Grammar Correction in L2 Writing Class

8/3/2019 Grammar Correction in L2 Writing Class

1/43

Language Learning 46:2, June 1996, pp

Review Article

The Case Against Grammain L2 Writing Clas

John Truscott National Tsing Hua Unive

The paper argues that grammar corre

ing classes should be abandoned, for the f(a) Substantial research shows it to be inefshows it to be helpful in any interesting stheoretical and practical reasons, one caineffective; and (c) it has harmful effectsand reject a number of arguments prev

favor of grammar correction.

In second language (L2) writing course

is something of an institution. Nearly all L

it in one form or another; nearly everyonsubject recommends it in one form or an

researchers hold a widespread, deeply en

grammar correction should, even must, be p


2/43

328 Language Learning

few serious attempts to justify the practice onthose that exist pay scant attention to the s

that has found correction ineffective or harmfu

the subject simply takes the value of gram

granted. Thus, authors often assume the pr

without offering any argument or citing any

someone cites evidence, it generally consists

token sources, with no critical assessment of

Researchers have similarly failed to lo

nature of the correction process. Work on considers the many practical problems inv

correction and largely ignores a number of

which, if taken seriously, would cast doubt on

Finally, researchers have paid insufficie

side effects of grammar correction, such as itsattitudes, or the way it absorbs time and energy

Commentators seem to feel that we cannot el

lems through limited adjustments in the correc

simply have to live with them. They assu

correction must be used in writing classesproblems it creates; this assumption is ver

seriously.

Grammar correction is too important to


3/43

Truscott

Before proceeding with the argumentclarify a few points. First, I do not deny the

accuracy; the issue is whether or not gram

contribute to its development. Nor do I gene

as a teaching method; I will have very little to

to the content, organization, or clarity of instance, and I certainly will not suggest tha

misguided. Finally, the key term needs so

grammar correction, I mean correction of gr

the purpose of improving a student's ability

This correction comes in many different fopurposes such distinctions have little significa

there is no reason to think any of the variatio

writing classes, and there is considerable rea

all misguided.

Grammar Correction Does No

A large number of studies have attempte

(or lack of effects) of grammar correction. Tstraightforward: The researchers compare th

who have received grammar correction over a

that of students who have not. If correcti


4/43


different types of grammar correction. They f

had little or no effect on students' writing a

difference who the students were, how m

corrected, which mistakes were corrected, how

ments were, or in what form they were presen

had no effect. The conclusion for LI, then, isnot helpful.

These studies on LI learning certainly

correction is ineffective in L2 language learn

technique that is not helpful in the one case co

other. But they certainly provide strong groview of their results, it would be folly to assu

evidence, that correction is useful in L2 learni

the effect of the LI research is to place the bur

on those who would claim that correction is h

So I turn now to the research on L2 learn

made that correction works? Clearly and una

fact, the L2 evidence fits very well with that

correction is clearly ineffective.

Hendrickson (1978) reviewed the avaiconcluded that little was known. He claimed

be corrected, but the work he reviewed did

view. His own work (1978, and in more det

1981) i di d h i ll


5/43

Truscott

but offered no reason that better-done co

helped.

Semke's (1984) large, 10-week study

produced similar results. She divided th

'; groups, each receiving a different type of

^received only comments on content, with nGroup 2 received only comments on errors. G

types of comments, and Group 4 had their e

were expected to make corrections themselv

significant differences among the groups in

writing. In addition, Group 1 (comments osignificantly better than all the others on flu

test. Thus, feedback on errors was not onl

harmful to learners. Those who received c

' plus correction were significantly inferior to

;! only comments on content. Semke also f

correction) inferior to all the other groups

against the use of a technique frequently r

S literature (but always with little or no supp

JBartram & Walton, 1991; Hendrickson, 1978?Hyland, 1990; Raimes, 1983).

Grammar correction's futility also sho

; Robb, Ross, and Shortreed (1986). They us


6/43


for grammar correction are reasonably clear n

thing, the amount of information contained in

so much among the four groups that one wou

differences among them if the information w

That there were no differences argues strongl

any value. Moreover, the practical difference

thetical fifth group and the actual fourth grou

small. In fact, Frantzen and Rissel (1987) fou

told the exact location of an error, learnersdetermine exactly what that error was; in vie

would be extremely surprising if the learners

gained any insights from their much more limi

one can reasonably treat these learners as a

lack of any contrast between them and the gmore informative feedback thus provides goo

ineffectiveness of grammar correction.

More evidence of this ineffectiveness c

(1991), who experimented with two forms of f

diate Spanish as a foreign language (FL) participants received comprehensive correctio

errors with brief explanations or statements

half received comments on content instead w


7/43

Truscott

errors (and nothing else) in conferences with

the other group, feedback and conferences de

the content of the students' writing. Thus, if e

helpful, the content group should have suffe

grammatical ability. However, Sheppard fouthe error-correction group, the results actuall

tent group. In accuracy of verb forms, there

between the groups, both improving signific

marking of sentence boundaries (through a

tion), the content group made significant ga

did not, and the difference was significant. F

of the complexity of students' writingthe

with which they used subordinate clausesth

no significant changes, although the error gro

worse (though there was no significant differe

groups on this measure). Sheppard attribute

an avoidance strategy on the part of the stud

frequently correctedtheir fear of making m

limit the complexity of their writing.

Thus Sheppard's (1992) work resembles

and Kepner (1991). Correction was not onl

studies but also actually hindered the learni

Fi ll f dditi l t di


8/43


pretations, though. The variable relevant her

nonuse of grammar correction, but a number of o

have influenced the results of the experiments.

obvious candidates can be discounted.

First, the results probably cannot be explain

ence between FL and SL learning, the ident

language, or the learners' LI. The studies that

ineffective included ESL, EFL, German FL, a

besides, the students' origins and LIs differed w

Another factor that can probably be dismiss

correction used. The studies varied between

(learners given correct forms for each error) a

(errors pointed out, usually by means of a code, b

not given). In addition, Robb et al. (1986) alo

different degrees of directness. The case is somew

the other major variable of this sortthe dif

comprehensive and selective correction. Moreviewed here relied on the former, but Hendrick

both types and found no difference between the

research described above found comprehensiven

irrelevant. Additional reasons to doubt the v

correction will be presented below.Another explanation of the results is that th

in these studies could have had a delayed effect t

up during the research. However, available


9/43

Truscott

1991; Hoekje & Linnell, 1994; Skehan, 1988,1

quite well. Second, they used a variety of meas

these included counts of all grammatical and l

studies (plus style in one of them), verb form

others, and an independent measure of sente

one. They also frequently included measur

complexity of writing, and one study added a c

of these measurements found any significa

students whose writing had been corrected. T

used in some of the studies did find significant d

groupsalways favoring the uncorrected stud

found significant gains (and occasionally loss

posttest. Clearly, these measures can detec

Thus, that none of them found any significant

sort for corrected students must be taken ser

Similar comments apply to differences in

tion used in the various studies. The auth

limited information, but this information s

variation. In Robb et al.'s (1986) research, mo

was devoted to correction practice and sentence-

(1991) described her classes as proficiency-b

concern for personal growth and the deve

Sheppard's (1992) students, in addition to

experience, read two novels and underwent

instruction (on topics overlapping the points


10/43


ability and was inferior to the uncorrected gr

and to all the other groups on fluency.) Hendri

use rewriting, but after each assignment was

class time for students to study the correction

It is also unlikely that the lack of benefi

by the students' proficiency level or ability. ranged from beginning to advanced levels of la

In addition, Hendrickson (1981) included co

ciency as one of his independent variables,

included verbal ability; neither found any eff

Of course, other learner variables could blearners differ from one another in an enormo

and the research discussed here considered

However, though such a possibility cannot be r

no more than speculation.

However, assume for the sake of argvariables are crucial to the effects of gramm

certain types of students do benefit. A new p

because the knowledge that such students exis

unless instructors can determine exactly who

now this is not realistic. The (hypothetical) those who benefit and those who do not could

of variables, such as gender, age, educationa

tude field independence tolerance for ambig


11/43

Truscott

which learners acquire certain grammatica

research has found these same sequences in

learning situations, in spite of instructional

counter to them. This raises the possibility

used in the research described above failed b

respect these sequences: Teachers corrected

mar points for which they were not yet ready

The research on developmental sequenc

morpheme studies of Dulay and Burt (1973,

den, and Krashen (1974), and Perkins an

(1975). This work has since become the sub

(Dulay, Burt, & Krashen, 1982; Larsen-Free

Freeman & Long, 1991; Rosansky, 1976) and

considered conclusive. However, subsequent w

languages (e.g., Cancino, Rosansky, & Sch

1984, 1988, 1989; Felix, 1981; Hyltenstam1984,1989; VanPatten, 1987; Weinert, 1987; W

little doubt that developmental sequences are

sion has met wide acceptance among SLA res

1993; Dulay et al., 1982; Ellis, 1990; Ha

Freeman & Long, 1991; Lightbown & Spad1984; VanPatten, 1986b). It signifies, for pr

grammar instruction (or correction) that doe

sequences will probably encounter problems


12/43


Nonevidence for Grammar Correction

The discussion of possible limitations on

arrived at the same conclusion reached pre

correction (at least in any form now available) is not enough, though, to show that many stu

negative results. A number of additional stud

presented as evidence favoring grammar corr

sary to look at these as well. However, none of

negative findings described above, primarilythem actually address the present issue: Doe

tion in writing classes make students better wr

sense)?

First, it is not unusual to find vague refere

seem, in the context of the discussion, to procorrection works, but actually do not even attem

examples will suffice: Higgs (1979) and Gau

former is simply a detailed description of Higgs

of correction. Similarly, Gaudiani simply provi

writing course along with guidelines for teacimplement it. Neither provided, or claimed

evidence for the effectiveness of correction; th

is effective.


13/43

Truscott

tion, this oral research's credibility is weaken

other studies that found oral (or in some case

oral and written) correction ineffective (EUis

Holley & King, 1971; Lightbown, 1983a; Pla

Fathman and Whalley (1990) studied thehaving one group of ESL students revise the

the benefit of comments from the teacher, whi

their revisions without such comments. N

former group produced better final drafts th

result, though interesting and valuable, doquestion: Does grammar correction make stu

Fathman and Whalley have shown that stu

better compositions when teachers help them

lar compositions. But will those students be

future because of this help? Nothing in thpositive answer.

Lalande's (1982) work appears more rel

the effects of correction procedures in writi

concerned with effects beyond the particula

considered. But it too actually dealt with a qthat being considered here. Lalande's pur

composition teaching method he developed, in

sive correction by means of a special code


14/43


15/43


16/43


17/43


18/43


19/43


20/43


21/43


22/43


23/43


24/43


25/43


26/43


27/43


28/43


29/43


30/43


31/43


32/43


33/43


34/43


35/43


36/43


37/43


38/43


39/43


40/43


41/43


42/43


43/43

Documents

Grammar Correction in L2 Writing Class