Abstract
Writing for academic purposes in a second/foreign language is a major challenge faced by many students at both secondary and tertiary levels. This suggests that displaying content knowledge and understanding of a subject through a second language is a very complex process. This article discusses the findings of a longitudinal intervention study that attempted to investigate the impact of writing strategy instruction on writing strategy use and writing performance of a group of undergraduate students following a course in English for Academic Purposes in Sri Lanka. The study used a pre-test post-test experimental research design and the data were collected using multiple methods. The results revealed that the students could be trained to use writing strategies effectively and their writing strategy use and writing performance increased significantly after strategy instruction.
I Introduction
There is no argument that writing is crucial to academic success of students in higher education. The task of writing becomes extremely challenging to the writer when the medium of writing is in a second language (L2). As Kellogg (1999) argues, successful writing depends on the ability of the writer to retrieve and apply relevant procedures, schemas, facts and episodes through working memory. The literature related to language learner strategies (LLS) discusses the involvement of strategies in working memory processing. According to Macaro (2006), clusters of strategies interact with each other in the working memory and when these combinations of clusters interact with the language processes, they contribute to enhancement of performance in L2. Hence, if L2 writers learn to use strategies effectively, it may enhance their working memory functions, which may result in successful writing. ‘Having our language learners more skilled at using their strategy repertoire […] can ideally have both a short-term and a long-term impact’ (Cohen & Macaro, 2007, p. 284).
Strategy instruction has also been found to be beneficial in studies that focused on different aspects of the skill of writing (e.g., Bishop, 2001; Olson & Land, 2007; Sasaki, 2000; Sengupta, 2000). However, as Carrell (1998) points out, strategy instruction and its effects could be context-dependent. In the present context, the learners are undergraduates studying in their L2 (English) and they find it difficult to cope with the writing demands. In a broader global context, Manchón (2009) stresses the importance of L2 writing research in different settings.
The present study hence attempts to see the impact of strategy instruction on students’ writing strategy use and writing performance in an English for Academic Purposes (EAP) setting at the Open University of Sri Lanka.
II Literature review
Language learner strategy research began with the pioneering work by researchers like Rubin (1975), Stern (1975) and Naiman, Frohlich, Stern, and Todesco (1978/1996) and became popular among many researchers (e.g., O’Malley & Chamot, 1990; Oxford, 1990; Wenden & Rubin, 1987). However, the field has been criticized for not been able to come to a consensus regarding a definition for LLS and for not having a sound theoretical foundation (Dörnyei, 2005; Dörnyei & Skehan, 2003; Ellis, 1994 ). Early researchers (O’Malley & Chamot, 1990) made use of McLaughlin’s (1987) information-processing model and Anderson’s (1985) Active Control of Thought (ACT) model to describe learner strategies. However, O’Malley and Chamot (1990) identified a number of limitations in these models. An area of debate within the theoretical models of LLS is how they differ from normal processes. Macaro (2006) attempts to address this issue by proposing a model based on the connectionist approach (Ellis & Schmidt, 1997) while others base their research on the theories of self-regulation in educational psychology (Tseng, Dörnyei, & Schmitt, 2006). Tseng et al. (2006) developed a new instrument to measure strategic learning supported by the theories of self-regulation. Ellis (2008), reviewing learner strategy research, claims that the absence of a sufficiently rigorous definition has made it impossible for researchers to ‘reliably determine which learning behaviours constitute strategies and which do not’ (p. 719).
In order to address these claims, Macaro (2006) proposes a definition for a strategy and using that definition and recent views on learner strategies (e.g., Cohen, 2007; Gu, 2005), a working definition for writing strategies was formed for the purposes of the present study.
Working definition: A writing strategy is a conscious mental activity, employed in pursuit of a goal, often with an aim to solve a problem in writing within a learning situation and an activity that is ‘transferable to other situations and tasks’.
1 Research on writing strategies
Looking specifically at EAP and thus of particular importance to the present study, Leki (1995) identified English as a second language (ESL) students’ strategies developed in order to face the writing demands of their main academic disciplines (i.e., focusing on problems, seeking help from teachers and peers, making use of past and current experience or feedback, looking for models and balancing competing demands). According to Canagarajah (2002), while an understanding of students’ preferred strategies is helpful for a teacher to encourage students to stick to their own learning styles, ‘using, alternately, the strategies that are uncomfortable to them enables students to gain different skills/competencies in language’ (p. 146). This view is important when selecting strategies for a strategy instruction programme. Cumming (2006) reports the findings of a multi-year research project that investigated the goals for writing improvement of students and their teachers. The research showed that students set goals for improving language, rhetoric, ideas and knowledge in their writing.
The influence of proficiency level on writing strategy use was investigated by Sasaki in two studies. Sasaki (2000) compared the writing processes of three paired groups of English learners using stimulated recall and found that experts in her study planned for a longer time and made more detailed plans of overall organization than did the novices. Sasaki reported that after six months of instruction, novices in her study had begun to use some of the strategies used by the experts. Another study by Sasaki (2002) investigated writing processes of two groups (12 experts and 22 novices). She focused mainly on six strategies: global planning, local planning, thematic planning, rereading, rhetorical refining and translating from first language (L1) to L2. She identified some patterns of different types of English as a foreign language (EFL) writers’ writing processes by analysing their written texts and drafts and using stimulated recall protocols. Sasaki stresses the importance of carrying out qualitative studies to compare the writing processes of individuals within a group. Sasaki (2000, 2002), however, did not group students based on their performance at a pre-test and she did not provide strategy instruction for the experts and did not have a comparison group to compare the effects of instruction on the novices.
The writing processes that have gained the attention of researchers are limited mostly to planning and formulation (Manchón & Roca de Larios, 2007; Roca de Larios, Manchón, Murphy, & Marin, 2008; Roca de Larios, Marín, & Murphy, 2001; Sasaki, 2000, 2002) and these studies have investigated whether writing processes are dependent on proficiency, language used (Roca de Larios, Manchón, & Murphy, 2006) and the stage of writing. Some studies have investigated the temporal issues in the writing process (Manchón & Roca de Larios, 2007; Roca de Larios et al., 2008). Manchón and Roca de Larios (2007) discovered that with increased proficiency, the writers spent more time planning before writing and incorporated what was planned into their writing. The distribution of planning throughout the writing process also varied across proficiency levels and the language of composing.
The difference between the problem-solving nature of L1 and L2 formulation processes from a temporal perspective was studied by Roca de Larios et al. (2006). The time spent by writers on solving L2-related problems while formulating was twice as much as the time they devoted for the same when writing in L1. This has also been observed in previous studies (e.g., Roca de Larios et al., 2001; Silva, 1993). The predominance of formulation from a temporal perspective was also investigated in a study by Roca de Larios et al. (2008). With increased proficiency, more balanced allocation of time to different writing processes (i.e., a gradual increase of the time spent on planning, evaluation and revision and a parallel decrease in the time devoted to formulation) was observed. The findings of Roca de Larios et al. (2008) shed light on the issues in theorizing composing processes as it revealed ‘(i) the predominance of certain processes over others; (ii) the proficiency-dependency of cognitive activity while writing; and (iii) the recursiveness of the writing process’ (p. 43).
2 Studies on strategy instruction
As shown above, the majority of studies on LLS have focused on descriptive studies on strategy use. This section considers how some of the themes identified in those studies manifest themselves in strategy instruction studies. Research on writing strategy instruction includes evaluation of the effects of training of individual writing strategies, such as dictionary use while writing (e.g., Bishop, 2001), revision strategy instruction (e.g., Sengupta, 2000), brainstorming (e.g., Rao, 2007), self-monitoring (e.g., Conti, 2004), planning the text structure (e.g., Kirkpatrick & Klein, 2009) and training of two or more writing strategies, such as planning and revision (Ching, 2002; Sasaki, 2000, 2002), planning, formulating and monitoring (Graham & Macaro, 2007) and a number of writing strategies (Macaro, 2001). The orchestration of strategy use was the focus of some studies. Olson and Land (2007) focused on both reading and writing while Graham and Macaro (2007) focused on both listening and writing in their strategy training research.
The impact of metacognitive strategy instruction on students learning French (L2) was investigated by Aziz (1995, as cited in Macaro, 2003). Aziz found that the group that received cognitive strategy training only showed improvement in grammatical agreement but not in overall quality in writing. The group that received both cognitive and metacognitive strategy training performed significantly better than the group that received cognitive strategy training only and produced writing with better overall quality and grammatical agreement.
The effect of metacognitive strategy training in self-monitoring on students’ writing was investigated by Conti (2004) using 10 university students learning Italian as a foreign language. The use of self-monitoring strategies, including error targeting, feedback handling, production monitoring and monitoring familiar errors and editing were investigated. This longitudinal study reported that the students showed significant reduction in the number of errors they made while retaining the quality of the content in their writing.
The orchestration of strategy use was the focus of a longitudinal strategy instruction study by Olson and Land (2007) that focused on both reading and writing skills of English Language Learners (ELLs). Olson and Land (2007) found consistent positive outcomes from multiple data sources and they claim that their findings prove the efficacy of the cognitive strategy instruction programme with ELLs. The strategies in their Tool Kit, however, consisted of cognitive strategies as well as metacognitive strategies (i.e., planning and goal setting, monitoring and evaluation).
Sasaki (2000) investigated the effects of strategy training on process writing, such as planning and revision, on eight novice writers. Sasaki found that the number of strategies used by novices decreased by almost half after six months of instruction. According to Sasaki, this decrease in the number of strategies used could be interpreted in a number of ways: application of prescribed ways they have learnt to approach the given task, improvement in their English proficiency, which enabled them to use less translation, or the students making less effort in thinking about writing while composing. Sasaki (2000) found that some students started using ‘skilled’ writers’ strategies such as ‘rereading’ and ‘global planning’ after instruction (p. 282).
Using Sasaki (2000) as a pilot study, Sasaki (2002) investigated Japanese learners’ processes of expository writing in English cross-sectionally and longitudinally using written texts, video-taped writing behaviours and stimulated recall protocols. The novices (N = 22) received process writing instruction for two semesters, which included writing strategy training such as planning and revising. The findings showed that experts (N = 12) used more strategies than the novices but the novices appeared to have used more strategies after the instruction than before.
The formulation stage of writing, which has gained relatively less attention in research, was the focus of Graham and Macaro (2007) and they believe that planning, formulating and monitoring are interrelated. The writers’ planning or not planning before writing and language used for planning, according to them, affect formulation. Time taken for strategy training, perceptions of teachers and students regarding strategy training, lack of familiarity some teachers had in interpreting strategy training instructions and material were the most visible difficulties faced by these researchers.
3 Analysis of the effects of strategy instruction
In terms of the effectiveness of strategy instruction, Hassan et al. (2005) carried out an in-depth review of 25 studies based on the clarity of intervention and the presence of a comparison group. Hassan et al. (2005) concluded that strategy training can be effective. According to them, raising awareness of strategy training interventions and training in implementing the strategies themselves can be beneficial to learners. However, long-term effects of strategy training were not seen.
Graham and Perin (2007) provide a meta-analysis of writing instruction studies, mostly in L1 writing contexts and using adolescents (Grades 4–12). They found that most of the strategy instruction studies yielded positive results. Strategy instruction in planning, revising and editing strategies when writing compositions had a mean weighted effect size of 0.82. The effectiveness of L2 strategy instruction in all four skills was the focus of another meta-analysis by Plonsky (2011). After calculating effect sizes for 61 primary studies, Plonsky found a small to medium overall effect of Strategy Instruction (d = 0.49). According to Plonsky (2011), in strategy instruction studies, it is important to select strategies to suit the level of the students and based on the assessment of strategy use prior to intervention (Chamot & Rubin, 1994; Grenfell & Harris, 1999; Harris, 2003). Plonsky also shows concern about the scarcity of studies that have controlled for proficiency. The need to conduct strategy instruction research in writing and in L2 contexts has also been stressed by Plonsky (2011) and hence makes the case for the present study.
The present article focuses mainly on two research questions.
Research Question 1a: To what extent can the learners be trained to use writing strategies?
1b: Does the impact of strategy instruction on writing strategy use differ by attainment level?
Research Question 2a: To what extent does EAP learners’ writing improve as a result of strategy instruction?
2b: Does the impact of strategy instruction on writing performance differ by attainment level?
III Design of the present study
This study had a pre-test post-test experimental design. It was an intervention study that provided writing strategy instruction to the experimental group over a period of six months. The sample consisted of 72 students registered for the Bachelor of Science degree programme at the Open University of Sri Lanka, following the English for Science Course Part 2.In this course, students are usually placed in groups according to their performance at an examination, which tests Reading and Writing Skills. The total population (446) was grouped into two groups, a high attainment group (those who scored 50 and above) and a low attainment group (those who scored 49 and below). These two sets were numbered serially and two groups of 36 students each were selected randomly from these two sets using a computer-generated list of random numbers. Each group was then re-grouped randomly into two groups of 18 students using the same computer program. These four new groups with 18 students each were randomly assigned to a control group and an experimental group to have 36 students each, with18 high attainment students and 18 low attainment students in each group. The scores of the high attainment group varied from 55 to 70, while the scores of the low attainment group were between 25 and 35. The students were told about the research process and the experimental group was told that they had the option of joining the other group but nobody wanted to do so. The experimental group and the control group gave their informed consent in writing to be research participants.
1 Instruments
The main instruments used in this study were pre and post strategy instruction questionnaires and a writing task battery. These instruments were designed by the researcher and were pilot tested with a similar sample. The instruments were used to compare the writing strategy use/writing performance of the experimental group prior to and after the strategy instruction and that of the control group at the beginning and at the end of the course. Stimulated recall was also used with a sub-sample of students to understand more about their strategy use while writing (see De Silva (2010) for more details).
a Questionnaire
The Pre-/Post-Strategy Instruction Questionnaire consisted of two parts. Part A comprised 21 items that provided the researcher with the respondents’ background information and their self-evaluation of the target language knowledge and their knowledge of L1 and their attitudes towards writing in English.
Part B of the questionnaire (see Appendix 1) had 32 statements that tried to elicit respondents’ L2 writing behaviour. The respondents were asked the degree to which they agree or disagree with each statement and were asked to tick the relevant box for each item. The points on the scale were subsequently assigned numbers from 1 to 6 (e.g., Strongly Agree = 6, Strongly Disagree = 1).
b Writing task battery
The writing task battery consisted of two writing tasks (description of a graph and an essay) and it was administered to both experimental and control groups at the beginning and at the end of the course. They were given a time limit (one hour) and it was administered as normal classroom tasks and not under strict examination conditions.
All writing samples were scored by two ESL specialists with over 10 years experience in teaching and marking ESL and EAP writing tests in the same context, using a detailed marking criteria. The written texts produced at pre-test and post-test were mingled after labelling them with an identification code known only to the researcher before they were handed to the two raters for assessment. This was done in order to conceal from the raters which texts were written first and which group they belonged to (Sasaki, 2002). All the scripts were marked at the same occasion and each script was marked by both raters. A conference marking session was held before the actual marking started and three pilot scripts from each rater were checked by another English language teaching (ELT) expert for standardization of marking. The inter-rater reliability was 0.92.
2 The strategy instruction programme
The cycle of strategy instruction was developed by the researcher using models for strategy training (Grenfell & Harris, 1999; Macaro, 2001; O’Malley & Chamot, 1990) and models for L2 writing (Badger & White, 2000; Hyland, 2003). Since the product, process and genre approaches to teaching writing have their own advantages and disadvantages, an eclectic approach that incorporates all three approaches was thought to be useful in the context of the present study. The main features of the genre approach (i.e., modelling, joint construction and independent construction) and those of the process approach (i.e., planning, drafting, input from the teacher, peers and texts) and some features of the product approach (i.e., using model texts) were incorporated (Hyland, 2007). This eclectic approach to teaching writing, in combination with strategy instruction models, was used in developing a model for writing strategy instruction for EAP students.
The experimental group was provided with a comprehensive writing strategy instruction programme designed by the researcher, and the control group had the same number of teaching hours and followed the same course content except the strategy instruction.
The control group was taught by a teacher who had qualifications and experience similar to the researcher. The strategy instruction was conducted by the researcher and was spread out over a 24-week period. The instruction was provided during class time as a series of (20) two hour workshops and in the form of homework assignments. The researcher provided explicit instruction of selected strategies during the first eight workshops and, thereafter, the instruction was embedded into their daily writing activities. The control group teacher too devoted the same amount of time for teaching writing and the researcher observed some of her lessons and had informal discussions with her to understand her approach to teaching writing and in order to find out whether she trained her students to use any writing strategies. It was found that the control group students received no proper training in the use of writing strategies using a strategy training cycle. The control group teacher mainly focused on the structure of the model essay in the course book and taught how an essay should have an introduction, body and a conclusion. She did some brief brainstorming with the students but never gave them training in using the strategy.
a The cycle of writing strategy instruction
The strategy instruction cycle consisted of 11 steps (see Figure 1, reproduced from De Silva, 2010). This instruction cycle was designed using models for strategy instruction and L2 writing (Badger & White, 2000; Bereiter & Scardamalia, 1987; Feez, 1998, as cited in Hyland, 2003; Grabe & Kaplan, 1996; Grenfell & Harris, 1999; Macaro, 2001; O’Malley & Chamot, 1990).
The cycle of strategy instruction.
The cycle of writing strategy instruction starts with Goal-Setting. The importance of goal-setting in language learning has been discussed by various researchers (Anderson, 1985; Cumming, 2006; Tudor, 1996). In the present study, goals were first set at a broader level (to be achieved at the end of the course) and then for each individual task they attempted. For example, in the Lab Report task the goal was to write a clear and concise lab report in past passive voice with appropriate sequence markers and using the format specified by the relevant faculty to obtain the highest possible marks from the subject lecturer.
The second step in the cycle is Task Analysis. Wenden (2002) discusses the importance of Task Analysis in planning. According to Wenden, learners determine the purpose of the task, ‘the nature of the problems the task poses’ and the task demands through task analysis (p. 47). In the present study, the students were trained to analyse the tasks (i.e., to pay attention to the key words in the task rubrics and decide on the aspects that need attention). Task Analysis helps the students in deciding which genre to use and who the audience will be. At this stage, the strategies the students already use when attempting similar tasks are discovered and modelling of new strategies is done. The next stage is Joint Construction of the text where the teacher guides the students in applying the strategies during different stages of the text and explores the possibility of using a combination of strategies and orchestrating them to fulfil the task successfully (Anderson, 2003; Graham, 1997; Macaro, 2001).
Then, the students’ use of strategies is reinforced with a similar task. The students are provided with scaffolding in the form of learner support sheets and resource sheets, which include vocabulary and sentence structures specific to a particular genre, genre templates (see Hyland, 2007, for a description) and guidance for self-monitoring and self-editing. In addition, the students are constantly reminded to use appropriate strategies during this stage. They are advised to self-evaluate their strategy use and peer evaluation is also encouraged. At this stage, extensive feedback is provided by the teacher to the whole class (on the common problems/strengths observed by the teacher) and to individual students as well (Graham & Macaro, 2007).
The next stage in the cycle is the independent construction of the text by the students. The scaffolding is gradually withdrawn before this stage (Macaro, 2001). If the task is successfully completed, the students are introduced to a new task (new genre) and taken through the cycle. If not, the students are taken through the relevant stages again.
The experimental group was taken through this strategy instruction cycle. Strategies selected for initial training included planning strategies (i.e., brainstorming, clustering, mind-mapping, outlining), formulating (i.e., approximating, translating), monitoring (i.e., problem identification, auditory monitoring, visual monitoring), evaluation and revision strategies. These strategies were selected through literature review and from the responses to pre-strategy training questionnaire and the stimulated recall protocols. Training was also given in selecting and using strategies appropriate to a task, monitoring strategy use and orchestrating various strategies (Anderson, 2003; see Appendix 2 for a sample lesson plan).
In the present study, tasks were selected to represent each type of text the students were expected to learn during their language course. These were then sequenced from less demanding tasks to more demanding tasks using Vähäpässi’s (1982, cited in Weigle, 2002) classification. The order in which the tasks were used is given below:
description of an object;
description of a process;
lab report;
description of a graph;
definition/classification;
essay – personal opinion;
essay – argumentative.
The strategies were introduced gradually using the above-mentioned tasks and after taking them through the cycle, students were advised to practise strategy use with similar tasks that were given as homework assignments. They were also advised to choose tasks on their own and practise the use of strategies. Strategy instruction was provided mainly in English but L1 was also used when necessary.
3 Data analysis procedures
In order to decide on the type of statistical analyses to be used for these comparisons, tests of normality (Shapiro–Wilk and Kolmogorov–Smirnov tests) were conducted for strategy groups and on the writing scores with experimental condition/attainment group as the independent variable. The results showed a normal distribution of data (p > .05) for some strategy groups but non-normal distribution (p < .05) for others at the pre-test and the post-test. Since some strategy groups showed non-normal distribution, the researcher decided to use non-parametric analyses (Mann–Whitney test) on all groups at pre-test and post-test.
While the total writing scores at the pre-test and the post-test showed normal distribution (p > .05) with the Kolmogorov–Smirnov and Shapiro–Wilk tests, most of the sub-scores for the experimental and control groups and for high and low attainment groups showed non-normal distribution. Hence, a non-parametric test, the Mann–Whitney U-test, was used for the analysis of writing task scores.
IV Results
1 Comparison of pre-post total strategy use
Descriptive statistics for the total strategy use at pre-test and the post-test were calculated for the experimental and control groups. The effect size (Cohen’s d) for both within and between groups was also calculated.
As shown in Table 1, the experimental group showed greater increase in the mean total strategy use than did the control group between the pre-test and the post-test.
Total strategy use in the experimental and the control groups.
A Mann–Whitney U-test on the pre-test data showed no significant difference between the experimental group and the control group in total strategy use (Z = .505, p = .61, Cohen’s d =.13). A Mann–Whitney U-test on post-test data showed significantly higher total strategy use (Z = 3.72, p < .01, Cohen’s d = 1.07) for the experimental group than that for the control group. The effect size Cohen’s d for the Experimental and Control groups at pre-test was .13, while it was 1.07 at post-test.
a Comparison of pre-post total strategy use according to attainment level
The total strategy use for the Experimental High Attainment group was very much higher than that for the Control High Attainment group at the post-test. The Experimental Low Attainment group showed a greater mean increase between the pre-test and the post-test than did the Control Low Attainment group (see Table 2).
Total strategy use according to attainment level.
A Mann–Whitney U-test on the pre-test data for total strategy use showed no significant difference between the experimental and control high attainment groups (Z = .34, p = .73). The results were the same for the two low attainment groups at pre-test (Z = .47, p = .64) A Mann–Whitney test on the post-test data showed significantly higher total strategy use for the Experimental High Attainment group than that for the Control High Attainment group (Z = 2.66, p = .01). Cohen’s d was 1.16, which shows a high effect. A Mann–Whitney U-test on post-test data showed significantly higher total strategy use (Z = 2.67, p = .01) for the Experimental Low Attainment group than that for the Control Low Attainment group. The effect size Cohen’s d was .206 between the two low attainment groups at pre-test while Cohen’s d was high (1.01) at post-test.
b Comparison of the use of strategy categories
As shown in Table 3, the experimental group showed a higher increase in the mean value between the pre-test and the post-test than did the control group for total strategy use in all strategy categories, that is, the Planning, Formulating, Resourcing, Monitoring, Evaluation and Revision.
Use of strategy categories by the Experimental Group, the Control Group, Experimental High and Low and Control High and Low Groups.
A Mann–Whitney U-test on the pre-test data showed that there was no significant difference (p > .05) between the experimental group and the control group in the total strategy use for Planning (Z = .79, p = .43, Cohen’s d = .17), Formulating (Z = .59, p= .55, Cohen’s d = .16), Resourcing (Z = 1.01, p = .31, Cohen’s d = .22), Monitoring (Z = 1.47, p = .14, Cohen’s d = .31), Evaluation (Z = .37, p = .71, Cohen’s d = .07) and Revision (Z = .36, p = .72, Cohen’s d = .12). Cohen’s d values show a very small effect size at pre-test.
A Mann–Whitney test on the post-test data showed a statistically significant difference between the experimental group and the control group for most of the strategy categories. Cohen’s d was also calculated. Total Planning strategy use (Z = 2.10, p = 0.03, Cohen’s d= .49), Formulating strategy use (Z = 3.08, p < .01, Cohen’s d = .81), Resourcing strategy use (Z = 2.28, p = .02, Cohen’s d =.61), Evaluation strategy use (Z = 2.42, p = 0.01, Cohen’s d = .69) and Revision strategy use (Z = 2.56, p = 0.01, Cohen’s d = .64) were statistically significant, while the Monitoring strategy used approached significance (Z = 1.89, p = 0.058, Cohen’s d. = .51). The effect size showed a medium to high effect at post-test.
Mann–Whitney tests were also done on pre-test and post- test data for strategy categories for two attainment levels in the two groups. The results for the Experimental and Control High Attainment students and the effect size Cohen’s d are as follows: Planning (Z = .66, p = .51, Cohen’s d = .17), Formulating (Z = 1.27, p = .21, Cohen’s d = .47), Resourcing (Z = .17, p = .86, Cohen’s d= .02), Monitoring (Z = .48, p = .65, Cohen’s d= .06), Evaluation (Z = .68, p = .49, Cohen’s d = .19) and Revision (Z = .49, p = .62, Cohen’s d = .17). At post-test, Mann–Whitney and effect size calculations on the Experimental and Control High Attainment groups showed the following: Planning (Z = 1.15, p = .13, Cohen’s d = .45), Formulating (Z = 2.47, p = .01, Cohen’s d = .94), Resourcing (Z = 1.75, p = .08, Cohen’s d = .63), Monitoring (Z = .95, p = .34, Cohen’s d = .43), Evaluation (Z = 1.19, p = .23, Cohen’s d = .62) and Revision (Z = 2.53, p = .01, Cohen’s d = .97). The results showed a high effect size for Revision and medium effect size for the others at the post-test.
The same tests on the Experimental and Control Low Attainment groups showed the following results: at the pre-test, Planning (Z = .45, p = .64, Cohen’s d = .24), Formulating (Z = .61, p = .53, Cohen’s d = .80), Resourcing (Z = 1.33, p = .18, Cohen’s d = .42), Monitoring (Z = 1.70, p = .08, Cohen’s d = .56), Evaluation (Z = .13, p = .89, Cohen’s d = .04) and Revision (Z = 1.45, p = .88, Cohen’s d = .05).
At the post-test, the Experimental and Control Low Attainment showed the following results: Planning (Z = 1.54, p = .12, Cohen’s d = .55), Formulating (Z = 1.72, p = .08, Cohen’s d = .65), Resourcing (Z = 1.66, p = .09, Cohen’s d = .61), Monitoring (Z = 1.63, p = .10, Cohen’s d = .59), Evaluation (Z = 2.14, p = .03, Cohen’s d = .85) and Revision (Z = 1.01, p = .31, Cohen’s d = .36). A high effect size was found for Evaluation at post-test and a medium effect for other strategy categories.
The answers to the open-ended question at the end of the questionnaire showed that students used few other strategies, such as building a story on what to write, combing sentences when the question demands a limited number of sentences and exploring websites and posting on the web for solving problems.
2 Comparison of pre-post writing performance
a Comparison of pre-post scores of the experimental group and the control group
A Mann–Whitney test on total scores at the pre-test showed no significant difference (Z = .92, p = .35) between the experimental group and the control group. A Mann–Whitney test on post-test scores also showed no significant difference (Z = .39, p = .69) between the two groups. The effect size (Cohen’s d) for total scores between the experimental and control groups at pre-test was .32 while at post-test it was .10 (see Tables 4–6).
Descriptive statistics for the pre- and post-writing task battery scores.
Descriptive statistics for the scores of the Experimental and Control High Attainment groups at pre-test and post-test.
Descriptive statistics for the scores of the Experimental and Control Low Attainment groups at pre-test and post-test.
b Comparison of the total scores of the sample according to attainment level
The total scores of the experimental group and that of the control group at the pre-test and the post- test were compared according to attainment level in order to find answers to research question 2b.
A Mann–Whitney test was also conducted on total pre-test scores to compare the Experimental High Attainment group scores with Control High Attainment scores. The Control High Attainment group scored significantly higher than the Experimental High Attainment group at the pre-test (Z = 2.49, p = .01). Hence, a Mann–Whitney test was conducted on the gain scores (post–pre). The results showed a significantly higher increase in performance by the Experimental High Attainment group (Z = 2.56, p = .01) than the Control High Attainment group between the pre-test and the post-test. Cohen’s d for Experimental High Attainment and Control High Attainment groups at pre-test was .87 and it was .58 at post-test.
A Mann–Whitney test was conducted on pre-test data to compare the total scores of the Experimental Low Attainment group with the scores of the Control Low Attainment group. The results showed no significant difference (Z = .06, p = .95) between the two groups at pre-test. A Mann–Whitney test on the post-test scores showed that the difference between the Experimental Low Attainment group and the Control Low Attainment group approached significance (Z = 1.95, p = .06). The higher mean rank was for the Experimental Low Attainment group. Cohen’s d for Experimental and Control Low Attainment groups at pre-test was .17 and at post–test it was .83.
As shown above, the intervention led to greater improvement in the total scores for the experimental group than for the control group, irrespective of attainment level. However, the effect was higher for the Experimental Low Attainment group than that for the Experimental High Attainment group in most of the sub-scores and the total score.
V Discussion
The findings of the present study show that writing strategy instruction is beneficial to EAP students. The experimental group showed a statistically significant increase in Total Strategy Use after receiving strategy instruction, a finding in contrast to Sasaki (2000). The study found significant improvement in Planning, an important finding given the importance of Planning within the models of writing (Grabe & Kaplan, 1996; Hayes, 1996). However, unlike Sasaki’s (2002) experts, who were highly qualified writers, the informants did not show much evidence in detailed planning. Formulation, which has gained importance in writing models (Kellogg, 1996; Macaro, 2006 ), also showed a significant increase at the post-test for the experimental group (Cumming, 1989; Roca de Larios, Marín, & Murphy, 2001) and Resourcing too showed similar results. Some Experimental group students became less dependent on the teacher towards the end of the strategy instruction programme and they used other resources to get their problems solved. Macaro (2001) found similar results in an intervention study, which showed that the experimental group used resourcing strategies (i.e., dictionary use) more selectively than they did at pre-test.
The increase in the use of Self-monitoring (Kellogg, 1996; Macaro, 2006) by the Experimental group in comparison to the Control group indicates the effectiveness of the present intervention. However, the Experimental High Attainment students benefitted more from self-monitoring than did the Low Attainment group, which could be attributed to their insufficient linguistic knowledge. Xiang (2004) too reported similar results. Evaluation and Revision strategy use also showed a higher increase for the Experimental group than that for the Control group at post-test (Sengupta, 2000).
The findings show that strategy instruction has a positive impact on writing performance of the sample. Even though the effect size was small for the total scores at post-test (.10), the written texts of the Experimental group at the post-test showed some improvement in overall organization and there was evidence of their attempts to improve the quality of their writing. Within subject Cohen’s d showed a medium effect (.69) on the total score of the Experimental group, while it was small (.26) for the Control group. Organization and Cohesion sub-scores had a large effect (1.4 and .89, respectively) for the Experimental group. When analysed for attainment groups, Cohesion, Organization and Grammar sub-scores and the total score were found to be significantly higher for the High Attainment students, while Cohesion and Organization were significantly higher for the Low Attainment students than that for the Control group students. Large effect sizes for Cohesion (1.16) and Organization (1.29) were reported. The researcher may have paid more attention to Cohesion and Organization in the Strategy Instruction programme, as students are usually weak in these aspects. The above results show that strategy instruction has a positive effect on learners’ writing performance and thus counters the claim that spending class time for strategy instruction may be of little use (Rees-Miller, 1993, as cited in Grenfell & Macaro, 2007).
The present study shows that the students can be trained not only to use writing strategies but also to use them effectively, which accords with the results of previous L1 and L2 studies (Bereiter & Scardamalia, 1987; Graham & Macaro, 2007; Macaro, 2001; Sasaki, 2002). The success of the strategy instruction programme may be a result of the effectiveness of the strategy instruction cycle designed specifically to suit the needs of the learners, the inclusion of both cognitive and metacognitive strategies and the task-specific nature of the instruction. The feedback provided on students’ strategy use and the freedom given to students to evaluate their strategy use and select ones that lead to their desired outcomes may also have contributed to the success of the intervention (Graham & Macaro, 2008). Even though the results are not reported here, the study found that strategy instruction led to an increase in self-motivation, determination and positive attitudes towards writing in English (see De Silva, 2010), which may lead to increased performance later.
The present study will have an important contribution to the field of LLS as relatively few studies have been conducted on the impact of strategy instruction on writing. Furthermore, longitudinal intervention studies that used EAP students as the sample and with a strategy instruction cycle specifically designed for EAP students studying in their L2 in an L1-dominated environment are rare. Moreover, most EAP studies have been done with heterogeneous populations (i.e., international students studying in English-speaking countries), while the present study used a relatively homogeneous sample.
The study has a number of implications for practitioners. The positive results of the study encourage trying out strategy intervention in EAP classrooms. However, amendments to the cycle will be necessary if it is to be adopted to train younger learners, such as secondary school learners who study in the medium of English or ESL students. If writing strategy instruction is to be tried out with younger learners, more elaborate and more frequent feedback and greater scaffolding may be necessary. Strategy instruction via online will also bring new insights to the field.
Footnotes
Appendices
Acknowledgements
I would like to thank Professor Suzanne Graham and Professor Brian J Richards of the University of Reading, UK, for their continuous support throughout the study. My thanks are due to the Open University of Sri Lanka for supporting me in conducting this research. I would also like to thank the students who participated in this study and the two anonymous reviewers who gave useful comments on earlier drafts of this article.
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.