Ellg lish Teaching, Yol. 67 , No.3, A utumn 20 12
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions
27
Eunkyung Hwang
(Sookmyung Women 's University)
Hwang, Eunkyung. (2012). Korean EFL learners ' language development across
p"oficiency levels in w.-itten productions. English Teaching, 67(3),27-50.
The present study was designed to investigate Korean EFL college learners ' language
development across writing proficiency levels in the ir written productions. For the
evaluation of their language development, 18 subsets of the syntact ic complexity,
accuracy, and fluency (CAF) were used. 86 uni vers ity students participated in the
study and wrote narrat ive essays about an assigned topic by using the Criterion®. The
writing samples were scored by e-rater® and divided into three different proficiency
levels to compare language development among groups. Then their essays were
analyzed by us ing L2 Sy ntactical Complexity Analyzel: The findings reveal significant
differences in 15 out of 18 measures of CAF and language learners' linguistic
developmental patterns across their writing proficiency levels. The results indicate that
there are distinctive features regarding CAF among the three groups. With respect to
fluency and accuracy, levels I and 2 showed a similar language developmenta l pattern,
and level 3 was much higher than those groups. In terms of syntactic compl exity,
levels 2 and 3 presented a similar developmental pattern, and level 1 was much lower
than those groups in general. This study shows the predictive potential of CAF across
wri ting proticiency levels of language learners.
I. INTRODUCTION
The term, interlanguage, first introduced by Selinker (1972) is considered to be a system
composed of a number of sub-systems, such as complexity, accuracy, and fluency (CAF).
Since the teml's introduction, second language acquisition (SLA) researchers have long
tried to examine second/foreign language development, supposing that it implies
di stinctive linguistic features of language learners' developing second language systems in
written/spoken contexts (Elli s, 1989; Norris & Ortega, 2009; Ortega, 2003). Certainly,
beginning with Crookes (1989), numerous studies have ShOWll the relationship between
second language learners ' proficiency and their language development in various contexts.
28 ElInkYlIng Hwang
To investigate linguistic development, a number of measures on CAF have been devised
by doing in-depth analysis of interlanguage (e.g., Crookes, 1989; Ishikawa, 2006; Ortega,
1999; Robinson, 1995; Skehan & Foster, 1999). Wolfe-Quintero, Inagaki, & Kim (1998)
indicate that this is a "developmental index study" which attempts to "judge the
development of learners at known proficiency levels" by using measures of complexity,
accuracy, and fluency.
Although the previous studies seemed to provide insights into the linguistic
development of second language learners in English in terms of complexity, accuracy, and
fluency (Homburgh, 1984; Ortega, 2003; Perkins, 1980; Robinson, 1995), most studies
have been conducted in ESL contexts, not in EFL contexts. Moreover, a limited number of
studies on the learners' language development in EFL contexts are lopsided and
concentrated on the subjects of a few countries, such as Japan (Casanave, 1994; Ishikawa,
1995, 2006; Kamimura, 1996; Robinson, 2001; Sasaki, 2000) and China (Ellis & Yuan,
2004; Larsen-Freeman, 2006; Lu, 20ll). Even the studies that deal with Korean subjects
(Kim, 2011; Lee, Oh, & Shin, 2007; Song, 2005a, 2005b) do not examine the linguistic
development of Korean EFL learners with respect to all the dimensions of CAF.
Furthermore, there are few studies that measure EFL learners ' language development in
relation to proficiency levels. Some of the studies used low-English language proficiency
level EFL learners (Ishikawa, 1995,2006); others focused on the intermediate levels as the
English language proficiency (Casanave, 1994; Larsen-Freeman, 2006). Based on the
critical analysis of the earlier scholarships, it seems crucial to investigate which measure is
the most valid and reliable in measuring the language development of Korean university
students as EFL learners across writing proficiency levels.
Therefore the present study was designed to examine the following research questions:
I. Which measures on syntactic complexity, accuracy, and fluency (CAF) show
significant correlation with writing scores? Which pairs have strong correlations
within a given subset?
2. Are there any significant language developmental patterns to be discovered across
writing proficiency levels in terms of syntactic complexity, accuracy, and fluency
(CAF)?
3. Which measure is the most effective predictor of syntactic complexity, accuracy,
and fluency (CAF) with respect to writing proficiency?
Korean EFL Learners' Language Development across Proficiency Levels in Wrinen Productions 29
II. LITERATURE REVIEW
1. Complexity, Accuracy, and Fluency (CAF)
Since the 1990s, syntactic complexity, accuracy, and fluency (CAF) have appeared
prominently as measures to assess second/foreign language development (Skehan, 2009).
Tlu'ough various studies, tllis emerging concept, CAF, is considered to comprise distinct
variables to measure the language learner's linguistic performance that can be separately
measured under various second language-Ieaming contexts (Housen & Kuiken, 2009).
The main reason for measuring L2 performance in tem1S of complexity, accuracy, and
fluency (CAF) is to describe "how and why language competencies develop for specific
learners and target languages, in response to particular tasks, teaching, and other stimuli,
and mapped against the details of developmental rate, route, and ultimate outcomes"
(Non-is & Ortega, 2009).
To assess syntactic complexity, accuracy, and fluency (CAF) together in a given context
is to defme "a global picture of language development" (Housen & Kuiken, 2009). These
measures focus more on morphological, syntactic, and lexical language development than
on characteristics of discourse (Housen & Kuiken, 2009). In most studies on L2 writing,
the main unit for analysis examined is the T-unit, defined as "one main clause plus any
subordinate clause or nonclallsal structure that is attached to or embedded in it" (Hunt,
1970, P. 4); the types of clauses used (e.g., independent, dependent, adjective, adverbial,
nominal clauses) is also considered. The key terms used in this study are defined as
fo llows:
I) Syntactic Complexity
Syntactic complexity is defined as "the range of forms that surfaces in language
production and the degree of sophistication of such [on11s" (Oltega, 2003), and is obvious
in second language writing in ten1lS of "how varied and sophisticated the production units
or grammatical structures are" (Foster & Skehan, 1996; Lu, 20 11 ; Ortega, 2003; Wolfe
Quintero et aI. , 1998). This dimension has been taken into account as a crucial construct in
second language teaching and research since some syntactic features of target language
indicate a second language leamer 's general development in a given context (Ortega,
2003 ; Lu, 201 I). For this reason, it is considered to be a val id and reliable index of second
language learners' developmental level or overall proficiency in the target language. A
nW11ber of various measures have been proposed for characterizing syntactic complexity in
second language writing. For measuring syntactic complexity, ratio or frequency of
coordination (e.g., coordination phrases per T-unit, T-llnits per sentence), subordination
30 Elinkyling Hwang
(e.g., clauses per T-unit, dependent clauses, dependent clauses per T-unit), surface syntactic
structures (e.g., clauses per sentence), and sophistication of particular syntactic structures
(e.g., complex nomina Is per T-unit, verb phrases per T-unit) is used (Lu, 2011; Non'by &
Hakansson, 2007; Ol1ega, 2003).
According to Wolfe-Quintero et al (1998), developmental pattern in L2 learners '
syntactic complexity changes as a reversed U shape as time goes by. In the beginning, L2
learners produce some strings of words, or sentence fragment, and then they produce a
simple sentence, then they combine some sentences by using coordinate conjunctions such
as ' and, or but'. Then they make more complicate sentences, that is, subordinate clauses .
In this period, L2 learners' syntactic complexity reaches the peak point. Then finally they
can produce reduced fornls like some types of phrases instead of subordination clauses.
2) Accuracy
Accuracy is based on the errors in grammar, vocabulary, and complexity referred to the
use of simple and complex clauses (Wolfe-Quintero et aI, 1998). It is also defined as
"lieedom from error" (Foster & Skehan, 1996) and "the degree of deviancy from a
particular nonn" (Hammeriy, 1991; Wolfe-Quintero et aI., 1998). Wolfe-Quintero et al.
mentioned that some of the good measures of accuracy are the number of error-free T-units
(EFT), error-free T-urnts per T-unit (EFT/T), and the number of errors per T-unit (E/T).
Although the first two measures may be useful for more advanced learners, it is not always
easy to find any error-free units (EFT) in the perfonnance of beginners and (low)
intennediate learners (Ishikawa, 1995). The number of errors per T-unit (E/T) tells us the
overall accuracy of the language users . The developmental pattern ofL2 language learners
is a U shape.
3) Fluency
Fluency is the "processing of language in real time" (Schmidt, 1992), with a focus on
the "primacy of meaning" (Foster & Skehan, 1996). Fluency means that "more words and
more structures are accessed in a limited time," whereas a lack of fluency means that "only
a few words or structures are accessed in a limited time" (Foster & Skehan, 1996). Fluency
is "not a measure of how sophisticated or accurate the words or structures are," but "a
measure of the total number of words or structural urnts" which a writer is able to include
in hislher writing within a given time (Wolfe-Quintero et aI., 1998). There are measures of
fluency such as total words in a text (TW), mean length of sentence (MLS), mean length
of T-urnt (MIT), and mean length of clause (MLC), etc. Like the pattern of syntactic
complexity, the developmental pattern ofL2 language learners' fluency can be depicted as
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions 31
a reversed U shape.
2. Previous Research on CAF
There are two well-known synthesized studies in this field . First, Wolf-Quintero et al.
( \998) investigated how second language development is evaluated in L2 writing research
through reviewing 39 writing studies and examining which measures of complexity,
accuracy, and fluency (CAF) are the best indictors ofL2 developmental levels in language
learners' written productions. They identified relationshjps among the numerous measured
CAF variables. Five years later, Ortega (2003) took over the work by standardizing 25
second language writing studies at the college level that used various criteria in each L2
writing study. However, Ortega only synthesized these studies, especially with respect to
syntactic complexity. Thjs is because many of the studies indicated a connection between
the level of syntactic complexity and the overaIl proficiency of a student (Ishikawa, 1995;
Lu, 20 11 ; Ortega, 2003).
In spite of the extended use and importance of these metrics in language development
studies, very little systematic research has been done to date in this area in Korea. Song
(2005a, 2005b) and Lee et al. (2007) conducted the task-based researches with Korean
college students ' narrative spoken performance in terms of complexity, accuracy and
fluency (CAF). Song (2005a, 2005b)'s studies investigated whether pre-planning time and
content familiarity affected the spoken perfonnance of 20 advanced learners in a narrative
task. Lee et al. also investigated the effects of pre-planning time on the spoken narrative
tasks of L2 low-intermediate learners in terms of complexity, accuracy and fluency. The
results showed that planned tasks produced more fluent and more accurate speech and that
there were no significant differences between pre-planned and non-planned groups. Up till
now, there has been only one study that has investigated Korean EFL learners' written
production in terms of CAF. Kim (20 11 ) 's study is a very recent work that examines the
effect of cognitive task complexity on 18 Korean college students' written performance
with respect to complexity, accuracy and fluency. However, Kjm 's study focuses not on
language development but on the influence of cognitive task complexity across language
proficiency levels.
3. Patterns of L2/FL Writing Development
Many of the previous researchers describe the relations among the three dimensions,
CAF, and claim patterns of second language development based on their studies (Ortega,
1999; Sasaki, 2000; Wolfe-Quintero et aI. , 1998). Complexity and accuracy are generaIly
connected to the current state of the learner 's second language knowledge and "the leve l of
32 EUllkyung Hwang
analysis of internalized linguistic information." In contrast, fluency is primarily related to
" learners' control over their linguistic L2 knowledge" (Housen & Kuiken, 2009; Wolfe
Quintero et aI., 1998). Ortega (1999) claims that the developmental rates do not take place
at the same degree. Fluency shows faster development among them and reaches higher
levels than complexity or accuracy in general. On the other hand, complexity seemed to
undergo the slowest development within the same period (Wolfe-Quintero et aI. , 1998).
Moreover, "interlanguage development of fluency and complexity are linear, while that of
accuracy is not" (Ortega, 1999).
To be specific, Wolfe-Quintero et al. (1998) identifies how granmlatical complexity in
written language develops, by noting that it develops from coordination to subordination to
reduced phrases, by referring to works of Monroe (1975), Cooper (1976), Sharma (1988)
and Hunt (1965). That proposal has been supported somewhat in case of adult second
language learners by Ishikawa (1995) who suggests that at the very beginner stages,
writers move from fragments to clauses to T-units. Furthermore, Shanna (1980) suggests
that although there is a stage at which relative clauses increase, at the more advanced
reduction stage they may decrease in favor of adjectives and prepositional phrases. More
advanced writers may tend to use more reduced forms (Cooper, 1976; Hunt, 1965; Monroe,
1975; Sharnla, 1980).
III. METHODOLOGY
1. Setting and Participants
The subjects in this study consisted of 86 Korean university students in Seoul, Korea,
who attended S university in the spring semester of 2012. At that time when the data for
the present study was collected, 35 out of 86 students took the general English program,
English Reading and Writing, and the 51 students took the English major class,
Introduction to English Language. Table 1 shows the descriptive statistics of subjects'
writing test scores by e-rater. The e-rater® engine provides a holistic score from one to six
for an essay about granIDlar, usage, mechanics, style and organization, and development
(Monaghan & Bridgeman, 2005). The results of e-rater's evaluations are given below.
Based on the subjects' scores on the writing test, a one-way ANOYA test was done on the
different proficiency groups and the three groups were determined to be significantly
different from each other statistically. They were divided into three levels according to
their writing scores as level 1 (N= 19, mean score: 1.89), level 2 (N=31 , mean score: 3.00),
and level 3 (N=36, mean score: 4.22) groups.
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions 33
TABLE] Descriptive Statistics of the Participants' Writing Test Scores
Level
Level 3 (High-intermediate)
Level 2 (Intermediate)
Level 1 (Beginner)
Total No. Mean Score (SO) Writing Score of Subjects (p < .01) Score Description
36 4.22 (0.48) 6 excellent 5 skillful 4 sufficient
31 3.00 (0.00) 3 uneven
19 1.89 (0.32) 2 insufficient unsatisfactory
Note. one-way ANOYA test: Data are expressed as Mean (SD). level 1 (Beginner); level 2 (Intermediate); level 3 (High-intermediate)
No. of Subjects
1 5
30
31
17 2
Table 2 displays the descriptive statistics with regard to subjects' writing levels and their
personal background related to English study as one of two sets: means and standard
deviations (SD); numbers and percentages. The participants consisted of 45 (52.3%)
freshmen, 21 (24.4%) sophomores (24.4%), 13 (15 .1%) juniors, and 7 (8.1%) seniors.
Their average age was 21.29. As shown in Table 2, the participants' writing proficiency
was not affected by their grade, major, or experience in studying English essays. On the
other hand, subjects who had an official English score or had experience abroad in
English-speaking countries seem to have higher writing proficiency than those who had
not. 38 (44.19%) students responded that they learned or studied how to write essays in
English for preparing some official English tests such as TOEFL, TOEle, or TEPS. More
than 50 percentages of students were freshmen and most of them had no experience in
writing and had never taken a formal English writing class. 32 (37 .2%) students responded
that they had experience traveling or going abroad to English speaking countries. However,
18 (56.3%) out of 32 students had visited English-speaking countries for less than three
months, whi le 7 (21.88%) out of 32 students had stayed in English speaking countries
between six months to one year for studying or family business. The last 7 (21.88%)
students also had lived in English speaking countries [or an average of2 years for studying
or family business.
34 Eunkyung Hwang
TABLE 2
Total Levell Level 2 (n= 19) (n=3l )
Writing score 3.28±0.97 1.89±0.32 3.00±0.00 Grade
I 45(52.33) 13(68.42) 17(54.84) 2 21 (24.42) 1(5.26) 9(29.03) 3 13(15. 12) 2(10.53) 3(9.68) 4 7(8.14) 3(15.79) 2(6.45)
Major Non-English 69(80.23) 17(89.47) 27(87.10)
English 17(19.77) 2(10.53) 4(12 .90) Study of English writing
Yes 38(44.19) 5(26.32) 13(41.94) No 48(55.8 1) 14(73.68) 18(58.06)
Official Engl ish test score Yes 38(44.19) 2(26.32) 7(41.94) No 48(55 .8 1 ) 17(73.68) 24(58.06)
Experience in English-speaking country Yes 32(37.21 ) 4(10.53) 8(22.58) No 54(62.79) 15(89.47) 23(77.42)
Note. x? test or Fisher exact test: Data are expressed as N(%) or Mean±SD.
2. Tasks and the Procedures
Level 3 (n=36)
4.22 ±0.48
15(41.67) 11 (30.56) 8(22.22) 2(5.56)
25(69.44) II (30.56)
20(55.56) 16(44.44)
29(55.56) 7(44.44)
20(80.56) 16(19.44)
<.01 .13
.13
. 11
<.01
.01
In the second week of the semester, the participants took a web-based writing test as a
course requirement. The writing test was conducted through Criterion® online writing
evaluation service in a language laboratory. Before taking the test, they listened to
explanations about the process of the test and completed a questionnaire designed to
survey their background in Korean (e.g., grade, major, English education background, etc.)
in 10 minutes. Then they wrote a narrative essay on the following writing prompt for 40
minutes: Think about the goals you have for your foture. Write an essay about what you
will do to reach your goals, whether that is to be a professional athlete, a famous scientist
or a happy Mom or Dad. Tell the story of how you will achieve your goals, being as
specific as possible in describing the obstacles and challenges you might face on the road
to success. While taking the test, they were not allowed to look up words in an English
dictionary and to use scratch paper. Submitted essays were scored from one to six based on
the scale by e-rater® engine on criterion® online writing evaluation service. Based on the
writing scores, the writing samples were divided into three different proficiency levels
(e.g., level l as a beginner level, level 2 as an intermediate level, and level 3 as a high
intermediate level) to compare writing development among groups.
Korean EFL Learners ' Language Development across Proficiency Levels in Written Productions 35
3. Measures and Dependent Variables
18 subsets of syntactic complexity, accuracy, and fluency (CAF) were used to measure
language development in written productions. Syntactic complexity was assessed by four
types of sub-variables as Table 3 shows: sentence complexity; coordination; subordination;
particular structures. To measure accuracy, error free T-unit (EFT), error-free T-unit per T
unjt (EFTIT), and errors per T-unit (EIT) were used. Fluency was established by means of
total words in a text (TW), mean length of sentence (MLS), mean length ofT-unit (MLT),
and mean length of clause (MLC).
Variable
Syntactic Complexity
Accuracy
Fluency
4. Analysis
TABLE 3
Summary of CAF Measures Investigated in the Study
Type Measure Sentence complexity
Coordination
Subordination
Particular structures
Accuracy
Length of production
clauses per sentence coordinate phrases per clause coordinate phrases per T-unit T-units per sentence clauses per T·unit complex T-units per T-unit dependent clauses per clause dependent clauses per T-unit complex nominals per clause complex nominals per T-unit verb phrases per T-unit error-free T-unit error-free T-units per T-unit errors per T-unit total words in text mean length of sentence mean length ofT-unit mean length of clause
Code CIS CP/C CP/T TIS CIT CT/T DCIC DC/T CNiC CNIT VP/T EFT EFTIT E/T TW MLS MLT MLC
Submitted essays were scored from one to six scales by e-rater® engine automatically.
However, 3 out of 86 essays scored as N/A (not applicable) because they were off-topic
essays. To give scores for the three N/A scored essays and count errors for measuring
accuracy, two experienced native speaking instructors of English with at least 3 years
experience working in a large university writing center rated these essays based on the
standardized rubric of e-rater®. The rubric was used to holistically assess the quality of
the essays and had a minimum score of I and a maximum score of 6. The raters were first
trained to use the rubric with 10 sample essays. The interrater reliability of at least r = .80
was significant (p < .01) for scoring the essays. After that, errors, defined as any
36 Eunkyung Hwang
morphological, syntactic, semantic errors excluding spelling and punctuation errors (Scott
& Tucker, 1974), were counted by the first rater for measuring accuracy (e.g., EFT, EFTIT,
and EIT). To secure interrater reliability, 10 writing samples were randomly selected and
analyzed by the second rater. lnterrater reliability scores for all measures were over 95
percent (p < .05). Finally, L2 syntactic complexity analyzer 2.4 (Lu, 2010) was used to
analyze the syntactic complexity and fluency of the essays.
A statistical analysis of the data was performed using SPSS version 18.0. Data are
reported as means and standard deviations (SD) for continuous variables and as numbers
or percentage for categorical variables. Basic characteristics among the three different
proficiency groups were compared using one-way analysis of variance for continuous
variables and the :x? test or Fisher exact test for categorical variables. A series of one-way
ANOVAs were performed on the measures followed by post-hoc Scheffe tests. The alpha
for achieving statistic significant was set at .05. Simple correlation analysis via Pearson
coefficient or Spearman coefficient was used in estimations of the strengths of correlations
between 18 measures of syntactic complexity, accuracy, and fluency (CAF) and three
different writing proficiency groups.
IV. RESULTS AND DICUSSION
The results, as shown in the Table 4, reveal correlation I between each writing score and
syntactic complexity, accuracy, and fluency (CAF). Each variable of CAF consisted of its
own subsets used to measure language development in this study. All subsets of fluency
are statistically significant with writing scores (p < .01). Among them, TW shows the
highest correlation with writing scores (r = .90), and MLS (r = .53) and MLT(r = .55)
present moderate correlation with writing scores. MLC (r = .35) indicates weak correlation.
For accuracy, all subsets are statistically significant with writing scores (p < .01). EFT (r
= .65) shows the strong positive correlation with writing scores. EFTIT (r = .47) correlates
moderately in positive direction and EIT (r = -.46) does in negative direction. However, in
case of the syntactic complexity, all but two measures (CPIC and CP/T) correlate with
writing scores at a statistically significant level (p < .05). Five measures (CIT, DCIC, Dc/T,
CNIT, VP/T) show moderate correlation, and four measures (CIS, CT/T, TIS, CN/C)
indicate weak correlation with writing scores. CPIC and CPIT as the measures for
coordination do not show correlation with writing scores.
I Following Wolfe-Quintero et al. (1998), correlations are character ized as high (r 2: .65), moderate (.45 :s r < .65), and weak (.25 :s r < .45).
Korean EFL Learners' Language Development across Proficiency Levels in Written Product ions 37
TABLE 4
Correlation Coefficient between Writing Scores and CAF
Fluency Accuracy Syntactic Complexity
TW .906 EFT .656 CIS .396 DCIT .476 CN/C .27" MLS .53b EFTIT .47b CIT .4Sb CP/C -.05 CNIT .53 h
MLT .5Sb EIT _.46b CTIT .42b CP/T .07 VP/T .Sl b
MLC .3Sb DCIC .4Sb TIS .24"
Note. Spearman correlation. a p < 0.05 , 6 P < 0.01 ; E/T, CTIT , DCIT: Log transformed.
Tables 5 to 7 summarize the correlations among each subset on syntactic complexity,
accuracy, and fluency (CAF). Shown in Table 5, the highest correlations on fluency are
between MLT and MLS (r = .92), moderate correlations are between MLC and MLS (r
= .53), MLC and MLT (r = .59), and week correlations are between TW and MLS (r = .43),
TW and MLT (r = .39), even though they are significant in statistics (p < .01). To compare
with Lu (20 11)'s result of EFL Chinese col1ege students' written productions, the highest
correlation is the same as Lu's result (MLS and MLT, r = .9 1). The result of correlations
between MLC and MLS (r = .57), and MLC and MLT (r = .66) are a little higher
numerical value than this study, but this is not a significant difference in statistics. In his
study, there was no TW measure, so with this measure no comparison was possible. The
overall results of my study support the previous study (Lu, 2011) that MLS and MLT
correlate strongly in measuring length of production. Moreover, among the four measures
of length of production, MLS and MLT are more strongly correlated than others such as
MLCandTW.
TW MLS MLT MLC
TABLES
Correlation Coefficients among Subsets on Fluency
TW MLS MLT .43 6 .396
I .92b
I
Note. Pearson correlation. a p < 0.05 , 6 P < 0.0 I
MLC .22" .S3b
.59b
I
According to the Table 6, the highest correlation in accuracy is between EFTIT and EfT
(r = -.80). Correlation between EFT and EFTIT (r = .67) is also high, and EFT and EfT (r
= -.60) also show moderate correlations. Al1 measures reach statistical significance and
seem to be strongly correlated with each other.
38
EFT EFT/T E/T
Eunkyung Hwang
TABLE 6
Correlation Coefficients among Subsets on Accnracy EFT EFTIT
Note. Pearson correlation. "p < 0.05, b P < 0.0 I; E/T: Log transformed.
E/T
In Table 7, the highest con-elation in syntactic complexity is between DCIC and DC/T (r
= .98) as a subordination measure. Lu (2010),s study also showed the highest con-elation
between these two measures (I' = .95). This result also supports Lu (2011)'s study. In
general, some subsets for sentence complexity (CIS), subordination (CIT, CT/T, DCIC,
DCIT), or part of particular structure (CN/T, VP/T) with DCIC or DC/T show high
con-elation with them. On the other hand, they do not show any con-elation with
coordination (CP/C, CP/T, TIS) and one type of particular structures (CN/C) . In case of
coordination measures (CP/C, CP/T, T IS), 12 out of 18 show no con-elation between any
pairs, 4 out of them (CP/C and CIT, TIS and CIS, TIS and CIT, TIS and CP/T) exhibit weak
con-elations, and only last two pairs (CP/C and CP/T, TIS and CP/C) show high
correlations between them. In the case of particular structure, CN/T and VPIT are more
strongly con-elated with other measures than CN/C with others. To be specific, CNIT and
VP/T measure show mostly above-average con-elation with all other measures except for
coordination measures (CP/C, CP/T, TIS), and CN/C seems to con-elate with coordination
measures (CP/C, CPIT, TIS) in contrast to CN/T and VP/T. The results of this study
support Lu (201 1)'s previous study except for con-elation between CNIT and CN/e.
TABLE 7
Correlation Coefficients among Subsets on Syntactic Complexity I. 2. 3. 4. 5. 6. 7. S. 9. 10. II.
CIS CIT CT/T Dc/C Dc/T CP/C CP/T TIS CN/C CN/T VP/T I .726 .726 .S26 .S66 -. I I .17 .24a .12 .566 .696
2 I .70b .SOb .SSb _.2Sb -.I S .37b .31 b .60b .74b
3 I .SOb .SOb .1 9 .1 3 -.1 4 .03 .SSb .49b
4 I .9Sb -.17 .07 .16 .06 .64b .69b
5 I -.15 .11 .13 .06 .67b .7S b 6 I .77b _.6Sb .2 1 a -.02 _.24" 7 1 _.46b .SS b . IS .OS S I _.29b .09 .34b 9 I .SOb .19 10 I .73b
II I Note. Pearson correlation. " p < 0.05, b P < 0.0 I; CT/T, Dc/T: Log transformed
Through Tables 8 to 10, three aspects of language use were investigated to see how the
participants in the three writing proficiency groups perfonned in the written task. Tables 8
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions 39
to 10 report the mean values and standard deviations (SD) separately in the subset of
syntactic complexity, accuracy, and fluency (CAF) in the three writing proficiency levels.
That is, they summarize the actual between-level differences that show a significant
difference statistically (p < .05) in a Scheffe test, a one-way analysis of variance (ANaYA)
post multiple comparison test to detennine whether such differences exist between any
two writing proficiency groups .
According to Table 8, in fluency measures, all four measures (TW, MLS, MLT, MLC)
are statistically-sigrtificantly different among groups (p:S .01). From the result of the post
hoc test, TW presents significant difference in the statistics among the three different
writing proficiency levels . This result supports the previous studies (Hirano, 1991;
Homburg, 1984; Larsen-Freeman, 1983) summarized in Wolfe-Quintero et al. (1998), all
of which "showed sigrtificant effect compared writers at different proficiency levels using
a composition with a time limit." The three groups are nonadjacent according to the TW
measure . In other words, level 3 is the most fluent group and level 2 perfonned more
fluently than level 1. However, in the case of the other three measures as fluency ratios, the
results are a little different. The results of the one-way ANaYA of MLS, MLT, and MLC
shown in Table 8 reveal that the scores on the ratios of them increase across the
proficiency levels and are statistically sigrtificant (p :s .01); however, the post hoc test
shows that there is no sigrtificant difference between the levels 1 and 2. In other words, the
leve ls 1 and 2 are adjacent groups; only level 3 is a non-adjacent group. Wolfe-Quintero et
al. (1998) summarized some average ranges of the three measures from the lowest level
learners to the most advanced learners and indicated that these three measures increase in a
linear relationship to proficiency in most studies : the means ofMLS is a range from 8.5 to
23.59 words per a sentence; the means of MLT is a range from 6.0 to 23.0 words per T
unit; the means ofMLC is a range from 5.20 to 10.83 words per clause.
TABLE 8
Group Means and Standard Deviations fo.· Each Measure on Fluency
Means (SO) Variable Measure Level I Leve l 2 Level 3 p -value
TW 168.84 (4 1.45) 255.65 (39. 15) 367.94 (83.03) <.0 I MLS 10.98 (3. 09) 12.81 (2.65) 15 .24 (2.75) <.01
Fluency ML T 10.1 7 (2.43) 11.73 (2.26) 13.86 (2.37) <.01 MLC 7.74 ( 1.36) 8.02 (0.94) 8.83 (1.12) .0 I
Note. a = level I (beginner); b = level 2 (i ntermediate); c = level 3 (h igh-intermediate)
post-hoc a<b<c a=b<c a=b<c a=b<c
Accuracy was measured in three ways; error-free T-unit (EFT), error-free T-units per T
unit (EFTff) and errors per T-unit (E/T) . Table 9 shows the results for this aspect of
production. All measures are shown to be statistically-significantly different among groups
(p < .0 1). EFT as one of the frequency measures appears to increase sharply as proficiency
40 Eunkyung Hwang
increases, and there is an overall statistically significant difference among the three groups.
This supports the previous studies (Hirano, 1991; Homburg, 1984). However, this result
seems to not support to Wolfe-Quintero et al. (1998)'s claims that EFT/T does not
necessarily increase with language learner proficiency. To be specific, the relationship
between accuracy and proficiency in second language writing doesn 't seem to be a linear.
However, Larsen-Freeman (1983) gives a clue to explain this phenomenon. She claims
that language learners peaked in their EFT/T at the intermediate level and then decreased
at more advanced levels. Actually, there is no advanced level in this study, so it is hard to
find the decreased phenomenon of the result of EFT/T. E/T is one of the measures that
appear to discriminate among proficiency levels (Flahive & Snow, 1980). EfT is shown to
discriminate writing proficiency levels in a negative direction. This is the reason error
ratios decrease across proficiency levels. According to Wolfe-Quintero et aJ. (1998), there
are some rates for E/T from a 0.035 low for advanced learners (Perkins and Leahy, 1980)
to a 1.51 high for students who failed a writing test (Perkins, 1980). The multiple
comparison tests of EFT/T and EfT reveal some differences between levels I and 3, and
between levels 2 and 3. However, there is no significant difference between levels 1 and 2.
That is, levels 1 and 2 are adjacent groups with regard to accuracy as well as fluency.
TABLE 9
Group Means and Standard Deviations for Each Measure on Accu.-acy
Variable
Accuracy
Measure EFT EFTfT EfT
Note. EfT: Log transformed
Level 1 9.16 (4.39) 0.51 (0 .1 4) 0.79 (0 .32)
Means (SO) Level 2
13 .10(4.62) 0.58 (0.13) 0.65 (0 .28)
Leve l 3 18.47 (6.26) 0.68 (0.12) 0.46 (0.22)
p-value <.0 1 <.01 <.01
a = level 1 (beg inner); b = leve l 2 (intermediate); c = level 3 (high-intelll1ediate)
post-hoc a<b<c a""b<c a""b>c
Table 10 indicates the mean values and standard deviations (SD) in the II subset of
syntactic complexity across the three different writing proficiency levels . 8 out of 11
measures show significant differences in statistics (p < .05). According to Wolfe-Quintero
et al. (1998), sentence complexity ratio (CIS) is one of the good indicators of
developmental growth of beginning level writers (Ishikawa, 1995) and adult learners
(Bardovi-Halig, 1992). According to Table 10, the ratio of CIS ratio seems to increase
across writing proficiency levels in a positive direction. It supports Ishikawa (1995)'s
study but does not fit to with the result ofLu (2011)'s study, which shows a decrease of the
ratio across the school levels. From the multiple comparison test of CIS, levels 1 and 3
show differences between two groups. However, there is no difference between levels I
and 2, and between levels 2 and 3. The only difference is between levels 1 and 3.
In case of the four subsets to subordination ratio (CIT, CT/T, DCfC, DCfT) in syntactic
Korean EFL Learners' Language Development across Proficiency Levels in Wrinen Productions 4 1
complexity, all measmes present statistical significances (p ::; .01). That is, levels 2 and 3
are adjunct levels and they do not significantly differ across groups. The only different
group in the subordination measmes is levell , which consists of writing scores 1 and 2.
When it comes to coordination, CP/C and TIS ratios decrease across writing proficiency
levels. CP/T ratio also declines from levels 1 to 2 and stays in level 3. To compare this
result with the results of Cooper (1976)'s study of relations between CP/T and school
levels, it supports his claims to some degree. From his study, the trend toward a relation
between CPIT and school level decreases from second-year to third-year students and
stays the same in fourth-year students and then increases in fifth-year students. However,
Cooper's study does not reach the same conclusion as Lu (2011)'s study. Even though the
results within respect to CP/C and CP/T fai l to show a statistically significant difference
among the three groups, a decreasing pattern of coordination is one of the indexes of an
ESL leamer's linguistic development. Simply because these measmes do not captme the
linguistic change among groups does not mean that they are not valid measmes. This
phenomenon could be explained by the fact that most of the language learners across the
three groups have already passed the acquisition of the coordination stages or are passing
such a stage at the moment of the study. According to Ishikawa (1995), some measures
may "reflect the development of writing proficiency better but change more slowly than
measures which change significantly. Also, change is not necessarily progress and may not
reflect improvement". The last coordination measure, TIS, is significant statistically (p
< .05). To compare with Ishikawa (1995)'s study, the ratios for levels are very similar. In
Ishikawa's study, the ratios of TIS of the two different proficiency groups of writers are
0.95 and 1.06, respectively. Ishikawa explains that early development proceeds from
uncoordinated clauses to coordinated clauses. Even though TIS is a good measure for
different proficiency level groups, it does not support for school levels or holistic rating
levels (Cooper, 1976; Homburg, 1984; Lu, 2011). From the post-hoc tests, there are
significant differences between levels I and 3. However, there is no difference between
levels I and 2, and between levels 2 and 3.
Within the particular structures, CNIT and VPIT show significant difference in statistics
(p < .01). These measures increase in a positive direction across writing proficiency levels.
This supports the previous studies that CNiC increased with the age of writers (Hunt,
1965) and with the grade of writers (Cooper, 1976). However, CN/C does not support
Wolfe-Quintero et al. (I 998)'s assun1ption that CN/C might yield even better results across
writing proficiency groups. This result does not support to Lu (20 11)'s result that CN/C
shows statistical significance between different levels across school levels, either. It might
be a good measure for different school levels, but not for different proficiency levels.
In the opposite case, VPIT shows statistical significance (p < .01) and group difference
between levels 1 and 3, and between levels 2 and 3. Even though VPIT fai led to show
42 Eunkyung Hwang
significant difference across school levels in a previous study (Lu, 2011), it is an effective
measure with which to examine language learners ' syntactic complexity across the writing
proficiency levels.
TABLE 10
Group Means and Standard Deviations for Each Measure on Syntactic Complexity
Means {SD2 Variable Measure Level I Level 2 Leve l 3 2 -value Qost-hoc
CIS 1.41 {0.332 1.60 {0.292 1.74 (0.332 <.01 a<c CIT 1.20 (0.4 1) 1.4 7 (0.23) 1. 58 (0.29) .0 1 a<b=c CTIT 0.25 (0.12) 0.41 (0.24) 0.44 (0. 15) .0 1 a<b=c DCIC 0.21 (0.08) 0.31 (0.09) 0.34 (0.09) <.01 a<b=c
Syntactic DCIT 0.29 {0.J72 0.47 {0.2 12 0.56 {0.272 <.01 a<b=c
Complexity CP/C 0.27 (0.23) 0.24 (0. 18) 0.21 (0.09) .46 CP/T 0.38 (0.37) 0.31 (0. 16) 0.33 (0. 14) .55 TIS 0.96 {0.272 1.06 {0.212 1.10(0.092 .04 a<c CNiC 0.85 (0.48) 0.76 (0.23) 0.89 (0. 18) .19 CN/T 0.94 (0.36) 1. 15 (0.35) 1.40 (0.36) <.0 1 a=b<c VP/T 1.80 (0.572 1.93 (0.422 2.36 (0.4 12 <.01 a=b<c
No/e. CTIT, DCIT: Log transformed a = level l(beginner); b = level 2 (intermediate); c = level 3 (high-intermediate)
Tables 8 to 10 reveal significant differences in the subsets on syntactic complexity,
accuracy, and fluency (CAF) among the three different writing proficiency groups. Raw
scores of each measure are not comparable because they are on different scales, so they
have been converted to z-scores, or standardized scores. These scores express the distance
from the mean for that measure and make it possible to place each syntactic complexity,
accuracy, and fluency (CAF) measure on the same graph. In Figures 1 and 2, it is clear that
there are some patterns that show the adjacent or nonadjacent groups across syntactic
complexity, accuracy, and fluency (CAF).
When it comes to fluency and accuracy in Figure 1, levels 1 and 2 seem to be adjacent
groups, and this is similar to the results shown in table 7 and 8. The frequency of measures
for fluency (TW) and for accuracy (EFT) shows a significant difference between any two
groups among the three groups. On the other hand, all other ratios of measures for fluency
(MLS, MLT, MLC) and accuracy (EFTIT, EIT) show the same pattern that places levels 1
and 2 as adjacent groups. In case of fluency, the most predictable measure of the four
subsets is TW to present among group differences. The next one is the MLT or MLS as
ratios of measures. It could be the alternative best predictable measure of fluency if the
written productions are not produced within the same assigned time, making it impossible
to use TW. For accuracy, EFT is the most variable for measuring accuracy, but EFT/T and
E/T are also good measures for judging accuracy across writing proficiency levels. It is
hard to say that one is better than the other because those two subsets, EFTIT and E/T,
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions 43
have the almost same value to capture the development of accuracy.
:" <S
FIGURE 1
Group Means of z Scores on Fluency and Accuracy
1.2 ~--------------------------
0,8 +-~~------------------------
0.4 +--------------"''------------\--+ --
~ 0 +---~---,---~--_r--_,---,_---¥~-_.
N ·o , ~ +---------------=0------------1-------\,--
.0 ,8 -I------".A!~=::::tIlL-------_.P""""~------
·1.2 -'---~'---------------------------
Fluency Accuracy
-+-Level .\
~Levelj
As shown in Figure 2, different patterns of syntactic complexity to compare with
fluency and accuracy can easily be recognized. In the case of sentence complexity (CIS),
subordination measures (CfT, CrfT, Dc/C, DCfT) and coordination measure (TIS), levels
2 and 3 are adjacent groups. On the other hand, in case of particular structures (CNfT,
VPfT) obviously levels I and 2 are adjacent groups.
FIGURE 2
Group Means of z Scores on Syntactic Complexity
1.2
0,8
-.... OA • • - • .. --Level 3 a
<.> ° Ir" ~ CIS ----Levell crr CTrr D C/e
·o , ~
--·0,8
·1. 2
Syntactic Complexity
As a matter of fact, all the results of the study in terms of CAF show the some stages of
44 Eunkyung Hwang
the interianguage, or language learner 's developmental stages, which is similar to what
was shown by the results of previous studies (Casanave, 1994; Wolfe-Quintero et a!. ,
1998). EFL writers have a tendency to write longer, more complex sentences and
subordination clauses, and to write more elTor-free T-w1its (EFT), when they come to
reach higher proficiency levels . Students who were in level 3 used less coordination
clauses than students who were in levels I and 2, but it does not seem to be statistically
significant. In general, the coordination measure has a predictive power for measuring
syntactic complexity at beginner levels of L2 development (NolTis & Ortega, 2009).
However, the participants in tillS study consisted of beginners to high-intermediate levels,
so this index does not seem to be a powerful measure for the three writing proficiency
levels. This is because all three writing proficiency level groups have already passed the
coordination stage, so there is no statistical difference among the groups. In this case, the
subordination index is a greater predictor for measuring syntactic complexity from
beginner to high-intelTnediate levels of L2 development.
Among the three writing proficiency groups, level 2 has undergone the most critical
language development (See Figures 1 and 2). Even though level 2 is adjacent to level 1 in
telTnS of fluency, level 2's ratios of MLT and MLS are considerably higher than those of
level 1. In syntactic complexity, level 2 is adjacent to level 3. This supports the previous
studies (Larsen-Freeman, 1983 ; Ortega, 2003; Wolfe-Quintero et a!., 1998) that language
learners in intermediate level tend to use complex grammar structures, in particular,
subordination, which they have learned. TillS is why they seem to be adjacent to group
with level 3. However, the use of complex nominal and verb phrases including non-finite
verb phrases is another story. These gran1mar fOlTns are acquired in a much later
developmental stage than subordination, so language learners in level 2 do not use them as
often as language learners in level 3. That is why they are adjacent to group with level 1
for using particular structure forms. With regard to accuracy, they still make quite high
frequency elTors for this reason because they try to use some of the complex structures that
they have not fully acquired yet. Thus, their accuracy in language development is still as
low as language learners in the beginner level.
To summarize, this study presents Korean EFL college learners' written developmental
stages with regard to syntactic complexity, accuracy, and fluency (CAF). Linguistic
development does not take place at the same rate (Ortega, 1999). As shown in Figures I
and 2, language learners in each proficiency group shows that its members are in
developmental stages somewhere between their L 1 and the target language and still are
engaged in an ongoing process leading to the next stages.
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions 45
v. CONCLUSIONS
The present study attempted to examine some different features of linguistic
development in written productions across language learners' proficiency levels. For
measuring them, 18 subsets of syntactic complexity, accuracy, and fluency (CAF) were
used for written productions. This study also allowed the researcher to investigate EFL
learners' language development and to examine the relationships between language
developmental patterns and writing proficiency. This study has critical advantages over
previous studies. Firstly, it reveals the relationships between Korean EFL writers' linguistic
development and their writing proficiency levels . Many of the previous studies were
conducted across grade, program, or school levels in ESL contexts (Bardovi-Harlig, 1992;
Flahive & Snow, 1980; Henry, 1996; Larsen-Freeman, 1978, 1983; Monroe, 1975) or EFL
contexts (Kawata, 1992; Lu, 2011 , Tomita, 1990). There are relatively few studies that
were conducted according to language learners' proficiency levels, particularly in EFL
settings (Casanave, 1994; Ishikawa, 1995). Moreover, this study has attempted to use
various measures of syntactic complexity, accuracy, and fluency (CAF), respectively to
observe the language learners' developmental stages from multiple angles with various
measures.
Within this fundamental frame, this research yielded several significant findings with
important implications for L2 writing development research. First, regarding the measure
of language development of Korean EFL college learners in terms of syntactic complexity,
accuracy, and fluency (CAF), there are some obvious linguistic features in Wlitten
productions across language proficiency levels. From a one-way ANOVA post hoc test
with the present data and from comparing results with previous studies, level 1 was
confirmed as the begilmer level. They seem to already have passed the stage of acquiring
coordination structure and are passing through the next stage, or subordination. They are
also starting to use subordination structures. In case of the level 2, the members of the
group seem [0 be somewhere in the intermediate level. The results concerning accuracy
and syntactic complexity indices reveal that they did not avoid the use of complex
syntactic structures that they had not yet fully acquired. Moreover, it turns out that level 3
in this study has not reached the advanced level yet. The members of level 3 are also in a
linguistic developmental stage. They have not totally passed the threshold to use reduction
fonns, or particular structures instead of subordination clauses. In the case of coordination
measures, CPIC and CP/T decrease according to the writing levels in a negative direction.
Even though the results are not statistically significant, the decreasing pattern of
coordination is one of the indices of an ESL learner 's linguistic development. Within
subordination measures, they use subordinate clauses more frequently than other
proficiency level groups. According to Lu (2011), CIT may be lower at advanced levels as
46 Eunkyung Hwang
a result of reduction from clauses to phrases. ESL advanced learners tend to use fewer
clauses than those in other proficiency levels because they use more reduction from
clauses to phrases (Lu, 2011; Ortega, 2003; Quintero et aI., 1998). They are in somewhere
at the high-intennediate level.
Next, this study offered evidence confirming previous studies' claims that there are
different patterns across language proficiency. With regard to syntactic complexity,
accuracy, and fluency (CAF), this study supports Ortega (1999)'s claim that development
does not take place at the same rate. Language learners in each proficiency group show
that they are in somewhere in particular language developmental stages and still involved
in an ongoing process leading to the next stages. These results enlighten us with respect to
the linguistic development of Korean learners of English. Students at higher proficiency
levels tend to produce longer clauses and T-units, not as a result of increased use of
coordinate phrases, but as a result of increased use of dependent clauses or complex T
units.
Finally, this research continned that syntactic complexity, accuracy, and fluency (CAF)
are effective predictors across the language learners ' writing proficiency levels. In
comparing three writing proficiency groups, the author found significant differences in
syntactic complexity, accuracy, and fluency (CAF) in 15 out of the 18 ratios used in the
study. As with frequency measure, TW is the best predictor. EFT is the most reliable
variable for accuracy within a limited time set. For syntactic complexity, subordination
variables are most predictable measures. However, these variables can give different
results in other settings such as school levels, grades, or program levels.
This study has pedagogical implications for ESLIEFL instructors to teach grammar and
writing for language learners who are in the various language proficiency levels. In other
words, grammar and writing are not separate as independent learning systems. Language
learners need some sets of knowledge of grammar to write with. In this case, ESLIEFL
instructors need to set reasonable ordering criteria for what grammatical features to teach
and how and when they teach some part of grammar (e.g., coordination structure,
subordination structures, and some reduction phrases). This process could be settled by the
learners' proficiency and enough practices to take them to the next developmental stage.
This study, however, has several limitations. Even though there were 86 students in total,
level 1 consisted of fewer than 30 subjects. Furthennore, all participants were female
university students, so the result might be affected by gender. Lastly, there was no
advanced level to examine the advanced language developmental stages. For this reason, it
is difficult to generalize from its results to typical second/foreign language development.
Thus, future studies should look into significant differences on syntactic complexity,
accuracy, and fluency. For instance, subordination measures on syntactic complexity
revealed significant differences among groups; however, this study did not deeply explore
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions 47
this phenomenon. That is, a deeper exploration would need to identifY significantly
dominant clauses stage such as adverbial, adjective, and nominal clauses in linguistic
development of the written productions in the writing proficiency levels. In addition, the
types of errors across writing proficiency levels could be observed more closely. Finally,
larger groups of participants in various writing proficiency levels are needed to measure
language learners ' interJanguage in language learning contexts.
REFERENCES
Bardovi-Harlig, K. (1992). The relationship of form and meaning: A cross-sectional
study of tense and aspect in the interlanguage of learners of English as a second
language. Applied Psycholinguistics, 13, 253-278.
Casanave, C. P. (1994). Language development in students' journals. Journal of Second
Language Writing, 3(3), 179-201.
Cooper, T. C. (1976). Measuring written syntactic patterns of second language learners of
German. Journal of Educational Research, 69, 176-183.
Crookes, G. (1989). Planning and interlanguage variation. Studies in Second Language
Acquisition, 11, 183-199.
Ellis, R. (1987). Interlanguage variability in narrative discourse: style shifting and use of
the past tense. Studies in Second Language Acquisition, 9, 1-20.
Ellis, R. (1989). Are classroom and naturalistic acquisition the same? Studies in Second
Language Acquisition, 11, 305-328.
Ellis, R., & Yuan, F. (2004). The effects of planning on fluency, complexity, and accuracy
in second language narrative writing. Studies in Second Language Acquisition, 26,
59-84.
Falhive, D. E., & Snow, B. (1980). Measures of syntactic complexity in evaluating ESL
compositions. In J.W. Oller & K. Perkins (Eds.), Research in language testing.
Rowley, MA: Newbury House.
Foster. P., & Skehan, P. (1996). The influence of planning and task type on second
language performance. Studies in Second Language Acquisition, 18, 299-323.
Hammerly, H. (1991). Fluency and accuracy: Toward balance in language teaching and
learning. Philadelphia: Multilingual Matters.
Hirano, K. (1991). The effect of audience on the efficacy of objective measures of EFL
proficiency in Japanese university students. Annual Review of English Language
Education in Japan, 2, 21-30.
Homburgh, T. J. (1984). Holistic evaluation of ESL compositions: Can it be validated
objectively? TESOL Quarterly, 18, 87-1 07.
48 Eunkyung Hwang
Housen, A., & Kuiken, F. (2009). Complexity, accuracy, fluency in second language
acquisition. Applied Linguistics, 30(4), 461-473.
Ishikawa, N. (1995). Objective measurement of low-proficiency EFL narrative writing.
l ournal of Second Language Writing, 4, 51-69.
Ishikawa, N. (2006). Syntactic complexity measures and their relation to oral proficiency
in Japanese as a foreign language, Language Assessment Quarterly, 3(2), 151-169.
Kamimura, T. (1996). Composing in Japanese as a first language and English as a foreign
language: A study of narrative writing. RELC Journal, 27(1), 47-69.
Kim, Y. S. (2011). The effect of cognitive task complexity on a language leamer's written
performance with respect to accuracy, complexity, and fluency. Korean l ournal of
Applied Linguistics, 27(2),285-3 13.
Konnos, J. (2011). Task complexity and linguistic and discourse features of narrative
writing performance. Journal of Second Language Writing, 20, 148-161.
Larsen-Freeman, D. (1978). An ESL index of development. TESOL Quarterly, 12(4), 439-
449.
Larsen-Freeman, D . (1983) . Assessing global second language proficiency. In H. W.
Seliger & M. Long (Eds.), Classroom-oriented research in second language
acquisition (pp. 287-304). Rowley, MA: Newbury House.
Larsen-Freeman, D. (2006). The emergence of complexity, fluency, and accuracy in the
oral and written production offive Chinese learners of Engli sh. Applied Linguistics,
27(4),590-619.
Lee, H., Oh, M., & S, Y. (2007). The effects of planning time on the second language
perfonnance in a narrative task. English Teaching, 62(1), 105-120.
Lu, X. (2010). Automatic analysis of syntactic complexity in second language writing.
International lournal of Corpus Linguistics, 15(4), 474-496.
Lu, X . (2011) . A corpus-based evaluation of syntactic complexity measures as indices of
college-level ESL writers' language development. TESOL Quarterly, 45(1), 36-
62.
Monaghan, W., & Bridgeman, B. (2005). E-rater as a quality control on human scores.
Retrieved February 3, 2012, from the World Wide Web: http://www.ets.orgIMedial
ResearchlpdflRD _ Connections2.pdj
Monroe, J. J. (1975). Measuring and enhancing syntactic fluency in French. The French
Review, 48, 1023-1031.
Norrby, c., & Hakansson, G. (2007). The interaction of complexity and grammatical
processability: The case of Swedish as a foreign language. International Review
of Applied Linguistics in Language Teaching, 45(1),45-68.
Norris, J. M., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis
and quantitative meta-analysis . Language Learning, 50,417-528.
Korean EFL Learners' Language Development across Proficiency Levels in Written Productions 49
Norris, 1. M. , & Ortega, L. (2009). Towards an organic approach to investigating CM in
instructed SLA: The case of complexity. Applied Linguistics, 30(4), 555-578.
Ortega, L. (1999). Planning and focus on form in L2 oral performance. Studies in Second
Language Acquisition, 21, 109-148.
Ortega, L. (2003). Syntactic complexity measures and their relationship to L2 proficiency:
A research synthesis of college-level L2 writing. Applied Linguistics, 24(4),492-
518.
Perkins, K. (1980). Using objective methods of attained writing proficiency to
discriminate among holistic evaluations. TESOL Quarterly, 14,61-67.
Perkins, K., & Leahy, R. (1980). Using objective measures of composition to compare
native and non-native compositions. In R. Silverstein (Ed.), Occasional Papers
in Linguistics, No.6 (pp. 306-316). Carbondale: Southern Illinois University.
Polio, C. G. (1997). Measures of linguistic accuracy in second language writing research.
Language Learning, 47(1), 101-143.
Robinson, P. (1995).Task complexity, and second language narrative discourse. Language
Learning, 45,99-1 40.
Robinson, P. (2001). Task complexity, task difficulty, and task production: Exploring
interactions in a componential framework. Applied Linguistics, 22(1),27-57.
Sasaki, M. (2000). Toward an empirical model of EFL writing processes: An exploratory
study. Journal o/Second Language Writing, 9(3),259-291.
Schmidt, R. (1992). Psychological mechanisms underlying second language fluency.
Studies in Second Language Acquisition, 14, 357-385.
Scott, M. S., & Tucker, G. R. (1974). Error analysis and English-language strategies of
Arab students. Language Learning, 24,69-97.
Sehnker, L. (1972). Interlanguage, IRAL, 10(3), 209-231.
Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford University
Press.
Skehan, P. (2009). Modeling second language performance: Integrating complexity,
accuracy, fluency, and lexis. Applied Linguistics, 30(4), 510-532.
Skehan, P., & Foster, P. (1999). Task type and task processing conditions as influences on
foreign language perfonnance. Language Teaching Research, 1, 185-21 2.
Song, 1. W. (2005a). Task-processing conditions as influences on spoken language. English
Teaching, 60(3), 117-137.
Song, 1. W. (2005b). The spoken performance of advanced learners in a narrative task.
Foreign Language Education, 12(4), 85-106.
Torras, M. R., & Celaya, M. L. (2001). Age-related differences in the development of
written production: An empirical study of EFL school learners. International
Journalo/English, 1(2), 103-126.
50 Eunkyung Hwang
Wolfe-Quintero, K., Inagaki, S., & Kim, H.Y. (1998). Second language development in
writing: Measures of fluency, accuracy & complexity. Honolulu: University of
Hawaii Press.
Applicable levels: university level
Key words: writing proficiency, L2 / EFL writing development, L2 / EFL language development, inter language,
Complexity, Accuracy, Fluency (CAF)
Eunkyung Hwang
Dept. of English Language and Literature
Sookmyung Women 's University
Cheongpa-dong, Younsan-gu
Seoul, 140-742, Korea
Fax: (02) 710-9380
Email: [email protected]
Received in June, 2012
Reviewed in July, 2012
Revised version received in August, 2012
Top Related