Upload
marija12990
View
235
Download
0
Embed Size (px)
Citation preview
8/10/2019 Rod Ellis a SLA Perspective
1/26
Teachers of English to Speakers of Other Languages, Inc. (TESOL)is collaborating with JSTOR to digitize, preserve andextend access to TESOL Quarterly.
http://www.jstor.org
Teachers of English to Speakers of Other Languages, Inc. TESOL)
Current Issues in the Teaching of Grammar: An SLA PerspectiveAuthor(s): Rod EllisSource: TESOL Quarterly, Vol. 40, No. 1 (Mar., 2006), pp. 83-107Published by: Teachers of English to Speakers of Other Languages, Inc. (TESOL)Stable URL: http://www.jstor.org/stable/40264512
Accessed: 11-08-2014 07:33 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available athttp://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of contentin a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.For more information about JSTOR, please contact [email protected].
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/http://www.jstor.org/action/showPublisher?publisherCode=tesolhttp://www.jstor.org/stable/40264512http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/40264512http://www.jstor.org/action/showPublisher?publisherCode=tesolhttp://www.jstor.org/8/10/2019 Rod Ellis a SLA Perspective
2/26
Currentssues n theTeaching f
Grammar:
An SLA
Perspective
ROD
ELLIS
Universityf
Auckland
Auckland,
New Zealand
Thestudyfhow earnerscquire second anguage SLA) hashelped
to
shape
thinking
bout how to
teach
the
grammar
f a second
language.
There
remain,however,
numberof
controversialssues.
This
paper
considers
ight
ey
uestions
elating
o
grammar edagogy
in the
light
f
findings
rom
LA. As
such,
this rticle
omplements
Celce-Murcia's
1991)
article n
grammar eaching
n the
25th
nniver-
sary
ssueof
TESOL
Quarterly,
hich onsidered
herole of
grammar
n
a communicative
urriculum
nd drew
predominantly
n a
linguistic
theory
f
grammar.
hese
eightquestions
ddresswhether
rammar
should
be
taught
nd if so what
grammar,
hen,
nd how.
Although
SLA does not
fford efinitiveolutions
o these
uestions,
t erves he
valuable
purpose
of
problematising
his
spect
of
anguagepedagogy.
This article
concludes
with a statement
f
my
own beliefs about
grammar eaching,
rounded
n
my
wn
understanding
f SLA.
article
dentifies nd
discusses a number
of
key
ssues
relating
o
the
teaching
of
grammar
n a
second
language
(L2)
and,
by drawing
on
theory
nd
research
n
SLA,
suggestsways
o address these
problems.
It
points
to
a number
of
alternative
olutions to
each
problem, ndicating
thatmore oftenthannot there are no clear solutionscurrentlyvailable.
The
aim, therefore,
s
not
to
identify
ew solutions to
existing
ontrover-
sies,
nor even
to
present
new
controversies.Rather it addresses within
the
compass
of a
single
article whole
range
of ssues
related to
grammar
teaching,
problematises
hese
ssues,
and
by
o
doing, provides
a
counter-
weight
to the
advocacy
of
specific,
but also
quite
limited,
proposals
for
teaching
grammar
that
have
originated
n
some
SLA
quarters.
However,
I
conclude
with
a
statement
f
my
own
position
on
these issues.
The
questions
that will
be addressed are
1. Should we
teach
grammar,
r
should we
simply
reate the conditions
by
which learners earn
naturally?
2.
What
grammar
hould we
teach?
TESOL QUARTERLY
Vol.
40,
No.
1
March
2006
83
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
3/26
3.
When should
we
teach
grammar?
s
it
bestto teach
grammar
hen
learners irsttart o earn n
L2
or
to
wait
until aterwhen earners
havealready cquired ome inguisticompetence?
4.
Should
grammar
nstructione massed
i.e.,
the
available
eaching
timebe
concentrated
nto a
short
period)
or distributed
i.e.,
the
available
eaching
ime
pread
over
longerperiod)?
5. Should
grammar
nstructione intensive
e.g.,
cover
singlegram-
matical
tructure
n
a
single
esson)
or extensive
e.g.,
cover
many
grammatical
tructures
n
a
single
esson)?
6.
Is there
ny
value n
teaching
xplicit rammatical
nowledge?
7. Is there
best
way
o teach
grammar
or
mplicit
nowledge?
8. Should
grammar
e
taught
n
separate
essonsor
integrated
nto
communicativectivities?
DEFINING GRAMMAR
EACHING
Traditionally,rammar eaching
s
viewed
s the
presentation
nd
practice
f
discrete
rammatical
tructures.his
s
theview
romulgated
in
teacherhandbooks.
Ur
(1996),
for
example,
n
her
chapter
itled
TeachingGrammar as sections n presentingnd explaining ram-
mar and
grammar ractice
ctivities.
edge
(2000)
in
her
chapter
titled
Grammar
imilarly nly
considers
presenting
rammar
nd
practising rammar.
his constitutesn
overly
arrow efinition
f
grammar
eaching.
t
s
certainly
rue hat
rammar
eaching
anconsist
of the
presentation
nd
practice
of
grammatical
tems.
But,
as will
become
pparent,
tneed
not.
First,
ome
grammar
essons
might
onsist
of
presentation
y
tself
i.e.,
without
nypractice),
while
others
might
entail
onlypractice
i.e.,
no
presentation).
econd,
grammar
eaching
can
involve earners
n
discovering rammatical
ules
for themselves
(i.e.,no presentationnd nopractice). hird, rammar eaching anbe
conducted
imply y
exposing
earners o
input
contrived o
provide
multiple
exemplars
of the
target
tructure.
ere,
too,
there is
no
presentation
nd no
practice,
t east
n
the enseof
eliciting
roduction
of
the
structure.
inally, rammar eaching
an be conducted
y
means
of correctiveeedback
n
learner
rrorswhen hese rise
n the context
of
performing
ome communicativeask.
The definition
f
grammar
teaching
hat nformshis rticle
s
a broad one:
Grammar
eaching
nvolves
any
instructional
technique
that
draws learners'
attentionto some specific grammaticalform n such a waythat thelps them
either to understand t
metalinguistically
nd/or
process
it n
comprehension
and/or
production
so that
they
an internalize
t.
84
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
4/26
SHOULD WE TEACH GRAMMAR?
This questionwas motivated yearlyresearch nto naturalistic2
acquisition,
which howed that earners
ppeared
to
follow natural
orderand
sequence
of
acquisition
i.e.,
they
mastered
ifferent
ram-
matical tructures
n
a
relatively
ixed and universal
rder and
they
passed
through
sequence
of
tages
f
cquisition
n route
o
mastering
each
grammatical
tructure).
his ed researchersike
Corder
1967)
to
suggest
hat
earners ad their wn built-in
yllabus
or
earning ram-
mar.
n
line with
his,
Krashen
1981)
argued
that
grammar
nstruction
played
no
role
in
acquisition,
view based on the
conviction hat
learners
including
lassroom
earners)
would
automatically roceed
along
their uilt-in
yllabus
s
long
as
they
had access to
comprehensible
input
nd were
ufficiently
otivated. rammar
nstructionould con-
tribute
o
learning
ut this
was of
imited alue because communicative
ability
as
dependent
n
acquisition.
There followed
number
of
empirical
studies
designed
to
(a)
compare
he
orderof
acquisition
f nstructednd
naturalisticearners
(e.g.,
Pica,
1983), b)
compare
he
uccess f nstructednd
naturalistic
learners
Long,
1983)
and
(c)
examine whether
ttempts
o teach
specific rammatical
tructuresesulted
n
their
cquisition e.g.,
White,
Spada, Lightbown, Ranta,1991). These studies howedthat, yand
large,
heorder
f
acquisition
was the
ame for nstructednd naturalis-
tic learners
although
herewere some
interesting
ifferences1),
hat
instructed earners
generally
chieved
higher
evels of
grammatical
competence
than
naturalisticearners
and that
instruction as no
guarantee
hat
earners
would
cquire
what
hey
ad been
taught.
hese
results
were
nterpreted
s
showing
hatthe
acquisitional
rocesses
f
instructednd
naturalistic
earning
were the same but
that nstructed
learners
rogressed
more
rapidly
nd
achieved
higher
evelsof
profi-
ciency.
Thus,
some researchers oncluded
(e.g.,
Long,
1988)
that
teaching rammar asbeneficial utthat obe effectiverammar ad to
be
taught
n
a
way
that
was
compatible
with he natural
processes
f
acquisition.
Subsequent
research,
uch as Noms and
Ortega's
(2000)
meta-
analysis
f49
studies,
as borneout theoverall
ffectivenessf
grammar
teaching.
urther,
here s evidence
hat,
ontrary
o Krashen's
1993)
continued
laims,
nstructionontributeso both
acquired knowledge
(see
Ellis,
002a)
as well s learned
knowledge.
here s also
increasing
1For
example,
Pica
(1983)
notes that some structures
e.g.,
plural-s)
were used more
accurately y
nstructed earners nd some
(e.g.,
Verb-ing)
by
naturalistic
earners.
n
other
structures
e.g.,
articles)
therewas no difference.
CURRENT ISSUES
IN THE
TEACHING
OF
GRAMMAR
85
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
5/26
evidence hat
naturalistic
earning
n
the classroom
as,
e.g.,
n immer-
sion
programmes)
oes not
typically
esult
n
high
evels f
grammatical
competenceGenesee, 1987). In short,here snowconvincingndirect
and direct vidence o
support
he
teaching
f
grammar.
evertheless,
doubtsremain
bout the nature f the research
vidence.
Many
tudies
(including
most of
those reviewed
y
Norris
and
Ortega)
measure
learning
n
terms
f
constrained
onstructed
esponses
e.g.,
fill
n
the
blanks,
sentence
oining,
r sentence
ransformation),
hich an be
expected
o
favour
rammar eaching.
here s
only
mixed
vidence hat nstruction
results
n
learning
when
it is
measured
by
means of
free
onstructed
responses
e.g.,
communicative
asks).
Also,
it remains
the case that
learners o not
always cquire
what
hey
ave been
taught
nd that
or
grammarnstructiono be effectivet needs to take accountof how
learners
evelop
heir
nterlanguages.
s we will
ee,
there s
controversy
regarding
othhow
nterlanguageevelopment
ccurs nd how nstruc-
tioncan
facilitate
his.
WHAT
GRAMMAR SHOULD
WE TEACH?
Assuming,
hen,
hat
rammareaching
an
contribute
o
nterlanguage
development,
he next
logical
question
concernswhatgrammarwe
should teach. This
question
can be broken
down into two
separate
questions:
1.
Whatkindof
grammar
houldwe base
teaching
n?
2.
Which
grammatical
eatures houldwe teach?
Linguistics
ffords broad
selection f
grammatical
models to choose
from,
ncluding
tructural
rammars,enerative
rammars
based
on
a
theory
f universal
rammar)
and functional
rammars.
raditionally
syllabuseshave been based on structuralr descriptive rammars.
Structural
yllabuses
raditionally
mphasised
he
teaching
f
form ver
meaning
e.g.,
Lado,
1970).
Though
the
nfluence
f
structural
ram-
mars
s
still
pparent oday,
modern
yllabuses
ightlyive
more atten-
tion to the functions
erformed
y grammatical
orms.
Thus,
for
example,
ess
emphasis
s
placed
on such
spects
f
grammar
s sentence
patterns
r tense
paradigms
nd more on
the
meanings
onveyed y
different
rammatical
orms
n
communication.
ome
attempt
asonce
made to
exploit
he
nsights
o be
gleaned
from
enerative
heories
f
grammar
see,
e.g.,Bright,
965),
but
n
general, yllabus
esigners
nd
teachers ave not found uchmodelsuseful nd havepreferredorely
on modern
descriptive rammars,
uch
as Celce-Murcia
nd
Larsen-
86
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
6/26
Freeman's
1999)
Grammar
ook.This
resource
s
especially
aluable
because t
not
onlyprovides comprehensive,
lear,
nd
pedagogically
exploitable escriptionfEnglish rammarut also identifieshekinds
of
errors hat
2
learners
re known o makewith ifferent
rammatical
structures.
uch information
s
important
ecause it
helps
to
identify
which structures
nd
which
aspects
of a structure
equire special
attention. he Grammar
ook
s also ideal
in
that t
presents
nformation
not
only
bout
inguistic
orm
ut lso aboutthe
emantic
nd discoursal
meanings
ealised
yparticular
orms. s
VanPatten,Williams,
nd Rott
(2004)
emphasise,
stablishing
onnections etween
orm
nd
meaning
is a fundamental
spect
of
language
acquisition.
hus,
any
reference
grammar
hatfails
to describe
the
form-meaning
onnections f the
targetanguagemustnecessarilye inadequate. n general, hen, he
choice of which
ype
f
grammar
o use
as
a
basisfor
eaching
s not a
major
ource
of
controversy;
escriptive
rammars
hatdetail
the
form-
meaning elationships
f the
anguage
re ascendant.
In
contrast,
he choice
of which
grammatical
tructures
o teach s
controversial.
wo
polar
positions
an
be identified
nd various
ositions
in
between.
At one
end of this continuum
s Krashen's
minimalist
position.
Krashen
(1982)
argues
that
grammar eaching
should be
limited o a few
imple
nd
portable
ules uch
as
3rd
person-j
nd
past
tense-dhat an
be used
to
monitor
utput
rom he
acquired ystem.He baseshis
argument
n the claimthatmost earners re
only apable
of
learning
uch
simple
rules thatmore
complex
rulesare
generally
not earnable
r,
f
hey
re,
re
beyond
tudents'
bility
o
apply hrough
monitoring.
rashen's
laim,however,
s
not warranted.
here
is now
ample
evidence
that
many
earners re
capable
of
mastering
wide
range
of
explicit rammar
ules.Green nd Hecht
1992),
for
xample,
found hat
university-level
tudents
f
English
n
Germany
ere ble to
produce
lear
explanations
or
5%
of
the
grammatical
rrors
hey
were
askedto
explain,
while verall
he earners n their
tudy
who
ncluded
secondarychoolstudents)managed atisfactoryxplanations or46%
of the errors.
Macrory
nd Stone
(2000)
reported
hatBritish
ompre-
hensive
chool students
ad a
fairly ood explicit nderstanding
f the
perfect
ense
n French
e.g., they
nderstood
ts
function,
hey
knew
that ome verbsused
avoir nd some
tre,
hey
were familiar
with he
forms
equired
y
different
ronouns,
nd
they
were wareof
the
need
for
final ccent
on the
past participle).
Hu
(2002)
found
that dult
Chinese
earners f
English
emonstratedorrect
metalinguistic
nowl-
edge
of
prototypical
ules
f six
English
tructures
e.g.,
for
he
definite
article
pecific
eference
onstituted he
prototypical
ule)
but were ess
clear bout heperipheralules or hese tructurese.g.,genericeference)
At
the
other
pole
is the
comprehensiveosition:
each the wholeof
CURRENT ISSUES
IN
THE
TEACHING
OF
GRAMMAR 87
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
7/26
the
grammar
f the
target
anguage.2
his is
the
position
dopted by
many
ourse book writers
e.g.,
Walter&
Swan,
1990)
or authors f
grammar racticematerialse.g.,Murphy,994).Such a positionwould
also seem
unwarrantedecause earners re
clearly apable
of
earning
substantialmount f the
L2
grammar
ithoutnstructionnd because
most
eaching
ontexts ave imited ime vailable
or
eaching rammar
so
some selection s needed.
What then
should selectionbe based on? The answer
would seem
obvious the
nherent
earning
ifficulty
f different
rammatical
truc-
tures. he
problem
rises
n
how
to
determine his.To
begin
with,
t s
necessary
o
distinguish
wo ifferentenses f
earning
ifficulty.
hiscan
refer o
(a)
the
difficulty
earners
ave n
understanding grammatical
feature nd (b) to thedifficultyhey ave ninternalisinggrammatical
feature o that
hey
re
able to use t
ccurately
n
communication.
hese
two
enses elate o thedistinction
etween
earning rammar
s
explicit
knowledge
nd as
implicit nowledge,
hich s discussed ater.
Clearly,
what s difficulto
earn
s
explicit
nowledge
nd as
implicit nowledge
is notthe ame.
For
example,
most earners aveno
difficulty
n
grasping
the rule
for
English
hird
erson-s
ut
they
aveenormous
ifficulty
n
internalising
his tructureo
they
an use it
ccurately.
hese two enses
of
learningdifficulty
ave not
always
been
clearly
distinguished
n
languagepedagogy,
ith
he
result hat venwhen he tated
oal
is
the
development
f
implicit
knowledge,
t is the
anticipated
difficulty
studentswillhave n
understanding
feature hat
guides
the selection
and
grading
f
grammatical
tructures.hird
person-5,
or
xample,
s
typicallyaught ery arly
n a course.
How then
has
learning
difficulty
een established?
raditionally,
factorsuchas the
frequency
f
pecific
tructures
n
the
nput
nd their
utility
o learners ave
been
invoked
Mackey,
976),
but thesefactors
would seem to have more to do
with se3thanwith
nherent
ognitive
difficulty.
ere
I
consider wo
pproaches
hathave
figured
n
attempts
to delineate ognitive ifficulty.
1. Teach those
forms hat
iffer rom he earners'
irst
anguage
LI).
2.
Teach marked ather hanunmarked
orms.
2
Of
course,
it is not
possible
to
specify
he whole
grammar
f a
language.
Though
the
grammar
f a
language
may
be
determinate,
escriptions
f t are
certainly
ot. The
Longman
A Grammar
f
Contemporarynglish
Quirk,
Greenbaum,Leech,
8c
Svartvik,
972)
ran to 1081
pages (excluding
ndex and
bibliography)
ut
doubtlessly
oes not account for
ll the known
facts f
English
grammar.
Nevertheless,
here
s
a
recognized
anon of
English
tructures
hat,
in
the
eyes
of
syllabus esigners
nd textbook
writers,
onstitutes
he
grammar
f
English.3Structuresike
English
rticles hat re
very
requent
n the
nput
an
impose
considerable
learning
ifficulty.
tructuresuch
as
English
onditionals
may
be
very
seful
o
earners ut
are
also
difficulto
learn.
88
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
8/26
The first
pproach
was,
of
course,
the
one
adopted
in
many arly
structuralourses
ased
on
a
contrastive
nalysis
f the earner's I and
the target anguage.Although he contrastivenalysishypothesiss
initially
ormulated
s
clearly
ottenable
see
Ellis, 985,
hapter
),
SLA
researchers
till
enerallygree
that earners
ransfer
t east omeofthe
features
f their
I
into the
L2.
For
example,
here s
ample
evidence
(Trahey
White,
993)
to show hat rench
earners f
English roduce
errors
fthe
kind
Mary
issed
assionately
ohn
ecauseFrench
ermits
n
adverb
to be
positioned
between the
verb and the direct
object.
Nevertheless,
ontrastive
nalysis
oes
not constitute sound basis for
selecting
rammatical
tructures.
n
many eaching
ontexts,
he earn-
ers come from
mixed
anguagebackgrounds
here t would be
impos-
sibleto use contrastivenalysiso tailor rammar eaching othe entire
group
because
the
earners
avedifferent
is.
Also,
we
simply
o not
yet
know
nough
about when difference
oes
and does not translatento
learning
difficulty,
nd
in
some
cases,
learning
difficulty
rises even
where
here s
no difference.
The
second
approach,
however,
s also
problematic.
Markednessas
been defined
n terms
f
whether
grammatical
tructure
s
in
some
sense
frequent,
atural,
nd basic or
infrequent,
nnatural,
nd deviant
from
regular attern
Richards,
latt,
Weber,
985).
Thus,
heuse
of
an
infinitiveithout
o
ollowingmake,
s
in
He mademe
ollow
im an be
consideredmarkedbecause makes one the fewverbs n
English
hat
takes
this kind
of
complement
nd because
this
pattern
ccurs
only
infrequently.
he
general
idea
is that we
should teach the marked
featuresnd
leavethe
earners o earn
heunmarked orms
aturallyy
themselves.
he
problem
s
that,
s
the definition
uggests,
markedness
remains somewhat
paque concept,
o
that
t s
often ifficult
o
apply
with
he
precision
eeded to determine hich
tructureso teach.
The selection
f
grammatical
ontent, hen,
remains
ery
roblem-
atic.
One solution
o the
kinds f
problems
have mentioned s
to
base
selection n theknown rrors roducedby earners.n this espect,ists
ofcommon
earner
rrors
uch s those vailable
n
Turton nd Heaton's
(1996)
Longman
ictionary f
Common rrorsnd Swan and
Smith's
2001)
Learner
English:
A
Teachers
Guide
to
Interference
nd
Other roblems re
helpful.
The
problems
f
selection
probably
xplain
whygrammatical
ylla-
buses re
so similar
nd
have
changed
o little
ver he
years;
t s safer o
followwhathas
been done
before.
Of course
the
selection f whatto
teach
will also
depend
on the learner's
tage
of
development.
he
problems
hat
he earner's
tage
f
development
nvolve
re
discussed
n
subsequentections.
CURRENT ISSUES
IN
THE TEACHING OF
GRAMMAR 89
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
9/26
WHEN SHOULD WE TEACH
GRAMMAR?
There retwo ompetingnswersothis uestion. ccordingo the
first,
t s
best o
emphasise
he
eaching
f
grammar
n
the
arlytages
f
L2
acquisition.
ccording
othe
econd,
t s
best
o
emphasise eaning-
focused
nstructiono
begin
with nd ntroduce
rammar
eaching
ater,
when
earners ave
lready
egun
o form heir
nterlanguages.
will
briefly
onsiderhe
rguments
or
oth
ositions.
A
key
remise
f behaviouristheories f
anguage earning
s
that
errorike in
needs
to be
avoided t
all costs
Brooks, 960).
This
premise
olds hat
nce
earners ave
ormed
ncorrect
abits,
hey
ill
have
ifficultyradicating
hem nd
replacing
hemwith orrect abits.
Thus,t snecessaryoensure hatearnerseveloporrect abitsnthe
first
lace.
Thiswas ne of he
key remises
f
he
udiolingual
ethod
(Lado,1964).
Other
rguments
anbe advanced
n favourf
beginning
to teach
grammar
arly.
he
alternative
o a form-focused
pproach
emphasises
eaning
nd
message
reation,
s
in task-based
anguage
teaching
Skehan, 998),
ut
many
eacherselieve hat
eginning-level
learners
annot
ngage
n
meaning-centred
ctivitiesecause
hey
ack
the
necessary nowledge
f the
L2
to
perform
asks.
hus,
form-
focused
pproach
s
needed
nitially
o
construct
basis f
knowledge
that earnersanthen se andextendn a meaning-focusedpproach.
Finally,
urrent
onnectionist
heories f
L2
learning,
hich
ive
pri-
macy
o
implicitearning rocesses
ased
on massive
xposure
o the
target
anguage,
lso
provide
basis or
eachingrammar
o
beginners.
N. Ellis
2005)
has
uggested
hat
earning
ecessarily
ommences
ith
an
explicitepresentation
f
inguistic
orms,
hich
re
then
eveloped
through
mplicitearning.
e
suggests
hat
eaching rammar
arly
s
valuable ecause t
provides
basis or
he eal
earning
hat ollows.his
seems o echo
Lightbown's
1991)
metaphor,
ccording
owhich
ram-
mar nstruction
acilitates
earning yproviding
earners ith
hooks
which hey angrabon to.The idea behind hismetaphors that
conscious
nderstanding
f
how
grammatical
eatures ork
acilitates
thekind f
processinge.g.,
ttentiono
inguistic
orm)
equired
or
developing
rue
ompetence.
The
argumentgainst
eaching rammar arly
n derives rom
researchn mmersion
rogrammes
e.g.,
Genesee,
987),
which
hows
that earners
n
such
programmes
re able to
develop
he
proficiency
needed or
luentommunicationithout
ny
ormalnstruction
n
the
L2.
For
xample,
earnersf
L2
Spanish
o notneed
to be
taught
hat
adjectives
ollow ouns
n this
anguage; hey
eem
o be able
to earn
thisnaturalisticallyromxposure o communicativenput Hughes,
1979).
imilarly,
earnersf
L2
English
an
master
imple
elativelauses
90
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
10/26
(e.g.,
clauses
where
the relative
ronoun
functions
s
subject
nd the
clause
s attached
o
a
noun
phrase
following
he
verb)
There
s
ample
evidence o show hat earners an anddo learn gooddeal ofgrammar
without
eingtaught
t.
This
being
o,
why
other
o teachwhat an be
learned
naturally?
second reason
for
delaying rammar eaching
o
later
stages
of
development
s that
early interlanguage
s
typically
agrammatical
Ellis,
1984;
Perdue
&
Klein,
1993).
That
s,
earners
ely
on a
memory-based
ystem
f
exical
equences,
onstructing
tterances
either
by
accessing
ready-made
hunks or
by simply oncatenating
lexical items
nto
simple
strings.
llis
(1984)
givesexamples
of
such
utterances
n
the
early peech
of
three lassroom
earners:
Meno = I don'thaveany rayons)
Me milkman
=
I
want
o be the
milkman)
Dinner
ime
ou
ut
=
It
is
dinner ime
o
you
have to
go
out)
Such
pidginised
utterances
ely heavily
n contextand the use
of
communication
trategies.
hey
are
very
ffective
n
simple,
ontext-
embedded
ommunication.
rguably,
t s this exicalised
nowledge
hat
provides
he
basis for
the
subsequent
evelopment
f the
grammatical
competence
needed
for context-free
ommunication.
his,
then,
s
a
strong rgument ordelaying heteaching fgrammar ntil earners
have
developed
basic
communicative
bility.
In
general,
have
favoured
he second
of these
positions
see
Ellis,
2002b)
Given
hat
many
lassroom
earnerswill
not
progress
eyond
he
initial
tages
of
language
learning,
t
seems
to me that
a task-based
approach
that
caters
to the
development
f a
proceduralised
exical
system
nd
simple,
aturallycquired
grammatical
tructures
ill
nsure
a threshold
ommunicative
bility
nd,
therefore,
s to be
preferred
o an
approach
hat nsists
n
grammatical
ccuracy
rom he tart nd
that,
s
a
consequence,
may
mpede
the
development
f this
communicative
ability.ask-basedanguage eachingspossiblewith omplete eginners
if
he
firstasks
mphasise
istening
and
perhaps
eading)
nd allowfor
nonverbal
esponses.
However,
t s
possible
hat
uch an
approach
can
be
usefully
omplemented
ith ne
thatdraws
eginners'
ttention o
some
useful
grammatical
eatures
e.g.,
past
tense-^d
n
English)
that
they
might
therwise
miss.This
is the aim
of
input-processing
nstruction
(VanPatten,
996,
2003),
which
s discussed ater.
CURRENT ISSUES
IN THE TEACHING
OF GRAMMAR
91
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
11/26
SHOULD GRAMMAR
TEACHING
BE
MASSED OR DISTRIBUTED?
This
uestion
s
ogicallyndependent
f he
recedinguestion.
hat
is,
irrespective
f
when
grammareaching
ommences,
e need to
consider hethert hould e concentratednto
short
eriod
ftime
or
pread
ver
longer
eriod. emarkably
ittle esearch
as ddressed
this
uestion.
The researchhat asbeenundertaken
eports
n the elativeffects
of
massed nd distributed
anguage
nstruction
n
general
anguage
proficiency
atherhan he ffectsn
grammarearning.
ollins, alter,
Lightbown,Spada 1999) ummarisehe vailable esearchs follows:
None
of the
anguageprogram
valuation esearch
as found
n
advantage
fordistributed
anguage
nstruction.
lthough
he
findings
hus
far ead to
the
hypothesis
hatmore oncentrated
xposure
o
English
may
ead to
better
student
utcomes,
he evidence s not conclusive,
p.
659)
Collins nd
olleagues
hen
eport
heir wn
tudy
f hree
ntensiveSL
programmes
n
Canada,
ne
(the
distributed
rogramme)
aught
ver
thefull 0 months f one school
year,
ne
(the
massed
rogramme)
concentrated
nto months ut
aught
nly
o above
verage
tudents,
andthe hirdthemassedlus rogramme)oncentratednto months,
supplemented
ith utofclass
pportunities
o use
English
nd
taught
to tudentsfmixed
bility
evels. he main
inding
as hat
hemassed
and
especially
he
massed-plus
tudents
utperformed
he
distributed
programme
tudentsn most f the measures
f
earning,
ncluding
somemeasures f
grammaticalbility,lthough
his
inding ight
n
part
be
explained y
the fact
hat he massed
rogrammes
rovided
more verallnstructionalime.
Collins tal. s
study
oints
othe
need
for
urther
esearch,
specially
throughtudieshat omparemassed nddistributed
nstructioni-
rected t
specific
rammatical
tructures.
deally
ucha
study
ould
compare
hort
eriods
f nstruction
n
a
particular
tructure
pread
over everal
ays
with he
ame
mount
f nstruction
ompressed
nto
one or two
essons.4
eceivedwisdoms that
cyclical
pproach
o
grammar
eaching
Howatt,
974)
s tobe
preferred
ecause
t llows or
thekind f
gradual cquisition
f
grammar
hat
s
compatible
ith hat
is
known bout
nterlanguage
evelopment.
owever,
he
results
f
4Given heproblemshat rise ncontrollingxtraneousariablesn evaluationsfentire
programmes,
t
might rove
much
asier oconduct
igorous
tudies f
massed nddistributed
learning
hen hese re focused
n
specific rammatical
tructures.
92
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
12/26
Collins
t al. s
study uggest,
t
the
very
east,
hat uch a
position
eeds
to be
investigated
mpirically.
ere, then,
s an
issue about which
nothing efinitivean be said at themoment.
SHOULD
GRAMMAR
TEACHING
BE
INTENSIVE
OR
EXTENSIVE?
Intensive
rammar
eaching
efers o nstructionver sustained
eriod
of time
which
ould
be a lessonor
a
series f
essons
overing ays
r
weeks)
concerning single
grammatical
tructure
r,
perhaps,
pair
of
contrastedtructurese.g., Englishpast continuous
vs.
past simple).
Extensive
rammar
eaching
efers o nstruction
oncerning
whole
range
of structures
ithin short
eriod
of time
e.g.,
a
lesson)
so that
ach
structure eceives
nly
minimal ttention
n
any
one lesson. t is the
differenceetween
hooting pistolrepeatedly
t the same
target
nd
firing shotgun
o
spray ellets
t a
variety
f
targets.
nstructionan
be
intensive
r extensive
rrespective
fwhether
t
s
massed r distributed.
The massed-distributed
istinction efers
o how a whole
grammar
course
is
staged,
while
the intensive-extensiveistinction efers o
whether
ach
single
esson
addresses
single
or
multiple
rammatical
features).
Grammar
eaching
s
typically
iewed s
entailing
ntensive
nstruc-
tion.The
present-practise-produce
PPP)
model of
grammar
eaching,
which
underlies
most discussions
f
grammar
eaching
n
teacher
handbooks
see,
e.g.,
Hedge,
2000;
Ur,
1996),
assumes
n intensive ocus
on
specific
rammatical
tructures.
lthough
uch
discussions
cknowl-
edge
that earners'
readiness
o
acquire
a
specific
tructureimits he
effectiveness
f
teaching
no
matter ow
ntensivet
s),
they
lso assume
thatwith ufficient
pportunities
or
practice,
earnerswill
eventually
succeed
n
automatising
he
structures
hey
re
taught.
s Ur
says,
the
aim ofgrammar ractice s to getstudents o learn the structureso
thoroughly
hat
they
will be able to
produce
them
correctly
n
their
own
p.
83).
Thus,
the dea
that
practise
makes
perfect
s
the
primary
justification
orthe
ntensive
pproach.
Practise,
owever,
must nvolve
bothdrills
nd tasks
i.e.,
opportunities
o
practice
he
target
tructure
n
a communicative
ontext)
It
is
perhaps
ess
easy
to see how
grammar
eaching
an
comprise
extensive
nstruction.
teacherwould
probably
ot electto
present
nd
practise
whole
range
of
grammatical
tructures
ithin
single
esson.
Extensive
rammar
nstructionf
kind,
owever,
as
always
ad a
placein
grammar
eaching.
ome30
years go,
while
eaching
na
secondary
school
in
Zambia,
I
regularly ave
lessons
where
I
illustrated nd
CURRENT
ISSUES
IN THE TEACHING
OF GRAMMAR 93
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
13/26
explained
ome of
the common rrors hat had observed
my
tudents
making
n
their
written
ompositions. imilarly,
n
the context
f task-
based teaching,ome teachershave been observed o note the errors
that earnersmake and then to address them when
the task s
over
(Basturkmen,
oewen,
&
Ellis,
2004).
However,
xtensive
grammar
teaching
an occur within
learning
ctivity,
ot
ust
as
some kind of
postscript.
eachers
provide
orrective eedback
n
the
context f both
form-focusednd
meaning-focused
essons,
nd
although
eedback
n
form-focusedessons
may
e directed
rimarily
t the
tructure
argeted
by
he
esson,
n
the
meaning-focused
essons t s
ikely
o be directed t
whateverrrorsearners
appen
to make.Studies
f
corrective
eedback
(e.g., Lyster
Ranta,
997;
Ellis,
Batsurkmen, Loewen,
001)
demon-
strate hatncommunicativeessons widevarietyfgrammaticalorms
are
addressed
ncidentallyhrough
orrectiveeedback.
There is
littledoubt now that ntensive
rammar
essons
can be
effective.
hough
earlierresearch howedthat earners
o not
always
learnwhat
hey
re
taught, specially
hen
earning
s
measured
n
terms
of
spontaneousproduction e.g.,
Kadia,
1987),
more
recentresearch
(e.g., Spada
&
Lightbown,
999)
indicates
hat ven
f
earners
re not
ready
o
learn
the
targeted
tructure,
ntensive
rammar
eaching
an
help
them
progress
hrough
he
sequence
of
stages
nvolved
n the
acquisition
fthat tructure.
n
other
words,eaching
marked
tructure
intensively
an
help
learners earn
associated,
ess marked tructures
even
f t
does notresult
n
acquisition
f
the
marked tructure.
ntensive
instruction lso
helps
learners to use structures
hey
have
already
partially cquired
more
accurately
e.g.,
White,
pada,
Lightbown,
Ranta,
1991).
There are also theoretical
rguments
nd some
empirical
vidence
n
favour f
an
extensive
pproach.
Cook
(1989)
has
argued
from he
perspective
funiversal
rammar
hat earners
equire
minimal vidence
to set a
particular arameter
or he
grammar
hey
re
learning.
ther
researchersave mphasisedhe mportancefnegativevidencehrough
corrective
eedback or
grammarearning
y
dults. oewen
2002)
has
shown hat ven
very
rief
pisodes
of correctiveeedback
re
related o
correctness n
subsequent
ests.
n
that
tudy,
oewen identified
he
errors hat eachers ddressed
ncidentally
n thecontext
f
communica-
tive
anguageteaching
nd then
developed
tailor-made
ests,
which
he
administeredo the earnerswho
made the
pecific
rrors
ither ne
day
or
two
weeks ater.These tests howed
that the learners
were
subse-
quently
ften ble
to
identify
nd correct
heir wn
errors.
There
are
pros
and cons forboth ntensive
nd
extensive
rammar
instruction.ome structures aynot be masteredwithoutheopportu-
nity
for
repeated practice.
Harley
(1989),
for
example,
found
that
anglophone
earnersof
L2
French failed to
acquire
the distinction
94
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
14/26
between
he
preterite
nd
imparfait
ast
tenses fter oursof
exposure
(and
presumably
omecorrective
eedback)
n an
immersion
rogramme
but were ble to improve heir ccuracyn usingthesetwo enses fter
intensivenstruction.
owever,
ntensivenstructions time
onsuming
(in
Harley's
tudy
he
targeted
tructures ere
taught
ver a 6-month
period),
and
thus,
time
will constrainhow
many
structures
an be
addressed.
Extensive
rammar
nstruction,
n the other
hand,
affords
the
opportunity
o attend
o
large
numbers
f
grammatical
tructures.
Also,
more
likely
han
not,
many
of
the structures
ill be addressed
repeatedly
ver
period
of time.
Further,
ecause thiskind
of nstruc-
tion nvolves
response
o the errors ach learner
makes,
t s individu-
alized
nd affords
he killed eacher eal-time
pportunities
or hekind
ofcontextualnalysishatCelce-Murcia2002) recommendss basisfor
grammar
eaching.
However,
t
is not
possible
to
attend
to those
structures
hat
earners
o not
attempt
o use
(i.e.,
extensivenstruction
cannot deal
effectively
ith
voidance).
Also,
of
course,
t does not
provide
he
n-depth ractise
hat
ome structures
ayrequire
before
they
an
be
fully cquired.
Arguably,
rammar
eaching
eedsto
be conceived f
n
terms fboth
approaches.
herefore,
rammar
eaching
eeds to be
reconceptualised
in
teacher
handbooks
to include the kind
of
extensive reatment
f
grammar
hat
rises
naturallyhrough
orrectiveeedback.
IS
THERE ANY
VALUE IN TEACHING
EXPLICIT
GRAMMATICAL
KNOWLEDGE?
The
distinction
etween
xplicit
nd
implicit nowledge
was men-
tioned
briefly
arlier.
xplicit
nowledge
onsists f
thefacts hat
peakers
ofa
language
have
earned.These facts re often
ot
clearly
nderstood
and
may
e
in
conflict
ith ach other.
hey
oncern ifferent
spects
f
language ncluding rammar. xplicit nowledgesheldconsciously,s
learnable nd
verbalisable,
nd is
typically
ccessed
through
ontrolled
processing
hen
earners
xperience
ome
kind
f
inguistic
ifficulty
n
using
the
L2.
A distinction eeds to be
drawnbetween
xplicit
knowl-
edge
as
analysed
nowledge
nd as
metalinguisticxplanation.
nalysed
knowledge
ntails
conscious wareness
fhow structuraleature
orks,
while
metalinguistic
xplanation
onsists
f
knowledge
f
grammatical
metalanguage
nd
the
ability
o understand
xplanations
f
rules.
n
contrast,
mplicit
nowledge
s
procedural,
s held
unconsciously,
nd can
only
be
verbalized
f
t s made
explicit.
t is accessed
rapidly
nd
easily
and thus s available oruse in rapid,fluent ommunication. ostSLA
researchers
gree
that
competence
n an
L2
is
primarily
matter
f
implicit
nowledge.
CURRENT ISSUES
IN THE TEACHING
OF
GRAMMAR 95
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
15/26
Whether here s
any
value
n
teaching
xplicit
nowledge
f
grammar
has been
and remains
oday
one of the most controversialssues
n
teaching rammar.o make sense of the differentositions elatingo
the
teaching
f
explicit
knowledge,
t is
necessary
o considerthree
separate
uestions:
1. Is
explicit
nowledge
f
any
value
n
and of tself?
2.
Is
explicitknowledge
f
value
in
facilitating
he
development
f
implicit nowledge?
3. Is
explicit
nowledge
est
taught
eductively
r
inductively?
I
partly
ddressed
he
first
uestion
when
consideredwhat
rammar
to teach.
noted that esearchers
isagree
ver
earners'
bility
o earn
explicit nowledge, ith ome (e.g.,Krashen, 982) seeingthis s very
limited nd
others
e.g.,
Green &
Hecht,
1992)
producing
vidence o
suggest
hat t s considerable. here
s,however,
separate
ssuerelated
to
thefirst
uestion.
his ssue oncerns heextent o
which earners re
able to
use
their
xplicit nowledge
whatever
hat
onsists
f)
in
actual
performance. gain,
ne
position
s that his
bility
s
limited. rashen
argues
hat
earners
an
only
use
explicit nowledge
hen
they
monitor,
which
equires
hat
hey
re focused n
form
as
opposed
to
meaning)
and have
sufficientime
to
access the
knowledge.
here
is
also
some
evidence that
teaching xplicitknowledge y
itself
i.e.,
without
ny
opportunities
or
ractising
he
target
eature)
s noteffective.tudies
y
VanPatten nd Oikennon
1996)
and
Wong
2004)
indicate hat
xperi-
mental
groups
that
received
xplicit
nformationlone
performed
o
differently
n
interpretation
nd
production
ests
han control
roup
did. But other
positions
re also
possible.
have
argued
that
explicit
knowledge
s used
in the
process
f
formulating
essages
s well
as
in
monitoring
nd that
many
earners re adroit
n
accessing
heir
xplicit
memories or
these
purposes,
specially
f the rules
are,
to
a
degree,
automatised.
However,
his
does
require
time.Yuan and
Ellis
(2003)
showed that earners'grammatical ccuracy mproved ignificantlyf
they
had
timefor
on-line
lanning
hile
performing
narrative
ask,
result most
readily explained
in
terms of
their
accessing
explicit
knowledge.
Irrespective
f whether
xplicit
knowledge
as
any
value
in
and
of
itself,
t
may
ssist
anguagedevelopment
yfacilitating
he
development
of
mplicit nowledge.
his
ssue
s
addressed
y
the
second
of the two
questions.
t
concernswhathas become known
s the
nterfaceypothesis,
which addressesthe role
explicitknowledge lays
n
L2
acquisition.
Three
positions
an
be identified.
ccording
o the
noninterface
osition
(Krashen,1981), explicit nd implicit nowledge re entirely istinct
with heresult hat
xplicit nowledge
annot
be converted
nto
mplicit
knowledge.
This
position
s
supported by
research
suggesting
hat
96
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
16/26
explicit
nd
implicit
memories
re
neurologicallyeparate
(Paradis,
1994).
The
interfaceosition
rgues
he exact
opposite.Drawing
n
skill-
learning heory, eKeyser 1998) arguesthatexplicitknowledge e-
comes
mplicit
nowledge
f
earners ave
the
opportunity
or
plentiful
communicative
ractice.
he weak
nterface
osition
Ellis, 1993)
claims
that
explicit
knowledge
an convert nto
implicitknowledge
f the
learner s
ready
o
acquire
the
targeted
eature
nd
that
his
onversion
occurs
by
priming
number f
key cquisitional rocesses,
n
particular
noticing
nd
noticing
he
ap
Schmidt, 990).
That
s,
explicit nowledge
of a
grammatical
tructure
akes t more
ikely
hat
earnerswill ttend
to the
structure
n the
input
and
carry
ut the
cognitive
omparison
between
what
they
bserve
n
the
input
nd their wn
output.
These
positionsontinue o be argued t a theoreticalevel.Althoughhere s
plentiful
vidence
hat
xplicit
nstruction
s
effective
n
promoting
2
learning
e.g.,
Noms
&
Ortega,
2000)
no
published tudy
as
directly
tested
whether
xplicit
nowledge
onverts
irectly
nto
mplicit
nowl-
edge
or
simply
acilitates
ts
development.
ne reason for the
lack
of
research
s the
problem
of
measurement,
hat
is,
the
difficulty
f
ascertaining
hich
ype
f
knowledge
earners
mploy
when
they er-
form
language
ask
r test.
The three
positions
upport
ery
ifferent
pproaches
to
language
teaching.
he noninterface
osition
eads
to a zero
rammarpproach,that
s,
t
prioritizes
eaning-centredpproaches
uch s immersionnd
task-based
eaching.
he interface
osition upports
PP
the dea that
a
grammatical
tructure
hould
be
first
resented xplicitly
nd then
practised
ntil
t s
fully
roceduralised.
he weak nterface
osition
lso
lends
upport
o
techniques
hat
nduce earners
o
attend o
grammati-
cal features.
t has been
used to
provide
basis
for
onsciousness-raising
tasks hat
require
earners o derive heir
wn
explicit rammar
ules
from ata
they
re
provided
with
Ellis,
1993;
Fotos,
994).
t s
ikely
hat
all three
approaches
will
continue to attract
upporters,
rawing
n
differentheoriesof L2 acquisition nd citingresearch that lends
indirect
upport
to
the
preferred pproach.
It is
unlikely
hat this
controversy
ill
be resolved
hrough
esearch
n
the near future.
The third
uestion
ssumes here s value
n
explicit nowledge
nd
addresses
how
best
to
teach it.
In
deductive
eaching, grammatical
structure
s
presented
nitially
nd then
practised
n
one
way
r
another;
this s the first
in
the
present-practise-produce
equence.
n
inductive
teaching,
earners re
first
xposed
to
exemplars
f the
grammatical
structure
nd are
asked to arrive
t a
metalinguisticeneralisation
n
their
wn;
here
may
r
may
notbe a final
xplicit
tatementf
the
rule.
A number f studies see Erlam, 003,for review) ave examined he
relative ffectiveness
f these wo
pproaches
o
teaching xplicit
nowl-
edge.
The results ave
been
mixed.
or
example,
Herron
nd
Tomosello
CURRENT
ISSUES IN
THE TEACHING OF GRAMMAR 97
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
17/26
(1992)
found clear
advantage
or
nductive
nstruction,
obinson
(1996)
found hat deductive
pproach
asmore
ffective,
hile osa
and O'Neill (1999) foundno significantifferencen effectiveness.
Erlam's
2003)
own
tudy
evealed
significant
dvantage
or he
group
receiving
eductivenstruction.
erhaps
hemain esson
o be
learned
from he
researcho date
s
theneedfor differentiated
pproach
o
both
esearching
nd
teaching
xplicit
nowledge.
t
s
ikely
hat
many
variables
ffect
hich
pproach
earners enefit
most
rom,
ncluding
the
specific
tructurehat s the
target
f the instruction
nd the
learners'
ptitude
or
grammaticalnalysis.imple
ules
may
estbe
taught
eductively,
hilemore
complex
ules
may
best
be
taught
inductively.
earners
killed
n
grammaticalnalysis
re
likely
o fare
better ith ninductivepproachhan hose ess killed.
IS THERE A
BEST WAYTO TEACH
GRAMMAR
FOR IMPLICIT KNOWLEDGE?
To
answer his
uestion
t
s
necessary
o
identify
he nstructional
options
or
eaching rammar.
have
ttempted
his
n
a number
f
publications
e.g.,
llis
1997,
998,
002b).5
will
onsider
ust
two:
he
differenceetweennput-basedndproduction-basednstructionnd
between
ifferent
ypes
fcorrectiveeedback.
The case
for he
nput-basedption
s based
on a
computational
model f
L2
acquisition,ccording
o
which
cquisition
akes
lace
s a
product
f
learners
omprehending
nd
processing
nput.
uch
ap-
proaches,
hen irectedt
grammar,
eek odraw
earners'ttentiono
the
argeted
tructure
s)
in
one or more
ways:imply
y ontriving
or
numerous
xemplars
f the structure
s)
to be
present
n
the
nput
materials,
y
highlighting
he
target
tructure
s)
in some
way e.g.,
by
using
oldor talics
n
written
exts)
or
by
means f
nterpretation
asks
(Ellis, 995)directed tdrawingearners'ttentionoform-meaning
mappings.
anPatten
1996,
003)
has
developed
version
f he
nput-
based
option
hat e calls
nput rocessing
nstruction.
his s directed
t
helping
earners o overcome he
defaultrocessing
trategies
hat re
a
feature f
interlanguagese.g., assuming
hat
the first
oun
in a
sentence
s
always
he
gent)
A
casefor he
utput-based
ption
anbe
found
n
both
kill-buildingheory
see
previous
iscussion)
r
in a
sociocultural
heory
f
L2
learning,ccording
o which
earning
rises
5
1 distinguishbetween psycholinguisticnd methodologicaloptions (cf. Ellis, 1998).
Psycholinguistic
ptions
re
related to some model of
L2
acquisition.
Methodologicalptions
re
evident n
instructionalmaterials
or
eachinggrammar.
Here
I
consider
only
psycholinguistic
options.
98
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
18/26
out
of social nteraction hich caffoldsearners'
ttempts
o
produce
new
grammatical
tructures
Ohta, 2001).
A
number
of
studieshave
compared herelativeffectivenessf nput-basedndproduction-based
instruction,
ith
mixed
results,
esulting
n
ongoing
debate
about the
relative
merits fthese wo
ptions
VanPatten,
002;
DeKeyser,
alaberry,
Robinson,
&
Harrington,
002).
It
may
be
that,
n
classrooms,
his
comparison
s
ultimately eaningless
ecause,
n
practise,
oth
options
are
likely
o nvolve
nput-processing
nd
production.
or
example,
t s
quite
conceivable
hat
n
an
input-basedpproach,
ndividual
tudents
silently roduce
the
target
tructure,
hile
n
a
production-based
p-
proach,
an utterance
produced by
one student erves as
input
for
another.
t
is,
therefore,
ot
surprising
hat both
options
have been
shown o resultnacquisition.6
There is a rich
descriptive
iterature n corrective
eedback
i.e.,
teacher
responses
o learner
rrors)
but
remarkably
ew tudies
have
investigated
he relative ffectsf different
ypes
f feedback
n
acquisi-
tion.
Key ptions
re
(a)
whether hefeedback s
mplicit
r
explicit
nd
(b)
whether he
feedback s
nput
r
output
ased.
mpliciteedback
ccurs
when
hecorrective
orce f the
response
o earner rror
s
masked,
or
example,
recast,
hich
reformulates
deviant tterance
orrecting
t
while
keeping
he same
meaning:
NNS:Why e svery nhappy?
NS:
Why
s he
very
nhappy
NNS: Yeah
why
s
very nhappy? Philp,
2003)
Or,
as
in
this contrived
xample,
a
request
for clarification:
NNS:
Why
he is
very nhappy?
NS:
Sorry?
NNS:
Why
s
he
very nhappy?
Expliciteedbackakes number fforms,uch as direct orrection r
metalinguisticxplanation.
here is some evidencethat
explicit
eed-
back
s more effective
n
both
eliciting
he earner's mmediate
orrect
use
of the tructurend
in
elicitingubsequent
orrect
se,
for
xample,
in a
post-test
Carroll
& Swain
993;
Lyster
004).
But
omeevidence
nd
6
There is
also
controversyegarding
how to measure the effectivenessf
these two
and
other)
instructional
ptions.
Norris nd
Ortega
(2000)
have shown that
the effectivenessf
instruction aries
depending
on whether t is measured
using
metalinguisticudgements,
selected
response,
onstrained
onstructed
esponse,
r freeconstructed
esponse.
Most
SLA
researchers and teachers, too, perhaps) would consider the last of these the most valid
measure.Ellis
2002a)
reviewed
number f
tudies hat xamined the effects f
different inds
of
instruction n learners'freeconstructed
esponses, eporting
hat
nstruction an have an
effect
n this
ype
f
anguage
use.
CURRENT ISSUES
IN THE TEACHING OF GRAMMAR
99
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
19/26
some
strong
heoretical easons xist o
support mplicit
eedback
see
Long
1996,
n
press).
ndeed,
this
ype
f feedback
s more
compatible
with hefocus-on-formpproachdiscussed arlier ecause tensures hat
learnersare more
likely
o
stay
focused on
meaning.
However,
s
Muranoi
2000)
notes,
mplicit
eedback
s
probably
more ffectivehen
it
is
targeted ntensively
t a
preselected
ormthan
when
it occurs
extensively
n
ncidental
ocus n form.
n
the
atter,
xplicit
ttention
o
form
may
be moreeffective.
Input-basedeedback
odels the
correct
orm
orthe earner
e.g., by
meansof
a
recast).
Output-based
eedback
licits
roduction
f
the
correct
form rom he learner
e.g., by
means of a
request
for
clarification).
Again,
here
s
disagreement
bout therelative ffectiveness
f
these
wo
feedback ptionsand no clear evidenceforchoosingbetween hem.
Some
descriptive
tudies ave
hown hat
utput-based
eedback
s more
likely
o lead to learners
orrecting
heirown initial
rroneous
utter-
ances
n
what
s referredo as
uptake.
owever,
ptake
s not the ame
as
acquisition.
In
short,
lthough
considerable
progress
has been
made toward
identifying
hose instructional
options
that are
likely
to
be
of
psycholinguistic
ignificance,
s
yet,
ew onclusions
an
be drawn bout
which nes are themost ffective
or
cquisition.
t
s
possible
o
point
o
studies nd theoretical
rguments
hat
uggest
hat
ach of the
major
options
iscussed an contributeo
acquisition.
SHOULD
GRAMMAR
BE
TAUGHT
IN SEPARATE LESSONS
OR INTEGRATED
INTO
COMMUNICATIVE
ACTIVITIES?
In Ellis
(2001)
I
considered
three broad
types
of form-focused
instruction,
s
shown
n Table
1.
Focus
on forms
efers o
instruction
involving structure-of-the-daypproach,wherethestudents' rimary
focus
s
on
form
i.e.,
accuracy)
nd where
the activities
re
directed
intensively
t
a
single
grammatical
tructure.
his
approach,
then,
involves
eaching rammar
n
a
series f
separate
essons.
Focus n
form
entails
focus
on
meaning
with ttention
o form
rising
ut of
the
communicative
ctivity.
hisfocus an
be
planned,
here
focused
ask
s
required
to elicit occasions
for
using
a
predetermined
rammatical
structure,s,
for
xample,
n
Samuda
2001).
In
this
pproach,
ttention
to
the
predetermined
rammatical
tructures
ill also
be
intensive.
Alternatively,
ocus
n form an
be
incidental
here
ttention
o form
n
thecontext fa communicative
ctivity
s not
predetermined
ut rather
occurs
n
accordancewith he
participants'
inguistic
eeds
s the
ctivity
100
TESOL
QUARTERLY
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
20/26
TABLE
1
Types
of
Form-Focused nstruction
Type
Primary
ocus
Distribution
1. Focus n forms
Form
Intensive
2.
Planned ocus n
form
Meaning
Intensive
3.
Incidentalocus
n form
Meaning
Extensive
Note. his
able s-
dapted
rom
llis
2001,
p.
17).
proceeds.
n this
pproach,
t
is
likely
hat ttention ill
be
given
o
a
wide
variety
f
grammatical
tructures
uring ny
one
task nd thus
will
be extensive. ocus on form
mplies
no
separategrammar
essonsbut
rather
grammar
eaching ntegrated
nto a
curriculum
onsisting
f
communicative
asks.
There
is
considerable
heoretical
isagreement egarding
whichof
these
ypes
f nstruction
s most
ffective
n
developing
mplicit
nowl-
edge. Long
(1988,
1991)
and
Doughty
2001)
have
argued
trongly
hat
focus
n form
s
best
equipped
to
promote nterlanguage
evelopment
because
the
acquisition
of
implicitknowledge
ccurs as a
resultof
learners
ttending
o
linguistic
orm
t the same time
hey
re
engaged
withunderstandingnd producingmeaningfulmessages.Other re-
searchers,however,
ave
argued
that a focus-on-forms
pproach
is
effective.
eKeyser
1998),
for
example,
has
argued
that
grammatical
structuresre learned
gradually hrough
he automatisationf
explicit
knowledge
nd that
his an be
achieved
by
means
of
a
focus-on-forms
approach.
This
approach
acknowledges
he value of
teaching
xplicit
knowledge
nd
subsequently roceduralising
t
by
means of
activities
(drills
nd
tasks)
that
practise
ehaviours
i.e.,
involve
meaning)
rather
than
structures.t is worth
oting,
owever,
ne
point
of
agreement
n
thesedifferent
ositions:
nstruction
eeds to ensurethat
earners re
able to connectgrammatical orms o the meanings heyrealise n
communication.
o
far,
hedebatehas addressed he
difference
etween
focus
n form nd focus n
forms.
herehas been ittle
iscussion fthe
relativemerits
f
planned
and incidental
ocus
on form. n
effect,
his
discussion
would nvolve consideration f whether
nstructionhould
be intensive r
extensive,
question
we have
already
onsidered.
CONCLUSION
Grammar asheldand continues o hold a central
lace
in
language
teaching.
he zero
grammar pproach
was flirted ith ut
never
eally
CURRENT SSUES
IN THE TEACHINGOF
GRAMMAR
101
This content downloaded from 147.91.1.45 on Mon, 11 Aug 2014 07:33:07 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp8/10/2019 Rod Ellis a SLA Perspective
21/26
took
hold,
s is evident
n
both he urrent extbookmaterials
manating
from
ublishing
ouses
(e.g.,
Whitney
White,
001)
and
in
current
theories fL2 acquisition. here sampleevidence o demonstratehat
teaching rammar
orks.
Although
here s
now
clear onvictionhat traditional
pproach
o
teaching rammar
ased on
explicit
xplanations
nd drill-like
ractice
is
unlikely
o result n
the
acquisition
f the
mplicit
nowledge
eeded
for
luent nd accurate
ommunication,
here
ontinues
o
be
disagree-
ment
egarding
hat hould
replace
this. t seems
ppropriate,
hen,
o
finish
with statement f
my
own beliefs bout
grammar
eaching,
acknowledging
hat
many
f them emain
ontroversial:
1. The
grammar aught
hould
be one that
mphasises
ot
ust
form
butalso the
meanings
nd uses of different
rammatical
tructures.
2.
Teachers hould ndeavour
o
focus
n those
grammatical
tructures
that
re known
o
be
problematic
o earners ather
han
ry
o teach
the
whole
of
grammar.
3. Grammars best
taught
o earnerswho
have
already cquired
ome
ability
o use the
anguage
i.e.,
intermediate
evel)
rather han to
complete beginners.
However,
rammar
an be
taughtthrough
corrective eedback s soon
as
learners
egin
to
use the
anguage
productively.
4. Afocus-on-forms
pproach
s valid s
long
as it ncludes n
opportu-
nity
or earners
o
practise
ehaviour
n
communicative
asks.
5. Consideration hould be
given
to
experimenting
ith a
massed
rather handistributed
pproach
to
teaching rammar.
6. Use should be made of both
nput-based
nd
output-based
nstruc-
tional
ptions.
7.
A
case
exists or
eaching xplicit
rammatical
nowledge
s a means
of
assisting
ubsequent
cquisition
f
mplicit
nowledge.
eaching
explicit nowledge
an
be
incorporated
nto
both
a focus-on-forms
and a focus-on-form
pproach.
In the case of a focus-on-forms
approach,
differentiated
pproach
nvolving
ometimes