ERIC EJ1152481: Student Use of Concordancers for Grammar Error Correction

2013·0.13 MB·English

by  ERIC

Student Use of Concordancers
for Grammar Error Correction

Jaci Mull, Research Institute of English Language Education
Susan Conrad, Portland State University
Emmy and Maria, two intermediate- investigate what upper-intermediate level
level ESL students, are given a peer essay ESL students with little prior training could
and asked to correct five errors highlighted do with a concordancer when correcting
in it. When they discuss the error, They can grammar errors. In this article, we first pro-
save money for better future, they can’t vide some background information about
agree on what to do. Emmy believes better student use of concordancers. We then de-
should be changed to best. Maria believes scribe the methods and results of the study,
they need to insert the before better. At this and discuss suggestions for future research
point, they reach for a concordancer and and how to teach concordancing to students.
search a corpus. They do a search for best
future and better future with no results. Then Background
Emmy suggests they search for better alone,
and they find many samples. Emmy notes Concordancers are software pro-
that several samples, including a better grams that, when connected to a corpus, al-
chance and a better understanding, use the low users to search for all occurrences of a
indefinite article. She suggests, "Put an arti- word or sequences of words. Early descrip-
cle in front of better. So, ‘a better’." The pair tions of the potential for student concordanc-
corrects the error and moves on to the next. ing were given by Tim Johns, who called it
Data Driven Learning (DDL). Johns (1991)
This scenario shows the potential explains DDL as an investigation allowing
power of students using concordancers to students to form their own questions about
correct written errors. In the example above, the language, ask for authentic language
the students would have falsely corrected the samples via a concordancer, and use those
error had they been editing the paper with language samples to help them find an an-
only their prior knowledge. Instead, the pair swer to their question. DDL relies on the stu-
conducted their own small research study of dents to form their own generalizations
English and noticed a gap between their hy- about the language once they are presented
pothesized corrections and the target lan- with samples of the target language. Johns
guage. They then used the authentic target strongly connects DDL to induction, in
language samples they had found to establish which the language “rule” students are trying
an accurate and appropriate language pattern to learn is unknown to them and they must
which they applied to the error in order to discover it for themselves by making gener-
correct it. No teacher was needed to guide alizations from target language samples. The
them through this process; the students were only difference Johns notes between DDL
in full control of their language learning. and traditional induction is that, for DDL,
the teacher need not know exactly what pat-
In fact, Emmy and Maria were real terns await, and may be equally surprised as
students in a small case study conducted to the students by the results. This sense of
Volume 30, 2013 5
mutual exploration gives the learners a sense words. Even so, several studies have found
of autonomy and power in their language that students can make generalizations about
learning. Showing similar enthusiasm for structure as well as lexis from concordance
students as researchers, Tribble and Jones results (Gaskell, 2002; Gaskell and Cobb,
(1990) suggested a variety of ways that 2004; Kennedy and Miceli, 2001 & 2010).
teachers could prepare concordance lines for Of course, when students are engaged in er-
students to analyze or allow students to con- ror correction with a concordancer, the task
duct their own searches for lexical or struc- is slightly different from what Johns (1991)
tural patterns. was describing. With error correction, the
teacher has already identified an error and
While some educators may still be therefore has more control over what the stu-
wary about focusing on form in more com- dent will discover.
municative or task-based classrooms, it is
important to note that DDL has nothing to do In order to guide students through an
with progressive grammar teaching. Johns inductive process and consolidate their
(1991) himself notes that educators should learning, Gaskell (2002) created a set of five
not be trying to teach students a structured steps for students to complete while correct-
set of grammar. Instead, he sees DDL as an ing grammar errors (see Figure 1). These
unparalleled opportunity to explore difficult five steps instruct students to first identify
areas of language that progressive grammar the error they wish to correct, and then to
teaching traditionally avoided. He champi- search with the concordancer for samples of
ons allowing students to find answers to English that show more accurate uses of the
their questions free of teacher intuition, grammatical form. Students are then asked
which may not be based on close analysis of to summarize the pattern they see emerging
data and can be fallible. Johns suggests that from the concordance samples, followed by
through DDL, students are able to come to a stating a grammar rule based on the samples
far more subtle understanding of the target they see and finally correcting the error.
language than through other methods. The hope with guiding steps like those in
Gaskell (2002) is that students will follow
Several studies following Johns the steps in order and by doing so induce the
(1991) and Tribble and Jones (1990) focused structure patterns needed to complete the
special attention on students’ ability to learn editing task. At the end of the project, stu-
lexical patterns with concordancers (Cobb, dents also have a record of the grammar er-
1997; Todd, 2001). Lexical patterns are fair- rors they have corrected. However, up to
ly easy to spot in a Key Word in Context this point there is little evidence about
(KWIC) concordance search. The student whether or not students actually follow this
simply types in one word and sees all in- process when they use a concordancer and
stances of that word lined up
with context on either side. It
1) Example of the error
is easy to see what words as-
2) How the word/structure was used in the concordancer
sociate with the search term.
3) What was learnt from the concordancer
For grammar structure pat-
4) How to fix the error
terns, however, the task is
5) Correction of the error
more challenging. Students
must be aware of parts of
Figure 1: Gaakell (2002) steps for correcting grammar errors
speech and not simply exact
6 ORTESOL Journal
whether or not students can actually general- intermediate English classes in the intensive
ize structure patterns solely from looking at English language program at Portland State
concordance lines. University, had achieved a score of 520 or
higher on the TOEFL, and hoped to enroll in
Research Question the university eventually. All the participants
were comfortable with computers, but none
As teachers, we know that students had used corpora or concordancers for Eng-
do not always complete tasks using the pro- lish language learning previously.
cesses that we anticipate. While we might
look at their end product and assume they The participants completed the pro-
have, for instance, induced patterns in the ject in pairs to encourage verbalizing their
target language, we cannot be certain with- thought process as they worked with the con-
out direct observation. They may, after all, cordancer. All pairs had studied together in
simply be guessing or relying on previous the same class and were familiar with each
knowledge. We were therefore interested to other before the study. The first pair consist-
investigate what students are actually doing ed of Edison, a 27 year-old male student
with the tools we gave them. With this in from South Korea and Y, a 23 year-old fe-
mind, we asked the following questions: male student from Thailand. The second pair
1. How do students with little training consisted of Maria, a 23-year-old female stu-
employ a concordancer to investigate and dent from South Korea, and Emmy, a 36-year
correct errors in written work? -old female student from Thailand (all names
2. To what degree does a worksheet help are pseudonyms). The diverse first language
guide the students through the inductive backgrounds encouraged students to verbal-
process for grammar error correction with ize their ideas in English.
a concordancer?
The study was broken up into four
Methods sessions (see Figure 2). In the first session,
the pairs of students were introduced to a
Four ESL students participated in concordancer and allowed 30 minutes to ex-
this case study. All were enrolled in upper- periment with it using a guiding worksheet.
Session Activity Materials Time
1 Training with concordancer and time to experiment Essay with 10 errors 30 min.
Journal writing Journal 10 min.
2 Interview about session 1 NA 10 min.
Work session: Error correction time with con- Essay with 5 errors 20 min.
cordancer and required worksheet Worksheet (required)
Journal writing Journal 10 min.
3 Interview about session 2 NA 10 min.
Work session: Error correction time with con- Essay with 5 errors 20 min.
cordancer and optional worksheet Worksheet (optional)
Journal writing Journal 10 min.
4 Final Interview NA 30 min.
Figure 2: Sessions in the study
Volume 30, 2013 7
The training session was kept short because priate for these students, incorrect sentences
average classrooms do not have an abun- were selected from diagnostic essays written
dance of time with which to train students on by students of the same class level from a
new technologies and we wanted to see what previous term and incorporated into the cre-
students could do given modest training. ated essays. In order to be selected for incor-
Students were taught searches only with poration into the new essays, the errors had
words and word sequences, not with gram- to satisfy two criteria. First, they had to fit
matically “tagged” sequences (which make one of five grammatical categories students
grammar searches easier but require more at this level typically struggle with: articles,
advanced training). In the second session, prepositions, tense, clause construction, or
students were briefly interviewed about the agreement. Finally, they had to be searchable
training session and then were given 20 in the corpus. Specifically, by searching for
minutes to edit grammar errors in a prepared words in the highlighted sentences, the stu-
essay using a concordancer. They were also dents had to be able to find at least 10 sam-
given a guiding worksheet to aid with induc- ples that could help them correct the error.
tion. The third session was set up exactly as The training essay included 10 of these er-
the second session, including the brief inter- rors to ensure students had plenty to practice
view and work time, but with the concordancer.
the worksheet was op- They had to determine for Five errors were then
tional. Finally, in the last incorporated into each
themselves what kind of error
session, students met of the work session es-
with the researcher for a they faced. says, and the sentences
lengthier interview. were highlighted. One
representative error from each grammar cat-
Materials egory was included in each work session es-
say. Students were aware that the highlight-
In their ESL class, the students were ed sentences had grammar errors, but aside
working on writing persuasive essays, so we from that, they had to determine for them-
designed the study to ask students to correct selves what kind of error they faced.
errors in similar essays. A corpus was spe-
cially created for them using 80 persuasive Rather than simply handing students
essays written by native speakers of English. the Gaskell (2002) steps as seen in Figure 1,
The essays were found in the free use sec- a guiding worksheet was written to walk stu-
tions of two essay sharing websites: dents through the five steps (Figure 3). It
www.123helpme.com, and posed the steps in the form of directions. For
www.allfreeessays.com. Students used a example, instead of “example of the error”
copy of MonoConc 2.2 (Barlow, 2002) as students were instructed to 1) Choose an er-
their concordancer to search through the es- ror to correct and 2) Write the error below.
says.
Data Collection Instruments and Analysis
To develop the error correction task,
three persuasive essays were created to During the first three sessions, video
match the style taught in upper-intermediate recorders were pointed at the students’ com-
writing classes. One essay was for the train- puter screens to capture the searches they
ing session, and the other two were for the made with the concordancer and to record
work sessions. In order to use errors appro- their voices as they negotiated how to cor-
8 ORTESOL Journal
rect the error and use the concordancer. At thought that some of the errors had obvious
the end of each error correction session, stu- corrections that they could make based en-
dents were asked to write a journal entry tirely on their previous knowledge. Feeling
about their experience using a concordancer. confident, they did not see a need to use the
Throughout the sessions the students were concordancer to confirm their knowledge of
also interviewed about their opinions and to how the structures worked. The fact that
clarify points made in their journals. time was short – only 20 minutes of work
time for correcting 5 errors in each work ses-
For the analysis, we qualitatively de- sion – might also have made them feel they
scribed the process used by each pair for needed to focus on the answers they felt least
each error. We used evidence in the journals sure about. In fact, all of the corrections
and interviews to interpret the processes made without the concordancer were accu-
more fully, and identified common opinions rate (3 for Emmy and Maria, and 6 for Edi-
and attitudes about using the concordancer. son and Y).

Results and Discussion How students employed the concordancer

Each student pair had 10 errors to In the end, a total of 11 errors were
correct (5 in each work session). Both pairs addressed using the concordancer (7 by Em-
Worksheet
Please answer the questions below, in order, while you work on the computer to find corrections
for the error you are investigating.

1. Choose an error from the essay.

2. Write the sentence containing the original error below:
_________________________________________________________
_________________________________________________________

3. Use MonoConc (the computer program) to find examples of better grammar to correct the
error in the essay.

4. Look for a grammar pattern in the examples you find. Write down three examples you see
that have a grammar pattern similar to the error you are correcting:
A. _________________________________________________________
B.._________________________________________________________
C._________________________________________________________

5. Based on the grammar pattern you have found, think of a rule to explain why these exam-
ples are correct (and the original error is incorrect).

Rule: __________________________________________________________
__________________________________________________________

6. Use the rule you just wrote to correct the original error that was in the essay. Write the
corrected sentence:

Corrected Sentence: __________________________________________________________

Figure 3: Student worksheet
Volume 30, 2013 9
my and Maria and 4 by Edison and Y). With a new generalization using evidence from
regards to the first question, “How do stu- their concordancing. Just like Maria, Y be-
dents with little training employ a con- lieved the correction should be they can save
cordancer to investigate and correct errors in money for the better future. When she found
written work?” we found only one instance no results to support her answer, she aban-
of a full induction process as set out by the doned her hypothesis, but she was unable to
worksheet, which asked students to look at conduct a new search and generate a new
examples before generating a rule about the hypothesis. Her partner Edison suggested a
structure. Instead, students used shorter pro- completely different but inappropriate cor-
cesses that focused on hypotheses they al- rection (discussed in point 3 below). Y did
ready held, with the pairs having mixed suc- not accept this correction, but she offered no
cess when they found a lack of evidence for alternative. This pair appeared unable to fig-
their hypotheses. While analyzing the stu- ure out how to search for results that would
dents’ processes, we found the following generate a new hypothesis rather than just
three points especially important. confirm what they hypothesized. They may
also have felt constrained by the short
1. A full induction process amount of time for the
Once they realized there
occurred for one pair but task. In any case, this
not the other. was no support for either of pair’s inductive process
their theories, they had to never fully developed, and
As described in they never successfully
reconsider.
the opening of this article, corrected the error.
Emmy and Maria began
one of their concordance searches to correct 2. Students used the concordancer to support
the error they can save money for better fu- or challenge their hypotheses from previous
ture with two incorrect theories on how to knowledge.
change the error. Once they realized there
was no support for either of their theories, The most common way that these
they had to reconsider. In an interview Em- students used the concordancer was to con-
my admitted she first doubted the evidence firm patterns that they expected but were not
from the concordancer, but, knowing the entirely confident about. For every error
corpus was composed of all native speaker correction attempted, students started with
essays, she decided to trust the results and their own hypothesis (or competing hypothe-
look for another solution. After searching for ses) about the correction and then consulted
better alone, she was able to spot the regular the concordancer. In most cases (7 out of
use of an indefinite article and create a gen- 11), students found support for the first or
eralization from that. second hypothesis they checked and could
then confidently correct the errors.
We were surprised that there was on-
ly this one case of induction. It occurred Two of the other four cases where
when Emmy and Maria acknowledged the students used the concordancer are covered
lack of evidence for their previous hypothe- by points 1 and 3 (concerning full induction
ses and had no other hypothesis to check, but and generalizations that change meaning).
found a useful search to gather new evi- The final two cases add more support to the
dence. Edison and Y found the same error interpretation that students can find it diffi-
challenging, but they were not able to create cult to search with a concordancer when they
10 ORTESOL Journal
do not already have a clear hypotheses to guide the students through grammar error
test. For example, Emmy and Maria could correction with a concordancer?” In general,
not figure out a useful search when they the answer appears to be that the worksheet
were working on a structure that had an was not helpful. Even when the worksheet
omitted relative pronoun: if you take two was required, the students usually completed
different people have the same commitment. the steps in a different order or back-filled
They did not know what the error was and the worksheet after they were satisfied they
did four searches that focused on the verb knew how to correct the error. During the
take, which did not help them. Eventually one full process of induction, Emmy and
Emmy remembered the need for a relative Maria had the guiding worksheet with them,
pronoun, based solely on her previous but - even though they eventually went
knowledge, not on anything she saw in the through all the steps in the order listed - they
searches. Both students then felt confident didn’t refer to the worksheet until after they
about the correction and did not feel the need had analyzed the concordance lines and de-
to confirm it. cided on a correction. During the third ses-
sion, when students were free to choose
3. Students could make generalizations that whether they used the worksheet or not, both
were grammatically accu- pairs ignored it com-
The students were keen to use
rate but changed the orig- pletely.
inal meaning. the concordancer in this way.
The usefulness
For the phrase for of a worksheet, of
better future, Edison decided the best way to course, depends on the design of the work-
correct the error was to use the chunk in the sheet. The worksheet in this study did pro-
future, which he had heard many times. He vide a structure for the training session and
did a search and found plenty of evidence in forced the students to record what they had
concordance lines to support use of this learned. In trying to encourage a full induc-
phrase. He therefore corrected the sentence tive process, however, the worksheet did not
to read they can save money in the future. allow enough of a role for students’ use of
While this sentence is grammatically accu- their previous knowledge. It did not allow
rate, the meaning had changed significantly them a quick way to state and confirm their
from the original text. This example illus- hypotheses when they were correct. At the
trates one of the limitations of concordanc- same time, it did not guide students through
ing: if a reader does not already understand what appeared to be the most challenging
meaning differences of similar wordings, it aspect of concordancing: determining new
is difficult to see the difference without more searches to try to identify patterns when they
extensive analysis of the context in the cor- didn’t already have a clear hypothesis to test.
pus. For students at this level, teacher input If a guiding worksheet is to be used in the
would likely be a more efficient means of future, it seems more appropriate to follow
helping the student understand the meaning the students’ natural inclinations. Rather
change than corpus analysis would. than start with the concordance lines, the re-
vised worksheet could, like many other cor-
The guiding worksheet pus-based materials, ask students to use pre-
vious knowledge to state a hypothesis, and
Our second research question asked, then prompt students to search the con-
“To what degree does a worksheet help cordancer for support (or a lack of support)
Volume 30, 2013 11
for the hypothesis. Students could revise In our study, the students themselves
their hypothesis as many times as they want- suggested that they might have been better
ed until they had found appropriate support. able to induce language patterns had they
At that point they could state a rule and give been investigating the corpus freely as op-
a final correction for their error. If the stu- posed to correcting errors. Both pairs of stu-
dents ran out of their own hypotheses to try, dents came to this conclusion independently
the worksheet could guide them through and with no prompting from the researchers.
steps for new searches that might reveal pat- Interestingly, Edison and Y, who never
terns they had not thought about previously. showed a tendency towards induction both
agreed that the concordancer would be more
Student enthusiasm useful when exploring it openly rather than
correcting grammar errors. Both pairs also
This study was not set up to investi- suggested during interviews that they
gate student use of concordancers during thought the concordancer would be best used
writing production, but the students in the as a reference at home where students have
study were keen to use the concordancer in more time to explore and feel more relaxed.
this way, even after only 90 minutes with the It is interesting to note that by following the
concordancer spread over three days. In in- students’ suggestions, teachers would be
terviews, all four students compared the con- able to use their valuable class time to focus
cordancer to other reference tools they al- on their regular lesson plans, while still al-
ready used at home and saw the concordanc- lowing students the chance for autonomous
er filling a gap in their resources. They trust- language research outside of class.
ed the concordancer and corpus because they
saw native speaker examples, and they felt it Conclusion
could give them ideas for language patterns
they might not find in a standard dictionary Concordancing and error correction
search.
The results of this study were mixed
Bernardini (1996) argues student with respect to using concordancing for error
concordancing should focus more on discov- correction. In the majority of cases, students
ery learning in which students search a target felt so confident correcting the given errors
language corpus openly in a more explorato- that they did not use the concordancer to find
ry fashion. Giving students concordancers as supporting evidence. The chosen errors thus
exploratory tools to inform their writing is appeared to be too easy for the students in
not unprecedented. Kennedy and Miceli the study. Concordancers are most useful
(2010) have produced two studies investigat- when the students need more evidence
ing just that and have found that students about how a structure works – not when stu-
were able to create personal strategies for dents just need more time to apply their de-
using the concordancer. Their students used clarative knowledge to correct an error – but
the concordancer to hunt openly for new this can be a difficult judgment to make
words and expressions that they could use in when planning an activity for numerous stu-
their writing, to generalize lexical and struc- dents. At the same time, when one pair of
ture patterns when they had a specific ques- students did have the opportunity to use new
tion in mind, and to find target language evidence to revise their inaccurate hypothe-
equivalents for what they wanted to say. ses, they did not use the concordancer effec-
tively.
12 ORTESOL Journal
Despite the shortcomings found in this cordancer, the more ideas they reported for
study, however, the students also demonstrated applying it to their regular studies and doing
some productive uses of concordancing that investigations when they had more time at
facilitated their autonomy as language learners. home. Thus, even with minimal training,
They used the concordancer to find evidence concordancing might be a useful activity for
that confirmed their hypotheses about how to some students. Even instructors who do not
correct errors. In some cases, they also recog- want to spend much time with concordanc-
nized evidence that their hypotheses were ing in the classroom might consider intro-
wrong, and in one case, induced a new, accu- ducing their students to concordancing at
rate generalization that allowed them to correct least as a reference tool. Students can then
an error. judge for themselves whether it is compel-
ling enough to pursue at home or in extra
Student ability to use the concordancer study time in a computer lab.

Many studies have argued in favor of Introducing students to concordancing
gradual, structured training activities in the
classroom before Introducing stu-
allowing students to The students demonstrated some dents to concordancing
work more inde- requires access to two
productive uses of concordancing
pendently with a things: a corpus and a
concordancer. For that facilitated their autonomy as concordancer. In this
example, Kennedy language learners. study the researchers
and Miceli (2010) chose to create a cor-
created an pus tailored to the type
“apprenticeship” program to train their students of essays the class was working on, and if
in concordancing and dedicated roughly 30% teachers are able to do the same, it appears to
of their writing class, to it. Chambers (2005) be quite beneficial. Simply being able to tell
argued against training students as if they were students “this is a collection of writing just
future corpus linguists, but still had students in like what you are trying to do” seemed to
her study complete 9 hours of corpus linguis- add a level of credibility and authority to the
tics training. corpus in the students’ eyes. There are now
a number of free or relatively inexpensive
Realizing that the average instructor is concordancers that can be used with your
not in a position to squeeze hours of concord- own corpus (e.g. AntConc is available free at
ancing training into their already tight curricu- www.antlab.sci.waseda.ac.jp, MonoConc is
la, this study took a different tack and exam- available for order at www.athel.com).
ined how students used a concordancer after However, teachers need not make their own
only 30 minutes of introduction to basic word corpus or invest in a concordancer. The
searches. While the students in this study ex- Corpus of Contemporary American English
pressed some frustration with the software ini- (corpus.byu.edu/coca) contains 450 million
tially, and would likely have benefitted from words of American English with a free
more training focused on how to design search- searchable interface. Searches are possible
es, they did not seem discouraged by their lack in categories such as academic texts, news-
of training. They used the concordancer when papers, and popular magazines – thus allow-
they felt unsure about how to correct errors. In ing students to focus on a more specific type
addition, the more the students used the con- of writing than, say, searching for a common
Volume 30, 2013 13
word using Google. The corpus also includes rors? Unpublished master’s thesis, Uni-
grammatical tags if teachers want to intro- versite du Quebec a Montreal, Quebec.
duce students to more advanced grammar- Gaskell, D., & Cobb, T. (2004). Can learners
related searches. use concordance feedback for writing
Student use of concordancers has errors. System 32(3), 301-319.
probably not yet lived up to the ideal Johns Johns, T. (1991). Should you be persuaded:
envisioned when he first wrote about DDL two samples of data-driven learning ma-
in the late 1980s and early 1990s. Much re- terials. English Language Research
search is still needed into the best ways to Journal 4, 1-16.
use concordancers, both when applied to er- Kennedy, C., & Miceli, T. (2001). An evalu-
ror correction and during language creation. ation of intermediate students' approach-
Longitudinal research into the development es to corpus investigation. Language
of student skills with concordancers is also Learning & Technology, 5(3), 77-90.
needed. It is still largely unknown whether Kennedy, C., & Miceli, T. (2010). Corpus-
students will slowly develop strategies on assisted creative writing: introducing in-
their own over time, or whether training is termediate Italian learners to a corpus as
the only way to improve students’ skills. a reference resource. Language Learning
Even while research continues, however, & Technology 14(1), 28-44.
there is reason to believe that students can Todd, R.W. (2001). Induction from self-
get some benefit from consulting a con- selected concordances and self-
cordancer, whether they fully induce patterns correction. System, 29(1), 91-102.
from the results or simply confirm what they Tribble, C., & Jones, G. (1990). Concord-
already hypothesize. ances in the classroom: A resource book
for teachers. Harlow, England: Long-
References man.
Barlow, M. (2002). MonoConc Pro (Version
2.2) [Computer software]. Houston, TX: Jacqueline Mull is the Vice Dean of the Re-
Athelstan. search Institute of English Language Educa-
Bernardini, S. 1(996). Corpora in the class- tion in Kobe, Japan and is a full-time in-
room: An overview and some reflections structor at Kobe Shukugawa Gakuin Univer-
on future developments. In J.McH Sin- sity. She is a native of Oregon and graduat-
clair (Ed.), How to use corpora in lan- ed from Portland State University with an
guage teaching (pp. 15-36). Amsterdam: M.A. in TESOL.
Benjamins.
Chambers, A. (2005). Integrating corpus Susan Conrad is a Professor of Applied Lin-
consultation in language studies. Lan- guistics at Portland State University, where
guage Learning and Technology 9(2), she regularly teaches the class Corpus Lin-
111-125. guistics in Language Teaching. She is co-
Cobb, T. (1997). Is there any measurable author of several books that use corpus lin-
learning from hands-on concordancing? guistics, including Real Grammar: A Corpus
System 25(3), 301–315. -Based Approach to English (Pearson,
Gaskell, D. (2002). Can and will elementary 2009).
ESL learners use concordance infor-
mation to correct their own written er-
14 ORTESOL Journal