Abstract
Previous research in ESL writing has shown that lexical features accounted for a large amount of variance in writing scores and many of the instances in the think-aloud protocols were looking for the right word. The importance of vocabulary in ESL wring has been documented. This study examined the extent to which the role of vocabulary (i.e., EFL writers' vocabulary ability and the vocabulary use in their writing) is crucial for Taiwanese students (366 senior high school students and 151 university English majors) to write a 30-minute-timed argumentative essay. Participants also took two Vocabulary Levels Tests (Nation, 1990; Laufer & Nation, 1995) that measure learners' receptive and productive abilities at different frequency levels. Lexical variables obtained from their essays - levels of word families, lexical richness and essay length - gauged writers' lexical performance in writing. Results of the Vocabulary Levels Tests suggest that mastery of the most frequent 2000-word productive level was likely to be the threshold level. Results of lexical variables found that essay length contributed most to the variance of writing quality. Qualitative analysis of essay length provides additional information about strategies for writers with limited vocabulary to write efficiently. The study concludes by discussing the pedagogical implications of EFL writing instruction and learning in light of the role of vocabulary.
Introduction
Writers must achieve certain language competence before they can control the language used in writing. Kirby and Kantor (1983) described linguistic competence as being composed of three measurable elements: vocabulary content (words as indicators of knowledge), vocabulary fluency (the ability to call forth words for appropriate expression), and syntactic fluency (syntactic processing ability). A number of studies have addressed the issue of lexical and syntactic features in L2 writing. Studies that compare the compositions of L2 writers with those of L1 writers have shown that the great disparity of linguistic features in native and non-native speakers' texts is largely associated with issues of familiarity with and access to lexical features (Silva, 1997, for an overview ). Vocabulary proficiency appears to be pertinent to L2 writing. L2 writers' ideas need to be translated into words just as L1 writers, but L2 writers' vocabulary size or lexical knowledge may not be as sufficient as that of L1 writers'. The importance of lexical resources has been documented in L2 writing process research (see Krapels, 1990, for a review). Many of the concerns or processes in the think-aloud protocols of L2 writers were related to vocabulary; that is, they were looking for the right word (e.g., Cumming, 1990).
Literature Review
The development of high-quality writing in L2 requires an ever-increasing vocabulary that is capable of expressing writers' thoughts (Angelova, 1999). From writers' standpoint, the importance of possessing a sufficient vocabulary for L2 writing has been substantiated. Leki and Carson (1994) surveyed 128 undergraduates of English as a second language (ESL) in the writing courses of English for Academic Purposes (EAP) and found that vocabulary was cited as the most important factor for writing within the category of language skills. The finding suggests that L2 writers may be primarily constrained by their ability to choose from a repertoire of lexical options.
From readers' or raters' standpoint, vocabulary proficiency is perhaps the best indicator of overall composition quality. Astika (1993) assessed foreign students' writing by using the ESL Composition Profile (Jacobs, Zingraf, Wormuth, Hartfiel, & Hughey, 1981). Of the five components (content, organization, vocabulary, language use and mechanics), variations in the ability to employ vocabulary were found to account for the largest amount (84%) of variance in the total writing scores. In a study of Koda (1993) who investigated descriptive and narrative texts of Japanese as a foreign language, there was a strong correlation between vocabulary knowledge and text quality. Her findings also show that vocabulary knowledge contributes substantially to the writing quality, while sentence complexity has relatively little independent influence on the writing quality (Koda, 1993, p. 337). Engber (1995) examined the extent to which lexical richness and lexical errors were related to the quality of written compositions by ESL learners of mixed language backgrounds. Her findings support the basic premise that vocabulary is an integral component of both constructing and interpreting of meaningful written texts. Laufer and Nation (1995) also showed that vocabulary size (use of words of different frequency levels) and composition rating were highly intercorrelated
It is not enough to have vocabulary knowledge available while writing. Writers need to be able to use words fluently as a part of vocabulary knowledge. Being able to access words quickly means that more processing time is available for concentrating on what to say rather than how to say it (Coxhead & Nation, 2001). Word retrieval, the number of words accessed in a given span of time, can be regarded as fluency in writing (Reid, 1990; Wolfe-Quintero, Inagaki, & Kim, 1998). A series of studies by Kaufer, Hayes, and Flower (1986) provides insights into the process involved in text generation by English speakers. They found that the average length of "bursts" (the sentence part identified either by a pause or a revision of earlier production) increases from 7.3 words for competent writers to 11.2 words for advanced writers. Using similar methodology, Friedlander (1989) found that Chinese ESL students produced very short bursts of English writing, approximately two words in length. These protocol data provide clear evidence of a consistent finding that L2 writers' transcribing is more laborious, less fluent, and less productive than that of L1 writers (see Silva, 1993, for a meta-analysis).
Essay length (the number of words in a text) is also an important text feature relevant to writing. Connor (1990) investigated 150 persuasive essays written by English-speaking high-school students from England, the USA, and New Zealand. Among eleven linguistic/rhetorical features, essay length was the second best predictor of essay quality. Similar results were also found in L2 writing. Ferris (1994), using a multidimensional approach to investigating a corpus of 160 timed compositions written by ESL students of mixed ethnic backgrounds, found that of the 28 text variables, essay length (number of words) contributed most significantly (R2 = .38) to holistic writing scores.
In summary, the availability or capability of vocabulary appears to be crucial for L2 learners to express or communicate in writing. Investigating writers' ability in vocabulary (linguistic competence) and lexical uses in their writing products (linguistic performance) will cast insights into the relation between vocabulary abilities and writing proficiency.
Methodology
Participants
….Participants were invited and recruited through their class teachers of English or English composition.
-----The samples of Taiwanese students were drawn from the following four groups of academic levels:
Group 1 (G1)
Level of first-year senior high school students (Grade 10)
Group 2 (G2)
Level of senior-high school students (Grade 12)
Group 3 (G3)
Level of second-year university English majors
Group 4 (G4)
Level of third-year university English majors
A total of 517 students - 191 in G1, 175 in G2, 104 in G3 and 47 in G4 - made up the participants in this study.
The senior high school students came from several prestigious schools known to have a relatively high level of academic achievement as measured by the government-administered Joint Senior High School Entrance Exam. As a result, it was very likely that the students were motivated to learn English for the future Joint University Entrance Exam (JUEE). Similarly, the university English majors were likely to be interested in English to have attained a fairly high score in the JUEE English test in order to be accepted into university English departments.
English educational backgrounds of the four groups are explained below. Students in Group 1 had four years of English in high school (three years of junior high school and one year in senior high). They made their first attempt to write English compositions when they did the English writing task for this study; thus they represented inexperienced EFL writers.
Students in Group 2 represented those who had six years of English in high school, including instruction in paragraph writing for the English composition test of the JUEE. Because senior high school graduates do not necessarily plan to enter an university English department, this study intentionally included the third-year high school students (one month before JUEE) and university entrants (one month after JUEE) in Group 2 because they represent a similar academic level for the purpose of the developmental analysis. The university entrants to English departments in Group 2 were considered to be of the optimal level that senior high school graduates could achieve in terms of English ability.
Students of Group 3 and Group 4 had completed either one or two years of university education, respectively, in addition to their six years of high school education. In terms of English writing courses (usually two hours a week), Group 3 had received one year of training and Group 4 had received two years. In brief, participants' experiences with L2 composition instruction, a potential explanatory factor of L2 writing quality (Becker, 1995; Hirose & Sasaki, 1994), were systematically varied by the selection of these four educational levels.
Of 32 teachers contacted, 20 teachers were willing to have their students do the tasks of this study in class. Data collected from the following students were excluded from the study: 12 non-English majors in the writing classes, 8 students repeating the course, and 3 overseas students. This procedure ensured that members of each group were of the same educational level, enhancing the reliability of assessing student performance as a group. The investigation was related to English writing, so 81 students who did not take the English writing task were also excluded. Thus, a total of 517 students - 191 in Group 1, 175 in Group 2, 104 in Group 3 and 47 in Group 4 - made up the participants in this study.
The numbers of participants in the four groups varied due to uncontrollable factor such as different class sizes, and due to unexpected occurrences such as two teachers changing class levels and one teacher failing to invite his students to participate. Nevertheless, all participants were drawn from the targeted educational level in terms of prior schooling and formal instruction received in English writing. This was important because, as Mohan and Lo (1985) and Carson (1992) have suggested, L2 writers' previous schooling is an important factor in their ability to write in L2.
In summary, participants were recruited and selected on the basis of their educational backgrounds. Methodologically, the cross-sectional design is crucial for observing changes and explaining the development in student performance.
Writing tasks
Participants were asked first to write an English composition for 30 minutes. This study employed two topics from the argumentative genre that have been used to evaluate composition writing in previous studies - Topic A, adopted from the IEA study (Gorman, Purves, & Degenhart, 1988), and Topic B, from the study of Laufer and Nation (1995):
Some people say that watching television makes independent thinking difficult. What do you think of the statement? Give your opinion and support it.
Some people say that a person cannot be poor and happy, because money is always needed to gain something that is important to that person. What do you think of the statement? Give your opinion and support it.
The scoring guidelines for English essays were based on the 6-point scale of the Test of Written English (TWE) in the TOEFL exam (ETS, 1990). Two experienced TWE raters scored the English essays independently.
Vocabulary tests
The vocabulary tests for participants included one matching format (Nation, 1990) and one blank-filling format (Laufer & Nation, 1995) of the Vocabulary Levels Test. For the purpose of measuring distinctive levels of lexical proficiency - high frequency vocabulary, low frequency and vocabulary for special purposes - the Vocabulary Levels Tests chose four frequency levels and a level of academic vocabulary. The four frequency levels are (1) the most frequent 2000 words (viz., 2000 word-level), (2) the third 1000 words (viz., 3000-word level), (3) the fifth 1000 words (viz., 5000-word level), and (4) the tenth 1000 words (viz., 10,000 word-level). The 2000 and 3000 word-levels contain high-frequency words that provide the basis of about 87 % of the words likely to be encountered in general English texts. The 5000-word level includes the boundary of high- and low-frequency words. The 10,000-word level accounts for much less than 1% of the running words in a text (Laufer & Nation, 1999).
In addition, a level of academic vocabulary, the University Word List (UWL) (see Xue & Nation, 1984), was included. The UWL excludes the first 5000 words from Thorndike and Lorge's list (1944), the first 2000 words from the General Service List (West, 1953), and technical vocabulary (i.e., vocabulary that is recognizably specific to a particular topic, field, or discipline). By omitting these higher frequency general words as well as technical words, the UWL is intended to be a list of general academic words that occur across a wide range of academic disciplines. The UWL contains over 800 words, giving an 8.5% coverage of academic texts (Nation & Hwang, 1995).
The matching format of the Vocabulary Levels Test (Nation, 1990) measured receptive ability. The following example illustrates an individual set in the Word Levels Test:
business
clock
horse
pencil
shoe
wall
A. part of a house
B. animal with four legs
C. something used for writing
Participants needed to match three out of the six words on the left with the appropriate definitions on the right. Within each set, all six words were the same part of speech.
The blank-filling format of the Vocabulary Levels Test (Laufer & Nation, 1995) gives a minimum number of cue letters to elicit the target word. For example, the four letters orig cue the target word original in the context of "They will restore the house to its orig state." This test was employed to measure L2 learners' productive ability, or the ability to call forth a word appropriate to the context.
For both tests, 18 testing items were used for each vocabulary level. The answers of the items were scored as correct or incorrect. Minor spelling mistakes and errors of inflection were not marked as incorrect for the blank-filling tests. Each correct answer was given one point. Percentage scores were calculated from the raw scores of each vocabulary level. For example, if a participant got 9 correct items out of 18 at the 3000 word-level, he or she had acquired roughly about 50 % of that level.
Lexical analysis
Vocabulary use in English compositions was measured to gauge writers' ability in lexical production. Each English essay was assessed by the computer program "Lexical Frequency Profile," a reliable and valid measure of lexical features in writing products (Laufer & Nation, 1995). Three kinds of lexical variables were obtained. First, word families in a text were calculated at four frequency levels -- the first thousand most frequent words, the second thousand, the University Word List (UWL) (Xue & Nation, 1984), and the off-list words, namely, words not found in the lists of the former three levels. A "word family" is defined as a base form with its inflected and derived forms (Bauer & Nation, 1993). For example, develop, develops, developed, and developing all belong to the "word family" of the base form develop, and the total number of their occurrences in a text is counted as one "word family." The number of word families is more revealing than words or word types, because it is an indication of lexical richness "to quantify the degree to which a writer is using a varied and large vocabulary" (Laufer & Nation, 1995, p. 307).
The percentage of non-basic words a composition contains is the second lexical variable. Non-basic words refer to vocabulary in the UWL level and the off-list level. The relative proportion of non-basic words used in a text provides an index of lexical complexity.
The third kind of lexical variable was essay length (i.e., the number of words written in a text). Essay length serves as an indicator of the ease of written expression (Reid, 1990) and as an indicator of the ability to call forth the word needed (Kirby & Kantor, 1983). Thus, essay length reflects written fluency in a timed condition. In term of this variable, a lack of fluency means fewer words are accessed in a given time. Fluency is an index of efficiency of retrieving productive vocabulary items during the composing process.
Results and Discussion
Quantitative analysis of vocabulary competence
Percentage scores were calculated from the raw scores of the Vocabulary Levels Tests, which tested 18 items in each level. For example, if a participant got 9 items correct out of 18 in a level, he or she received a score of 50%. The percentage score for each level represents the proportion of words known at that level. In addition, analyses of variance were performed to compare percentage scores across academic groups at each word level. Results of ANOVA analyses and descriptive data for each word level are shown in Table 1 and Table 2 for the receptive and productive vocabulary respectively.
Overall, students increased their vocabulary at each word level as their level of English study increased. The measures of vocabulary competence, in both the receptive and the productive capabilities, show a reliable, consistent, and linear progression as the level of English learning increases, demonstrating their validity as measures of language development.
Within each academic group, the proportion of words known generally declined as the word levels increased. A comparison between the two kinds of vocabulary measures shows that students' productive ability generally lagged behind their receptive ability. This finding concurs with previous studies showing that vocabulary size, as measured by the receptive test, is larger than that measured by the productive test (Laufer, 1998; Laufer & Paribakht, 1998).
Table 1 Means of percentage scores of receptive vocabulary
R2000
R3000
RUWL
R5000
R10000
Group (N)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
G1 (191)
66.4 (17.3)
32.5 (14.4)
29.8 (16.5)
19.5 (11.8)
11.5 (10.7)
G2 (168)
90.9 (11.4)
61.6 (18.4)
63.4 (19.1)
49.9 (20.5)
16.3 (14.1)
G3 (99)
94.8 ( 8.3)
72.2 (18.7)
71.3 (17.5)
55.0 (20.4)
21.0 (14.1)
G4 (45)
99.9 ( 0.8)
93.2 ( 9.2)
91.9 ( 8.4)
83.8 (11.9)
40.9 (18.3)
F (3, 499) =
174.76**
248.26**
251.0**
234.45**
61.64**
RUWL: receptive university word level; SD: standard deviation; **p < .01
Table 2 Means of percentage scores of productive vocabulary
P2000
P3000
PUWL
P5000
P10000
Group (N)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
G1 (191)
34.7 (13.9)
9.8 (8.3)
1.2 (3.2)
0 (0.7)
0 (0.0)
G2 (168)
65.3 (21.2)
33.7 (18.1)
9.1 (11.4)
8.5 (9.5)
0.5 (1.6)
G3 (101)
75.0 (17.2)
44.7 (21.7)
19.0 (16.9)
11.6 (12.3)
0.2 (1.3)
G4 (46)
91.1 (7.5)
78.4 (16.2)
52.4 (20.5)
36.6 (21.8)
1.9 (5.1)
F (3, 502) =
224.26**
276.39**
243.14**
164.49**
13.41**
PUWL: productive university word level; SD: standard deviation; **p < .01
A main goal of the Vocabulary Levels Tests was to find the threshold level at which an L2 learner is able to write an acceptable English composition in 30 minutes. In order to identify a vocabulary level as the most effective predictor for estimating students' English writing scores, a structural equation model was employed by the statistical program AMOS (Analysis of Moment Structures). AMOS statistical procedures involve constructing a model identifying interrelationships among the predictor variables, the identification of variables that have significant effects, and a measure of the extent that the resulting model fits the available data.
In this model, the R10000 and P10000 levels were not included in the collective set of the predictors, because a negligible number (see Table 1 and Table 2) shows a few percent of the participants had reached the developed stage of the 10000 word-level. Figure 1 shows the predictor variables (vocabulary levels), the correlations (the bi-directional arrows), the dependent variable (English writing score, EWS), the beta coefficients (the values above the paths) and the R2 value for the English writing score (appearing at the right front corner of the EWS rectangle).
The R2 value of 0.46 indicates that the model explains 46% of the variance of the English writing score. Beta coefficients indicate that only the productive 2000 (P2000) and productive university word list (PUWL) levels were statistically significant, because their absolute t-values were more than 2.0. Surprisingly, receptive 3000 level (R3000) has a negative relationship (negative beta value) with the English writing score. The negative beta value obtained may be result of being suppressed by the other highly interrelated vocabulary levels.
Each vocabulary level was further examined for its degree of association with English writing quality. Table 3 shows the correlation coefficients of each vocabulary level as the independent variable. Results indicate that the magnitude of the P3000 level has the largest association (r = .635). It is possible that the statistically insignificant contribution of P3000 to variance in English writing shown in Figure 1 could be a result of suppression of the P3000 variable by the highly interrelated P2000 (r = .84) and PUWL (r = .84) variables.
An analysis of the structural equation model (Figure 1) and correlation coefficients (Table 3) shows that the P2000, PUWL and P3000 levels are the most effective vocabulary-related measures for estimating English writing proficiency in this sample of writers. Using these word levels individually can predict about 38% to 40% (R2 as correlation coefficients squared) of the variance of English writing scores. The better a participant was at these productive word-levels, the better his or her English writing proficiency was likely to be. Therefore, an examination of the relationship between vocabulary levels and English writing scores suggests that mastery of the productive 2000 word-level was the minimum level (the threshold level) needed for the EFL students in this study to produce an English composition under the 30-minute time constraint. It should be noted that this conclusion is based on an analysis in which there were relatively few students studying at the advanced level (G4).
Notes: E1 is the residual or error variance that also causes variances in the English writing score (EWS).
*The absolute t-value is greater than 2
Figure 1 Correlations among vocabulary levels and paths to English writing scores
Table 3 Correlations of English writing scores as the dependent variable
R2000
R3000
RUWL
R5000
P2000
P3000
PUWL
P5000
.501**
.560**
.582**
.597**
.613**
.635**
.613**
.587**
** p < .01
Lexical use in English writing
Three kinds of lexical use in the English compositions were assessed using the Lexical Frequency of Profile (Laufer & Nation, 1995). First, word families (base forms with their inflected and derived forms) in a text were calculated at four frequency levels: the first 1000 most-frequent words, the second 1000 most-frequent words, the university word list (UWL), and the words not in the lists of the former three levels (off-list). Secondly, lexical complexity was measured as the frequency of use of non-basic word families (i.e., the more advanced frequency levels of the university words and the off-list words). Essay length (total words in a text) was the third kind of lexical variable.
Analyses of variance were performed to compare lexical use in a text across academic groups. Table 4 show the results of the ability to use word families at each level, as well as results of lexical richness and essay length at each academic level. Results of ANOVA analyses indicate that there were significant differences across academic groups for each lexical variable. Overall, the data show that the lexical productive ability increased as the level of writers' experiences with English study increased. This finding suggests that as students progress to a higher level of EFL learning, they are likely to produce more words in terms of word families, complex words (non-basic words), and the number of words in a text within the time frame for the English writing task.
Table 4 Means of lexical variables across groups
Word families
Lexical complexity
Length
1st 1000
2nd 1000
UWL
Off-list
Non-basic word
Group (N)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
Mean % (SD)
Mean (SD)
G1 (191)
63.3 (16.8)
4.2 (2.7)
.85 (1.2)
1.6 (1.4)
3.3 (2.4)
150.1 (57.6)
G2 (175)
72.3 (15.9)
4.9 (3.2)
2.7 (2.3)
2.5 (2.1)
5.9 (3.2)
167.7 (47.6)
G3 (104)
77.8 (18.3)
6.1 (3.2)
3.4 (3.0)
3.4 (3.7)
7.0 (4.2)
191.4 (59.4)
G4 (47)
90.6 (18.9)
7.7 (4.0)
5.9 (3.7)
4.9 (3.5)
9.9 (5.7)
241.7 (41.1)
F(3, 513) =
39.33**
20.38**
72.39**
27.39**
58.10**
42.23**
UWL: university word level; Off-list: word families not in 1st 1000, 2nd 1000 or UWL
SD: standard deviation; **p < .01
Similar statistical procedures to those used for vocabulary were undertaken for finding the most effective lexical predictors of English writing quality: regression analyses (Table 5) and an AMOS equation (Figure 2).
Notes: E1, e2, and e3 represent residuals or error variances.
*The absolute t-value is greater than 2.
Figure 2 Correlations among word-family levels, paths of word-family levels to lexical variables, paths of lexical variables to English writing scores (EWS)
Table 5 shows the correlations between lexical use and English writing proficiency. All of the lexical variables significantly correlated with the quality of English writing, suggesting the role of lexical use as an important factor in holistic scoring. Results also show that essay length was the strongest predictor, explaining 53% (correlation coefficient squared) of the English writing variance. The second strongest predictor was the first 1000-word family. As the first 1000 word families also highly correlated with essay length (r = .89, p < .01, 2-tailed), the implication is that skill in using the 1000 word families also contributes greatly to essay length of the text.
Table 5 Correlations of English writing scores as the dependent variable
1st 1000
2nd 1000
UWL
Off-list
Lexical richness
Length
.689**
.432**
.476**
.363**
.331**
.731**
** p < .01
The bi-directional arrows in Figure 2 show that the word families are significantly correlated with each other (r ≥ .46. p < .01). The R2 values, as shown at the top right corners of the length rectangle, indicate that the levels of word families together accounted for 50% of the variance of essay length; the levels of UWL and the off-list word families explained 79% of variance in lexical richness; and the variables of essay length and lexical richness together predicted 54% of the variance in the English writing score.
Qualitative analysis of the variable of essay length
Qualitative analysis of the length variable, which correlated highly with writing quality, was used to look for strategies of fluency by examining long writing samples of good quality. In terms of fluency as an ease of expressing in words (Reid, 1990) and an ability to call forth the word needed (Kirby & Kantor, 1983), it was of interest to see how writers with limited vocabulary overcame this likely barrier and were able to produce a long essay with good quality.
Writing samples to be examined were the long (at the upper-third) and good (middle or high scoring) essays written by students whose productive 2000 word-level was not fully developed. There were 18 essays that fitted these criteria - the middle scoring essays written by the least advanced group (G1). None of the writers at the higher academic levels (G2, G3, or G4) in the middle or high scoring samples met the criterion whereby the productive 2000 word-level was undeveloped.
Analyses of the 18 selected samples showed several strategies that these writers used to produce a long text of good quality. In Sample A, it is clear that the writer explicated one idea in each short paragraph (264 words in 6 paragraphs). This strategy reflected a "knowledge-telling" (Scardamalia & Bereiter, 1987) approach that is capable of producing a long text without the complicated demand of organizational planning.
Sample A
I don't know why people have that kind of thinking. Remember! When we born to the world, we have nothing, even clothes. Beside, when we leave the world, we still have nothing. Although you have a lot of money, you still can't bring it with you.
In old times, many people think money is everything. Someone has money which can everything thing he want, do everything he want. But I think although money is very important in our life, it isn't everything "That's why" money you will ask. Lend me tell you:
First: Money can't buy true love, real friendship, family. Those all the most important, valuable things in life, But money can't buy.
Second: Money can't buy sunny, moon, star… many sight all over the world.
You know? That national thing can make us happy, satisfied, and surprised. When you enjoy it your heart must be peaceful and they want be war any more. This is also important. But money still can't buy it.
Third: Money can't buy times. Times is always pass, and can't go back. You just can complain why times is so fast. But in that time, time still pass…… there are still many thing that money can't buy, I think you have think of it by yourself.
Happy is very easy. Know what you want to get, what you want to do. Have a plan, then just to do. Although you don't have money, you still can dream. Go down what road you want, then you will happy. The most fun in our daily is to make your own choice. (264 words, score = 3.5)
Another strategy used in Sample A is the repetition of the opening "Money can't buy…" at the beginning of three body paragraphs. Such repetitious sentences at the beginning of body paragraphs served as a topic sentence. This strategy was also found in an essay (240 words, score = 4.0) which used "If you want…," and similarly employed in another essay (219 words, score = 4.0) which used the phrase "If you don't have money…".
The third strategy to produce longer texts is found in Sample A, where the writer relied on referring to the reader in the second person. There are 19 occurrences of "you" or "your" in the text. This strategy was also found in an essay (283 words, score = 4.0) which had 30 occurrences of "you" or "your". It is likely that the strategy of talking to the reader in the second person helped the writer to better simulate a verbal communicative activity.
The fourth strategy to produce longer and more fluent texts is found in essays like Sample B, where the writers used descriptive personal experience to support personal opinions. There were eleven occurrences of the use of the first person "I" in both of the essays.
Sample B
In my opinion, I don't think that watching television makes independent thinking difficult. The following is my feeling about this statement and I would like to show it with you.
As some TV programs are good, so some are bad. If people only like to watch the bad ones which can't make you better, watching TV is just to waste your time. But there are more good programs than bad ones. Why not choose those which makes you better and better?
I am a student. I must go to school every day. Studying and learning is very hard. And after school, I always feel tired. So I often watch TV to relax myself. But watching TV doesn't make me independent thinking difficult. Sometimes when I am very very tired, I would watch some funning programs which are not very good. But I will makes me forget the things which are bad to me. I think movies are the best choices without those which are adult movies or about fighting. I see the movies and think of what will happen. It's a good chance to help me think.
Maybe sometimes we could think independently difficultly. But that's because we didn't choose the good programs. We should choose the programs carefully. We have rights to choose our own programs. Everyone should choose good programs to help no think. (226 words, score = 3.0)
A fifth fluency strategy involved keeping the flow of ideas going by omitting a word or phrase that could not be retrieved or expressed, as shown in an essay that left a blank space for unknown words. Not pausing for the word needed may allow the writer to keep on writing more fluently. It has been found that young children, when allowed to invent spellings and structures, write coherent long texts earlier and more quickly than expected (Valdes, Haro, & Echevarriarza, 1992, p. 340). This may explain the case of the writer who did not allow the inability to recall a single word to interfere with whole-text fluency.
Conclusion and Implications
Results of Vocabulary Levels Tests show that writers' abilities at each vocabulary-word level were highly correlated with writing proficiency. Mastery of the productive 2000-word level appeared to be the vocabulary threshold necessary for the Taiwanese EFL students to produce an English composition in 30 minutes. This implies that a lack of the minimal requirements of the productive 2000-word level may impose psychological limitations on EFL writers' abilities to conceptualize their intended meanings and to organize their thoughts (Freedman, Pringle, & Yalden, 1983), because language learners require "sufficient levels of second-language proficiency to be able to sustain the self-regulated behavior that writing performance in a second language requires" (Cumming, 1989, p. 126). The minimum level of vocabulary proficiency also applies for L1 writers but it is seldom a worry for native speakers because they are very likely to have a vocabulary size of more than 10,000 words by the time they are 14 years old (Nation, 1990). This huge difference in vocabulary proficiency between EFL students compared to native speakers is sometimes forgotten in discussions of L1 - L2 differences for ESL/EFL writers (Grabe, 2001).
Results of vocabulary use in the English essays found that essay length was the strongest predictor for estimating the variance of the English writing quality. The qualitative analysis of essay length provides additional information about strategies for writers with limited vocabulary (like those in the G1 level) to produce longer essays and achieve better writing quality than their peers. These fluency strategies reflect "strategic competence" that enabled those writers to make more effective use of their available language abilities in performing (Bachman, 1990). The strategies that they used to help them overcome the difficulties involved in composing should be of interest to teachers in the field of L2 writing.
As fluency can be defined as the number of words produced in a given span of time, it can be postulated that more fluent writers access a greater number of words more efficiently; and less fluent writers access fewer words less efficiently. This implies that writers need to access their lexical resources efficiently while engaging in production. Proficiency in vocabulary can be measured both in terms of an ease of expressing ideas with words (Reid, 1990) and an ability to access the word needed (Kirby & Kantor, 1983). Nevertheless, the importance of access to lexical resources in academic writing is rarely addressed, while the current trend in L2 writing instruction is mainly to focus on the teaching of the writing process (Hinkel, 2002).
Access to vocabulary is crucial to writing fluency. Most of the EFL students in this study had not developed even the minimum productive level of the 2000 most frequent words. These EFL learners, particularly, need to make the best use of their limited productive vocabularies.
Vocabulary abilities should not be seen so much as concerned with correctness as with the need to be as efficient as possible in retrieving words required to successfully fulfill academic writing tasks (Leki & Carson, 1994). The amount of attention-process resources required to create a text also influences writing fluency (de Beaugrande & Dressler, 1981). Excessive concern with accuracy or correctness in a text may have worked against their writing efficiency. That is, as they wrote, they might have tried to find the most appropriate words. This emphasis may be misplaced and is likely to be the result of stress on vocabulary in Taiwan's EFL courses that present the students with a limited concept of writing (Teng, 1990).
As discussed above, a lack of lexical resources to access or an inappropriate attention to processing resources expended on the production tends to work against EFL writing fluency. The ability to produce longer compositions under time pressure is likely to be a manifestation of a more advanced level of composing competence for EFL learners (Sasaki, 2000).
Pedagogical implications
Building up vocabulary competence, especially at the productive level of the 2000 most frequent words, is critical to writing. This can be the essential lexical level to enhance the development of L2 writing. Lower-level EFL writers may constantly struggle to find the right word while their ideas are slipping away in the meantime. If such students have sufficient vocabulary, they will not feel as frustrated when writing in a foreign language. The implication is that, as the guidelines of vocabulary levels of frequency have suggested (Nation, 1990, p.24), learners of English as a foreign language need a productive vocabulary of around 2000 high-frequency words in addition to strategies for how to generate the low-frequency words they need to produce in writing.
However, the teaching of vocabulary should be separated from the concerns of vocabulary in the writing process. Too great an emphasis on linguistic concerns in writing will interfere with the flow of ideas. With this study's evidence that writers who produce more text do not suffer a decline in writing quality, the development of fluency as part of the writing processes should be given importance in EFL writing instruction. In doing so, the competent skills which combine to create fluency will suggest a high priority for instruction in L2 writing.
Given the relationship of access to vocabulary and to fluency, it is important to address an L2 students' need for efficiency in vocabulary retrieval. Speed in language processing only develops through extensive and repeated acts of language processing (Chenoweth & Hayes, 2001; Leki & Carson, 1994). The implication is that prewriting activities for developing efficient vocabulary retrieval may enhance the ease by which a writing task is accomplished. For example, a prewriting activity like brainstorming on the topic-related words may help students access their inner storehouse of words (i.e., their receptive vocabularies). Likewise, a speaking activity like reporting on topic-related materials can bring relevant vocabularies into productive use.
In summary, facilitating the skill of fluency must become a priority in L2 writing instruction, particularly in foreign language classes. L2 writing instruction based can focus on efficiency in content and organization at the first stage and effectiveness in forms and grammar at another stage. Instruction to help students develop fluency and to ease the process of composition is particularly relevant to the educational context, where writing in a second or foreign language usually involves "high-stakes writing" practices (e.g., an argumentative essay for demonstrating writing proficiency) that are rigidly assessed for soundness of content and clarity of presentation (Elbow, 1997, 2002). Fluency-aimed writing exercises may empower L2 learners to express ideas and information without the pressure of producing polished, finished products. This notion has important implications for the instruction of L2 students, who typically receive few opportunities for extensive writing. The amount of extended writing necessary to facilitate effective writing remains an issue for further research.
