	group mean Z-score
1	80.5	0.96
2	66.125	-0.22
3	59.625	-0.74

Table 9 presents the mean T-score and mean Z-score of each group. It is clear from the table that for both the T-score and Z-score, Group 1 is the highest, Group 2 in the middle and Group 3 the lowest. The scores confirm the hypothesis that the writing proficiency of the junior English major students is the highest, freshman English major students in the middle, and sophomore non-English major students the lowest.

4.2.4 Correlation analysis
One of the research questions is to find out the relations among the composition scores and other variables, such as frequency of each writing strategy, frequency of all the writing strategies, frequency of pauses, mean pausing length, writing time span, and the proportion of total pausing length to writing time span. In the present study, correlation analysis was used to get the answer. The variables mentioned above were first recorded by SPSS 10.0 and then analyzed by Bivariate Correlation Tests. The correlation between each 2 variables was analyzed. Since the size of the sample is 12, smaller than 30, Spearman Correlation Coefficient for non-parametric analysis is used to indicate the significance of correlations. The result of the correlation analysis is presented in Appendix H.

From Appendix H, we could see that the correlation of several variables is significant (at the 0.01 level or the 0.05 level). The results are presented in the following table.

Table 10: Significant correlation

Variable 1	Variable 2	Correlation coefficient
Score	Translation	0.583*
Reviewing	Correction	0.761**
Transition	Expression	0.820**
Mean pausing length	Translation	0.643*
Strategy frequency	Planning	0.582*
	Translation	0.629*
Pause frequency	Planning	0.732**
	Expression	0.669*
	Transition	0.618*
	Strategy frequency	0.935**
Writing time span	Strategy frequency	0.662*
	Pause frequency	0.769**
Pausing length / writing time span	Planning	0.606*
	Strategy frequency	0.651*
	Pause frequency	0.713**
	Mean pausing length	0.685*

**	Correlation is significant at the 0.01 level (2-tailed).
*	Correlation is significant at the 0.05 level (2-tailed).

Table 10 shows that all the significant correlation is positive correlation. Composition score only positively correlates with the strategy of translation. It indicates that EFL learners (with either high or low writing proficiency) have to use translation in writing if they want to get high scores.
The strategy of reviewing correlates with correction at the 0.01 level. It means that the two strategies are used together most of the time. The reason is simple. Most of the subjects who reviewed their writing would always find some mistakes or inappropriate expressions, so they corrected them after reviewing, which makes these two strategies closely correlate with each other.

Two other closely correlated strategies are transition and expression. During the video-watching process, the researcher found that most of the subjects who paused for better expressions also said that at the same time, they were trying to figure out how to transit to the next part. As a result, the strategy of expression and transition are always used together.

Mean pausing length correlates with the strategy of translation. It means that the longer the mean pausing length is, the more times the strategy of translation is used by the subjects. The reason is not hard to understand. All the subjects are EFL learners and the process of translating their own ideas from Chinese to English needs to be done in pauses. Usually the translation process needs time. As a result, the more they translate, the longer they pause. 
The frequency of all the strategies correlates with two other variables: planning and translation. It indicates that planning and translation are the top 2 strategies used most frequently by all the subjects. For all the subjects, the most important strategies they need in EFL writing are planning and translation, which also positively correlates with composition scores. It means that if the subjects want to get higher scores in writing, they have to plan carefully and frequently translate from Chinese to English.

The frequency of pauses correlates with four other variables: planning, expression, transition, and frequency of all the strategies, in which planning and frequency of all the strategies significantly correlate with frequency of pauses at the 0.01 level. Since all the strategies were used in the pauses, the frequency of pauses naturally correlates with the frequency of all the strategies. And as mentioned above, planning is one of the most frequently used strategies used by the subjects, so it in turn correlates with the frequency of the pauses. As for the other two strategies, expression and transition, as analyzed before, these two strategies significantly correlate with each other. Both of them correlate with the frequency which indicates that subjects paused mostly for better expressions and the expressions are always about how to transit.

Both the frequency of all the strategies and pauses positively correlate with writing time span. It means that the more frequently the subjects paused, the more strategies they used, and the longer time they spent in writing. The result is also consistent with the researcher’s observation. Subjects who spent longer time in writing always paused more frequently to think and in turn used more strategies.
The last variable to analyze is the proportion of total pausing length to writing time span. It is clear that it positively correlates with planning, frequency of strategies, frequency of pauses and the mean pausing length. It reveals a simple math phenomenon. Pausing length is the result of mean pausing length times the frequency of pauses, so it of course positively correlated with the latter two variables. And as mentioned before, frequency of pauses correlates with frequency of all the strategies, among which planning is one of the most frequently used. As a result, given the same writing time span, planning, frequency of all the strategies, frequency of pauses and mean pausing length all correlate with pausing length.

5. Conclusion and Implications
5.1. Research conclusion
Two different types of instruments are employed in the present study, in which a questionnaire is used as a quantitative instrument and Stimulated Recall Protocol as a qualitative instrument. The subjects are Chinese college EFL learners with different levels of writing proficiency. The findings of the present study are summarized as follows:
Firstly, as for writing strategy use, it is also confirmed that there are significant differences in strategy use between English majors and non-English majors. What is more, there are more significant differences between junior English majors and sophomore non-English majors than between freshman English majors and sophomore non-English majors. However for junior and freshman English majors, there is no significant difference in strategy use.

Secondly, English majors hold a more positive attitude towards English writing. Most of the English majors like English writing, while none of the non-English majors like English writing. The reason behind this phenomenon is worth searching for.
Thirdly, both English majors and non-English majors cannot handle the strategy of coherence well. It reveals that in writing, EFL learners still do not pay much attention to coherence, which is extremely important in English writing.
Fourthly, English majors pay attention to more “global” strategies, such as planning, while non-English majors pay attention to more “local” strategies, such as grammar and spelling. It indicates that more proficient writers write more systematically and consider more wholly than the less proficient ones. Both the questionnaire and SRP experiment confirm this result.

Fifthly, quantitative and qualitative instruments should be integrated in research to draw conclusions. In the present study, the result of the correlation tests shows that composition scores correlate positively with only one writing strategy, translation. It is also shown that the composition scores also positively correlate with the writing proficiency of the writers, but the questionnaire shows that the most proficient writers seldom use translation in writing. The two results contradict each other. The reason lies in the different nature of the quantitative and qualitative instruments. What the subjects choose in the questionnaire may just be what they consider right in theory, but they may not put it into practice in real writing. Therefore, combining the data from the quantitative instrument with the first-hand material from the qualitative instrument is a scientific method for the research.
Finally, the reliability and validity of instruments in the present study are confirmed. SRP is proved to be an effective qualitative instrument in English writing research and the ESL Composition Profile is a classic EFL composition evaluation system.

5.2. Pedagogical implications
As an empirical study, the present study can serve many purposes. The most important purpose, however, is putting the findings of the study into practice of EFL writing and improving the writing proficiency of Chinese college EFL learners. Some suggestions on teaching and learning EFL writing are listed below.

Firstly, it is necessary to change the attitude of non-English majors towards English writing. It is commonly believed that English writing is the most boring course in all EFL courses, especially for non-English majors. The teachers could provide more interesting materials and create an active atmosphere in writing class and change the situation gradually.
Secondly, the importance of coherence should be emphasized. The result of the present study shows that both English majors and non-English majors have not realized the importance of coherence. EFL writing teachers could provide more authentic writing materials to help the students get familiar with the coherent writing pattern of native writers.
Thirdly, the translating ability of the students should be improved. The result of the present study shows that the strategy of translation positively correlates with composition scores, which proves that in a non-English atmosphere, such as in China, it is hardly possible for EFL learners to think in English during writing, even for the most proficient writers. The better they translate their ideas into English, the higher the scores they get. This implies that if EFL learners want to improve their writing proficiency, their translating ability should be improved first. EFL could pay more attention to translation training in class, especially for non-English majors who do not have access to translation classes at present.
Last, both global and local writing strategies should be emphasized in EFL writing teaching. These two different types of strategies play an equally important role in EFL writing. Only if the writers handle both types of strategies can they improve their writing proficiency.

References
Anderson, N. J. (2002). The role of metacognition in second language teaching and learning. Eric Digest, EDO-FL-01-10. Retrieved April 10, 2007 from http://www.cal.org/resources/digest/0110anderson.html

Baak, E. (1997). Portfolio development: An introduction. Forum, 35(2), 38. Retrieved September 10, 2006 from
http://exchanges.state.gov/forum/vols/vol35/no2/p38.htm

Bosher, S. (1998). The composing processes of three southeast Asian writers at the post-secondary level: An exploratory study. Journal of Second Language Writing, 7(2), 205-241.

Daoud, S. A. (1998). How to motivate EFL learning and teaching of academic writing by cross-cultural exchanges. English for Specific Purposes, 17(4), 391–412.

English, J. A. MOO-based metacognition: Incorporating online and offline reflection into the writing process. Retrieved September 10, 2006 from http://english.ttu.edu/kairos/3.1/features/english/metacog.html

Flower, L. & Hayes, J. R. (1981). The cognitive process model of the composing process. Retrieved October 20,2005 from
http://faculty.goucher.edu/eng221/Flower_and_Hayes_Cognitive_
Process_Model_of_Composition.htm

Furneaux, C. (1998). Process writing. In Johnson, K & H Johnson (Eds.),Encyclopedic Dictionary of Applied Linguistics (pp.257-260). Oxford: Blackwell. Retrieved December 20,2005 fromhttp://www.rdg.ac.uk/app_ling/process.htm

Halter, J. Metacognition. Retrieved October 20,2005 from http://coe.sdsu.edu/eet/articles/metacognition/start.htm,

Hansman, C. A. & Wilson, A. L. (1998). Cognition and practice: Adult learning situated in everyday activity. Retrieved June 20,2006 from http://www.edst.educ.ubc.ca/aerc/1998/98hansman.htm

Jacobs, H. L., Hartfiel, V. F., Hughey, J. B., & Wormuth, D. R. (1981). Testing ESL composition: A practical approach. Boston, MA: Newbury House.

Jiménez-Jiménez, A. F. (2005). The analysis of compensatory strategies applied to language attrition research. The Proceedings of 2nd International Conference on First Language Attrition, 2005-8.

Levine, A,& Reves, T. (1998). Data-collecting on reading-writing strategies: A comparison of instruments: A case study. TESL-EJ, 3(3).Retrieved June 20,2006 fromhttp://www-writing.berkeley.edu/TESL-EJ/ej11/a1.html

Myers, J. L. (2001). Self-evaluations of the ‘stream of thought’ in journal writing .System,29, 481–488.

Petric,´B. & Cza´rl, B.(2003). Validating a writing strategy questionnaire .System,31,  187-215.

Piolat, A. & Roussey, J. (1996). Students' drafting strategies and text quality. Learning and Instruction, 6(2), 111-129.

Sasaki, M. (2000). Toward an empirical model of EFL writing processes: An exploratory study. Journal of Second Language Writing,9(3), 259 -291.

Shao, H. W. & Fang, W. L. (2003). Affective filter in second language acquisition and teachers' responses to learners' performances in class. Journal of Yangzhou University (Higher Education), 7(2), 90-93.

Schlig, C. Improving foreign language writing competence. Retrieved June 20,2006 from http://education.gsu.edu/ctl/FLC/Schlig_FLWriting_revised.htm

Slavin, R. E. (2004). Educational psychology: Theory & practice. Boston, MA: Allyn & Bacon.

Søvik, N. & Flem, A. (1999). The effects of different tasks on children’s process and product variables in writing. Learning and Instruction,9, 167–188.

Wu, H.Y.& Liu, R. Q. (2004). A metacognitive framework for L2 writing: A factor analysis approach. Foreign Language Teaching and Research, 36(3), 187-195.

Yang, S. X. (2002). The differences of writing strategies employed by successful EFL learners and less successful EFL learners. Foreign Language World,89(3), 57-64.

Yanga, L.& Shib, L.(2003). Exploring six MBA students’ summary writing by introspection. Journal of English for Academic Purposes, 2, 165–192.

Winckel, N. V. (1982).The role of preconscious thought in the composing process. JAC, 3(1/2).Retrieved October 20,2005 from
http://www.jacweb.org/Archived_volumes/Text_articles/V3_Winckel.htm

Appendix See PDF