Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233140262

The Process Writing Approach: A Meta-analysis

Article  in  The Journal of Educational Research · October 2011


DOI: 10.1080/00220671.2010.488703

CITATIONS READS
147 5,315

2 authors, including:

Steve Graham
Arizona State University
352 PUBLICATIONS   21,622 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

National Survey: Writing to Learn in Middle School View project

Strategy Instruction View project

All content following this page was uploaded by Steve Graham on 10 April 2019.

The user has requested enhancement of the downloaded file.


The Journal of Educational Research, 104:396–407, 2011
Copyright 
C Taylor & Francis Group, LLC
ISSN: 0022-0671 print / 1940-0675 online
DOI:10.1080/00220671.2010.488703

The Process Writing Approach:


A Meta-analysis
STEVE GRAHAM
KARIN SANDMEL
Vanderbilt University

writing methods, ranging from explicitly teaching strategies


ABSTRACT. The process approach to writing instruction
is one of the most popular methods for teaching writing. for planning, revising, paragraph and sentence construction,
The authors conducted meta-analysis of 29 experimental and word processing as a tool for writing, and studying and em-
quasi-experimental studies conducted with students in Grades ulating models of good writing.
1–12 to examine if process writing instruction improves the The instructional method that is probably best situated
quality of students’ writing and motivation to write. For stu- to be implemented broadly in any effort to reform writing
dents in general education classes, process writing instruction
resulted in a statistically significant, but relatively modest im- practices in the United States is the process writing
provement in the overall quality of writing (average weighted approach (sometimes referred to as Writers’ Workshop;
effect size [ES] = 0.34). Variation in ES was not related to Atwell, 1987; Calkins, 1983; Graves, 1983). A sizable
grade, reliability of the writing quality measure, professional minority of elementary and secondary teachers presently
development, genre assessed, or quality of study. The process use this approach exclusively when teaching writing, with
writing approach neither resulted in a statistically significant
improvement in students’ motivation nor enhanced the qual- a majority of teachers combining process writing with
ity of struggling writers’ compositions. other instructional procedures, such as more traditional
writing skills instruction (Cutler & Graham, 2008; Kiuhara,
Keywords: meta-analysis, writing, writing process Graham, & Hawken, 2009). Perhaps even more impor-
tantly, the National Writing Project (see Nagin, 2006)
provides professional development in how to use the process

A ccording to findings from the 2007 National


Assessment of Educational Progress (NAEP;
Salahu-Din, Persky, & Miller, 2008), only 33%
of eighth-grade and 24% of 12th-grade students perform at
writing approach to more than 100,000 teachers a year.
No other instructional approach in writing comes close
to having such an infrastructure for scaling up (see also
commercial materials such as those produced by Calkins &
or above the proficient level in writing (defined as solid aca- Colleagues, 2008). If writing becomes a central element in
demic performance). Students who score below this level the national reform movement, then the writing process
are classified as obtaining only partial mastery of the literacy approach is likely to play a major role in this effort.
skills needed at their respective grade. If partial mastery is Although there is no universally agreed-on definition for
interpreted as performing below grade level, then 67% of the process approach to writing, there are a number of un-
eighth-grade and 76% of 12th-grade students can be consid- derlying principles that are common to it (e.g., Graham &
ered as writing below grade level. These and previously disap- Perin, 2007; Nagin, 2006; Pritchard & Honeycutt, 2006).
pointing findings from past NAEP assessments (Greenwald, Students engage in cycles of planning (setting goals, gener-
Persky, Ambell, & Mazzeo, 1999; Persky, Daane, & Jen, ating ideas, organizing ideas), translating (putting a writing
2003) as well as concerns about students’ readiness for the plan into action), and reviewing (evaluating, editing, revis-
writing demands of college and the world of work (Achieve, ing). They write for real purposes and audiences, with some
Inc., 2005; Bates, Breslow, & Hupert, 2009; Greene, 2000) of their writing projects occurring over an extended period
have led to calls to improve the teaching of writing in Amer- of time. Students’ ownership of their writing is stressed, as is
ican schools (National Commission on Writing, 2003). self-reflection and evaluation. Students work together col-
The call to reform writing instruction is based on the as- laboratively, and teachers create a supportive and nonthreat-
sumption that there are effective practices for teaching this ening writing environment. Personalized and individualized
complex skill. The results from several meta-analyses con-
ducted during the last three decades supports this contention
(Bangert-Drowns, 1993; Goldring, Russell, & Cook, 2003; Address correspondence to Steve Graham, Vanderbilt University,
Graham & Perin, 2007; Hillocks, 1986; Rogers & Graham, Peabody College, Box 228, Nashville, TN 37023, USA. (E-mail:
2008). These reviews verified the effectiveness of a variety of steve.graham@vanderbilt.edu)
The Journal of Educational Research 397

writing instruction is provided through minilessons, writing zation of the process writing approach (e.g., Alloway et al.,
conferences, and teachable moments. 1979), this was not the case in all instances (e.g., Wienke,
There are many potential advantages to the process 1981). In addition, the two studies with the weakest effects
writing approach (Graham & Harris, 1997). First, students involved college students (average effect size [ES] = −0.14).
are encouraged to plan, draft, and revise. The cognitive As a result, the relatively small average ES of 0.18 for writing
activities involved in these writing processes account for quality reported by Hillocks may not provide a true estimate
close to 80% of the variance in the quality of papers of the impact of process writing for children in Grades 1–12.
produced by adolescent writers (Breetvelt, Van den Bergh Second, a larger average weighted ES of 0.32 for writing
& Rijlaarsdam, 1994, 1996; Rijlaarsdam & Van den Bergh, quality was reported for the process approach in a meta-
2006; Van den Bergh & Rijlaarsdam, 1996). Second, analysis of the writing intervention literature with students
instruction in writing through minilessons, conferences, in Grades 4–12 (Graham & Perin, 2007). Because there
and teachable moments should result in improved quality of was considerable variability in ES, moderate analyses were
writing. These teaching tools also provide mechanisms for undertaken to determine if specific study features were as-
addressing the instructional needs of individual students. sociated with the obtained ES. Graham and Perin found
Third, motivation for writing should be enhanced as that when teachers received professional development in
collaboration, personal responsibility, personal attention, applying the process approach model (most often provided
and a positive learning environment are stressed. These through the NWP) effects were larger (ES = 0.46) than
types of activities are thought to facilitate the value that when teachers were not provided with such preparation
students place on specific academic tasks (Wigfield, 1994). (ES = 0.03). However, in a follow-up analysis involving
Despite possible advantages, the process approach to writ- studies in which teachers did not receive professional devel-
ing is not without its critics (e.g., Baines, Baines, Stanley, & opment (considerable variability in ES still existed in these
Kunkel, 1999). Some have charged that the instruction pro- studies), the process writing approach had a positive and sta-
vided in process writing classrooms is not powerful enough tistically significant impact on writing quality for students
to ensure that students, especially students experiencing dif- in Grades 4–12 (ES = 0.27), but not for students in Grades
ficulty with writing, acquire needed writing skills and pro- 7–12 (ES = −0.05).
cesses (for a discussion of potential strengths and weaknesses, Although Graham and Perin (2007) selected studies that
see Graham & Harris, 1997). Critics argue that not enough met the description of process writing presented previously,
attention is devoted to mastering foundational skills, such as their review excluded studies conducted with younger
handwriting, spelling, and sentence construction (although students (Grades 1–3) as well as studies in which the
the National Writing Project does emphasize systematic reliability of the outcome measure (i.e., writing quality)
sentence-combining instruction; Nagin, 2006). They also was not established. Moreover, in four of the 21 studies
contend that very little time is devoted to explicitly teach- reviewed, the process approach to writing was compared to
ing students strategies for carrying out basic writing processes strategy instruction in which students were taught strategies
such as planning and revising (for data consistent with this for carrying out specific writing processes, such as planning
viewpoint, see Cutler & Graham, 2008). (Moye, 1993; Troia & Graham, 2002) or planning and
Given the popularity of the process writing approach and revising (Gamelin, 1996; Yeh, 1998). In each of these
the possibility that it may likely be applied even more broadly studies, strategy instruction was the treatment condition
in the future, it is important to determine its effectiveness. and process writing was the control condition.
Two previous meta-analyses have addressed this issue, with In this article, we report the findings of a new meta-
varying degrees of success. First, a seminal meta-analysis of analysis examining the effectiveness of the process approach
writing intervention research conducted by Hillocks (1986) to writing. Similar to Graham and Perin (2007) and Hillocks
examined the effectiveness of the natural process mode in (1986), we limited our review to experimental and quasi-
nine studies with students in Grade 6 through freshman experimental research. Such designs include mechanisms
year in college. Hillocks defined this approach as involving for eliminating alternative explanations for the claim that
general objectives for writing, free writing using self-selected treatment was responsible for observed changes in behavior
topics, writing for and receiving generally positive feedback (Phye, Robinson, & Levin, 2005). However, it is impor-
from peers, opportunities to revise written work, and high tant to note that we extend the two previous reviews in six
levels of student interactions. Although the natural process important ways.
mode and process writing as described previously share a First, younger students were not included in the two pre-
number of commonalities (e.g., student revising and high vious meta-analyses. For example, Graham and Perin (2007)
levels of student interaction), they are not identical. For limited their review to studies conducted only with students
example, free writing (Fox & Suhor, 1986) and teacher in Grades 4–12. It is possible that their analyses underesti-
as facilitator (Stotsky, 1988) received more emphasis in mated the effects of process writing on school-age students
Hillocks’s natural process mode than is typical in most in general. They found that ES for younger students (Grades
process-oriented approaches. Although Hillocks’s review 4–6) were larger than ES for older ones (Grades 7–12). Thus,
included studies compatible with our previous characteri- it is possible that the inclusion of studies conducted with
398 The Journal of Educational Research

primary-grade children (Grades 1–3) may increase the av- this approach places on critical writing processes such as
erage ES for all process writing studies. It is also possible planning and revising as well as the inclusion of instruc-
that differences in the effectiveness of process writing be- tional components (e.g., minilessons) aimed at improving
tween younger and older students are accentuated by the students’ writing skills. It was further expected the process
inclusion of studies conducted with primary-grade children. writing would enhance motivation because it includes in-
Consequently, in this review we examine the effectiveness structional components, such as a positive learning environ-
of process writing from Grades 1–12. ment, peer collaboration, and personal attention, theorists
Second, similar to Graham and Perin (2007) and Hillocks contend heighten the value students place on academic tasks
(1986), we were interested in the impact of the process writ- (Maehr & Midgley, 1991; Wigfield, 1994).
ing approach on the overall quality of students’ writing. Fifth, in contrast with Graham and Perin (2007) and
However, we did not limit our analyses just to studies where Hillocks (1986), ES for quasi-experimental studies in this
scoring reliability was established, as was done in the two meta-analysis were adjusted for possible pretest differences
previous meta-analyses. The assumption underlying this de- between the process writing group and comparison group.
cision was that the assessment of writing quality involves Because students were not randomly assigned to conditions
some degree of subjectivity, making it important to consider in such studies, their performance may not be equivalent at
reliability of scoring procedures (Graham & Perin, 2007). the start of the investigation, biasing the obtained outcomes.
Although we do not disagree with the subjectivity assump- Last, we examined if the process writing approach had a
tion, there are two ways to approach the problem. One is positive impact on improving writing quality for struggling
to exclude studies where scoring reliability is not demon- or at-risk writers. This analysis was done separately from our
strated, and the other is to include all studies and determine analysis with more typical students, whereas such an analysis
if scoring reliability is related to variation in ES. We chose was not undertaken by Graham and Perin (2007) or Hillocks
the latter approach, as the exclusion of studies can bias anal- (1986). We decided to compute a separate average weighted
ysis and limit inference (Herbison, Hay-Smith, & Gillespie, ES for struggling writers because of critics’ concerns that the
2006). process writing approach is not powerful enough for these
Third, in contrast with Graham and Perin (2007), we did students (see Graham & Harris, 1997). However, we did not
not calculate ES for studies that compared the process writ- make a prediction concerning the effects of process writing
ing approach to strategy instruction for planning, revising, for these students. It is possible that this approach to writing
or both (such studies were not reviewed by Hillocks [1986]). enhances these students’ performance because of its empha-
Pritchard and Honeycutt (2006) placed strategy instruction sis on the processes of writing (planning, translating, and
under the process approach umbrella, arguing that process reviewing), the inclusion of instructional components de-
writing includes direct instruction in writing strategies. As a signed to enhance writing skills (e.g., minilessons and writ-
result, we decided to limit the comparison condition in the ing conferences), and methods for fostering motivation (e.g.,
studies we reviewed to more traditional writing instruction collaboration and positive learning environment). In addi-
(e.g., teaching writing skills). Process writing developed, in tion, several case studies (Cousin, Aragon, & Rojas, 1993;
part, as a reaction and alternative to such instruction (Nagin, Foulger & Jimenez-Silva, 2007; Zaragoza & Vaughn, 1992;
2006). Nevertheless, we did not treat strategy instruction Zucker, 1993) reported that the process writing approach
as synonymous with process writing instruction for two had a positive impact on struggling writers, including stu-
reasons. First, the level of systematic and explicit instruction dents with learning disabilities (LD) and English language
provided in strategy instructional studies exceeds what is learners (ELL). Nevertheless, it is possible that this approach
evident in the programs developed by process advocates is not effective with these students, as not enough attention
(e.g., Atwell, 1987; Calkins, 1983; Graves, 1983). Second, and intensity is directed at teaching basic text transcription
strategy instruction, unlike the process writing approach, skills, sentence construction skills, or strategies for planning
does not constitute a full writing program, and many studies and revising. Struggling writers typically have difficulty in
investigating its effectiveness concentrate on teaching one or more of these areas (Graham, 2006).
strategies for a restricted range of processes, such as planning Similar to Graham and Perin (2007), we tested if profes-
(see Graham & Perin, 2007; Rogers & Graham, 2008). sional development, type of genre assessed, and quality of
Fourth, Graham and Perin (2007) and Hillocks (1986) the study was associated with higher ES for process writing
did not examine the effects of the process writing ap- instruction. We expected that training would mediate the
proach on students’ motivation. Proponents of the process impact of this approach to writing, as it is a complex inter-
approach contend that this method of teaching enhances de- vention. As noted previously, Graham and Perin reported
veloping writers’ attitudes toward writing (e.g., Pritchard & that process writing studies that involved professional de-
Honeycutt, 2006). Consequently, writing quality and mo- velopment had a larger average effect than studies in which
tivation served as outcome measures in this review. Based no training was provided. We did not anticipate that genre
on the findings from the previous reviews, we anticipated would be associated with magnitude of ES. Graham and
that process writing would have a positive impact on writ- Perin did not find such a relationship, and there is no ev-
ing quality. We also based this prediction on the emphasis ident reason why process writing should be more effective
The Journal of Educational Research 399

with one genre than another. Likewise, we did not expect writing instruction plus word processing to a control condi-
that study quality would be associated with writing quality, tion, we only compared the process writing approach alone
as Graham and Perin did not find a statistically significant to the control, as word processing is not commonly used
association between magnitude of ES and study quality for in writing programs at either the elementary or secondary
the broader writing intervention literature. levels (Cutler & Graham, 2008; Kiuhara et al., 2009).
In summary, this meta-analysis addressed the following Third, only quasi-experimental studies that collected data
questions: at pretest and posttest were included. The pretest data had
to measure the same construct as the posttest measure. This
Research Question 1: Does the process approach to writing
allowed us to adjust for possible pretest differences. A pretest
improve the quality of students’ writing in general educa-
was not required for experimental studies, as random assign-
tion classrooms in Grades 1–12?
ment is designed to ensure that students in different groups
Research Question 2: Does this approach improve the qual-
are equivalent at the start of a study.
ity of writing produced by struggling or at-risk writers in
Fourth, only studies that measured writing quality, moti-
Grades 1–12?
vation, or both were included. Writing quality is based on
Research Question 3: Are effects for writing quality related
readers’ judgment of the overall merit of a paper, taking into
to grade, reliability of the outcome measure, professional
account factors such as ideation, organization, vocabulary,
development, genre assessed, or study quality?
sentence structure, and tone. These attributes are assessed
Research Question 4: Does this approach enhance student
singularly (analytic scale) or together (holistic scale) on
motivation?
a numerical Likert-type rating scale (Diederich, 1966).
We also considered scores from norm-referenced writing
tests, such as the Test of Written Language–3 (TOWL-3;
Method Hammill & Larsen, 1996), as a measure of overall quality, as
Selection of Studies long as the score was based on a sample of students’ writing.
Measures of motivation involved self-report measures where
The search for articles in this meta-analysis was influenced students quantified their attitudes on Likert-type scales.
by seven factors. First, only studies examining the effec- Fifth, a study was only included if it contained sufficient
tiveness of the process writing approach were included. We information to calculate an ES (Lipsey & Wilson, 2001).
applied Graham and Perin’s (2007) definition of the process Sixth, studies that were conducted with students in Grades
writing approach, and it involved the following elements: 1–12 were included, whereas studies involving kindergarten
extended opportunities for writing; writing for real audiences or college students were excluded. Seventh, studies con-
and purposes; emphasis on the cyclical nature of writing, ducted with students attending regular public or private
including planning, translating, and revising; student own- schools were included. We did not include studies deliv-
ership of written compositions; interactions around writing ered in special schools for children with deafness, autism,
between peers as well as teacher and students; a supportive severe emotional disturbance, and so forth. Although we
writing environment; and students self-reflection and eval- believe that writing instruction is an important part of the
uation of their writing and the writing process. If a process curriculum for these students (and that they should be edu-
intervention focused on just one component of the process cated in their neighborhood schools whenever possible), the
writing approach (e.g., planning or revising), the study was purpose of this review was to draw recommendations on the
excluded. effectiveness of the process writing approach within regular
Second, only studies that employed an experimental school settings. This included studies conducted with stu-
or quasi-experimental design were included. Studies that dents with disabilities or ELL in separate settings, as long as
employed a correlational, descriptive, single-subject, or qual- instruction occurred in regular school settings.
itative design were excluded. Thus, all selected studies had
to include a process writing treatment group and a control or Strategies for Identifying Appropriate Studies
comparison group. Acceptable control or comparison condi-
tions included traditional skills instruction (e.g., grammar- Four specific techniques were used to locate possible stud-
based instruction) or a district or school writing curriculum ies for inclusion in this review. First, an extensive search of
(as long as it was not a process writing program). Studies were published and unpublished studies was conducted using the
excluded when the comparison condition was another form following five databases: ERIC (Education Resources Infor-
of process writing instruction. We also excluded studies in mation Center), Education Abstracts (Education Full Text),
which the comparison condition involved strategy instruc- Dissertation Abstracts, PsycINFO, and ProQuest. The ini-
tion in one or more writing processes. If a study compared tial search included the following terms: free writing, free
process writing and process writing plus an additional com- writing and elementary, free writing and middle school, free writ-
ponent to a comparison group, then only the process writing ing and high school, process approach to writing, process approach
condition and control group were compared. For example, to writing and elementary, process approach to writing and mid-
if a study compared process writing instruction and process dle school, process approach to writing and high school, national
400 The Journal of Educational Research

writing project, national writing project and elementary, national disabilities or ELLs, general education population excluding
writing project and middle school, national writing project and high students with disabilities or ELLs, only students with
school, process writing, process writing and elementary, process disabilities, or only ELLs), (g) genre of writing (narrative,
writing and middle school, process writing and high school, writing expository, persuasive, combination of writing genre or
process approach, writing process approach and elementary, writ- student choice, and nonidentifiable), and (h) environment
ing process approach and middle school, writing process approach in which treatment was delivered (general education class-
and high school, writer’s workshop, writer’s workshop and ele- room, self-contained special education classroom—students
mentary, writer’s workshop and middle school, writer’s workshop spend more than 60% of the day in the classroom, pull-out,
and high school, attitude and writing, self-efficacy and writing, or resource classroom—student spends less than 60% of the
belief and writing, and perception and writing. The ending date day in the classroom, or classroom for ELLs).
for these searches was August 2009. The eight study quality indicators (taken from Gersten
Second, reference lists of previous meta-analyses of the et al., 2005) were (a) type of experimental design (experi-
writing intervention literature (Graham & Perin, 2007; ment with random assignment and correct unit of analysis =
Hillocks, 1986) were examined to identify potentially appro- score of 1.0; experiment with random assignment and incor-
priate studies. Third, a hand search of Research in the Teaching rect unit of analysis = score of .75; quasi-experiment with
of English and Written Communication was conducted from correct unit of analysis = score of .50; or quasi-experiment
1986 to 2009. In addition, Research in the Teaching of English with incorrect unit of analysis = score of .25), (b) ceiling or
periodically publishes bibliographies of scholarly papers in floor effects for measures (if the mean for a measure for any
the area of composition studies, and these sources were also condition was more than one standard deviation away from
examined to identify possible studies. Last, the reference lists the highest score of the measure and less than one standard
in all collected studies were examined to locate additional deviation away from the lowest score possible = score of 1.0;
articles and papers. if these conditions were not met = score of 0), (c) descrip-
Once all studies were collected, each was read and as- tion of the control group (traditional approach or a district
sessed by Karin Sandmel to determine if it was appropriate or school curriculum described and applied = score of 1.0;
for inclusion in this review, using the seven criteria reviewed activities not described = score of 0), (d) attrition (90% or
in the previous section. These were treatment involved the more of students who began the study completed it = score
process writing approach as defined; study was an experimen- of 1.0; if this condition was not met = score of 0), (e) equiv-
tal or quasi-experimental design; quasi-experimental design alent attrition (if overall attrition 10% or less and attrition
study included a pretest; control condition was appropriate equivalent across conditions = score of 1.0; if these con-
as defined; writing quality, motivation, or both were out- ditions were not met = score of 0), (f) treatment integrity
come measures; subjects were in Grades 1–12 in a regular (data showing that at least 70% or more of the treatment de-
school setting; and an ES could be computed. To be in- livered as intended = score of 1.0; lack of such data = score
cluded in the present review, a study had to meet all of these of 0), (g) teacher effects were controlled (teachers randomly
criteria. assigned to conditions, teachers taught both conditions, or
A random sample of one third of all collected studies researchers matched the teachers = score of 1.0; no control
were independently rescored by a second rater to establish for teacher effects = score of 0), and (h) reliability of writing
reliability of the study inclusion procedures. Reliability was quality measure (reliability was reported at .60 or higher =
calculated by determining the number of agreements (ac- score of 1.0; no reliability score or score below .60 = 0). A
cepts or rejects) between the two raters and dividing by total total quality score was computed by summing the separate
number of agreements plus disagreements. Reliability was scores for the eight indicators.
.93. If a disagreement occurred, the study was reviewed a Karin Sandmel coded each study for the descriptive and
second time to determine if it should be included in the study quality indicators. A trained second rater indepen-
review. dently scored a random sample of 33% of all studies. Inter-
rater reliability between the two coders was .91.
Coding the Studies
Calculating Effect Sizes
Each study included in this review was coded for eight
descriptive and eight indicators of study quality. The For experimental studies, an ES was calculated for writing
descriptive variables scored were: (a) publication type quality, motivation, or both by calculating the mean differ-
(e.g., journal article, dissertation, other), (b) professional ence between the process writing and the control group (i.e.,
development versus no professional development, (c) the Ȳtx − Ȳctrl ) and dividing by the pooled standard deviation
length of professional development in days when profes- (i.e., d). For quasi-experimental studies, d was calculated by
sional development was provided, (d) length of treatment determining the mean difference between the treatment and
in weeks, (e) grade level of students (elementary school, control group (i.e., Ȳtx − Ȳctrl ) after first adjusting for pretest
middle school, high school, or a combination), (f) student differences by subtracting the mean difference at pretest
type (general education population including students with from posttest or estimating the posttest mean-difference
The Journal of Educational Research 401

statistic from covariate-adjusted posttest means. The ad- than no effect. A random-effects model was applied, as it
justed posttest mean difference was standardized by the allows generalizability to studies not included in this meta-
reported or estimated posttest standard deviation (What analysis (Lipsey & Wilson, 2001). Although great care was
Works Clearinghouse, 2007). If standard deviations were undertaken to locate all studies, it was possible that some
not reported, they were estimated from summary statistics studies were missed, making a random-effects model prefer-
or by estimating residual sums of squares to compute a root able.
mean squared error (RMSE; Shadish, Robinson, & Congx- For each average weighted ES calculated, we conducted
iao, 1999; Smith, Glass, & Miller, 1980). All computed ef- a test of homogeneity (using a fixed-effects model) to de-
fects were adjusted for small-sample size bias (Hedges, 1982). termine if the various ES weighted and averaged together
As noted previously, measures of writing quality take into in a treatment estimated the same population ES. We fur-
account factors such as ideation, organization, vocabulary, ther conducted a moderator analyses with writing quality
sentence structure, and voice. For holistic measures, all of for students in the general education setting (there were
these factors are considered simultaneously when assigning not enough ES for writing quality for students who strug-
a single score. For analytic measures, a score is assigned to gle with writing or the motivation measure to do this) to
each factor separately. If a holistic score was available, we determine if variability in effects were related to identifi-
calculated the ES with this measure. If holistic and analytic able differences between studies (e.g., grade level). Using
scores were available, only the holistic score was used. If a random-effects model (Lipsey & Wilson, 2002), ES were
just an analytic scale was available, we first calculated an grouped into two mutually exclusive categories (i.e., ES cal-
ES for each attribute separately, and then averaged these culated from studies where professional development was
separate ES to obtain a global measure of writing quality provided vs. studies where no training was provided), and
(similar to a holistic score). For the most part, the holis- the homogeneity of ES within each category was tested, as
tic and analytic measures used in the studies reviewed here was the difference between the levels of the two mutually
concentrated on the attributes of writing quality described exclusive categories (i.e., the independent variable). Pre-
previously. planned contrasts focused on grade, reliability of the writing
For two studies (Gauntlett, 1978; Pritchard & Marshall, quality measure, professional development, genre tested, and
1994), the population samples were large and could po- study quality. The statistical package used to conduct these
tentially influence the direction of the average ES. These analyses was MetaWin (Version 2; Rosenberg, Adams, &
two population samples were winsorized using the Tukey Gurevitch, 2000).
method. A box plot was created to look at the distribution
of the ES. The interquartile range was calculated by subtract-
ing the 25th percentile value from the 75th percentile value. Results
This interquartile range was then multiplied by 1.5. If an ES Description and Quality of Studies
was greater than 1.5 times the interquartlie range plus the
75th percentile value or less than 1.5 times the interquartile We located 29 studies that met our inclusion criteria (see
range subtracted from the 25th percentile value, the ES was Table 1). Most of these studies (n = 24) were conducted
winsorized so that it was equivalent to the one of the two ob- with typical students in general education classes (7,865
tained values just described (i.e., the value it was closest to). students). All but one of these studies assessed the impact
In addition, the ES for writing quality for one study (Kelley, of the process writing approach on quality of writing, with
1984) was almost 5 times the size of the interquartile range. six studies assessing its effect on motivation. An additional
Consequently, this ES was winsorized using the previous for- five studies were conducted with struggling or at-risk writers
mula. All ES were calculated separately by Steve Graham (285 students). This included four studies that involved
and Karin Sandmel. Any disagreements between the two students with disabilities (Clippard & Nicaise, 1998; Croes,
scorers were resolved. This was only necessary in 12% of the 1990; Curry, 1997; Weiss, 1992) and one investigation with
cases (n = 4). ELL students (Green, 1991). All of these studies assessed
writing quality, and one of them also assessed motivation
Statistical Analysis for Effect Sizes (Clippard & Nicaise, 1998).
Almost one half of the studies were conducted exclusively
Our meta-analysis employed a weighted random-effects with elementary students (n = 14), 21% with high school
model. A separate weighted effect and confidence interval students (n = 6), 14% with middle school students (m = 4),
was calculated for writing quality for students in general ed- and 17% with students at multiple levels (n = 5). All of the
ucation classrooms as well as struggling or at-risk writers. studies conducted with struggling or at-risk writers (n = 5)
A separate weighted effect and confidence interval was also involved children in the elementary grades (see Table 1). In
calculated for motivation. Weighted ES were computed by 55% of the studies, teachers were taught how to apply the
multiplying each ES by its inverse variance. It should be process writing approach (n = 16), with relatively similar
noted that when a confidence interval does not intersect percentages for students in general classes and for strug-
zero, then the average weighted ES is significantly greater gling or at-risk writers. When researchers reported length of
402 The Journal of Educational Research

TABLE 1. Studies, Population, Grades, Sample Size (SS), Genre, Professional Development, Publication Type, Quality, and
Attitude Effect Sizes (ESs)

Professional
Study SS Grades n Genre development Publication ES quality ES attitude

Adams, 1971 GE hs 135 ? N D 0.45 −0.27


Alloway, 1979 GE ms/hs 225 ? Y C 0.57 —
Beachem, 1984 GE ele 87 C/SC N D 0.62 —
Clippard, 1998 SE ele 27 C/SC Y J 0.39 0.86
Croes, 1990 SE ele 157 N Y D 0.34 —
Curry, 1997 SE ele 45 N Y D 0.41 —
Dougans, 1993 GE ele 86 E N D 0.25 —
Eads, 1989 GE ele/ms/hs 502 C/SC Y D 0.22 —
Ewing, 1992 GE ms 235 P Y D −0.04 —
Fleury, 1988 GE es 74 ? Y D 0.34 —
Ganong, 1974 GE hs 135 ? N D 0.06 —
Gauntlett, 1978 GE hs 791 ? N D 0.05 —
Green, 1991 ELL ele 32 E N D −0.54 —
Hamilton, 1992 GE ele 124 N N D 0.77 —
Hansen, 1985 GE ele 205 C/SC N D 0.82 —
Hayes, 1984 GE ms 70 C/SC N D 0.11 —
Kelley, 1984 GE ms 101 N N D 1.04 —
Marker, 2000 GE ele 46 ? Y D 0.09 —
Moore, 1983 GE hs 40 C/SC N D 0.28 —
Olson & DiStefano, 1980 GE ms 201 E Y J 0.40 —
Pantier, 1999 GE ele 29 N N D −0.24 0.17
Pollington, 1999 GE ele 130 Y D — 0.37
Pritichard, 1987 GE ele/ms/hs 383 ? Y J 0.38 —
Pritchard & Marshall, 1994 GE ele/ms/hs 3919 E Y J 0.45 —
Reimer, 2001 GE hs 60 E Y D −0.09 —
Roberts, 2002 GE ele/ms/hs 132 E Y D 0.37 0.16
Scannella, 1982 GE hs 95 E Y D 0.35 0.27
Umbach, 1990 GE ele 60 C/SC Y D 0.07 —
Weiss, 1992 SE ele 24 ? N D 0.89 —

Note. GE = general education; SE = special education; ELL = English language learners; ele = elementary school; ms = middle school; hs = high
school; C/SC = combination of genres or student choice; E = expository writing; N = narrative writing; P = persuasive writing; ? = unknown genre;
Y = yes, participated in professional development; N = no, did not participate in professional development; D = dissertation; J = journal article.

professional development (n = 9), it ranged from 2 to 40 Overall, the quality of studies was not strong, but also
days. Length of the process writing treatment for students not that dissimilar from the quality of writing intervention
ranged from 3 to 45 weeks in general education, and from studies reviewed by Graham and Perin (2007). For the eight
8 to 40 weeks for struggling or at-risk writers. Writing gen- quality indicators assessed (see Table 2), only 45% were
res emphasized in treatment varied from 25% of studies in evident for studies involving students in general education
which a combination of genres or self-selection were empha- classes (quality score range = 1.25–7.25), whereas 58% were
sized (n = 7), 25% of studies focusing on expository writing met for studies with struggling or at-risk writers (quality
(n = 7), 18% concentrating on narrative (n = 5), and 4% on score range = 2.5–6.25). For all 29 studies, four of the qual-
persuasive writing (n = 1). Writing genre was not identified ity indicators were met in the typical study: writing quality
in 29% of the investigations (n = 8). was reliably assessed (72% of studies; n = 21); attrition was
All of the studies included in this meta-analysis were equivalent across groups (62% of studies; n = 20), control
selected so that the following quality indicator was present: condition was described (62% of studies; n = 20); attrition
Students in the process writing condition were compared to was less than 10% (62% of studies; n = 18), and no ceiling
students in a control or comparison condition. Only three or floor effects for measures were evident (52% of studies;
studies involved random assignment, with the researchers n = 15). However, teacher effects were controlled in only
conducting statistical analysis using the correct unit of 34% of studies (n = 10), random assignment with the cor-
assignment. In three additional studies, teachers were ran- rect unit of analysis occurred in just 10% of studies (n = 3),
domly assigned to treatments, but student-level data were and researchers provided evidence that the treatment was
analyzed. Thus, 90% of all studies were quasi-experimental. delivered as intended in just 7% of all studies (n = 2).
The Journal of Educational Research 403

TABLE 2. Total Quality Score and Percentage of Studies in Which Quality Indicator Was Present, by Treatment

Total quality score Attrition


Assignment above Mortality Ceiling Control Treatment Teacher Interrater
(random) 90% equivalent or floor type fidelity effects reliability
Treatment M SD n (%) (%) (%) (%) (%) (%) (%) (%)

Process writing 3.95 1.64 24 12.5 58.3 66.7 45.8 62.5 04.2 33.3 75.0
(general
classes)
Process writing 4.70 1.59 5 0.0 80.0 80.0 80.0 100.0 20.0 40.0 60.0
(struggling
or at-risk
writers)

Does Process Writing Improve Writing Quality for Students in established ES = .26), Q(between) = 0.43, p = .51; nor did
General Education Classes? professional development (professional development ES =
.28; no professional development ES = .42), Q(between) =
As expected, the process approach to writing instruction 1.05, p = .31. We also examined if type of writing moder-
improved the overall quality of writing produced by students ated ES. This involved four categories of writing: expository
in general education classes. The mean average weighted (expository or persuasive; n = 6), combination of genres or
ES computed with a random-effects model was 0.34. This self-selection (n = 6), narrative (n = 3), and writing genre
was statistically different than no effect, as the confidence not specified (n = 7). Genre of writing did not account for
interval did not cross zero (confidence interval ranged from excess variability in ES (expository ES = 0.24; combina-
.22 to .47). Eighty-three percent of the comparisons resulted tion or self-selection ES = 0.37; narrative ES = 0.73; not
in a positive effect for the process writing approach. ES specified ES = 0.27), Q(between) = 7.29, p = .06.
varied substantially, Q(22) = 87.90, p < .001, indicating It is interesting to note that studies of the highest quality
that the studies were not from a single population of studies yielded a slightly lower average weighted ES than did studies
(see Table 1 for specific ES). of lower quality. For the three experimental studies (i.e.,
random assignment with the correct unit of analysis), the
Does Process Writing Improve Writing Quality for Struggling average weighted ES was 0.29, whereas it was 0.35 for the 20
and At-Risk Writers? quasi-experimental studies. Likewise, for the 11 studies that
met one half or more of the quality indicators, the average
The process approach to writing instruction did not im- weighted ES was 0.33. For the remaining 12 studies, it was
prove struggling or at-risk students’ overall writing quality. 0.35. Study quality did not moderate ES for writing quality;
Although the average weighted ES was 0.29, this average the Q statistic for the regression model was .08 and not
weighted effect did not statistically differ from no effect, as statistically significant, Q(1) = 0.08, χ 2(23) = 0.78, p = ns.
the confidence interval crossed zero (the confidence interval
ranged from −0.24 to 0.81). Nevertheless, the ES appeared
Does Process Writing Enhance Motivation?
to be from a single population, as the test for homogeneity
was not statistically significant, Q(4) = 7.62, p = .11. Contrary to expectations, the process writing approach
did not enhance students’ motivation. The average
Is Variation in Writing Quality Effects Related to Professional weighted ES was 0.19, and it was not statistically different
Development, Grade, Scoring Reliability, Genre Assessed, or from no effect, as the confidence interval crossed zero (the
Study Quality? confidence interval ranged from –.16 to .53). In addition,
the ES appeared to be from a single population, as the test
As noted previously, ES for writing quality varied con- for homogeneity was not statistically significant, Q(5) =
siderably for students in general education classes. Conse- 10.35, p = .07.
quently, we examined if specific study features moderated
average weighted ES and accounted for excess variability. Discussion
None of the preplanned comparisons was statistically signif-
icant. Grade level did not moderate ES (Grades 1–6 ES = The primary purpose of this meta-analysis was to deter-
0.37; Grades 7–12 ES = 0.31), Q(between) = 0.31, p = .58; mine if the process approach to writing is an effective method
nor did the establishment of scoring reliability for the qual- for teaching writing to students in Grades 1–12. This is an es-
ity measure (reliability established ES = .36; reliability not pecially important question, as the process writing approach
404 The Journal of Educational Research

is likely to play a prominent role in any efforts to reform writ- classrooms. It is possible that process writing teachers in
ing instruction in the United States. It is already a relatively these studies were made to feel special, enhancing their effort
popular approach to writing instruction (Cutler & Graham, and performance. However, it is worth noting that the data
2008; Kiuhara et al., 2009), with a substantial infrastruc- from this review are not consistent with this hypothesis,
ture for making it even more widely available, ranging from as the average weighted ES for studies with and without
the work of the National Writing Project (Nagin, 2006) to professional development did not differ statistically. Fourth,
commercial materials developed by advocates such as Lucy the overall quality of these 24 studies was not strong. On
Calkins (e.g., Calkins & Colleagues, 2008). Moreover, the average, each study met slightly less than one half (45%)
findings from two previous meta-analyses of experimental of the eight quality indicators assessed. This raises concerns
and quasi-experimental research (Graham & Perin, 2007; about the confidence that can be placed in the claim that the
Hillocks, 1986) provided evidence that this approach im- process writing approach improves the writing of students in
proves the writing of students in Grades 4–12. typical classrooms. These concerns are alleviated somewhat
This review builds on the two previous meta-analyses by by the finding that the average weighted ES for higher quality
extending the range of studies analyzed. This includes studies studies was generally similar to the average weighted ES for
conducted with primary-grade children (Grades 1–3), inves- lower quality studies.
tigations with struggling or at-risk writers at any grade level, The fifth factor that must be taken into consideration
and studies assessing the impact of process writing on student when interpreting the obtained effect is magnitude. A widely
motivation. In contrast with prior reviews, studies did not used rule of thumb is that an ES between 0.20 and 0.49 is a
have to establish reliability of outcome measures to be in- positive but small effect. Lipsey and Wilson (2001) argued
cluded, but a pretest was required for all quasi-experimental against using such a formulaic approach, as it is important
studies. This allowed us to provide a more general and less to interpret an ES within the context of a given field. This
biased answer to the previous question. better contextualizes the magnitude of an ES.
Based on the available scientific evidence from experi- To contextualize the findings from the present study, we
mental and quasi-experimental intervention studies, the an- draw on Graham and Perin (2007). They obtained aver-
swer to our question about the effectiveness of the process age weighted ES for writing quality for 10 other practices
writing approach depends on who is assessed and on what for teaching writing, with average weighted ES for prac-
outcome. As expected, when studies were conducted in gen- tices ranging from −.32 to .82. Within this context, the
eral education classrooms, students receiving process writing magnitude of the average weighted ES for the process writ-
instruction were better writers at the end of the experiment ing approach fares relatively well when considered against
than were students in the control condition. The average teaching writing through inquiry (ES = 0.32), using prewrit-
weighted ES of 0.34 for writing quality in 24 studies was ing strategies to gain and organize ideas for writing (ES =
statistically greater than no effect. This was almost identical 0.32), and studying and emulating models of writing (ES
to the ES of 0.32 reported by Graham and Perin (2007), = 0.25) or traditional grammar instruction (ES = −0.32).
even though there was only a 59% overlap in the studies However, it fares less well when considered against strat-
used to compute these two summary statistics. This was due egy instruction (ES = 0.82); summary writing instruction
to us including nine investigations not analyzed in prior (ES = 0.82); collaborative planning, drafting, and revis-
reviews (e.g., ones without established reliability), and ex- ing (ES = 0.75); goal setting (ES = 0.70); word process-
cluding seven studies from it (e.g., no pretest provided for ing (ES = 0.50); and sentence combining (ES = 0.50).
quasi-experimental studies). Viewed within this context, the process approach to writing
Interpretation of this positive effect must be tempered by instruction is an effective, but not particularly powerful ap-
five factors. First, the obtained effect is based on comparisons proach for teaching writing to students in general education
to classes in which students were mainly taught writing skills classrooms.
or the writing instruction received by control students was When the focus of the analysis narrows to just weaker
not described (39%). Thus, no claims can be made about writers, the evidence from this meta-analysis does not sup-
the effectiveness of process writing in comparison to other port the claim that the process writing approach is an effec-
types of writing treatments. Second, definitions of process tive method for improving quality of writing. The average
writing have changed over time (Stotsky, 1988), and there is weighted ES in five studies was 0.29, and not statistically
presently some disagreement as to what constitutes a process different than zero. This finding is at odds with several case
approach (see Pritchard & Honeycutt, 2006). We applied studies (Cousin et al., 1993; Foulger & Jimenez-Silva, 2007;
the same definition as Graham and Perin (2007), and any Zaragoza & Vaughn, 1992; Zucker, 1993), where the pro-
conclusions about the effectiveness of process writing must cess writing approach reportedly improved struggling writ-
be made in reference to this characterization. ers’ compositions. Additional research is needed to more
Third, the obtained ES may be inflated due to Hawthorne fully examine if process writing is an effective approach to
effects. For example, professional development was provided teaching weaker writers. This research needs to expand the
to process writing teachers in 12 of the 23 studies (52%). age and type of writers tested. The five studies analyzed
Such training was not provided to teachers in the control here concentrated just on elementary grade students, and
The Journal of Educational Research 405

participants were either students with learning disabilities lective and no attempt is usually made to rule out competing
(four studies) or ELLs (one study). explanations (Graham, 2010).
The findings from this meta-analysis were also not sup- We are not suggesting that the process approach to writ-
portive of the process writing approach when the focus ing as it was characterized in this review be abandoned.
shifted from writing quality to motivation. The average First, we think that this is unlikely to happen. Second, there
weighted ES for motivation in six studies was 0.19, and this is much to like about the process approach. This includes
was not statistically different from zero. Thus, the avail- its emphasis on the critical role of process in writing, col-
able evidence from experimental and quasi-experimental laboration, personal responsibility, authentic writing tasks,
research does not support the contention that the process and a supportive learning environment. Instead, we sug-
approach to writing enhances students’ motivation, at least gest that advocates of process writing instruction integrate
when compared with more traditional and undefined writing other effective writing practices into this approach. There
interventions. We anticipated positive motivational effects, is some empirical evidence that this is a fruitful avenue to
as this approach includes a variety of instructional compo- pursue. For example, impressive improvements in the writ-
nents thought to enhance motivation (Wigfield, 1994), in- ing of average and struggling writers were obtained when
cluding peer collaboration, personal responsibility, and cre- the amount of explicit and systematic instruction provided
ation of a positive learning environment. It should be noted in process writing classrooms was increased (Curry, 1997;
that motivation is difficult to measure, and the methods used Danoff, Graham, & Harris, 1993; MacArthur, Schwartz, &
to assess it differed from one study to the next. This may have Graham, 1991). These studies involved teaching strategies
contributed to the obtained outcome. Additional research for planning and revising. Other studies are needed to deter-
is needed to assess possible motivational effects of process mine if incorporating other evidence-based practices, such
writing. as sentence combining (Graham & Perin, 2007) or spelling
Contrary to expectations, we did not find a statistically and handwriting instruction (Graham, 2010), into the pro-
significant relationship between variability in writing qual- cess writing instruction further enhances the power of this
ity ES and study characteristics, such as receiving profes- approach.
sional development, grade level of students (1–6 vs. 7–12), Our recommendation that the process writing approach
or establishing reliability of the outcome measure. Likewise, needs to undergo experimentation and change is not a
genre assessed and study quality did not account for excess radical idea. In fact, process writing has evolved since the
variability in ES (as predicted). The finding that scoring 1980s, when pioneers such as Graves (1983) and Calkins
reliability was not associated with magnitude of ES pro- (1983) began promoting this method for teaching writing
vides some support for our decision to include studies where (Pritchard & Honeycutt, 2006). Groups such as the National
reliability for writing quality was and was not established. Writing Project have also taken a flexible stance over time,
Previous reviews (Graham & Perin, 2007; Hillocks, 1986) recommending the use of evidence-based practices such as
excluded studies when reliability was absent. However, the inquiry learning and sentence combining (Hillocks, 1986) as
first two findings were unanticipated, as each (professional part of the process writing approach (Nagin, 2006). Perhaps
development and grade) had accounted for significant vari- the greatest experimentation takes place in schools, where
ance in process writing effects in Graham and Perin (2007). more teachers combine process writing and traditional skills
The reasons for the obtained differences between this meta- instruction together than just teach the process writing
analysis and the previous one are unclear, but it is likely due, approach alone (Cutler & Graham, 2008; Graham, Harris,
at least in part, to differences in the criteria for including Fink, & MacArthur, 2002). High-quality research is needed
and excluding studies. to examine the effectiveness of the most promising hybrids.
We indicated previously that the process writing approach
is likely to play a significant role in any serious effort to re-
form writing instruction in the United States. Even if writ- REFERENCES
ing remains mostly absent from national reform efforts, it is
likely that this approach will continue to play a key role in References marked with an asterisk (∗ ) indicate studies included in the meta-
analysis
how writing is taught due to its popularity and the available
infrastructure for promoting its implementation and contin- Achieve, Inc. (2005). Rising to the challenge: Are high school graduates prepared
ued use. If this is the case, this review raises an important for college and work? Washington, DC: Author.
∗ Adams, V. A. (1971). A study of the effects of two methods of teaching com-
concern. Although the process approach is effective in im- position to twelfth graders (Unpublished doctoral dissertation). University
proving the writing of typical students, it is not a particularly of Illinois, Champaign-Urbana, IL.
∗ Alloway, E., Carroll, J., Emig, J., King, B., Marcotrigiano, I., Smith, J., &
powerful approach relative to other writing treatments, and
Spicer, W. (1979). The New Jersey Writing Project. New Brunswick, NJ:
its impact with those that are most vulnerable educationally, Rutgers University, Educational Testing Program Service, and Nineteen
ELLs and children with learning disabilities, are unproven New Jersey Public School Districts.
outside of a few case studies. Unfortunately, case studies do Atwell, N. (1987). In the middle: Reading, writing, and learning from adoles-
cents. Portsmouth, NH: Heinemann.
not provide the strongest evidence for establishing the value Baines, L., Baines, C., Stanley, G., & Kunkel, A. (1999). Losing the product
of an instructional practice, as the data presented is often se- in the process. English Journal, 88, 67–72.
406 The Journal of Educational Research

Bangert-Drowns, R. (1993). The word processor as an instructional tool: Goldring, A., Russell, M., & Cook, A. (2003). The effects of computers on
A meta-analysis of word processing in writing instruction. Review of student writing: A meta-analysis of studies from 1992–2002. Journal of
Educational Research, 63, 69–93. Technology, Learning, and Assessment, 2, 1–51.
Bates, L., Breslow, N., & Hupert, N. (2009). Five states’ efforts to improve Graham, S. (2006). Writing. In P. Alexander & P. Winne (Eds.), Handbook
adolescent literacy. Washington, DC: Institute of Education Sciences. of Educational psychology (pp. 457–478). Mahwah, NJ: Earlbaum.
∗ Beachem, M. T. (1984). An investigation of two writing process interven- Graham, S. (2010). Teaching writing. In P. Hogan (Ed.), Cambridge ency-
tions on the rhetorical effectiveness of sixth grade writers (Unpublished doc- clopedia of language sciences (pp. 848–851). Cambridge, UK: Cambridge
toral dissertation). Rutgers, the State University of New Jersey, New University Press.
Brunswick, NJ. Graham, S., & Harris, K. R. (1997). Whole language and process writing:
Breetvelt, I., Van den Bergh, H., & Rijlaarsdam, G. (1994). Relations Does one approach fit all? In J. Lloyd, E. Kameenui, & D. Chard (Eds.),
between writing processes and text quality: When and how. Cognition Issues in educating students with disabilities (pp. 239–258). Hillsdale, NJ:
and Instruction, 12, 103–123. Erlbaum.
Breetvelt, I., Van den Bergh, H., & Rijlaarsdam, G. (1996). Rereading and Graham, S., Harris, K. R., Fink, B., & MacArthur, C. (2002). Primary
generating and their relation to text quality: An application of mutilevel grade teachers’ theoretical orientations concerning writing instruction:
analysis on writing process data. In G. Rijlaarsdam, H. Van den Bergh, & Construct validation and a nationwide survey. Contemporary Educational
M. Couzjin (Eds.), Theories, models and methodologies on writing research Psychology, 27, 147–166.
(pp. 10–21). Amsterdam, The Netherlands: Amsterdam University Press. Graham, S., & Perin, D. (2007). A meta-analysis of writing instruction for
Calkins, L. (1983). Lesson from a child: On the teaching and learning of writing. adolescent students. Journal of Educational Psychology, 99, 445–476.
Portsmouth, NH: Heinemann. Graves, D. (1983). Writing: Teachers and children at work. Exeter, NH:
Calkins, L., & Colleagues (2008). Units of study for primary writing: A Heinemann.
yearlong curriculum (K–2). Portsmouth, NH: Heinemann. ∗ Green, L. C. (1991). The effects of word processing and a process approach to
∗ Clippard, D., & Nicaise, M. (1998). Efficacy of writers’ workshop for writing on the reading and writing achievement, revision and editing strategies,
students with significant writing deficits. Journal of Research in Childhood and attitudes towards writing of third-grade Mexican American students (Un-
Education, 13, 7–26. published doctoral dissertation). University of Texas at Austin, Austin,
Cousin, P., Aragon, E., & Rojas, R. (1993). Creating new conversations TX.
about literacy: Working with special needs students in a middle-school Greene, J. (2000). The cost of remedial education: How much Michigan pays
classroom. Learning Disability Quarterly, 16, 282–298. when students fail to learn basic skills. Midland, MI: Mackinac Center for
∗ Croes, M. J. (1990). The efficacy of employing a writing process approach Public Policy.
for the instruction of language arts with learning disabled elementary students Greenwald, E., Persky, H., Ambell, J., & Mazzeo, J. (1999). National assess-
(Unpublished doctoral dissertation). Baylor University, Waco, TX. ment of Educational progress: 1998 report card for the nation and the states.
∗ Curry, K. A. (1997). A comparison of the writing products of students with Washington, DC: U.S. Department of Education.
learning disabilities in inclusive and resource room settings using different writ- ∗ Hamilton, A. C. (1992). Performance assessment of personal correspondence
ing approaches (Unpublished doctoral dissertation). Florida Atlantic Uni- on the development of written language use and functions in traditional and pro-
versity, Boca Raton, FL. cess writing second-grade classrooms (Unpublished doctoral dissertation).
Cutler, L., & Graham, S. (2008). Primary grade writing instruction: A University of Alabama, Birmingham, AL.
national survey. Journal of Educational Psychology, 100, 907–919. Hammill, D. D., & Larsen, S. (1996). Test of written language–3. Austin,
Danoff, B., Harris, K. R., & Graham, S. (1993). Incorporating strategy TX: Pro-Ed.
instruction within the writing process in the regular classroom: Effects ∗ Hansen, T. L. (1989). A comparison of the effects of two process writing
on the writing of students with and without learning disabilities. Journal programs and a traditional program on the writing development of first-grade
of Reading Behavior, 25, 295–322. children (Unpublished doctoral dissertation). University of South Dakota,
Diederich, P. (1966). How to measure growth in writing ability. English Vermillion, SD.
Journal, 55, 435–449. ∗ Hayes, B. L. (1984). The effects of implementing process writing into a seventh
∗ Dougans, J. Y. (1993). Teaching writing to fifth-grade children: The pro- grade English curriculum (Unpublished doctoral dissertation). Delta State
cess/conference approach versus the traditional method (Unpublished doc- University, Cleveland, MS.
toral dissertation). Temple University, Philadelphia, PA. Hedges, L. V. (1982). Estimation of effect size from a series of independent
∗ Eads, V. A. (1989). A study of the effects of teacher training in writing on experiments. Psychological Bulletin, 92, 490–499.
students’ writing abilities (Unpublished doctoral dissertation). Baylor Uni- Herbison, P., Hay-Smith, J., & Gillespie, W. (2006). Adjustment of meta-
versity, Waco, TX. analyses on the basis of quality scores should be abandoned. Journal of
∗ Ewing, S. C. (1992). The effects of time and instruction on writing perfor- Clinical Epidemiology, 59, 1249–1256.
mance of eighth-grade students in writing assessment situations (Unpublished Hillocks, G. (1986). Research on written composition: New directions for teach-
doctoral dissertation). Purdue University, West Lafayette, IN. ing. Urbana, IL: National Council of Teachers of English.
∗ Fleury, B. (1988). Cumulative editing writing instruction: Implications for ∗ Kelley, K. R. (1984) The effect of writing instruction on reading comprehension
staff development (Unpublished doctoral dissertation). Northern Arizona and story writing ability (Unpublished doctoral dissertation). University
University, Flagstaff, AZ. of Pittsburgh, Pittsburgh, PA.
Foulger, T., & Jimenez-Silva, M. (2007). Enhancing the writing devel- Kiuhara, S., Graham, S., & Hawken, L. (2009). Teaching writing to high
opment of English language learners: Teacher perceptions of common school students: A national survey. Journal of Educational Psychology, 101,
technology in project-based learning. Journal of Research in Childhood 136–160.
Education, 22, 109–124. Lipsey, M., & Wilson, D. (2001). Practical meta-analysis. Thousand Oaks,
Fox, D., & Suhor, C. (1986). Limitations of free writing. English Journal, CA: Sage.
75, 34–36. MacArthur, C., Schwartz, S., & Graham, S. (1991). Effects of a reciprocal
Gamelin, Y. M. A. (1996). The effects of Cognitive Strategy Instruction in peer revision strategy in special education classrooms. Learning Disability
Writing (CSIW) on the writing skills of severely learning disabled students and Research and Practice, 6, 201–210.
their peers in an inclusive classroom (Unpublished master’s thesis). Simon Maehr, M., & Midgley, C. (1991). Enhancing student motivation: A
Fraser University, Burnaby, British Columbia, Canada. schoolwide approach. Educational Psychologist, 26, 399–427.
∗ Ganong, F. L. (1974). Teaching writing through the use of a program based ∗ Marker, E. S. (2000). The impact of the writer’s conference on the writing
on the work of Donald M. Murray (Unpublished doctoral dissertation). achievement of at-risk fourth-grade students (Unpublished doctoral disser-
Boston University, Boston, MA. tation). Widener University, Chester, PA.
∗ Gauntlett, J. F. (1978). Project WRITE and its effect on the writing of high ∗ Moore, M. A. (1987). The effect of word processing technology in a develop-
school students (Unpublished doctoral dissertation). Northern Arizona mental writing program on writing quality, attitude towards composing, and
University, Flagstaff, AZ. revision strategies of fourth and fifth grade students (Unpublished doctoral
Gersten, R., Fuchs, L., Compton, D., Coyne, M., Greenwood, C., & In- dissertation). University of South Florida, Tampa, FL.
nocenti, M. (2005). Quality indications for group experimental and Moye, M. J. (1993). The impact of a cognitive strategy on students’ composing
quasi-experimental research in special education. Exceptional Children, skills (Unpublished doctoral dissertation). College of William and Mary,
71, 149–164. Williamsburg, VA.
The Journal of Educational Research 407

Nagin, C. (2006). Because writing matters: Improving student writing in our Shadish, W. R., Robinson, L., & Congxiao, L. (1999). ES: A computer
schools. San Francisco, CA: Jossey-Bass. program for effect size calculation. Memphis, TN: University of Memphis.
National Commission on Writing. (2003). The neglected R: The need for a Smith, M. L., Glass, G. V., & Miller, T. I. (1980). The benefits of psychother-
writing revolution. New York, NY: CollegeBoard. apy. Baltimore, MD: Johns Hopkins University Press.
∗ Olson, M. C., & DiStefano, P. (1980). Describing and testing the effec- Stotsky, S. G. (1988). Commentary. Research in the Teaching of English, 22,
tiveness of a contemporary model for in-service education in teaching 89–99.
composition. Engineering Education, 12, 69–76. Troia, G., & Graham, S. (2002). The effectiveness of a highly explicit,
∗ Pantier, T. F. (1999). A comparison of writing performance of fifth-grade stu- teacher-directed strategy instruction routine: Changing the writing per-
dents using the process writing approach and the Shurley method (Unpublished formance of students with learning disabilities. Journal of Learning Dis-
doctoral dissertation). Oklahoma State University, Stillwater, OK. abilities, 35, 290–305.
Persky, H. R., Daane, M. C., & Jin, Y. (2003). The nation’s report card: ∗ Umbach, B. T. (1990). A comparison of two methods of teaching written
Writing 2002 (NCES 2003–529). Washington, DC: U.S. Department of language to low-performing fourth graders in two rural schools (Unpublished
Education, Institute of Education Sciences, National Center for Educa- doctoral dissertation). Auburn University, Auburn, AL.
tion Statistics. Van den Bergh, H., & Rijlaarsdam, G. (1996). The dynamics of composing:
Phye, G., Robinson, D., & Levin, J. (Eds.). (2005). Empirical methods for Modeling writing process data. In C. Levy & S. Ransdell (Eds.), The
evaluating educational interventions. San Diego, CA: Elsevier. science of writing (pp. 207–232). Mahwah, NJ: Erbaum.
∗ Pollington, M. F. (1999). Intermediate grade writers self-perception: A com- ∗ Weiss, D. H. (1992). The effects of writing process instruction on the writ-
parison of the effects of writing workshop and traditional instruction (Unpub- ing and reading performance of students with learning disabilities (Unpub-
lished doctoral dissertation). Brigham Young University, Provo, UT. lished doctoral dissertation). Florida International University, Miami,
∗ Pritchard, R. J. (1987). Effects on student writing of teacher training in FL.
the National Writing Project Model. Written Communication, 4, 51–67. What Works Clearinghouse. (2007). Technical details of WWC-
Pritchard, R. J., & Honeycutt, J. (2006). Process writing. In C. MacArthur, conducted computations. Retrieved from http://www.whatworks.ed.gov/
S. Graham, & J. Fitzgerald (Eds.), Handbook of Writing Research (pp. reviewprocess/conducted computations.pdf
275–290). New York, NY: Guilford. Wienke, J. W. (1981). Strategies for improving elementary school students’
∗ Pritchard, R. J., & Marshall, J. C. (1994). Evaluation of a tiered model writing skills. Retrieved from ERIC database. (ED209679)
for staff development in writing. Research in the Teaching of English, 28, Wigfield, A. (1994). The role of children’s achievement values in the self-
259–285. regulation of their learning outcomes. In. D. Schunk & B. Zimmerman
∗ Reimer, M. (2001). The effect of a traditional, a process writing, and a com- (Eds.), Self-regulation of learning and performance: Issues and educational
bined talking and writing instructional approach on the quality of secondary applications (pp. 101–124). Hillsdale, NJ: LEA.
English students’ written response (Unpublished master’s thesis). Univer- Yeh, S. (1998). Empowering education: Teaching argumentative writing
sity of Manitoba, Winnipeg, Manitoba, Canada. to cultural minority middle-school students. Research in the Teaching of
Rijlaarsdam, G., & Van den Bergh, H. (2006). Writing process theory: English, 33, 49–48.
A functional dynamic approach. In C. MacArthur, S. Graham, & J. Zaragoza, N., & Vaughn, S. (1992). The effects of process writing instruc-
Fitzgerald (Eds.), Handbook of writing research (pp. 41–53). New York, tion on three 2nd-grade students with different achievement profiles.
NY: Guilford. Learning Disabilities Research and Practice, 7, 184–193.
∗ Roberts, C. A. (2002). The influence of teachers’ professional development at Zucker, C. (1993). Using whole language with students who have language
the Tampa Bay Area Writing Project on student writing performance (Un- and learning disabilities. Reading Teacher, 46, 660–670.
published doctoral dissertation). University of South Florida, Tampa,
FL.
Rogers, L., & Graham, S. (2008). A meta-analysis of single subject de- AUTHORS NOTE
sign writing intervention research. Journal of Educational Psychology, 100,
879–906.
Rosenberg, M., Adams, D., & Gurevitch, J. (2000). MetaWin: Statistical Steve Graham is the Currey-Ingram Professor of Special
software for meta-analysis (Version 2). Sunderland, MA: Sinauer Ass. Education and Literacy at Vanderbilt University. His re-
Salahu-Din, D., Persky, H., & Miller, J. (2008). The Nation’s Report Card: search interests include writing development, writing in-
Writing 2007 (NCES 2008–468). Washington, DC: U.S. Department of
Education, Institute of Education Science, National Center for Education struction, and writing-to-reading connections.
Statistics. Karin Sandmel is a doctoral student at Vanderbilt Uni-
∗ Scannella, A. M. (1982). A writing-as-process model as a means for improving
versity in the Department of Special Education. Her inter-
compositions and attitudes toward composition in the high school (Unpublished
doctoral dissertation). Rutgers, the State University of New Jersey, New ests involve writing and students who are second-language
Brunswick, NJ. learners.
Copyright of Journal of Educational Research is the property of Taylor & Francis Ltd and its content may not
be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.

View publication stats

You might also like