ArticlePDF Available

Teaching Expressive Writing to Students with Learning Disabilities: A Meta-Analysis

Authors:

Abstract and Figures

We present results of a mete-analysis on writing interventions for students with learning disabilities and draw implications for practice. 13 studies designed to teach students with learning disabilities to write better expository or narrative text were analyzed. Results indicated that the interventions used in the research studies consistently produced strong effects on the quality of students' writing as well as students' sense of efficacy and understanding of the writing process. Findings suggested that 3 components should be part of any comprehensive instructional program. Explicit teaching of (a) the steps of the writing process and (b) the critical dimensions of different writing genres should be provided, as well as (c) structures for giving extensive feedback to students on the quality of their writing from either teachers or peers.
Content may be subject to copyright.
Teaching Expressive Writing to Students with Learning Disabilities: A Meta-Analysis
Author(s): Russell Gersten and Scott Baker
Source:
The Elementary School Journal,
Vol. 101, No. 3, Special Issue: Instructional
Interventions for Students with Learning Disabilities (Jan., 2001), pp. 251-272
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/1002247 .
Accessed: 31/10/2014 17:18
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://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
content in 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 support@jstor.org.
.
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The
Elementary School Journal.
http://www.jstor.org
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
Teaching
Expressive
Writing
to Students
with
Learning
Disabilities:
A
Meta-Analysis
Russell Gersten
Scott
Baker
Eugene
Research
Institute
University
of Oregon
The
Elementary
School
Journal
Volume
101,
Number 3
C
2001
by
The
University
of
Chicago.
All
rights
reserved.
0013-5984/2001
/10103-0002$02.00
Abstract
We
present
results
of a
meta-analysis
on
writing
interventions
for students
with
learning
disabil-
ities and draw
implications
for
practice.
13 stud-
ies
designed
to teach
students
with
learning
dis-
abilities
to
write better
expository
or narrative
text
were
analyzed.
Results
indicated that the
in-
terventions
used
in
the research
studies consis-
tently
produced
strong
effects
on the
quality
of
students'
writing
as well as students'
sense of
efficacy
and
understanding
of the
writing pro-
cess.
Findings suggested
that 3
components
should
be
part
of
any
comprehensive
instruc-
tional
program.
Explicit teaching
of
(a)
the
steps
of the
writing
process
and
(b)
the critical dimen-
sions
of
different
writing
genres
should be
pro-
vided,
as well
as
(c)
structures
for
giving
exten-
sive
feedback
to students
on the
quality
of their
writing
from either teachers
or
peers.
Over
the
past
15
years,
innovative
research
in
special
education
has
developed exciting
new
methods for
teaching
students with
learning
disabilities to write
essays.
This re-
search
has
explored
ways
to
teach students
how to
analyze
material learned
in
the
classroom
and to write both
personal
nar-
ratives,
based on
students'
own
interpreta-
tions of life
experiences,
and
persuasive
es-
says,
in
which students
take
positions
on
topical
social
and
political
issues.
Initially
stimulated
by
studies
in
reading
compre-
hension,
research
on
teaching expressive
writing increasingly
has taken on a
life
of
its own. In
the
past
10
years,
it
has
sur-
passed
research
topics
such
as
reading
com-
prehension
in
breadth
of
coverage
as well
as instructional
implications
for
improving
classroom
practice
(see,
e.g.,
Gersten,
Wil-
liams,
Fuchs,
&
Baker, 2000;
and
Mastropi-
eri,
Scruggs,
Bakken,
&
Whedon,
1996,
for
contrast).
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
252
THE
ELEMENTARY
SCHOOL
JOURNAL
There are several reasons for the
emer-
gence
of this
body
of
high-quality
research.
Interest
in
developing
methods for
teaching
students with
learning
disabilities to
ex-
press
their own ideas was influential from
the
beginning.
Writing,
of
course,
is
an
ex-
cellent vehicle
for
expressing personal per-
spectives
and
opinions
and
serves
a
legiti-
mate academic
purpose.
Even more factual
expository writing
allows
students
to dem-
onstrate
their
unique perspectives
on,
and
understanding
of, social,
political,
and his-
torical issues.
Early
researchers viewed instruction
in
expressive writing
as a
way
to
expand
spe-
cial education
teaching
to include activities
both
cognitively demanding
and intrinsi-
cally motivating.
Indeed,
many
of
the
in-
structional
approaches
developed by spe-
cial
education
researchers,
such as
Englert,
Raphael,
Anderson,
Anthony,
and
Stevens
(1991)
and Graham and Harris
(1989),
have
played
a
major
role
in
shaping
classroom
practice
in
general
education.
Clearly,
there is
a
strong
need for
in-
structional
strategies
that
help
students
with
learning
disabilities write
more effec-
tively.
On
every
conceivable measure of
writing
performance-including
both
mea-
sures
of
writing
quality
and
quantity,
and
occurring
across narrative and a
range
of
expository
text
structures-students
with
learning
disabilities write much
more
poorly
than do
students without
disabilities
(Englert
et
al., 1991;
Graham &
Harris,
1997).
In
summarizing
their
writing
skills
compared
to their
peers,
when
students
with
learning
disabilities
write,
"little
atten-
tion is
directed
to
the
needs
of
the
audience,
the
organization
of
text,
the
development
of
rhetorical
goals,
or the
constraints
imposed
by
the
topic"
(Graham
&
Harris,
1997,
p.
414).
A
shift
in
thinking
also
helped
spur
the
rapid
growth
of
research
in
expressive-writ-
ing
instruction.
Early
evidence
suggested
that
the academic
strengths
of
students with
learning
disabilities could be
elicited more
effectively
if
writing
instruction focused
less
on
mechanics
(e.g.,
legibility
and
punc-
tuation)
and
more on
content.
Goldman,
Hasselbring,
and the
Cognition
Technology
Group
at Vanderbilt
University
(1997)
sum-
marized
this shift
by
noting,
"Research
studies indicate
that
when asked to write
about
complex
ideas,
students
with
learn-
ing
disabilities often demonstrate
concep-
tual
performance
that far
exceeds
what
would be
predicted
based
on
their
perfor-
mance
on
lower level
skills such as
capital-
ization,
punctuation
and
spelling" (p.
203).
It was
special
education
research
in
expres-
sive
writing
that first
clearly
demonstrated
this
phenomenon.
Also,
writing
instruction was viewed as
a
means to
help
students
understand
the
linkages
between
reading
and
writing
(En-
glert, Raphael,
&
Anderson, 1992;
Graves,
1978,
1983).
As students
began
learning
how
different
types
of texts are
structured,
this
helped
them better
understand what
they
were
reading.
Therefore,
it
made
sense
to
encourage
them to use
their
growing
awareness of text
structure
in
their own
writing assignments.
Students
learned,
for
example,
that
they
could use
the
major
ele-
ments of
the
stories
they
read--characters,
settings, problems-in
their
own
stories to
make them more
enjoyable
to read and to
write.
A
final
reason for
the shift from com-
prehension
research to
expressive
writing
research
may
be that
expressive
writing,
though
quite
difficult to
assess
in
a valid
fashion,
may
be
somewhat easier to
assess
than
reading
comprehension
(see,
e.g.,
Kucan &
Beck,
1997).
For
example,
prod-
ucts
of
writing
ability
are
easier to
collect
than
products
of
reading comprehension,
thus
making
it
easier
for
assessment
experts
to work on
developing
tasks
that
not
only
meet traditional
criteria for
reliability
and
validity
but
also
satisfy
current
conceptu-
alizations
regarding
assessment
task au-
thenticity.
In
fact,
the measures used
in
ex-
pressive
writing
research
often
have
served
as models for
means to
validly
assess writ-
ing samples
used
in
state assessments.
Not
JANUARY
2001
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
EXPRESSIVE WRITING 253
surprisingly,
students
with
disabilities
are
increasingly
included
in
high-stakes
writ-
ing
assessments in
many
states.
By
contrast,
developing
valid
measures of
reading
com-
prehension
has
remained
an
elusive
goal
(Garcia
&
Pearson, 1994;
Kucan &
Beck,
1997).
The
purpose
of
the
present synthesis
was to
summarize
research conducted
on
interventions
in
expressive
writing
for
stu-
dents
with
learning
disabilities. We
believe
there is
a
need
for this
type
of
synthesis,
as
interventions have
incorporated
different
components,
used
different
expressive
writ-
ing
text
structures,
and
employed
a
variety
of
approaches
for
teaching
students
the es-
sentials
of
writing.
This
fruitful
area of
re-
cent
research has
incorporated
intervention
components
associated
with
both
the
effec-
tive
teaching
literature
of
the
1980s
(Brophy
&
Good,
1986;
Gersten,
Woodward,
&
Darch,
1986),
as well
as
insights
from
more
cognitive-oriented
breakthroughs
of
the
1990s
(Bransford, Brown,
&
Cocking,
1999;
Englert
et
al.,
1991;
Harris &
Pressley,
1991;
Kucan &
Beck,
1997).
For
this
meta-analysis,
we
defined
ex-
pressive
writing
as
writing
for
the
purpose
of
displaying
knowledge
or
supporting
self-
expression
(Graham
&
Harris,
1989).
In
the
studies
we
analyzed,
students
used
writing
to
demonstrate
their
knowledge by
de-
scribing,
informing,
and
convincing.
Self-
expression
was
promoted
by
giving
stu-
dents
opportunities
to
select
their own
topics
for
journal
writing, personal
narra-
tives,
and
stories.
In
the
words
of
Isaacson
(1994,
p.
40),
expressive
writing
"can
be
considered a
product
of
the
interaction
among
the
task,
the
learner,
and
the
instruc-
tional
environment."
Butler,
Elaschuk,
and
Poole
(in
press)
provide
a
vivid
description
of
the
goals
of
expressive
writing
instruction.
They
note
that
"effective writers
are
self-regulating...
They
analyze
task
requirements,
articulate
writing
goals,
and
then
select,
adapt,
or
even invent
strategic
approaches
to
achieve
their
objectives.
They
monitor
the
success
of
their efforts
as
they
engage
recursively
in
planning,
text
production,
and
editing
ac-
tivities.
If obstacles
are
encountered,
effec-
tive writers
adaptively adjust goals
...
or
writing strategies."
As
will be
seen,
each
of
the studies
in the
meta-analysis
attempts
to
help
students
become
more effective
writers
according
to the
dimensions
outlined
in
the
Butler
et al.
(in
press)
definition.
Our
criteria
for
including
studies
in
this
synthesis
were
that
they
focus
on
students
with disabilities
and
on
instruction
target-
ing
the
writing process
(i.e.,
composing,
ed-
iting,
revising)
rather
than
writing
mechan-
ics. In
other
words,
for
the
purposes
of
this
meta-analysis,
we
intentionally
exclude
studies
that address
writing
mechanics
only.
Rather,
we follow
the
conceptualiza-
tion of Butler
et al.
(in
press)
and
Wong
(1994).
Method
Search Procedures
Our search
procedure
was
organized
around
the
following question:
Of
those
studies
designed
explicitly
to
improve
the
writing
of
students with
learning
disabili-
ties,
which interventions
and
intervention
components
are most
effective,
and
what
is
the
strength
of
those
effects?
We
employed
three
procedures
to
iden-
tify
studies for
the
meta-analysis.
First,
us-
ing
the
reference list from a
recent
study
in-
vestigating
the
effectiveness of
all
types
of
interventions
on the
academic,
social,
and
cognitive
outcomes of
students with
learn-
ing
disabilities
(Swanson,
Hoskyn,
&
Lee,
1999),
we
identified
those
studies that in-
cluded
at least
one
dependent
measure of
expressive
writing.
Swanson
et al.
searched
all
relevant
on-line
databases,
including
PsycINFO,
MEDline,
and
ERIC,
from
1963
to
1997,
and
contacted
numerous
research-
ers
and
state
directors
of
special
education
to
locate
relevant
studies.
After
reading
these
potential
studies,
we
selected
for
closer
scrutiny
those in
which
the main
pur-
pose
of
the
intervention
was to
improve
the
quality
of
student
writing.
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
254
THE ELEMENTARY
SCHOOL
JOURNAL
The second
search
procedure
we
em-
ployed
was
to
manually
explore
all of
the
major
special
education
journals
(e.g.,
Jour-
nal
of
Learning
Disabilities,
Journal
of Special
Education,
Learning
Disability
Quarterly,
Learning
Disabilities
Research and
Practice,
Ex-
ceptional
Children)
and other
relevant
jour-
nals
(e.g.,
Journal
of
Educational
Psychology)
from
1998 and 1999.
For
the third
search
procedure,
we
invited
three
prominent
scholars in
the
area of
expressive
writing
to
review
our list of
studies and
assess it for
comprehensiveness;
we also
asked
them
to
suggest
other
published
studies,
disserta-
tions,
and
unpublished
papers
that
might
be included.
A
number of
potential
studies,
which we
otherwise would
have
missed,
turned
up
this
way,
including
two that
met
our
criteria for
inclusion.
Eligibility
Criteria
We included
both
experimental
studies
(i.e.,
those with
random
assignment
of stu-
dents to
instructional
conditions)
and
quasi-experimental
studies
(where
assign-
ment of
students to
conditions
was
not
ran-
dom,
but
groups
were shown to
be
equiv-
alent on
relevant
pretest
writing
measures)
if
they
met all of
the
following
criteria:
1. An
independent
variable
pertaining
to
written
expression
was a
primary
focus
of
the
study.
2.
The
intervention
lasted at
least
45
minutes and
occurred
across at
least 3
days
of
instruction.
3.
At
least
one
dependent
measure of
student
writing performance
was
collected.
4.
The
intervention
focused
on
improv-
ing
the
writing
disabilities of
students with
identified
learning
disabilities.
That
is,
stu-
dents
in
the
sample
were
assessed
by
the
school
district and
met the
district
and
state's
criteria
for a
learning
disability,
making
them
eligible
for
special
education
services.
5.
At
least
66%
of
the
sample
comprised
students
with
learning
disabilities.
If
the
percentage
was
less than
that,
separate
out-
come
data for
the
sample
of
students
with
learning
disabilities
were
presented.
6. The
study
included
a
comparison
group,
and students
in the
comparison
group
were
students
with
learning
disabil-
ities.
Coding
of
Studies
Thirteen
studies
met
the criteria
for
in-
clusion
in the
meta-analysis.
These
studies
and
major
descriptive
information
are
pre-
sented
in Table
1. Initial
coding
involved
determining
the
following
information:
(a)
whether
the
study
used
random
assign-
ment
of students
to
treatment,
(b)
number
of students
per
condition,
(c)
grade
level(s)
of
students,
and
(d)
whether
the focus
was
on
narrative
or
expository
writing.
The more
detailed
coding
scheme
for
the
set
of
studies
was
developed
over
the
course
of
rereading
each
study
a
number
of times
to
determine
major
categories
and
subcategories.
The
following
section
is
or-
ganized
around
the
two
major
categories
identified:
the
nature
of
the
instructional
intervention
and
the
measures used
to
as-
sess
performance.
Nature
of
the
instructional
intervention.
Three
studies
investigated
the effects of
cur-
riculum
programs
that
taught
students
to
think
and write
creatively.
Creativity
stud-
ies
were
included
in
the
meta-analysis
be-
cause
the intent
of these
studies,
like
the
in-
tent
of
the other
included
studies,
was
to
help
students
generate
quality
writing
through
the
use
of
effective
instructional
practices.
Though
these
studies
are
recog-
nized
within
the
meta-analysis,
they
have
been
categorized
as
a
separate
subset
of
studies
from
those
that
dealt
with
more
contemporary
approaches
to
writing
in-
struction.
The
expressive
writing
studies
were
coded
to
reflect
specific
aspects
of
the
in-
structional
intervention
that
were
utilized
in
the
experimental
and
comparison
groups.
We
classified
the
nature of
instruction
ac-
cording
to two
dimensions,
the
major
focus
of the
instructional
intervention,
and
the
more
detailed
instructional
components
central
to the
intervention.
The
way
the
JANUARY
2001
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
EXPRESSIVE WRITING
255
TABLE 1. Studies
in
Meta-Analysis
Number in
Experimental
and
Control
Groups
References
by
Random
Grade
Writing Category Assignment
Levels
E
C
Expository writing:
Englert
et
al.,
1991
No
4-5
33
22
Englert
et
al.,
1995 No
1-4
E,
= 22
31
E2
=
35
MacArthur et
al.,
1995 No
Elementary
110
58
Reynolds,
1986
No
6-8
E,
=
18
18
E2
= 17
Welch,
1992
No 6 7
11
Wong
et
al.,
1994
No
8-9
E,
=
10a
13
E2
=
8a
Wong
et
al.,
1996
No 8-9 18b
20
Narrative
writing:
De La Paz &
Graham,
1997a
Yes 5-7 11
10
Englert
et
al.,
1995
No 1-4
E,
=
22
31
E2
= 35
MacArthur
et
al.,
1995
No
Elementary
110
58
MacArthur
et
al.,
1991
No
4-6 13
16
Sawyer
et
al.,
1992
Noc
5-6
El
=
11
10
E2
= 11
E3
= 11
Creative
writing:
Fortner,
1986
No
3-6
25 24
Jaben,
1983
Yes
Intermediate 25
24
Jaben,
1987
Yes
Intermediate 25
25
aLD
means
not
analyzed
separately.
However,
78% of
students
in
experimental
condition had
LD,
as did
75%
of
students
in
the control
group.
bLD
means
not
analyzed
separately.
However,
90%
and
75% of
students
in
the
two
experimental
conditions
had
LD,
as did
100% of control
group.
cRandom
assignment
to three
experimental
conditions,
nonrandom
assignment
for
comparison
group.
studies were
coded on each
dimension
is
presented
in
the
appendix.
Major
focus of
the
intervention:
Each
study
was coded
to
determine
whether two
pri-
mary
features of
writing
instruction
were
included.
The first
was
text
structure,
which
was coded
as
present
if
students
received
instruction
in
how
to
write
following
a
spe-
cific
type
of
text
structure. For
example,
in
learning
to
write
in
a
narrative text
struc-
ture,
students
might
have
been
taught
to
in-
clude a
description
of
the
setting,
charac-
ters,
or
attempts
to
solve the
story's
main
problem.
Instruction in
expository
text
was
coded if
students were
taught
to write
ac-
cording
to
specific
expository styles.
Seven
studies
included
writing
instruction
that
targeted
a
specific
type
of text
structure.
The second
primary
feature was
whether
the
study emphasized
instruction that ad-
dressed
the
writing
process,
including
in-
struction
that
helped
students
plan,
orga-
nize,
and
actually
carry
out
the task of
writing.
These
"plans
of
action"
are
essen-
tially
an
encapsulation
of
the
procedures
skilled writers
use when
they
compose
(e.g.,
Englert
et
al., 1991;
Graham &
Harris,
1989).
Teaching
these
plans
of
action
has
been a
key
feature of
expressive
writing
research,
from
the first
seminal
studies
(e.g., Englert
et
al.,
1991;
Graham
&
Harris,
1989),
to the
most
recent
(e.g.,
De La
Paz
&
Graham,
1997a,
1997b;
Wong,
Butler,
Ficzere,
& Ku-
peris,
1997).
Ten of
the 13
studies
included
instruction
that
emphasized
the
writing
process.
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
256
THE ELEMENTARY
SCHOOL
JOURNAL
Detailed
instructional
components:
We
coded each
study
on nine features that
rep-
resented salient
aspects
of
writing
instruc-
tion
(see
the
appendix).
In
some cases these
features
were
part
of the
writing process
de-
scribed
above,
but
because their
presence
or
absence varied
considerably
among
studies,
and
they
were
considered
essential
compo-
nents of
the intervention
in
and of
them-
selves,
we
included
them in
separate
coding
categories.
Instruction
for
the
comparison group
was
coded
in
one
of two
ways:
(a)
writing
instruction was
provided,
but more tradi-
tional
methods
were
used than that
used
in
the
experimental
group,
or
(b)
no
writing
instruction was
provided,
that
is,
a
practice-
only
comparison
group
was
used. Students
in
the
no-instruction
comparison group
practiced
their
writing
skills
in a
variety
of
contexts set
up
by
the researchers and
spent
as much time on
writing
activities as stu-
dents
in
the
experimental
group.
When the
comparison
group
received
writing
instruc-
tion,
we coded that
instruction on
the same
dimensions as the
experimental
group.
Dependent
measures.
We
analyzed
two
types
of
dependent
measures for
writing
in-
struction:
(a)
actual
measures
of
student
writing
and
(b)
measures
that
examined stu-
dents'
understanding
of
the
process
of writ-
ing,
including
students'
views of
themselves
as
writers
(a
measure of
metacognitive
awareness).
These two
types
of
dependent
measures
were divided
into a
number of
categories
for
the
meta-analysis.
Student
writing:
All
13
studies
included
at
least one overall
measure of student writ-
ing quality.
Each
measure
was coded
as ei-
ther:
(a)
a
holistic,
qualitative
rating
of
the
total
writing
sample (e.g., qualitative
rating
on
a
Likert
scale)
or
(b)
a
total
score of
the
writing sample
based
on
the sum of
indi-
vidual
rubrics,
such as
organization,
coher-
ence,
and
content. In
the
latter
category
(i.e.,
the total
score of
the
writing
sample,
based
on
rubrics)
we
also
included
three
studies
that
used a
standardized
measure
of
writ-
ing,
such
as the Test of
Written
Language
(TOWL;
Hammill &
Larsen,
1978).
Stan-
dardized
measures
were included
in
this
category
because
their total
score is the
sum
of
individual
rubrics
(e.g., fluency,
flexibil-
ity,
and
originality).
Eight
studies
also assessed
student
writ-
ing performance
on
more molecular
aspects
of
the
writing sample.
These were
coded
ac-
cording
to one
of
three
categories:
1.
Quality
of written
language
conven-
tions,
including
traditional
elements,
such
as
punctuation
and
capitalization,
as well
as
length
and
reader
sensitivity.
2.
Thought
units,
or
T-units,
which
as-
sessed
the
number of
independent
pieces
of
information
in
the
sample.
3.
An
assessment
of
the
degree
to
which
specific
features of text
structure
(e.g., story
grammar
elements)
were
present
in
the
written
sample.
This
genre-specific
measure
fit
either
the
story
grammar
elements
of
nar-
rative text
structure
(Sawyer,
Graham,
&
Harris,
1992)
or some
aspect
of
expository
text
structure,
such as
compare/contrast
(Englert
et
al.,
1991).
We
calculated the
mean
of all
measures
of
writing performance
to determine an
ag-
gregate
writing
score. This
score was used
in
the
majority
of
calculations in
which an
overall effect
size was
computed
across
multiple
studies.
Attitude and
metacognition:
We also
coded
studies
according
to
students'
atti-
tudes
and
understandings
about
writing.
Although
these
categories
were not
based
on
actual measures of
student
writing, they
reflected
outcomes
of
writing
instruction
that
researchers
and
practitioners
consid-
ered
important.
Six
studies
contained
these
types
of
measures,
which we
coded on
the
following
three
dimensions:
1.
Student at-
titudes toward
writing;
2.
Students'
strat-
egy
use
while
writing,
measured via
meta-
cognitive
surveys;
and 3.
Measures of
self-efficacy.
Meta-Analysis
Calculations
The
basic index of
effect size
used
in
this
meta-analysis
was
Cohen's
d,
defined
as
the
JANUARY
2001
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
EXPRESSIVE
WRITING
257
difference between the
treatment and
com-
parison
group
means divided
by
the
pooled
standard
deviation
(Cooper
&
Hedges,
1994).
For
studies
that
reported
pretest
and
posttest
scores,
we
calculated
posttest
effect
sizes
adjusting
for
pretest
performance,
us-
ing
the
following equations
(Wortman
&
Bryant,
1985):
adjusted
effect size
=
unad-
justed
d -
pretest
correction;
pretest
correc-
tion =
(ME[pretest]
-
Mc[pretest])/SDpooled
[pretest].
Results
Effect
sizes are
summarized in
Tables
2,
3,
and 4.
Table 2
presents
the
effect sizes
re-
lated
to the
nature of
the
independent
vari-
able.
In
this
table,
we
present
both
weighted
and
unweighted
mean
effect sizes
for
each
feature.
Confidence
intervals
are
presented,
following procedures
in
Cooper
and
Hedges
(1994).
Effect
sizes for
the
dependent
measure
categories
are
presented
in
Tables 3
and 4.
Table 3
presents
the
effect
sizes and
sum-
mary
statistics for
the
writing performance
measures.
Table 4
provides
a
summary
of
weighted
effect
sizes
for
dependent
mea-
sures
regarding
student
attitudes and
un-
derstanding
of
writing.
Summed
across
all 13
studies,
the
mean
effect
size on
the
aggregate
writing
measure
was
0.81.
In
educational
research,
an
effect
size
of this
magnitude
is
typically
consid-
ered
a
strong
effect
(Cohen,
1988).
The
95%
confidence
interval
was
0.65-0.97,
provid-
ing
clear evidence
that
the
writing
interven-
tions
had
a
significant
positive
effect
on
the
quality
of
students'
writing.
The
weighted
and
unweighted
effect
sizes
were
0.81
and
0.99,
suggesting
that
the studies
with
larger
sample
sizes
resulted
in
somewhat
smaller
effects.
Positive
effect
sizes
were
found
in
each
of
the
13
studies
(see
Table
3),
ranging
in
magnitude
from
0.30
(MacArthur,
Gra-
ham,
Schwartz,
&
Schafer,
1995)
to
1.73
(Wong,
Butler,
Ficzere,
&
Kuperis,
1996).
Two factors
support
the
validity
of
the
findings
of
this
meta-analysis.
First,
weighted
and
unweighted
effect
sizes
for
the
aggregate
and
the
three
major
categories
(qualitative
rating
of the
sample,
trait
or
ru-
bric
score,
and
standardized
test
score)
were
all moderate
to
high.
Second,
the
confidence
intervals
around
each
of the
weighted
effect
sizes
consistently
indicated
that
the
writing
interventions
were at
least of
moderate
ef-
fect across
the different
writing
measures.
Taken
together,
these
effect-size
calcula-
tions
suggest
that
the
innovative
instruc-
tional interventions
in
writing produced
moderate
to
strong
effects.
TABLE 2.
Summary
of
Mean
Effect
Sizes and
Confidence
Intervals
(CI)
Related to
the
Nature
of
the
Independent
Variable
95%
CI
for
Mean
Effect
Sizes
Weighted
ES
N
k
Weighted
Unweighted
Lower
Upper
Experimental
study
feature:
No-instruction
control
3
8
1.39
1.34
.92
1.85
Minimal-instruction
control
10
49
.76
.94
.59
.94
True
experiment
3
14
1.19
1.19
.83
1.55
Quasi
experiment
10
45
.71
.93
.53
.89
Instructional
feature:
Collaborative
practice
with
teacher
5
23
.76
.85
.48
1.03
Collaborative
practice
with
peers
6
33
.70
.98
.49
.90
Teacher
modeling
of
strategy
use
9
36
.69
.90
.51
.88
Use of
procedural
prompts
8
33
.86
.96
.63
1.09
Use of
computers
4
24
.64
1.06
.38
.90
NOTE.-N
=
number
of
studies;
k
=
number of
effect
sizes
aggregated.
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
E
CC)
Cl)Cl
~-ON
>>
u
u
10
t:
w
-)
T
.4.%4-4
'~-~ cncq
Cl)
-
Cl)
z
(
a-
(V cz
C,
co
Co
t
CD
0
U Cl
o
)a\ co
D
0
N.
o
oqz
-0
%0
C
~ccu
C
N1%
mxm
z
N
>a4-
0
U
u
aN
(VO~
aNN
ON
aNN
ON4N-1
-
N
;:2
z
aN aN
wON
?
W
W
r--4
- ON ONO Q
O-tl el l Q4
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
EXPRESSIVE
WRITING
259
TABLE 4.
Summary
of Effect
Sizes
(Weighted)
for Measures
of
Students'
Regard
of the
Writing
Process
Attitudes
Metacognitive Survey
Study
toward
Writing (Self-Report)
Self-Efficacy
Englert
et
al.,
1991
1.05
Englert
et
al.,
1995
.53
MacArthur
et
al.,
1991
.23
Sawyer
et
al.,
1992
-.11
Welch,
1992
1.53
1.48
Wong
et
al.,
1994
-.13
.27
1.12
Wong
et
al.,
1996
.41
.62
.68
Unweighted
average
.60
.70
.56
Weighted average
.40
.64
.61
Study
Characteristics
Influencing
Magnitude
of
Effect Size
The difference
in
magnitude
of
effect
sizes found
between
the studies
by
Mac-
Arthur
et al.
(1995;
d
=
0.30)
and
Wong
et
al.
(1996;
d
=
1.73)
may
be due
less to the
quality
of the
intervention and
more to
the
measures
collected and
the
nature of
the
comparison
group.
The
comparison
group
in
the
MacArthur
et al.
(1995)
study
re-
ceived
more
extensive
instruction
in
writing
than the
comparison
group
in
the
Wong
et
al.
(1996)
study.
For
example,
comparison
students in
the
MacArthur
et
al.
(1995)
study
received
instruction in
the
writing
process,
whereas
the
comparison
group
stu-
dents in
the
Wong
et
al.
(1996)
study
did
not.
Also,
the
aggregate
effect size in
the
MacArthur et
al.
(1995)
study
was the
av-
erage
of
the
global
qualitative
rating
(0.49)
and
conventions
(0.11).
The
aggregate
effect
size
was
reduced
considerably
by
the
very
small
effect
the
intervention
had on
writing
conventions,
a less
important
aspect
of
the
intervention
than
the
overall
rating
of
the
writing
sample.
In
the
first row
of
Table
2,
we
contrast
effect
sizes
related to
types
of
comparison
groups.
Ten
studies
had
a
comparison
group
that
received
at
least
a
minimal
amount of
writing
instruction.
In
three
studies,
students in
the
comparison
groups
received
practice
in
writing
but
did
not
re-
ceive
any
actual
writing
instruction.
The
mean
weighted
effect
size for
the
10
studies
with
comparison
groups
that received
at
least minimal
writing
instruction
was
0.76.
The mean
weighted
effect
size
for the
three
studies
that included
comparison
groups
that received
practice
only
was
1.39
(un-
weighted
effect
size
=
1.34).
As
one
would
expect,
the
effect
size
was lower
when
the
comparison
group
received
at least
some
writing
instruction.
A
second
contrast
that
emerged
from
the
combination
of
specific
studies
was that
of
true
experimental
studies
(i.e.,
those
that
randomly
assigned
students
to
treatment
conditions)
versus
quasi
experiments
(those
that
did
not).
Three
studies
used
random
assignment,
and
10 did
not.
(In
the
study
by
Sawyer
et
al.,
1992,
students
were
randomly
assigned
to three
experimental
groups,
but
a
nonrandomized
procedure
was
used
to
assign
students
to the
comparison
group.
Therefore,
we
coded
the
study
as a
quasi
experiment.)
As shown
in Table
2,
true
experiments
had an
effect
size of 1.19
(unweighted,
1.19)
and
quasi
experiments
had
an
effect
size
of
0.71
(unweighted,
0.93).
Most
meta-analy-
ses
(e.g.,
Swanson
et
al.,
1999)
have
shown
lower
effect
sizes
when
true
experiments
are
used.
The
fact
that
this
is not
the
case
here
indicates
the
high
quality
of
the
quasi
experiments.
Also,
all
three
of
the
true
ex-
periments
were
creativity
studies
by
Jaben
(1983,
1987)
and
Fortner
(1986),
which
were
conducted
before
the
1990s,
when
writing
interventions
became
more
process
oriented.
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
260
THE ELEMENTARY
SCHOOL
JOURNAL
Effects
on
Overall
Writing
Quality
or
Performance
Each
study
included either
a
global
rat-
ing
of the
quality
of
posttest writing
sam-
ples
or
a total score based
on
the sum of
scoring
rubric
(including
traits
such
as
co-
herence
or
organization).
The
study by
En-
glert
et al.
(1995)
included both
types
of
scores.
In
three
cases,
standardized
writing
measures were
used,
and their
scoring pro-
cedures were followed to
obtain a total
score.
Thus,
for
each
study,
we have an
overall index
of
the
quality
of
the
students'
writing
sample.
Six
of
the
13
studies had an
overall-qual-
ity
score. The
weighted
mean effect size
was
0.67,
with
a
range
of
0.49
(MacArthur
et
al.,
1995)
to 1.33
(MacArthur, Schwartz,
& Gra-
ham,
1991).
The
effect size
representing
the
95% confidence
interval
was
0.47-0.88,
in-
dicating
that students
in
experimental
groups
did
significantly
better than stu-
dents
in
comparison
groups.
In
eight
studies we
calculated a
total
score based on the
mean
of
the rubrics
used
to score the
writing samples.
The
weighted
mean effect size was
0.98,
with
a
range
of
0.12
(Reynolds,
1986)
to
1.73
(Wong
et
al.,
1996).
The 95%
confidence
interval
was
0.76-1.20,
indicating
that students
in
exper-
imental conditions did
significantly
better
than students
in
comparison
conditions.
One
explanation
for the low
effect size
in
the
Reynolds
(1986)
study
supports
the
in-
tegrity
of
the
meta-analysis:
the
difference
in
instruction
between
experimental
and com-
parison
groups
in
that
study
was consider-
ably
less
than
in
any
of
the other 12
studies.
The low
effect
of
0.12
reflects
this
minimal
difference.
In
the
Reynolds
study,
the
differ-
ence
between
experimental
and
comparison
group
instruction was
reflected
by
the
mag-
nitude
of
the
effect size
in
the
conventions
category
(ES
=
0.79).
In
other
words,
the in-
tervention
did have
a
strong
effect on
aspects
of
writing
outcomes
that reflected
the
focus
of the
writing
intervention.
Three
studies used
standardized
writing
measures
to
assess
outcomes,
and 10
studies
used
experimenter-developed
measures.
The
primary
difference between the
two
is
that the standardized
measures
are
pub-
lished and include normative data on
per-
formance.
Both
rely
on
subjective
ratings
of
writing
quality
to
assess
performance.
All
of the
studies
implemented
training proce-
dures to
ensure that raters achieved
accept-
able levels
of
reliability
before
they
scored
students'
writing samples.
The mean
weighted
effect size of the
experimenter-developed
measures
of
overall
writing performance
was
0.74,
with
a confi-
dence
interval
of
0.55-0.92;
the
weighted
ef-
fect size
for
standardized
measures was
1.17,
and
the confidence
interval was
0.82-1.52.
That the mean
effect size was
higher
on
stan-
dardized
measures than
experimenter-
developed
measures is
important,
especially
in
the context of
previous
research
using
meta-analyses.
Larger
effect sizes
are
typi-
cally
found with
experimenter-developed
measures
versus
standardized
measures
(e.g.,
Swanson
&
Hoskyn,
1998),
and
the
concern is that
intervention
effects
may
be
inflated when
interpreting
performance
on
experimenter-developed
measures. Previ-
ous
meta-analyses
have shown
that
the de-
gree
of
alignment
between
independent
and
dependent
variables,
typically
reflected
in
experimenter-developed
versus
stan-
dardized
measures,
significantly
influences
the
magnitude
of
the effect
sizes
(Gersten
&
Baker,
in
press;
Swanson &
Hoskyn,
1998).
The fact
that
standardized
measures
pro-
duced
a
larger
effect
strongly
indicates
that
the
expressive
writing
studies
employing
experimenter-developed
measures
did not
have
artificially
inflated
effect sizes.
Creativity
Training
Studies
We
included
three
studies,
conducted in
the
1980s,
with
some
reluctance because
they
only
partially
addressed
writing
abilities.
Each
examined
the effects of
an
instructional
intervention
designed
to
promote
creative
thinking
and
creative
writing
among
stu-
dents
with
learning
disabilities
(Fortner,
JANUARY
2001
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
EXPRESSIVE WRITING
261
1986;
Jaben,
1983,
1987).
Because
writing
was a
key
medium of instruction
and
the
key dependent
measure,
we decided
to
in-
clude
them. The two studies
by
Jaben
(1983,
1987)
used the Torrance
Test of Creative
Thinking
(Torrance,
1974)
as the
primary
outcome
measure. Student
writing
on
this
measure was evaluated on the basis of
flu-
ency, flexibility,
and
originality.
In
the
Fort-
ner
study
(1986),
the Test of
Written
Lan-
guage
(TOWL;
Hammill &
Larsen,
1978)
was used
along
with
the
Torrance test.
Although
it is
easy
statistically
to
inte-
grate
these
three studies with
the
others,
they
are
quite
different
conceptually.
Each
used
a
packaged
creativity
curriculum but
provided
little
explanation
about what the
curriculum
entailed or
specification
about
how
it was
implemented
in
the
study.
Con-
sequently,
it
was difficult
to
code these
studies in
terms of
the
independent
vari-
able.
It
is
likely
that the
creativity
studies
used
instructional
procedures
quite
differ-
ent from
those used in
the
other 10
studies.
Yet,
the
creativity
studies did
produce
con-
sistent
positive
effects.
Because
they
used
standardized
measures
for
assessing
effects,
the
weighted
mean
effect size
was 1.17.
It
may
well
be that
the
types
of
instructional
approaches
used
in
the
"packaged"
creativ-
ity
curricula
include
important
components
that
help
students with
learning
disabilities
become
better
writers.
Individual
Writing
Components
Writing
conventions.
Four
studies
as-
sessed
aspects
of
writing
that we
classified
as
writing
conventions.
These
assessments
included
conventions
such
as
standard
punctuation
and
spelling.
A
more
contem-
porary
assessment of
conventions
was
used
by
De
La Paz
and
Graham
(1997a),
who
rated
papers
on a
combination of
grammat-
ical
errors
and
the
way
sentences
were
linked
together.
T-units.
"Thought
units,"
or
T-units,
measure
the
number of
independent
pieces
of
information in
a
writing
sample.
There-
fore,
the number
of clauses
written
in
a
sam-
ple
is counted
rather
than
the number
of
sentences
(Hunt,
1965).
Four
studies
in
our
meta-analysis
included
writing
measures
that we
coded as T-units. The
mean
weighted
effect
size
of these
four
studies
was
0.69,
with
a
confidence
interval
range
of 0.42-0.96.
Text structures.
A third
writing
compo-
nent relates
to the use of text
structures.
Two of
the 13 studies included a
score of
how
well students
wrote
their
posttests
fol-
lowing
the
text structure
they
were
taught
in the
intervention
(Englert
et
al.,
1991;
Saw-
yer
et
al.,
1992).
In
the
Englert
et
al.
(1991)
study,
text structure
scores
were
recorded
for two
expository
writing samples-either
a
compare/contrast
sample
or
an
explana-
tion
sample.
For
example,
the
explanation
papers
were rated
on
the
following
ele-
ments:
(a)
introduction
of the
topic,
(b)
in-
clusion
of
a
sequence
of
steps,
(c)
inclusion
of
key
organization
words,
such
as
first,
sec-
ond,
and
third,
and
(d)
adherence to
an
or-
ganizational
technique
acceptable
for an
ex-
planation
essay.
In
Sawyer
et al.
(1992),
the
text
structure
was
narrative. For
each
of
eight
story
grammar
elements-main
char-
acter,
locale,
time,
starter
event,
goal,
action,
ending,
and
reaction-a
score
was
re-
corded.
In both
studies,
we used
the
sum
of
individual text
structure
scores
as
a
mea-
sure
of
genre-specific
writing.
The mean
weighted
effect
size of
the
two
studies was 1.11.
Individual
effect
sizes
in
the two
studies
(1.07;
Englert
et
al.,
1991;
and
1.21;
Sawyer
et
al.,
1992)
were
very
close,
indicating
strong
consistency
in
out-
comes
related
to text
structures
specifically
taught.
Also,
the
effect
sizes
are the
largest
of
any
of
the
aggregations.
However,
be-
cause
only
two
studies
comprised
this
cate-
gory,
inferences
about
meaning
should
be
made
cautiously.
Studies
That
Employed
More
than
One
Experimental
Condition
We
now
present
a
series of
analyses
of
five
studies
that
included
more
than
one ex-
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
262
THE ELEMENTARY
SCHOOL
JOURNAL
perimental
condition.
De
La Paz
and
Gra-
ham
(1997a)
used two
treatment
groups
to
determine effects of the
physical
demands
of
writing
on
learning
outcomes. The
two
treatment
groups
were identical
except
that
students
in
one
group
dictated the
essays
instead of
writing
them.
The
overall effect
size was
0.58,
favoring
students in
the
dic-
tation
group.
This
finding
lends
credibility
to the
idea that for students
with
learning
disabilities,
measures of
the overall
quality
of their written
products
may
underesti-
mate the
quality
of
content
they present.
This sentiment
has been
expressed
recently
by
Goldman et al.
(1997).
A
seminal
study
of
writing
instruction
conducted
by
Graham and
Harris
(1989)
was
not
included
in
our
meta-analysis
be-
cause there
was
no
comparison
group.
We
want to
discuss the
study
briefly,
however,
because it
provides
critical
background
to
our
discussion of the
Sawyer
et al.
(1992)
study
that
follows.
In
the Graham and
Harris
(1989)
study,
two
groups
of fifth-
and
sixth-grade
stu-
dents with
learning
disabilities received
the
primary
intervention:
instruction in
the
use
of a
writing
strategy.
One
group
also re-
ceived
instruction
in
(a)
self-monitoring
their
performance
(recording
and
graphing
the
number and
kind of
story
grammar
ele-
ments
they
included
in
their
stories)
and
(b)
setting goals
for the
number
of
elements
they
would
include.
Overall,
students in
both
groups
made
considerable
gains
in
writing
performance,
which
the
authors
were
able to
attribute
to
the
strategy
in-
struction
training,
the main
focus of
instruc-
tion.
However,
contrary
to
expectations
that
self-monitoring
and
specific
goal-setting
would
lead to
additional
positive
effects,
the
opposite
seems to
have
occurred
(Gra-
ham &
Harris,
1989).
Students who
received
the
strategy
training
but
not
the
self-
monitoring
and
goal-setting
component
made
greater
gains.
The
effect
size
was
0.58
favoring
the
strategy-only
group.
Although
the effect
size
was
moderate,
it
was
not
sig-
nificant.
Still,
it
remains
a
puzzling
finding.
It is not
immediately
clear
why
explicit
self-
regulation
and
goal
setting
would
have
a
negative
effect,
but Graham
and Harris
of-
fer
a
plausible explanation.
They
point
out
that
those students
who
did not receive
ex-
plicit
instruction
in
self-regulation
and
goal
setting
probably
used
at least
some
implicit
self-regulation
to
complete
the
writing
tasks.
In
a
subsequent
study,
Sawyer
et
al.
(1992)
further
investigated
the influence
of
specific
goal-setting
and
self-monitoring
on
the
writing
performance
of fifth-
and
sixth-
grade
students
with
learning
disabilities.
This
time there
were
three
conditions,
two
of
which
were
the same as
in the
Graham
and
Harris
(1989)
study:
one
group
received
writing
strategy
instruction
only,
and
an-
other
group
received
both
writing
strategy
instruction
and
self-regulation.
The
third
condition
was
the
comparison
group,
which
the
authors
defined
as direct
teaching.
In
the
comparison
conditions,
students
were
taught
the
components
of the
writing
strat-
egy,
but instructional
procedures
that
may
have
induced
implicit
self-regulation
were
omitted.
These
eliminated
components
in-
cluded
teacher
modeling,
collaborative
practice
with
the
teacher,
and
student
per-
sonalization
of the
strategy
for
use
in
actual
writing
tasks.
As we
expected,
there
was
a
strong
effect
when
the
two
primary exper-
imental
groups-those
who
received
full
instruction
in
the
writing
strategy
either
with
or
without
self
regulation-were
com-
pared
to
students
in
the
direct
teaching
comparison
condition
who
received a
streamlined
version
of
the
strategy
instruc-
tion.
The
effect
size
when
the
two
primary
experimental
conditions
were
compared
to
the
comparison
condition
was
0.55.
When
the
two
strategy
groups
were
compared
to each
other
(i.e.,
with
and
without
self-regulation),
the
addition
of
self-regulation
produced
a
positive
effect
(d
=
0.66),
which
was the
opposite
of
what
Graham
and
Harris
(1989)
found.
Sawyer
et
al.
(1992)
speculated
that
the
apparent
JANUARY
2001
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
EXPRESSIVE
WRITING
263
difference
in
the studies
was that
students
were
provided
with
feedback
on
pretest
story
performance
in the
Graham and
Har-
ris
study
but not
the
Sawyer
et al.
study.
Sawyer
et
al.
suggested
that
discussing
feedback
with
students at
pretest
may
have
induced
implicit
goal-setting
on
their
part.
Recent
research
in
the area of
goal-set-
ting
and
self-regulation
has
expanded
the
implicit
versus
explicit
goal-setting
inter-
pretations
of the
Graham
and
Harris
(1989)
and
Sawyer
et al.
(1992)
studies.
Graham,
MacArthur,
and
Schwartz
(1995),
for
ex-
ample,
found that
goal-setting
strategies
in-
volving
skills that
were
well within
the
range
of
student
capabilities
did not
con-
tribute to
successful
goal
attainment.
For
in-
stance,
students
with
learning
disabilities
in
the
Graham et al.
(1995)
study
were
asked
to
add
three
things
to their
compositions
when
revising.
Interpreting
the
findings,
Graham et al.
(1995)
suggested
that
adding
three new
ideas to
a
composition
was
a
su-
perfluous
task.
This
goal
did
not
extend
stu-
dents'
potential
contributions to
their
writ-
ing beyond
what
normally
might
have
been
expected
or
prescribed.
In
a
recent
study
ex-
amining
student
goal-setting
in
the
context
of
essays
incorporating
refutations and
counter
arguments,
Page-Voth
and
Graham
(in
press)
discuss how
challenging goals
that
extend
beyond
prescribed
expectations
can
assist
students in
writing
longer
and
better
essays.
Research
conducted
by
Wong
et
al.
(1994)
also
included
two
experimental
groups
and
one
comparison
group.
Stu-
dents
in
the
two
experimental
groups
dif-
fered
in
the
types
of
interactive
dialogue
they
had
about
their
writing
drafts and
how
they
should be
revised.
Both
experimental
groups
were
provided
with a
complete
in-
structional
intervention
that
included
both
a
focus on
the text
structure
and
the
writing
process.
Both
conditions
also
included
dia-
logical
and
compositional
collaboration
be-
tween
two
class
members. In
the
first
con-
dition,
this
collaborative
practice
involved
the
teacher
and
the
student.
The
second
condition
involved two
students
collaborat-
ing.
Our
prediction
was
that
the
collabora-
tion
between
the
teacher
and
student
would
result
in
better-written
products
than
two
students
collaborating together.
However,
the
effect
size
favored the
student
collabo-
ration
condition
over the
student-teacher
condition
(0.53).
This
finding
has
important
implications
for
the
potential
widespread
application
of
peer
tutoring
to
improve
the
quality
of
students'
writing.
In all
of the studies
with
more
that
one
experimental
condition,
we
expected
to
find
at
least small
differences
between
the
ex-
perimental
groups
because
of
explicit
vari-
ations
in the
independent
variable.
In
the
Reynolds
(1986)
study,
however,
we
ex-
pected
no
differences
between
the
two
con-
trasting
conditions
because
they
were
so
similar.
In this
study,
variations
in
the
order
of the
revising
and
editing
strategies
used
by
students
in
groups
were
the
primary
contrast.
The
two
components
of
revising
and
editing
were
(a)
the
use
of
a
specific
strategy
to address
mechanical
aspects
of
the
paper
and
(b)
specific
editing
strategies
students
could
use
to
modify
what
they
wrote
in their
first drafts.
The
difference
be-
tween
the two
experimental
groups
was
the
order
in which
these
two
strategies
were
used.
The
effect
size
was
0.04,
supporting
our
expectations
that
the
ordering
of
the
in-
tervention
components
would
not
have
an
effect
on
writing performance.
Intervention
Effects
on
Attitudes
and
Understandings
Seven
of the
13
studies
assessed
the
ef-
fects
of
instructional
interventions
on
stu-
dent
attitudes
and
understandings
(see
Ta-
ble
4).
The
most
frequent
measure in
this
category
was
metacognition.
Across
six
studies,
the
weighted
effect
size
on
meta-
cognitive
surveys
was
0.64.
We
interpret
this
moderate
effect
to
mean
that
students'
understandings
of
the
writing
process,
the
purpose
of
text
structure in
writing,
and
other
essential
goals
of
the
interventions
were
positively
influenced
by
the
interven-
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
264
THE ELEMENTARY
SCHOOL
JOURNAL
tions.
There
is a
great
deal
of
variability
in
the
effect
sizes
among
the
studies, however,
so the conclusion is
tenuous. It
may
be
that
writing
is
an
excellent
subject
area for the
development
of
metacognitive
knowledge
over brief
periods
of
time,
as some
of
the
studies indicated
(e.g., Englert
et
al., 1991;
Welch,
1992).
Or,
it
may
be
that,
as
Wong
et
al.
(1996)
point
out,
metacognitive
knowl-
edge
is
slower
to
develop
and
typically
oc-
curs after
changes
in
actual
measures
of
stu-
dent
performance
are
evident.
A
great
deal of
variability
also charac-
terized
changes
in
students'
attitudes
to-
ward
writing.
As
with
changes
in
metacog-
nition,
conclusions
are
also tenuous
for
this
reason.
In
addition,
there are
only
three
studies
in
this
category,
so
drawing
firm
conclusions is
problematic
for
this reason as
well. This lack of
clarity surrounding
inter-
vention effects for
metacognition
and atti-
tudes toward
writing
is
likely
attributable,
in
part,
to the
difficulties
involved
in
mea-
suring
both
concepts accurately.
The overall effect size on
attitudes was
positive
(0.40)
and
relatively
small,
sug-
gesting
that
writing
interventions
of the
type
in
the current set of
studies
may
not
result
in
strong changes
in
the
attitudes stu-
dents
have
about
writing.
The overall
effect
size for
self-efficacy
was somewhat
larger
(0.61),
but
drawing
firm
conclusions is
also
hampered by
the small
number of
studies.
Overall,
the
findings supporting changes
in
students'
attitudes toward
writing,
their
feelings
about
themselves as
writers,
and
their
understandings
of how
to communi-
cate
through writing
are
tenuous,
at
best.
Writing
interventions
designed
to
improve
quality
of
writing appear
to result
in
posi-
tive
changes
(albeit
perhaps
small)
in
stu-
dents'
attitudes and
understandings.
Discussion
Wong
et al.
(1994)
noted
that
teaching cog-
nitive
strategies
in
expressive
writing
should
not be a
difficult
instructional
objective.
Our
meta-analysis
suggests
that
achieving
it will
result
in
considerable
benefits for
students
with
learning
disabilities.
Indeed,
all of
the
interventions
developed
to
improve
the ex-
pressive
writing
skills of students
with
learning
disabilities
had
positive
effects.
The
quality
of their written
products
im-
proved,
the effect sizes were
consistently
moderate to
large,
and effects were
rela-
tively
consistent across studies. Interven-
tion
effects
were
also consistent across
writ-
ing genres (e.g., comparison essays
vs.
narrative
stories)
and
across
procedures
used
to
assess
quality (e.g., global
rating
of
quality
vs.
the sum of
scores
for individual
writing
traits
or
rubrics).
It is
encouraging
that the
findings
from
the small
body
of more clinical
multiple-
baseline studies
on
writing
interventions
for
students
with
learning
disabilities have
supported
the
findings
of
this
meta-analy-
sis.
Overall,
the
multiple-baseline
studies
suggest
that
writing
interventions for
stu-
dents
with
learning
disabilities are effective
and
feasible
(see,
e.g.,
Danoff, Harris,
&
Graham, 1993;
De La
Paz, 1997;
Graham,
MacArthur,
Schwartz,
&
Page-Voth,
1992;
Montague
&
Leavell, 1994;
Wallace &
Bott,
1989;
Zipprich,
1995).
We
also found
evidence of
positive
ef-
fects on
students'
sense of
efficacy,
that
is,
their
sense of
being
able to write.
Although
the
number of
studies is not
extremely
large,
it
is
large
enough,
and the
quality
of
the
research is solid
enough,
to
allow
infer-
ences to be made about
the
improvement
of
classroom
practice.
Most
important,
the
meta-analysis
high-
lights
a
range
of
research-based
instruc-
tional
approaches
that
educators should
use
when
teaching
written
expression
to
stu-
dents
with
learning
disabilities. Ten
of the
13
interventions
provided
relatively
de-
tailed
descriptions
of
the
instructional
ap-
proach
used. Our
analysis
of
the
instruc-
tional
interventions
indicates that
they
share
many
features.
Virtually
all
of
the
interventions were
comprised
of several
components.
We have
organized
these
components
according
to
three
areas: the
writing process,
awareness
JANUARY
2001
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
EXPRESSIVE
WRITING
265
of text
structures,
and feedback. These
areas
are
discussed
below.
Explicit
Teaching
of the
Critical
Steps
in
the
Writing
Process
Most
interventions
adhered
to the
frame-
work
of three
basic
steps
in
the
writing
pro-
cess:
planning,
writing,
and
revising.
Invar-
iably,
explicit
teaching
of
each
step
was
provided
by
the
teacher
through
several
ex-
amples,
often
supported
by
a
"think
sheet,"
a
prompt
card,
or a
mnemonic.
Teaching
students to
write
requires
showing
them how
to
develop
and
organize
what
they
want
to
say
and
guiding
them
in
the
process
of
getting
it down
on
paper.
The
plans
students
develop
for this
are
critically
important.
They
must entail
structures
that
encourage
and
prompt
the
overlapping
and
recursive
processes
required
in
writing.
In
order to
help
students conceive
and
effec-
tively
use a
"plan
of
action,"
teachers,
or
peers
competent
in
reading
and
writing,
ver-
balize
the
steps they
take when
they
read
or
write.
These
plans
of
action,
sometimes re-
ferred to
as
"procedural
facilitators,"
assist
the
teacher
(or
peer)
in
the
unfamiliar
task of
verbalizing
how
one
actually
composes
a
piece
of
writing.
Procedural
facilitators
also
serve
two
related
purposes.
First,
they
provide
stu-
dents with
a
map
for
engaging
in
the
writ-
ing process,
providing
suggestions
for
what
to
do
when
the
student
feels
"stuck"
or
overwhelmed. In
other
words,
they
give
students a
permanent
reminder
of
the con-
tent
and
structure
of
the
writing
task.
Sec-
ond,
they
provide
a
common
language
that
teachers and
students can
use to
focus and
enhance the
quality
of
dialogue
they
might
have
about
expressive
writing.
We
believe
this
dialogical
function
represents
a
major
advance of
writing
instruction
over
tradi-
tional
methods,
which
required
students
to
work
in
relative
isolation
as
they
wrote.
Well-developed
plans
for
writing
result
in
better
first
drafts.
Revising
and
editing
skills are
also
critical
to
the
writing
process.
Developing
methods
to
help
students
with
learning
disabilities
refine
and
edit
their
work
has been
difficult,
but
a
few
research-
ers have
begun
to
develop specific
strate-
gies
that
appear quite
promising
(e.g.,
En-
glert
et
al., 1991;
Harris
&
Graham,
1992;
MacArthur
et
al., 1991;
Schumaker,
Nolan,
&
Deshler,
1985;
Wong
et
al.,
1996).
The
use
of
peer
editing,
for
instance,
is an
integral
and recurrent
strategy
in the research
of
En-
glert,
Wong,
and
MacArthur.
Explicit
Teaching
of
the
Conventions
of
a
Writing
Genre
Different
types
of
writing
are based
on
different
structural
elements.
A
persuasive
essay,
for
example,
contains
a
thesis
and
supporting arguments-elements
that
dif-
fer
considerably
from
those
found
in
nar-
rative
writing.
In
narrative
writing,
for
ex-
ample,
some
writers
like
to
begin
with
the
climax
of
a
story
and
then
proceed
with
character
development,
while
others
like
to
develop
their characters
before
developing
the
plot.
The
organizational
approach
used
to
construct
a
narrative
is
not,
in
and
of
itself,
what
makes
a
story
engaging.
Rather,
good
writing
involves
utilizing
what
Englert
et
al.
(1991)
call
"overlapping
and
recursive
pro-
cesses"
(p.
364).
These
processes
do
not
pro-
ceed
in
a
particular
order,
and one
process
may
inform
another
in such
a
way
that
the
author
returns
to
previous
steps
for
update
or revision
on
a
regular
basis.
Explicit
teaching
of text
structures
pro-
vides
a
useful
guide
for
undertaking
the
writing
task,
whether
it
is
a
persuasive
es-
say,
a
personal
narrative,
or
an
essay
com-
paring
and
contrasting
two
phenomena.
Instruction
typically
includes
numerous
explicit
models
and
prompts.
Greater
lev-
els of
specificity provided
by
the
prompts
appear
to
be
associated
with
better
written
products.
Procedural
facilitators
make
writing
us-
ing
text
structures
visible
to
students and
help
demystify
the
writing process.
Figure
1
presents
a
procedural
facilitator
called
a
This content downloaded from 128.223.86.31 on Fri, 31 Oct 2014 17:18:30 PM
All use subject to JSTOR Terms and Conditions
266
THE ELEMENTARY
SCHOOL
JOURNAL
Name
Charles
Date
11-9
Topic
How
you
tack care
of
a kitten
WHO: Who am
I
writing
for?
Peple
who
want
to have a kittin
WHY:
Why
am
I
writing?
So
peple
know
how to take care
of
a kittin
WHAT: What do
I
know?
(Brainstorm)
1.
how to
feed
it
-
kinds
of food
2.
how to
change
the cat litter
3.
how to
love
it
4.
how to
picke
it
up
5.
veterinarian
visits
HOW: How can
I
group
my
ideas?
food
Food
care
kittin
-
soft food
Feed
twise
a
day
yung
kittin
-
Mix
soft
and
hard Throw
old
away
care
Equipment
change
cat litter
scratch
pad
take to vet
litter
and
box
Love and
play
with it
bed
food
How
will
I
organize my
ideas?
Comparison/Contrast
Problem/Solution
Explanation
Other
FIG.
1.-Example
of a
completed
plan
think-sheet.
Source:
Englert
et
al.
(1992).
What is
being
How to take
care
of
cat
-and kittin
explained?
Materials/things you
need?
scratch
board
litter
food
toys
Setting?
house
Where
to
put
things
in the house
What
are
First.
the
steps?
Feed
it
food
and watter
(evryday)
and
play
with
it
Next.
Change
kittin litter
every
week.
Put littr-box where
it likes to
go
Third.
Take it to vet
for
shots
When
its sick and
for
check
ups
Then.
Give it
attenshun
like
play
with
it
petting
cat
Last.
play
and have
a
good
time with
the cat or the kittin
FIG.
2.-Example
of
a
completed
explanation
or-
ganization
form. Source:
Englert
et
al.
(1992).
"planning
think
sheet,"
which
was used
in
several studies
by Englert
and
colleagues
(e.g., Englert
et
al.,
1991,
1992).
The
think
sheet
helps
students
plan
their
writing
through
a series of
sequential
and struc-
tured
prompts. Figure
2
presents
a
proce-
dural facilitator
that would
be used
specif-
ically
for
an
explanation
text
structure
(Englert
et
al.,
1992).
The use of
targeted
procedural
facilita-
tors for
specific
writing
tasks
helps
students
recognize
the
sameness,
or
reoccurring
text
patterns,
in
order
to
guide
their
"efforts to
produce
well-organized
texts
and fuel
their
ability
to
interpret,
monitor,
and talk about
texts"
(Englert
&
Mariage,
1991,
p.
333).
Similarly,
in
the context of
dictating
ideas
for written
compositions,
De La
Paz and
Graham
(1997a)
showed
clearly
that con-
crete structures for
advanced
planning
were
critical
to the
completeness
and over-
all
quality
of
expository
essays produced
by
students
with
learning
disabilities.
Guided
Feedback
The third
component
common to
all
in-
terventions
was
guided
feedback.
Either
teachers
or
peers provided
frequent
feed-
back
to students
on the overall
quality
of
writing,
missing
elements,
and
strengths.
When feedback
was
combined with instruc-
tion
on the
writing process
or
text
structure,
a common
vocabulary
was created
that
gave
teachers and students
a
meaningful
way
to
engage
in
dialogue,
which resulted
in
improved
written
products.
The