Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Students often claim that in timed writing tasks they have no opportunity to go
back and reread what they have written so that they can revise and edit. This
raises the questions: 1) When given more time specifically for revision, can
and do second language writers make sentence-level’ changes in their essays?
and 2) To what extent is their ability to edit during revision limited by their cur-
rent developmental level (i.e., their current interlanguage system)? The answers
to these questions have implications for three (not mutually exclusive) areas:
second language writing pedagogy, second language writing assessment, and
second language acquisition (SLA). This paper examines the extent to which
ESL learners can, in the absence of feedback, improve the linguistic accuracy
of their writing and whether or not additional instruction on editing helps this
process.
Direct all correspondence to: Charlene Polio, Michigan State University, A714 Wells Hall, East
Lansing, MI 48823.1027 <polio@pilot.msu.edu>
43
POLIO, FLECK & LEDER
The current second language writing pedagogy literature has moved away from
the goal of having students write error-free sentences on first drafts, arguing,
probably appropriately, that such a concern for correctness may stifle the genera-
tion of ideas. Such arguments in second language pedagogy texts where empiri-
cal support is provided (e.g., Ramirez, 1995) are based in part on Zamel’s (1983)
study of the behavior of unskilled and skilled ESL writers and similar studies of
Ll writers. Whether or not it is helpful to teach unskilled writers techniques
favored by skilled writers, expecting error-free first drafts is clearly unrealistic.
Even during revision2 activities such as conferences and peer review, students
are generally asked to focus on global revisions and not linguistic corrections.
(See, for example, peer response worksheets in Reid, 1993 and conferencing
worksheets in Hedge, 1988.) Editing tasks seem best left until later. However,
while editing is often postponed, second language writing textbooks generally
include sections on editing and proofreading (e.g., Bates, 1993; Leki, 1989;
Raimes, 1987; Tickle, 1996), and texts specifically for the editing stage have
been published (e.g., Fox, 1992; Lane & Lange, 1993). Other literature gives
teachers advice on how to teach editing (e.g., Ferris, 1995: Hedge, 1988). What is
rarely addressed in the writing literature is 1) whether or not some of these edit-
ing tasks and materials found in textbooks and the pedagogy literature actually
result in writing that has fewer errors and 2) the extent to which learners can edit
their own writing without any feedback.
While few studies have specifically examined the effects of teaching students
how to edit, several studies have examined the effects of feedback and error cor-
rection on learners’ linguistic accuracy3 in second language writing over time. For
example, Robb, Ross and Shortreed (1986) studied the effect of different types of
feedback on second language writing over the course of a year but found no sig-
nificant difference on learners’ essays with regard to linguistic accuracy. Kepner
(1991) studied second-year university Spanish students. Types of feedback on
journals (message-related or surface-error correction) as well as verbal ability
were the independent variables. Students’ journals were examined for higher-
level propositions4 and surface-level errors. Students receiving message-related
feedback had significantly more higher-level propositions, but there was no dif-
ference between the groups in terms of surface-level errors. Several more studies
were discussed in a recent, comprehensive review article by Truscott (1996). He
concluded that empirical research on grammatical feedback has shown that error
correction is not beneficial to second language writing.5
Frantzen (1995) examined not only error correction, but also supplemental
grammar instruction. Participants in her study were two groups of students leam-
ing Spanish, one of whom received additional grammar instruction which
ESSAY REVISIONS 45
To summarize the findings most relevant to this study, the literature seems to
indicate that grammatical feedback and additional grammar instruction are prob-
ably not, in a practical sense, helpful to second language writers in the long-term.
Whether or not learners can correct their own errors in the short-term, without
feedback, is still an open question.
Finally, whether or not students have the ability to correct their own language
should be of interest to researchers in second language acquisition. On one hand,
the ultimate concern of much SLA research is development, that is, the long-term
effects of tasks (e.g., types of interaction, pedagogical activities) on acquisition.
Thus, whether or not students can correct their own language over the short term
may not be of interest. On the other hand, the written language provides a way
for language learners to stop and look at their own language and focus on form,
ESSAY REVISIONS 47
The extent to which learners can correct their own language has been
addressed to some extent in the SLA literature. Many studies have shown that
learners modify their speech upon a signal of noncomprehension (implicit feed-
back) in spoken language and give examples of such modifications (e.g., Kasper,
1985; Gass & Varonis, 1989; Pica, Holliday, Lewis, & Morgenthaler, 1989), but-
generally, these modifications by nonnative speakers are not quantified for target-
like accuracy. Pica (1988) actually examined whether modifications were more
target-like and found that they were.
48 POLIO. FLECK & LEDER
One study that specifically examined learners reviewing their own language
without feedback is Gass (1983). She found that when students were given a sam-
ple of their own ungrammatical sentences from their essays (mixed in with other
sentences), they were able to identify their own ungrammatical sentences as
ungrammatical about 68 percent of the time. An intermediate and advanced group
were then able to correct those sentences 2 1.2 percent and 40 percent of the time
respectively, by, she argued, applying their explicit knowledge.
Swain and Lapkin (1995) examined students composing and editing their writ-
ing. Working in an SLA framework (as opposed to writing pedagogy), they exam-
ined whether or not adolescent learners of French as a second language were
aware of gaps in their linguistic knowledge and, if so, how they tried to solve their
linguistic problems. They found that students were often aware of gaps in their
knowledge, even without external feedback. Swain and Lapkin provided a taxon-
omy of language-related episodes found in the think-aloud protocols collected.
They did not specifically quantify how many errors were found or the kinds of
changes made; they were more interested in the thought processes of the learners.
However, their study, we believe, suggests the possible benefit of having learners
review their own written language.
Another potentially relevant study in the realm of SLA is Crookes (1989) who
investigated the effect of planning on various aspects of second language learner
speech. Crookes found that learners who were given planning time did not pro-
duce more linguistically accurate utterances. Apparently, the opportunity to think
about, and presumably rehearse what one is going to say, does not result in fewer
grammatical errors. (Crookes did, however, find significant differences on other
measures such as syntactic complexity.) One may conclude that, given planning
time, learners do not, or cannot, focus on producing more accurate language. We
suspect that reviewing written language, however, gives learners a chance to
focus on the accuracy of their language and remedy the mismatch between
explicit knowledge and production.
This study examines whether or not ESL students can revise for grammatical
errors when given additional time to do so. It was inspired by Kroll’s (1990)
study which did not find a significant difference in linguistic accuracy between
essays written under a time constraint and those not written under a time con-
straint. She made the point that we do not really know what the students in her
study were doing with the extra time, if they used it all, or if they were aware of
the various stages in producing a good essay:
ESSAY REVISIONS 49
It might be suggested that perhaps the students who wrote the essays analyzed here
did not know enough about what constitutes good writing or about the writing pro-
cess. Such students would attack every task with the same lack of skill regardless of
the conditions they were writing under. In fact, as noted previously, no class instruc-
tion was directed toward assisting the subjects in this study to maximize their extra
time available to them for preparing the essays written outside of class. Under such
conditions, their performance might have been accidental rather than calculated,
haphazard rather than planned. In fact, they might have spent less time on their out-
of-class papers. Without any mental formulation of what constitutes good writing or
an awareness of the steps involved in producing it, students cannot know how to
proceed in the task of writing and time could not buy them anything. (p. 152)
Because Kroll concluded that it is not certain how much time students were
using for the at-home essay, we decided to examine the effect of additional time in
a more controlled manner, giving learners 30 minutes for writing a draft and then
60 minutes to revise. Second, because she also stated that students may not have
been aware of how to use the extra time (e.g., saving time for revision), we also
wanted to have time specifically designated for revision after students had fin-
ished a preliminary version of their essay. And third, because she said that stu-
dents may not have been instructed in how to use their extra time, we planned to
examine the effects of editing instruction on revision.
HYPOTHESES
The null hypotheses and the predictions based on the literature are as follows:
We expected to reject this hypothesis. It may seem obvious that students’ lan-
guage would improve because of both instruction and exposure to input in a sec-
ond language environment, but it was necessary to test this hypothesis in order to
validate the measure of linguistic accuracy. If no difference from the beginning to
end of the semester was found, it might have been necessary to reconsider the
measure used.
The findings by Kroll(l990) and Crookes (1989) suggest that learners may not
be able to improve their linguistic accuracy with additional time, but two studies
reviewed above (Gass, 1983; Makino, 1993) indicated that students may be able
50 POLIO, FLECK & LEDER
to correct their errors when told explicitly to do so and when given specific sen-
tences to focus on. While this study asked students to revise their entire essay and
not just individual sentences, it also explicitly gave them time in which to make
improvements. Furthermore, Swain and Lapkin’s (1995) study suggests that
learners can notice problems without feedback, a first step in correcting the prob-
lem. Thus, we predicted that this hypothesis would be rejected.
Because the writing classes from which the subjects of this study are drawn
require students to write multiple drafts, we suspected that these students would
be able to revise more for language after instruction, and thus, this hypothesis
would be rejected.
METHOD
Participants
The participants in this study were 6.5 undergraduate and graduate ESL students
enrolled in an English for academic purposes composition course at Michigan
State University. All except 2 students were concurrently enrolled in academic
courses in their fields. These students had not met the university’s minimum
requirement on the writing portion of the university’s placement exam, which
they would have been exempt from had they acheived a TOEFL score of greater
ESSAY REVISIONS 51
TABLE 1
Data collection schedule
65 subjects
260 essay
than 5.50. The majority of students were Asian with some students from other
continents. A more detailed profile of the participants can be found in Appendix
A. The students were enrolled in one of four classes, each with a different
teacher. Essays only from those students who completed all four of the writing
tasks and who agreed to participate in the study were used.
Data Collection
Treatment
Students were asked if they wanted to participate in the study. They were told
that if they agreed, they had the chance of being assigned to the experimental
group where they would get additional grammar exercises and feedback on their
journals. About 90 percent of the students agreed to participate. Half were ran-
domly assigned to the experimental group.
Those in the control group wrote four journal entries each week for seven
weeks and received no feedback. Instead of this, the experimental group did the
52 POLIO, FLECK Sz LEDER
TABLE 2
Intercoder reliabilities for EFT/TT and WEFT/TW
Analysis
TABLE 3
Results of EFT/TT measure
TABLE 4
Results of WEFT/TW measure
Each essay was coded blindly by two of three coders. Any essay that differed
in EFT/TT by more than .10 was given to a third coder, and that coder’s score was
used. The intercoder reliabilities for the two measures are given in Table 2.
The results of the EFT/TT measure can be found in Table 3 and the results of the
WEFT/TW can be found in Table 4. On both measures, the subjects’ mean scores
increased in the following order: 30-minute pretest, 60-minute pretest, 30-minute
posttest, 60-minute posttest. A mixed-design three-way analysis of variance was
performed, between subjects for the experimental/control (exp/con) groups, and
repeated measures for the time (30/60) and pretest/posttest (pre/pos) variables.
Results were the same for both measures and can be found in Table 5 and Table 6.
Main effects were found for 30160 and prelpos, but not explcon. No interaction
effects were found among any of the variables. Thus, null hypotheses 1 and 2 were
rejected and null hypotheses 3 and 4 were not. Each is discussed in turn below.
This hypothesis was rejected. The fact that the two measures indicated a signif-
icant improvement from pre- to posttest is important because it shows that they
were able to measure some kind of linguistic change that occurred after 15 weeks
54 POLIO, FLECK & LEDER
TABLE 5
Results of ANOVA for WEFTRW
Within subject
prelpost .I7 1 .17 11.11 .001*
pre/post by E/C .oo 1 .oo .oo ,996
within group .94 63 .Ol
TABLE 6
Results of ANOVA for EFT/TT
Within subject
pre/post .22 1 .22 13.29 ,001s
prelpost by E/C .oo 1 .OO .oo .984
within group 1.02 63 .02
This hypothesis was also rejected. The ratios of EFT/TT and WEFT/TW are
significantly better on the 60-minute revisions than on the 30-minute essays. This
might mean that the students are able to edit their essays, to some extent, for sen-
tence-level grammar and vocabulary errors. One possible conclusion is that stu-
dents edited their original sentences. Another possibility is that they wrote
substantially different essays and that the extra time simply gave them a chance to
compose more linguistically accurate sentences in the first place.
If we examine the essays, we see that there is a range of revision behaviors,
from strict sentence-level editing to a complete reformulation of the essay. Some
students’ 30-minute essays and 60-minute revisions were so similar that they
could be compared side by side and corresponding sentences could be noted;
grammatical and lexical changes could be qualified and quantified. Two exam-
ples of this kind of revision behavior can be found in essays A30 and A60, and
B30 and B60 shown in Figures 1 and 2 respectively.
In other cases, students wrote completely different essays as in essays C30 and
C60 shown in Figure 3. Obviously, these students represent a continuum of revi-
sion behaviors, from almost exclusively sentence-level editing to writing a sub-
stantially different essay. In essays A30 and A60, we can actually see sentence-
level changes. Examining essays C30 and C60, we do not know, without some
kind of stimulated recall, how the writer used his or her extra time with regard to
language issues. However, we can say that overall, extra time, whether used to
focus on sentence-level changes or content, results in more linguistically accurate
essays.
Figure 1. Example uf original and revised essay with mostly sentence-level revisions
830 I360
1. Thereare many holidaysin my country. 1. There are many holidays in my countq, for instance, Chinese New Year,
Chinese Moon Fe.stival,L%agonBoat Festival,and so on.
2. Among those Chinese New Year is the most importantfestivaJin my 2. Ammtg these, Chiaese New Year is the most importantfestival in my
cmmtry. CouIItQ‘
Thus, thereare five days off in Chinese New Year.
3. Chinese New Year is on Januaryfirst of lunarcalender. 3. Chinese New Year is on Januaryfirst of the lunarcalendar
4, Most of festivals in Taiwan are counted by lunar claender because 4. In Taiwan, most big f&val are countedby the hmarcalendarbecause we
Taiwan is the countrychangedfrom agricultmeso&&y, have a five-thousand-yearhistoryand our rxmatrychangd from agricultum
society.
5. Therefore,we use lunar calendarmoreoften than solar calendar. 5, Therefore,we use lunar calendarmorethan solar calendar.
There-are five days off from the Eve of Chinese New Year.
It~tbatn~needstogoforworkufltilthcfiAhdayofChilles
Year.
6. In the Eve of Chinese New Year everyone should get together with his 6. In the Eve of Chinese New Year, everyoneshould get togetherwith his
family and have a big dinner. family and have a big dinner.
7. Besides, people has ta pay his f&n& a courtesy call on New Year‘s Day 7. Besides, people have to pay their tiends a coarteq call on Chinese Now 5
shortly thereafter, usually with a present. Year shortly thereafter, usually with a present. x
8. In addition, people always worships his ancestors on the same day. 8. In addition, people always worship their ancesWs on the same day. s
9. What‘s more: children will be given some money in red envelop by 9. What’s more? ChMren will be given money in red envelop by parents or 8
parents or adults. adults.
It sign%es luc@.
Thus, this is the happiest holiday for children.
LO.On the second day of New Year women who have married should come lo. On the second day of Chinese New Year, womea who have ma&d
back to their parent‘s home to reune. should come back to their parents’ home to wane.
I 1, After this day, people usually has a plan to travel to relax themselves. I 1. AtIer this day, people always travel everywhere to rehtx themselves.
12. in conclusion, Chinese New Year is the biggest baliday in my country. 12. Generally speaking, there will be a lot of activities to celebmte during
these days becam Chinese New Year is the biggest in Taiwan.
13. I hope I can come back to my country to spend this big holiday with 13. Finally, I hope I can come back to my cxnmttyto spend this big
my family holiday with my family next year.
Figure 2. Example of original and revised essay with mostly sentence-level revisions
tl:
c30 CT60
MSWis a big and completeduniversity, which rxntents me in most While MSU is a comfortable and beautiful envirement to live and
aspectbut still has som shortcoming study here, it still has two serious shortcomingto be improved. One is parking
There is a river, &led “Grand river”, cross the middle ofthe campus problem, and another*housmg on campus
from east to we& which is the sepamting line of two di&rent style of campus The parking lots an campus are obviously insoBcient Only faculty
building groups. The buildings on northern campus are usually classic style, and staffcan park their cars au northern campus, and only TA, RA and above
while the buildings on southernrzimpus~more modern style. can park near the shaw LANE. If you are just a student, neither TA, nor RA.,
The athehte ~u~prn~t on campus are sufKcient. There are many and you go to s&o1 by driving, you are required,as far as I know, to park your
gym, like 1.U East,IN. West,1-M. Circle, and one stadiom on campus,where car in the parldng lot which is locatedat the very southern end of campus. And
provide the field and ~uip~n~ to play various kind ball games and swim. then you can transfer to a campusbus to get into the main domain of campus,
&sides, there are also many autdoorfields of basketball, tennis and volIeybal1. Liing in campus is also not very amazing. Pii the cable TV system
Commutingin campusis convenient. The road net is designed dense and Telephon system has been set up and you have no choice but acceptthem.
and easy to use. Peoplecan commuteon campus via campus bus, driving Of course, it nsually costs you more expensive than it could be if you can
theirselves,riding bicycle. choose the system yourseE This point is especially serious for foreigen
studen&becaus they often make a long distance call to their home country. The
telephoneErnst is a big load for them persons. Another shooting is
the qtality ofwateris very bad. It cann not used for drinking directly, even it
is boiled. Many students by water for drinking when they go slmping weekly.
Compare to off campus, the dorm and appartment on campus are
smaller and more expensive. The room spaceof a dorm is aboutjust four times
as big as a car. And tvvoroomsmust share one bathroom thereforeyou can not
possessa compIetelyprivate space.
In conclusion, MgU should pay more attention to its parking and
housing problems, That will be directly beneficial for many stndents.
Since there was no interaction between prelpost and 30160, we cannot reject
this hypothesis. Perhaps students were already using the extra time to revise their
language at the beginning of the semester, or the instruction given in the writing
classes was not helpful in this respect. Yet another possibility is that the students
started at a higher level at the end of the semester and that editing may get pro-
gressively more difficult as linguistic complexity increases. We know that the
essays were better with regard to linguistic accuracy at the end of the semester and
thus, the students simply may have had less editing to do.
No difference was found between the two groups and thus the hypothesis was
not rejected. While we have simply shown that no effect was found, we have not
demonstrated that additional editing and grammar instruction is useless. Nonsig-
nificant results should never be seen as disproving a hypothesis since such results
could be related to a variety of factors including the use of a measure that was not
able to detect differences. Furthermore, the treatment simply may not have been
strong enough, or administered for along enough period of time. Another possibility
is that the instruction the control group received in their classes was so effective that
the additional practice for the experimental group did not make a difference.
IMPLICATIONS
Writing Pedagogy
The fact that students were able to improve the linguistic accuracy of their essays
with extra time may have implications for writing pedagogy. Most instructors
give feedback on essays and require multiple drafts. We suspect, however, that
most do not give grammatical feedback on all assignments, particularly journals.
Giving feedback on all assignments is time-consuming, and may not even be
helpful in the long-term (Robb, Ross, & Shortreed, 1986; Truscott, 1996), partic-
ularly if learners do not understand or process this feedback. It may, however, be
helpful to have students, as part of their journal assignments for example, revise
their writing without feedback. We cannot be certain that this will have long term
benefits and, in fact, the participants in our experimental group did not appear to
benefit from revising one journal entry a week for seven weeks. Our study, how-
ever, did not set out to examine the long-term benefits of such an activity; rather,
we had students revise to apply the editing techniques that they were learning.
60 POLIO, FLECK & LEDER
That there was no interaction effect between prelpost and 30160 (or with exp/
con) shows that our writing instruction may not be helping students revise for
grammatical errors; they were not any better at editing at the end of the semester
than at the beginning. We did not do classroom observations for this study, though
we do have a writing curriculum that includes some grammar instruction. It is not,
however, heavily grammar-oriented. Students are taught that editing is a step in
writing, but more emphasis is placed on revising for content and organization.
Also, as stated earlier, editing in a way that results in fewer EFTs may become
more difficult as linguistic complexity increases.
Finally, we have to consider the implications of the fact that the experimental
group did not do any better than the control group. On one hand, as mentioned
above, nonsignificant results should not be taken as proof, and we do not want to
dwell on this part of our study. On the other hand, this study can be added to a list
of studies that have failed to show improvement in students’ linguistic accuracy in
writing as the result of feedback (Kepner, 1991; Robb, Ross, & Shortreed, 1986;
and those discussed in Truscott, 1996) or additional grammar instruction
(Frantzen, 1995). As we have suggested earlier, these studies should not be taken
as evidence that grammar correction is theoretically ineffective; rather, these stud-
ies merely show that grammar correction as practiced is ineffective. Nevertheless,
we do concede that those who claim editing instruction or corrective feedback is
useful have the burden of demonstrating such effectiveness.
Testing
We have confirmed that learners are able to improve their language when given
more time, but with regard to testing, a more pragmatic question needs to be
addressed and that is: how much of an improvement is made? The improvement
from 30-minute pretest to 60-minute pretest was 1.6% more EFTs. From 30-
minute posttest to 60-minute posttest, it was 4.1% more EFTs.) ’ (Average essay
length was 26.87 T-units.) For the WEFT/TW measure, there was a 2% differ-
ence between the essays on the pretest and 4% on the posttest. From a testing per-
spective, this probably would not make much of a difference. We suspect that a
holistic measure of accuracy would not be able to pick up such minor differences.
(Again, we were not able to achieve high reliability on a subset of these essays
using a holistic measure of linguistic accuracy [Polio, 19971.)
We stated at the beginning of this article that having students write drafts and
then returning the essays to them for revision, may be a useful way to test writing
in that it more closely resembles writing done outside the classroom and it tests
revision skills. And while it may not be practical for large scale placement tests, it
may be possible for testing done in writing courses. Despite the fact that the dif-
ference in linguistic accuracy may be small, this method of testing writing should
ESSAY REVISIONS 61
still be considered because one needs to consider matters of content and organiza-
tion, which are beyond the scope of this article.
We also need to consider the extent to which the 30-minute essays and the 60-
minute revisions are measuring the same construct. The correlation between the
two sets of essays was .77 on EFT/TT and .8 1 on WEPT/TW.t2 With correlations
hovering around .80, one needs to carefully consider whether or not the two types
of essays are measuring the same thing. With regard to language, it may not be
worth the trouble to have students write and then revise.
Once again, this study was not intended to address issues related to SLA, but
rather issues of revision. We feel, however, that this study has added to some of
the prior research that shows that learners can and do correct their own language
without feedback. Swain and Lapkin (1995) showed that learners can notice gaps
in their written language and modify that language. We have shown that those
modifications do result in (somewhat) more linguistically accurate target lan-
guage. The long-term effects of this type of activity are unknown. Had we found
no significant improvement in the revised essays, we would have argued that
feedback (implicit or explicit) was necessary for learners to focus on the structure
of their language and make corrections. Instead, this study supports Cumming’s
(1990) view that writing may help learners’ second language development.
This study examined writing in a more controlled setting than in Kroll’s study,
but a less controlled setting than Gass’s or Makino’s studies, which were limited
to individual sentences. By requiring the students to hand in complete essays and
then providing more time, we were attempting to force the students to use the
extra time to revise. But because of the different revision behaviors, it was not
simply a study of editing (similar to Gass’s and Makino’s), but rather of editing
within the context of revision. Telling the students to limit themselves to sen-
tence-level editing would have increased the internal validity of the study but
reduced the external validity; restricting students to sentence-level editing may
be both unrealistic and pedagogically unsound.
This study examined only linguistic accuracy and not other aspects of the
essays. Thus, we have not addressed the question, more linguistic accuracy at the
expense of what? Perhaps some students had fewer errors because they spent their
entire time correcting grammar, while other students’ grammar did not improve
62 POLIO, FLECK & LEDER
because they were spending more time on adding material or changing the orga-
nization. Our next step is to examine these issues.
This is a product-oriented study. After examining the essays, we wanted to
find out why students were making the changes they did: Why did they believe
that such changes improved their essays? What sources of knowledge were they
tapping? A second study we have begun is a stimulated recall study of revision.
We are giving students a similar task but having them revise on the computer. All
the changes they make are videotaped. We then sit down and interview the stu-
dents immediately afterwards about the changes they have made in their essays.
Bygate (1996) argued that simply having a learner repeat a (speaking) task,
could result in the learner paying greater attention to form. In a case study, he
found that when a second language learner redid a speaking task a week later, she
used more accurate lexical items and more complex syntax, perhaps because she
was able to shift her attention from content to form. These results led us to ask
what would happen if we had students redo the writing task without giving them
their first drafts back; would we get the same results? It might be that simply hav-
ing the students redo the task can lead to improvements in accuracy for reasons
put forth by Bygate. An empirical question is: Does redoing an essay lead to
greater linguistic accuracy than revising an essay? One possibility is that there
were limited changes in the essays because students are bound (consciously or
unconsciously) to their original essays. Perhaps not giving them their drafts back
would result in greater accuracy.
And finally, one needs to consider the long-term effects of self-correction
without feedback on both second language writing and SLA. While this study,
Gass’s, Makino’s, Pica’s and Bygate’s all found short-term linguistic improve-
ments, and Swain and Lapkin’s found that learners were able to notice gaps, none
of these studies has shown any long-term changes in development.
In conclusion, this study has shown that learners can and do make improvements
in their grammar and lexical choice without feedback and that additional time does
lead to self-correction. Previous studies had been inconclusive. However, because
the amount of change that they make is quite small, we remain cautious about the
implications for writing pedagogy and second language writing assessment.
Acknowledgments: We would like to thank Susan Gass and Alison Mackey for their
helpful comments on this paper.
NOTES
2. We are using the term revision to mean any changes in what was origi-
nally written. Revision can include changes in content, organization, or language.
Editing refers specifically to sentence-level changes and is included in revision.
3. For a review of the ways to measure linguistic accuracy, see Polio
(1997). This term can refer to different constructs depending on the measure used.
Generally, it refers to the correct use of syntax, morphology, the lexicon, and
punctuation in connected discourse.
4. As operationalized in Quellmalz (1985) and Stiggens, Rubel, and
Quellmalz (1988).
5. Although the dismal results he reports may be accurate, we disagree
with Truscott because he claims that grammatical feedback is not beneficial in
theory. A myriad of practical problems can limit the benefits of grammatical feed-
back. Furthermore, error correcton that gets learners to notice a gap in their inter-
langauge should be beneficial according to claims made by several SLA
researchers including Schmidt (1990).
6. Of eighteen measures of grammatical accuracy, one revealed a signifi-
cant gain for the grammar group and one a significant gain for the control group.
7. Although we did try to keep all 13 prompts similar, we are not claiming
that all the prompts are of equal difficulty.
8. The essays at the end of the semester constituted a much larger portion
of the student’s grade. The first set of essays were counted as only a portion of
their assignment grade while the second set was their final exam.
9. The materials can be obtained by contacting the authors of this study.
10. Robb, Ross and Shortreed also used EFT/number of clauses and EFT/
number of words to measure what they called linguistic accuracy. The correla-
tions between these measures and EFT/TT were .95 and .94 respectively. The cor-
relation between WEFT/TW and EFT/TT was still high at .91. We felt, however,
that WEFTITW measure might be better at picking up changes in revision if stu-
dents were adding words to already error-free units.
11. Statistically, there is no difference between the 30/60 minute difference
on the pretest and the 30/60 minute difference on the posttest.
12. This is the correlation taking into account the intercoder reliabilities of
the 30-minute essays and 60-minute revisions. The correlations before correcting
for attenuation were .69 and .68 on the EFT/TT and WEFTITW measures respec-
tively.
REFERENCES
Bates, L. (1993). Transitions: An interactive reading, writing, and grammar text. New
York: St. Martin’s Press.
64 POLIO, FLECK & LEDER
Brown, J. (1989). Improving ESL placement tests using two perspectives. TESOL Quar-
terly, 23, 65-83.
Bygate, M. (1996, March). Task familiarity and task demand: Some implications from the
psycholinguistics of second language development. Paper presented at AAAL
Annual Meeting, Chicago.
Crookes, G. (1989). Planning and interlanguage variation. Studies in Second Language
Acquisition, 11, 367-383.
Cumming, A. (1990). Metalinguistic and ideational thinking in second language compos-
ing. Written Communication, 7, 482-5 11.
Ferris, D. (1995). Teaching students to self-edit. TESOL Journal, 4, 18-22.
Fox, L. (1992). Focus on editing: A grammar workbook for advanced writers. White
Plains, NY: Longman.
Fran&en, D. (1995). The effects of grammar supplementation on written accuracy in an
intermediate Spanish content course. Modern Language Journal, 79, 329-344.
Gass, S. (1983). The development of L2 intuitions. TESOL Quarterly, 17, 273-291.
Gass, S., & Varonis, E. (1989). Incorporated repairs in NNS discourse. In M. Eisentein
(Ed.), Variation and second language acquisition (pp. 71-86). New York: Plenum.
Hacker, D., Plumb, C., Butterfield, E., Quathamer, D., & Heineken, E. (1994). Text revi-
sion: Detection and correction of errors. Journal of Educational Psychology, 86,
65-78.
Hedge, T. (1988). Writing: Resource books for teachers. Oxford: Oxford University Press.
Hull, G. (1987). The editing process in writing: A performance study of more skilled and
less skilled college writers. Research in the Teaching ofEnglish, 21, 8-29.
Kasper, G. (1985). Repair in foreign language teaching. Studies in Second Language
Acquisition, 7, 200-2 15.
Kepner, C. (1991). An experiment in the relationship of types of written feedback to the
development of second-language writing skills. Modern Language Journal, 75,
305-313.
Kroll, B. (1990). What does time buy? ESL student performance on home versus class
compositions. In B. Kroll (Ed.), Second language writing: Research insights for the
classroom. (pp. 140-154). Cambridge: Cambridge University Press.
Lane, J., & Lange, E. (1993). Writing clearly: An editing guide. Boston: Heinle and Heinle.
Leki, I. (1989). Academic writing: Techniques and tasks. New York: St. Martin’s Press.
Makino, T. (1993). Learner self-correction in EFL written compositions. English Lan-
guage Teaching Journal, 47, 337-341.
Pica, T. (1988). Interlanguage adjustments as an outcome of NS-NNS interaction. Lan-
guage Learning, 37, 47 l-493.
Pica, T., Holliday, L., Lewis, N.. & Morgenthaler, L. (1989). Comprehensible output as an
outcome of linguistic demands on the learner. Studies in Second Language Acquisi-
tion, II, 63-90.
Pienemann, M. (1984). Psychological constraints on the teachability of language. Studies
in Second Language Acquisition, 6, 186-214.
Polio, C., & Glew, M. (1996). ESL writing assessment prompts: How students choose.
Journal ofSecond Language Writing, 5, 35-50.
Polio, C. (1997). Measures of linguistic accuracy in second language writing research.
Language Learning, 47, 101-143.
ESSAY REVISIONS 6.5
Quellmalz, E. (1985). Needed: Better methods for testing higher-order thinking skills.
Educational Leadership, 43, 29-35.
Raimes, A. (1987). Exploring through writing: A process approach to ESL composition.
New York: St. Martin’s Press.
Ramirez, A. (1995). Creating contexts for second language acquisition. White Plains, NY:
Longman.
Reid, J. (1993). Teaching ESL writing. Englewood Cliffs, NJ: Prentice Hall.
Robb, T., Ross, S., & Shortreed, I. (1986). Salience of feedback on error and its effect on
EFL writing quality. TESOL Quarterly, 20, 83-96.
Schmidt, R. (1990). The role of consciousness in second language learning. Applied Lin-
guistics, II, 274-282.
Stiggins, R., Rubel, E., & Quellmalz, E. (1988). Measuring thinking skills in the class-
room. Washington: NEA Professional Library.
Swain, M. (1985). Communicative competence: Some roles of comprehensible input and
comprehensible output in its development. In S. Gass & C. Madden (Eds.). fnput
and second language acquisition. (pp. 235-256). Rowley, MA: Newbury House.
Swain, M., & Lapkin, S. (1995). Problems in output and the cognitive processes they gen-
erate: A step towards second language learning. Applied Linguistics, 16, 371-391.
Tickle, A. (1996). The writing process: A guide for ESL students. New York: HarperCol-
lins College Publishers.
Truscott, J. (1996). The case against grammar correction in L2 writing classes. Language
Learning, 46, 327-369.
Zamel, V. (1983). The composing processes of advanced ESL students: Six case studies.
TESOL Quarterly, 17, 165-187.
APPENDIX A
Description of Participants
South Middle
Male Female UG Grad Asia America East Europe
Exp 13 21 18 16 27 2 3 2
con 16 15 18 13 22 0 6 3
Total 29 36 36 29 49 2 9 5
66 POLIO, FLECK & LEDER
APPENDIX B
Essay Topics
APPENDIX C
T-Units
Clauses
a. A clause equals an overt subject and a finite verb. The following are only one
clause each:
Error Guidelines
g. Count errors that could be made by native speakers (e.g., between you and I).
h. Do not count register errors related to lexical choices (e.g., lots, kids).
i. Disregard an unfinished sentence at the end of the essay.
Word Count