The researcher chose English majors as the participants of this current study

CHAPTER THREE METHODOLOGY

This study was designed mainly to examine Taiwanese English majors’ writing anxiety, writing strategy use, and the relationship between these two factors. The relationship of writing anxiety and strategy use to the students’ grade level was also investigated in the study. This chapter presents the research method of this study.

Participants

A total of 231 English majors from a national university located in northern Taiwan and one private university in central Taiwan were invited to participate this study. The participants were all undergraduates recruited from the last three grades of the English department (sophomores, juniors, and seniors). An intact class was selected from each of the three grades of the two universities. At the time of study, all of the participants had taken their required English writing courses for at least one year. The seniors had received three years’ writing instruction. The juniors had finished two years’ writing courses. The sophomores had completed one year’s writing course.

The researcher chose English majors as the participants of this current study

because they had more sufficient opportunities for writing in English than senior high

school students and non-English majors. English writing is a required course in every

English department in Taiwan. As a result, English majors would have more

opportunities to use writing strategies than other students while writing their papers in

English.

Instruments

One questionnaire that consists of two inventories was used to collect information on language learners’ writing anxiety and writing strategy use (see Appendix A). The two inventories included Cheng’s (2004a) Chinese version of the Second Language Writing Anxiety Inventory (SLWAI) and a modified Chinese version of Petric and Czarl’s (2003) Second Language Writing Strategy Questionnaire (SLWSQ). The questionnaire is composed of 3 sections and a total of 75 items.

Section 1 consists of 7 items that elicit the participants’ background information.

Section 2 contains a total of 22 items that concern participants’ writing anxiety.

Section 3 comprises a total of 36 items that concern participants’ writing strategy use.

Second Language Writing Anxiety Inventory

Students’ self-reported English writing anxiety was measured by Cheng’s (2004a) original Chinese version of SLWAI (see Appendix A, Part 2). Before the development of SLWAI, the Daly-Miller Writing Apprehension Test (WAT; Daly & Miller, 1975a) was the most commonly used measure of second language writing anxiety. The researcher chose the SLWAI because it appears to be more appropriate than the WAT for the present study. Firstly, the WAT was originally developed for first language learners, especially for English native speakers; as a result, it may not appropriately scrutinize second language writing anxiety. In contrast, the SLWAI was originally developed for L2 learners and written in Chinese; therefore, this inventory is much more suitable to the Taiwanese EFL subjects in this study. Secondly, the construct validity of the WAT has been questioned by a number of researchers (e.g., Burgoon &

Hale, 1983; McKain, 1991; Shaver, 1990). According to these researchers, the items

on the WAT did not deal with the feeling of anxiety specifically; components such as

self-confidence, outcome expectancies, and other constructs broader in scope than

writing anxiety were embedded within the WAT (Shaver, 1990). In contrast, when factor analysis was conducted on the SLWAI items and the items from the English Writing Self-efficacy Scale, no item from the SLWAI was found to load on the self-efficacy factor and vice versa (Cheng, 2004a). The results suggest that the SLWAI had better evidence of validity than the WAT. Thirdly, the WAT has been used as a unidimensional measure of writing anxiety (Cheng, 2004a). Since previous research has suggested that various facets of anxiety would cause different effects on language learners (Deffenbacher, 1977; Morris & Liebert, 1973, cited in Cheng, 2004a), findings obtained through the unidimensional WAT may be more limited than those obtained through the SLWAI, which adopts Lang’s (1971) tripartite framework to conceptualize anxiety.

The SLWAI is a 22-item self-report instrument used to scrutinize L2 students’

writing anxiety. It consists of three subscales: (1) Somatic Anxiety, composed of items measuring students’ physiological reactions (e.g., pounding heart, trembling body, and perspiration) (2) Avoidance Behavior, comprising items dealing with students’

avoidance behavior towards L2 writing; and (3) Cognitive Anxiety, consisting of items on students’ perceived worry and fear of negative evaluation. The SLWAI measures the degree of writing anxiety with a 5-point Likert response format (1 = strongly disagree; 2 = disagree; 3 = neither agree nor disagree; 4 = agree; 5 = strongly agree).

During the two administrations (test-retest) of the SLWAI, a reliability estimate

of .91 (Cronbach alpha) was obtained for both of these two tests, and the test-retest

reliability estimate was .85. For the reliability estimates (Cronbach alpha) of the three

subscales in the SLWAI, an average of about .88 was obatined for the Somatic

Anxiety subscale, .87 for the Avoidance Behavior subscale, and .83 for the Cognitive

Anxiety subscale. The test-retest reliability estimates of the three subscales were all

higher than .81 (.82 for the Somatic Anxiety subscale, .83 for the Avoidance Behavior subscale, and .81 for the Cognitive Anxiety subscale). As a result, the SLWAI as well as its three subscales can be seen as instruments with satisfactorily internal consistency and test-rest reliability. Regarding the validity of the SLWAI, the results of factor analysis of the SWLAI with the self-efficacy measure, as summarized in the previous paragraph, and the correlation analysis between the SLWAI and other criterion-related measures proved the convergent and discriminate validity of this inventory as well as its three subscales.

Second Language Writing Strategy Questionnaire

The frequency of students’ use of English writing strategies was measured by the Second Language Writing Strategy Questionnaire (SLWSQ) developed by Petric and Czarl (2003). The SLWSQ was translated into Chinese and modified for use in this present study (see Appendix A, Part 3). The development of the SLWSQ was mainly motivated by a need to create a survey instrument that covers “the fullest range possible of strategies employed” by second language writers (Leki, 1995, cited in Petric & Czarl, 2003, p. 188).

The format of the SLWSQ was taken from Oxford’s Strategy Inventory for

Language Learning (SILL, 1990) and it originally consists of 38 items that measure

the frequency of the students’ use of writing strategies with a 5-point Likert response

format (5 = always true; 4 = usually true; 3 = somewhat true; 2 = usually not true; 1 =

never true). These 38 items are “sequenced following the structure of the writing

process” (Petric & Czarl, 2003, p. 190): (1) planning strategies (8 items), (2)

while-writing strategies (14 items), and (3) revising stages (16 items). The items of

the SLWSQ came from the researchers’ personal experiences as non-native writers in

English and writing teachers, informal interviews with students, and from the

literature on similar issues.

The SLWSQ was validated by both quantitative and qualitative methods. A total of 30 students in 2002 took part in the study: 22 fourth-year secondary school students in a Hungarian-English bilingual school and 8 second-year English majors at a Hungarian university. According to Petric and Czarl (2003), the participants had taken a writing course as a compulsory part of their education and had had several years of extensive and varied experience in ESL writing. In their study, test-retest method was chosen as the main reliability check method. The questionnaires were administered twice at the students’ educational settings. Spearman correlations were calculated for all the 38 questionnaire items. Based on the suggestions given in Reid (1990) and Sakui and Gaies (1999), the items were classified as acceptable when satisfying the following two conditions: the test-retest correlation is above .60 and at least 60% of the subjects’ responses are the same on the two tests. It was found that 15 items satisfied the set correlation condition, 14 items met the condition of 60% matching responses, and only 9 items satisfied both of the conditions. Think-aloud procedure was chosen as the qualitative method to validate this questionnaire. Students’

responses were transcribed and summarized item by item. Based on the think-aloud

protocols, 19 items were regarded as having no validity problems. However, only 5

items were acceptable based on both validity and reliability measures. The analyses

also showed that quantitative and qualitative criteria could complement each other in

only 20 out of the 38 items. Based on these findings, Petric and Czarl (2003) provided

suggestions for improving the items in the SLWSQ. These suggestions were

incorporated into the modified version of the SLWSQ in the present study. Items from

other related questionnaires were also reviewed to help modify the SLWSQ. The

SLWSQ has been modified twice: the first time for the use in the pilot study and the

second time for the formal study. The modified SLWSQ (see Appendix B) used in the

pilot study consisted of 46 items (see Appendix C for details about the rationales for modifications). Moreover, in the pilot study, 2 open-ended questions were added at the end of the questionnaire to find out other writing strategies used by the students and problematic items on the questionnaire. Further modifications in the formal study were made based on the results of the pilot test (see Appendix D for the details about the modifications made in the formal study).

Data Collection Procedures Pilot Study

Before the formal study, a pilot study was conducted in order to investigate the test-retest reliability of the SLWSQ and to discover problematic items. Test-retest reliability was used here instead of internal consistency because it is still debatable whether all of the writing strategies could represent a single construct (Petric & Czarl, 2003). A total of 41 English majors who were juniors in an intact class of a national university located in northern Taiwan were invited to participate this pilot study in October 2006. A 52-item questionnaire which contained the modified Chinese version of the SLWSQ (46 items) and 6 items on background information (see Appendix B) were administered to the participants twice, with an interval of two weeks.

To measure the test-retest reliability of the modified SLWSQ, Pearson

product-moment Correlation was performed for each subscale and each item. The

test-retest reliability of each subscale was as follows: prewriting strategies (.73), while

writing strategies (.64), revision strategies (.61). In addition, the calculation of the

test-retest reliability of each item showed that the coefficient of 6 items (item 1.6,

item 1.9, item 2.14, item 3.9, item 3.11, and item 3.15; see Appendix D) was

non-significant (p < .05). After removing these items, the reliability of this writing

strategy inventory was improved. Meanwhile, the participants’ feedbacks to this inventory were also adopted by the researcher. Based on the participants’ responses, some items which did not correspond to real writing situations were removed (i.e., the original item 3.1 and item 3.2; see Appendix D for the details). Moreover, some items were added (i.e., the modified item 1.1, see Appendix D for the details) and combined (i.e., the original item 2.12, item 2.13, item 2.19 and item 2.20 were combined as one item, see Appendix D for the details) by the researcher in order to make this inventory more applicable to this present study.

Formal Study

In the early January of 2007, six classes of 231 English majors from a national university located in northern Taiwan and one private university in central Taiwan were invited to participate in this current study. Within these two universities, one intact class was recruited from each of the last three grades (sophomores, juniors, and seniors) of the English department. The researcher firstly asked the instructors of the students’ classwide courses for help, explaining to them the purpose of this study and administration procedures. After getting their permission, the researcher went to each class to administer the questionnaires herself. The participants were all informed that:

(1) there would be a 15-minute session for the survey; (2) their responses to these

questionnaires would not influence their course grades, but only served as research

data; (3) their honest and forthright answers would be a great help to this study. After

completing the questionnaires, the students turned them in to the researcher

immediately. All of the questionnaires were collected by the researcher and keyed into

computer for statistical analyses.

Data Analysis Procedures

The data of the questionnaires were analyzed through descriptive and inferential statistics using SPSS statistic package for window 13.0. Descriptive statistics for each variable were firstly analyzed and summarized. One-way repeated-measures ANOVA was used to examine whether there were significant differences in the participants’

responses to the three dimensions of writing anxiety and to the three phases of writing strategy use. Moreover, Pearson product-moment correlation was computed among the SLWAI total scores, the SLWSQ total scores, the SLWAI subscales, and the SLWSQ subscales to determine the relationship between writing anxiety and writing strategy use. A significance level was set at .05.

In addition, the effects of the students’ grade level on their perceived writing

anxiety (including the total scale and the three subscales) and writing strategy use

(including the total scale and the three subscales) were examined by performing a

series of one-way ANOVA.