You are on page 1of 19

1027537

research-article2021
ECXXXX10.1177/00144029211027537Exceptional ChildrenDatchuk et al.

Research Syntheses
Exceptional Children

Effects of Writing Interventions


 ­
2022, Vol. 88(2) 145–162
© The Author(s) 2021
Article reuse guidelines :
https://doi.org/10.1177/00144029211027537
on the Level and Trend of Total sagepub.com/journals-permissions
DOI: 10.1177/00144029211027537
Words Written: A Meta-Analysis journals.sagepub.com/home/ecx

Shawn M. Datchuk1, Derek B. Rodgers2, Kyle Wagner3,


Bridget O. Hier4, and Christopher T. Moore5

Abstract
We estimated effects of writing interventions on the level and trend of writing fluency—rate
of total words written over time—by students with and without disabilities. Using mixed-
effects regression and an information-theoretic ranking of competing models, we synthesized
results of 42 single-case experimental design studies with a total of 205 students. A variety of
acquisition and fluency interventions were used across studies, such as self-regulated strategy
development and timed practice with performance feedback. We found acquisition and fluency
interventions produced an increase in level and a gradual increase in trend of total words
written per minute. Students receiving fluency intervention tended to have higher levels of
performance across experimental phases (i.e., baseline, intervention, and postintervention), but
students receiving acquisition had steeper upward trends. In addition, we found higher levels of
total words written per minute on writing tasks with brief time limits (i.e., ≤10 min) and in the
writing of older students.

Writing is important to academic success. text is transcribed and generated with ease and
Across content areas, students are expected to automaticity (Ritchey et al., 2016). Writing
complete a variety of writing tasks to learn fluency plays an important cognitive role in
new content and express their understanding proficient writing. According to the simple
(Ray et al., 2016). Unfortunately, many stu- view of writing (Berninger & Amtmann,
dents with and without disabilities struggle to 2003), proficient writing results from at least
develop proficient writing. For instance, three composite areas: text generation, tran-
approximately 96% of eighth-grade students scription, and executive functions. Text gen-
with disabilities and 71% of eighth-grade stu- eration is the composition of connected text
dents without disabilities scored basic or into phrases, sentences, and composition.
below—displaying only a partial mastery of Transcription is the handwriting or typing of
fundamental writing skills—on the most letters into correctly spelled words. Execu-
recent administration of the writing subtest of tive functions help coordinate transcription
the National Assessment of Education Prog- and text generation within the writing process
ress in 2011 (U.S. Department of Education
Institute of Education Sciences, National Cen- 1
University of Iowa
ter for Educational Statistics, 2011). In addi- 2
University of Nebraska–Lincoln
3
tion to disability status, writing proficiency University of Findlay
4
Syracuse University
has been found to vary by other demographic 5
Minneapolis Public Schools
variables, such as age and gender (Fearrington
Corresponding Author:
et al., 2014; Keller-Margulis et al., 2015). Shawn Datchuk, University of Iowa, 254N Lindquist
Writing difficulties stem in part from issues Center, Iowa City, IA 52242, USA.
with writing fluency—the degree to which Email: shawn-datchuk@uiowa.edu
146 Exceptional Children 88(2)

(i.e., planning, translating, and revising) and (Ritchey et al., 2016). Although early
entail aspects of self-regulation, attention, research established a relationship between
and writing strategies. When transcription or TWW and writing quality (Deno et al.,
text-generation skills are fluent (i.e., pro- 1982), recent research demonstrates that of
duced with ease or automaticity), they are the common writing metrics—such as cor-
transferred to long-term memory, freeing up rect writing sequences (CWS) and words
working and short-term memory to attend to spelled correctly (WSC)—TWW has the
other aspects of a writing task and overall weakest criterion validity (r = .37; Romig
writing development (Graham, 2018). Indeed, et al., 2017) and some of the lowest sensitiv-
research suggests writing fluency in tran- ity and specificity to identify struggling writ-
scription and text generation is related to ers (Furey et al., 2016). In contrast, writing
gains in writing quantity and quality (Datchuk metrics based on CWS, which accounts for
& Kubina, 2013; Kim et al., 2018). spelling, capitalization, grammar, syntax,
and punctuation, typically have stronger psy-
Writing Fluency chometric properties, such as criterion valid-
ity (r = .51 to .60; Romig et al., 2017).
Measurement Despite the limitations of its technical fea-
Measurement of writing fluency is a recom- tures, TWW is a time-efficient, simple met-
mended aspect of schoolwide systems, such as ric to score with little training (Gansle et al.,
response to intervention, to prevent and iden- 2004); in fact, students can even learn to
tify writing difficulties (Shapiro, 2011). Argu- score TWW, cutting down on teachers’
ably, the most well-researched and widely assessment time (e.g., Koenig et al., 2016).
used writing fluency task is curriculum-based Performance on CBM-WE, as well as any
measurement of written expression (CBM- writing task with a standard time limit for
WE; McMaster et al., 2020). Following stan- writing, can be summarized in two ways: (a)
dard CBM-WE procedures, students are level, which represents either a single datum
provided with a writing prompt (e.g., story point of performance or an average of data
starter or picture) and then are given specific points across time, and (b) trend, which is a
time limits for planning and writing. A time slope summarizing growth across time (Riley-
limit of 1 min is typical for planning, and a time Tillman et al., 2020). Both level and trend can
limit of 1.5 to 10 min is typical for writing be compared with set criteria (e.g., goal or
(Romig et al., 2020). Brief time limits increase aim) in order to judge the adequacy of instruc-
the likelihood that student performance is tion or intervention. For instance, the trend of
indicative of writing fluency—automaticity writing performance can be compared with a
or speed of transcription and text generation national norm: An average student is likely to
(Ritchey et al., 2016)—as students are likely to gain approximately 0.17 to 0.47 TWW per
engage in sustained writing and devote little week on 3-min CBM-WE probes (AIMSweb,
time to executive functions or self-regulation, 2016). Although the limited growth students
such as stopping to consider plans and revision. make in TWW on CBM-WE probes may sug-
gest insufficiency with the measure or metric,
Measurement of writing fluency is a it is also possible that universal writing
recommended aspect of schoolwide instruction insufficiently targets writing pro-
systems, such as response to duction (Cutler & Graham, 2008).
intervention . . .
Writing Fluency Intervention
One commonly used writing fluency met-
ric is total words written (TWW), which is a Writing interventions can be categorized based
broad indicator of transcription and text-gen- on their intended outcome of acquisition and
eration speed. It is the number of words pro- fluency (Codding et al., 2019). Categories of
duced regardless of whether they follow acquisition and fluency have typically been
conventions of spelling, semantics, or syntax applied to math and reading interventions
Datchuk et al. 147

(e.g., Maki et al., 2020; Parker & Burns, 2014); addition to improvements in self-regulation of
however, there is an emerging line of research the writing process. In graphic organizer
on writing interventions (Parker et al., 2011). instruction, students engage in deliberate
Acquisition interventions are intended to planning of a writing genre by completing a
increase an area of low proficiency (e.g., low visual scaffold (e.g., fill in the blanks of a per-
level of TWW) through instruction that relies suasive essay). Procedures of SRSD are simi-
heavily on modeling, prompting, initial prac- lar but broader in scope; it entails (a) strategy
tice, and error correction (Codding et al., instruction on how to plan, compose, and
2019). Alternatively, fluency interventions are revise a genre of writing (e.g., narrative story)
intended to increase the speed or rate of per- and (b) self-regulation components of goal
formance (e.g., increase level or trend of setting, self-monitoring, self-instruction, and
TWW) through practice opportunities and self-reinforcement. As part of acquiring writ-
performance feedback (Parker & Burns, ing skills through SRSD, students set goals
2014). Fluency interventions often incorporate for writing (e.g., genre elements or holistic
acquisition instructional techniques, but a quality) and self-monitor their progress.
defining characteristic of fluency interventions Effects of fluency intervention have typi-
is the use of timed practice with an explicit cally not been reported in prior reviews of
focus on increasing automaticity, such as writing interventions. Results from several
10-min practice trials composing persuasive studies do suggest timed practice involving
essays (Mason et al., 2013). Because single- text generation and self-regulation can improve
case experimental designs (SCEDs) feature writing fluency (Hier & Eckert, 2016; Hoover
student-level data across time, results from et al., 2012; Koenig et al., 2016; Mason et al.,
these studies can be used to determine level 2013). Timed practice involves writing during
and trend effects of acquisition and fluency a brief time limit (e.g., 3 min or 10 min),
interventions on writing fluency or TWW receiving performance feedback (e.g., TWW
(Riley-Tillman et al., 2020). or CWS), goal setting, and graphing of writing
Based on prior reviews of the SCED writ- performance. In addition, timed practice can
ing intervention literature (e.g., Ciullo & be paired with initial instruction, such as
Reutebuch, 2013; Graham, 2006; Graham SRSD, to increase speed or rate of text genera-
et al., 2013; Rogers & Graham, 2008), more is tion of genre elements and writing quality
known about effects of intervention on writ- (Hoover et al., 2012; Mason et al., 2013).
ing length (i.e., count of TWW on writing Compared with intervention effects on
tasks without a specified time limit) than writ- writing fluency of TWW, more is known
ing fluency (i.e., rate of TWW on writing regarding level and trend effects on CWS. In a
tasks with a specified time limit). Further- recently published meta-analysis (Datchuk,
more, prior reviews (e.g., Rogers & Graham, Wagner, et al., 2020), researchers reported
2008) have primarily used overlap effect size that writing interventions produced little to no
metrics, such as percentage of nonoverlap- change in level and a gradual change in trend
ping data, that are insensitive to changes in of CWS. On average, students gained approx-
level and trend (Wolery et al., 2010). Results imately 0.39 CWSs per minute for each inter-
from prior reviews do suggest two types of vention session or four CWSs per minute for
acquisition interventions—graphic organizer every 10 intervention sessions. Writing inter-
instruction (Ciullo & Reutebuch, 2013) and ventions included several acquisition and flu-
self-regulated strategy development (SRSD; ency interventions, such as SRSD, explicit
Graham, 2006; Graham et al., 2013; Rogers & instruction, and timed practice with perfor-
Graham, 2008)—can improve writing length. mance feedback. Older students had higher
Graphic organizer instruction and SRSD levels of CWS per minute compared with
are primarily used to improve aspects of text younger students, but no significant differ-
generation, such as genre elements (e.g., per- ences were found for disability status, gender,
suasive essay parts) and holistic quality, in and race. Because CWS is scored differently
148 Exceptional Children 88(2)

than TWW, CWS summarizes unique aspects were excluded. Fifth, several demographic
of text writing not accounted for by TWW variables were reported: age or grade, disabil-
(e.g., semantics, spelling, and syntax). Thus, it ity status, and gender. Sixth, TWW data were
is important to examine if findings extend to presented in a graph or table for each student.
TWW. If studies met all criteria except for the sixth,
then we sent an email to lead authors asking
for their data.
Present Meta-Analysis A total of 42 articles met the inclusion crite-
We sought to extend prior research by esti- ria. To locate additional studies, we used four
mating level and trend effects of writing inter- search strategies: electronic, ancestral, forward,
ventions on the writing fluency (i.e., TWW on and hand search. The search began July 2019
timed writing tasks) of students with and and ended February 2021. The first author con-
without disabilities. We included only SCED ducted an electronic search of ERIC and Psy-
studies that reported TWW on timed writing cINFO databases in July 2019 using search
tasks with brief (i.e., ≤10 min) or extended terms similar to prior reviews and meta-analy-
(i.e., >10 min) time limits. We included both ses (Datchuk & Kubina, 2013; Datchuk, Wag-
types of timed writing tasks to have a broad ner, et al., 2020). We created two sets of search
estimate of writing fluency applicable to brief terms. The first set of search terms related to
tasks used for screening and progress moni- dependent variables and modes of transcrip-
toring (Romig et al., 2020) and to extended tion: “total words written,” “TWW,” “writing
tasks used for typical classroom instruction production,” “writing quantity,” “writing flu-
(Ray et al., 2016). We had three research ques- ency,” “written expression,” “curriculum-
tions. First, what are the effects of writing based measurement,” “CBM,” “handwriting,”
intervention on the level and trend of TWW? “typing,” or “keyboarding.” The second set of
Second, do effects differ by intervention type search terms related to independent variables:
(e.g., acquisition or fluency)? Third, do effects “academic intervention,” “treatment,” “fluency
differ by student demographics (e.g., age or practice,” “strategy instruction,” “self-regu-
gender) or writing task (e.g., brief or extended lated strategy development,” “SRSD,” “direct
time limits)? instruction,” “explicit instruction,” “perfor-
mance feedback,” “goal setting,” “error correc-
tion,” “word processing,” “computer use,” or
Method “timed practice.” Each search term was joined
Inclusion Criteria and Search by OR (e.g., “direct instruction OR explicit
instruction”), and the two sets were joined by
Strategies AND (i.e., “[all terms in Set 1] AND [all terms
We had six inclusion criteria. First, we included in Set 2]”). The search revealed 4,471 results in
published and unpublished (i.e., dissertations) ERIC and 6,107 results in PsycINFO.
studies written in English. Second, the study The first author downloaded all search
featured a SCED with repeated measurement results into a spreadsheet. Duplicates were
and multiple opportunities to detect an experi- noted but not removed. We used a two-stage
mental effect. Third, the independent variable screening process. First, titles and abstracts
was instructional in nature: An intervention were screened for eligibility. Second, if the
agent (e.g., student, teacher, or computer) pre- title or abstract met one or more of the inclu-
sented writing stimuli and prompted student sion criteria, then the full text was downloaded
responding. We excluded independent vari- and screened. At the end of the two-stage pro-
ables that were accommodations to support cess, a total of 14 articles met the inclusion
writing, such as word prediction or text-to- criteria. To check reliability of the screening
speech software. Fourth, TWW was measured process, 33% of all search results were ran-
on timed writing tasks that prompted text writ- domly selected and screened by the second
ing; handwriting and spelling assessments author. No discrepancies were noted.
Datchuk et al. 149

Following the electronic search, the first disabilities into two broad categories of high-
author located an additional 16 articles meet- and low-incidence disabilities based on their
ing the inclusion criteria from an ancestral and prevalence in K–12 schools in the United
a forward search conducted with Google States (McFarland et al., 2019). High-inci-
Scholar and Web of Science. Then, the first dence disabilities were the four most preva-
and second authors found one additional arti- lent disability types (i.e., each account for at
cle meeting the inclusion criteria through a least 10% of students receiving special edu-
hand search of eight journals (i.e., journals cation services): learning disabilities (LD),
with two or more articles meeting the inclu- speech and/or language impairment (S/LI),
sion criteria): Behavioral Disorders, Educa- other health impairment (OHI), and autism.
tion and Training in Autism and Developmental Low-incidence disabilities encompassed all
Disabilities, Exceptional Children, Exception- remaining disability types. Given little racial
ality, Journal of Behavioral Education, Jour- diversity in our sample (i.e., the number of
nal of Special Education Technology, Learning White students was almost double the next
Disabilities Research & Practice, and Learn- highest category), we also coded race as
ing Disability Quarterly. We hand searched White or non-White.
issues published since 1980, roughly corre- Independent variables were coded into one
sponding to the introduction of TWW in the of two categories: acquisition or fluency.
CBM literature (Deno et al., 1982). We Acquisition was defined as the delivery of
attempted to contact lead authors of studies instruction (e.g., SRSD) to introduce a skill or
that met all the inclusion criteria except for strategy without timed practice. Fluency
presentation of TWW as a table or figure. A involved the delivery of timed practice sepa-
total of nine lead authors responded to our rate from or in addition to instruction (e.g.,
email requests with their unpublished data. We SRSD with timed practice). Because SRSD
received the last data set December 2019. As a procedures were used in the majority of
final step, we searched the reference lists of included studies, we also coded for SRSD sta-
several related reviews of the writing interven- tus (i.e., SRSD or non-SRSD). Dependent
tion literature (Ciullo & Reutebuch, 2013; variables were coded for time limit (i.e., brief
Datchuk, Wagner, et al., 2020; Graham, 2006; or extended) and transcription mode (i.e.,
Graham et al., 2013; Rogers & Graham, 2008) handwritten or typed). Because writing flu-
and located two additional articles in January ency tasks typically range between 1.5 min
2021. We had a final total of 42 studies that and 10 min (Romig et al., 2020), we coded
met the inclusion criteria. any writing task between these time limits as
brief. Conversely, we coded any writing task
with a time limit greater than 10 min as
Data Coding and Extraction extended. When possible, we removed desig-
We created a coding manual, and the first nated planning time from time limits (e.g., 10
author served as the primary coder. We posted min of planning and 30 min of composition
the coding manual and all coded data online was coded as 30 min) or coded writing dura-
(Datchuk, Rodgers, et al., 2020). Several tion (i.e., time spent by students engaged in
aspects of student demographics, indepen- sustained writing) to better estimate writing
dent variables, dependent variables and meth- fluency; this was the case for five studies, or
odology were coded. Student demographics 12% of included studies (Asaro-Saddler &
included age, disability status, gender, and Saddler, 2010; Miller & Little, 2018; Saddler,
race. When only grade level was reported, we 2006; Shen & Troia, 2018; Southall, 2011).
imputed age: We assumed kindergarten stu- Similar to prior reviews (Datchuk & Kubina,
dents were 5 years old and increased age by 1 2013; Datchuk, Wagner, et al., 2020), we
year for each additional grade. We coded all coded for several aspects of the What Works
reported disability types. Given the low Clearinghouse (WWC; 2020) standards of
occurrence of some disability types, we coded methodological quality: experimental design,
150 Exceptional Children 88(2)

number of phases, data points per phase, fidel- information-theoretic approach to rank com-
ity, and interobserver agreement (IOA). peting regression models (Burnham et al.,
The first and second authors coded all 2011). There were three nested levels: writing
studies. The first author taught a graduate stu- observations (n = 2,965) at Level 1 nested
dent coding procedures related to student within students (n = 205) at Level 2 nested
demographics, independent variables, depen- within studies (n = 42) at Level 3. As part of
dent variables, and methodology. The gradu- our model-fitting process, we compared two-
ate student served as an independent observer and three-level models. We tried to fit a model
and coded a randomly selected 33% of arti- with no fixed parameters, a random effect of
cles meeting inclusion criteria (i.e., 14/42), sessions at Level 2, and random intercepts at
specifically, 1,554 cells within the coding Levels 2 and 3, but this model did not con-
spreadsheet. An exact agreement formula was verge. Ultimately, we found a three-level,
used for spreadsheet cells—number of agree- mixed-effects model with random intercepts
ments divided by the number of agreements at Levels 2 and 3 best accounted for variance,
plus disagreements multiplied by 100—and and it served as our basic model.
IOA was 100%. We populated the basic model with vari-
To extract data from each SCED figure, we ables thought a priori to influence TWWpM.
used WebPlotDigitzer (Rohatgi, 2018). It has At Level 1 were variables related to observa-
been used in prior reviews (e.g., Datchuk, tions across time (i.e., session and intervention
Wagner, et al., 2020), and it has been found to iteration). We coded session and intervention
have high rates of reliability (r = .99) between iteration as sequential data points (e.g., the
effect sizes calculated from extracted data and first observation was Session 0 and the second
those reported within primary studies (Drevon observation was Session 1), and we used indi-
et al., 2017). The first and second authors cator coding to specify experimental phase
extracted all data and checked each other’s (i.e., baseline, intervention, or postinterven-
accuracy, meeting to discuss any disagree- tion). We coded all postintervention sessions
ments until 100% agreement was reached. as occurring on the final intervention iteration
Given that administration time varied across (e.g., if there were four intervention iterations,
studies (e.g., 3 min to 90 min), TWW was then all postintervention data were coded as
converted to TWW per minute (TWWpM) to occurring on the fourth intervention iteration).
make comparisons across studies more appro- Level 2 had variables related to student demo-
priate. For multiple-baseline and multiple- graphics of age (i.e., we centered the intercept
probe designs, we partitioned data into on average age of students: 12 years old), dis-
baseline, intervention, and postintervention ability type (i.e., high incidence, low inci-
phases. For studies that used reversal (i.e., dence, or none), gender (i.e., female or male),
ABAB) designs (Bauernschmidt, 1991; Wolfe and race (i.e., White or non-White). Level 3
et al., 2000), we partitioned the first phase into had variables related to studies: intervention
baseline, then subsequent phases into inter- procedure (i.e., acquisition, fluency, SRSD, or
vention and postintervention (i.e., the second non-SRSD), writing task time limit (i.e., brief
A phase was treated as not a return to baseline or extended), and transcription mode (i.e.,
but rather a postintervention phase). handwritten or typed). Then, we contrasted
regression models of varying combinations of
those variables until a model of best fit was
Data Analysis determined (i.e., the regression model that
Similar to a prior meta-analysis that esti- accounted for the most variance in TWWpM).
mated level and trend from SCED studies Parameters in the top model had accompa-
(Datchuk, Wagner, et al., 2020), we used (a) nying level or trend estimates, and parameters
mixed-effects linear regression (Raudenbush in lower-performing models had no accompa-
& Bryk, 2002); (b) binary, categorical, and nying estimates. We calculated p values for
continuous variables (Long, 2012); and (c) an each parameter to ease interpretation; however,
Datchuk et al. 151

we did not rely on them to determine the dler & Saddler, 2010; Caroll, 2018; Hashey,
regression model of best fit as linear mixed- 2015; Levy, 2018; Lewandowski, 2011;
effect models may not meet common assump- MacArthur & Philippakos, 2010; Mourgkasi
tions of p values (Bates et al., 2015). We & Mavropoulou, 2018; Nordness et al., 2019;
performed the analysis with R (R Core Team, Ray et al., 2019; Saddler, 2006; Southall,
2020) and several packages, such as AICcmo- 2011; Werunga, 2018). Several SRSD studies
davg (Mazerolle, 2017). We posted the full list included multiple strategies for planning and/
of packages online (Datchuk, Rodgers, et al., or revising (McKeown et al., 2015; Shen &
2020). Troia, 2018), or SRSD was supplemented with
visual supports, such as student-drawn pic-
tures for planning (Dunn, 2012, 2015; Dunn &
Results Miller, 2016) and production of video self-
There were a total of 42 studies and 205 stu- models (Miller & Little, 2018). Besides SRSD,
dents. Table 1 shows parameter estimates, the remaining acquisition studies included
standard errors, t values, and p values of the graphic organizer instruction (Bishop et al.,
top model. Tables 2 and 3 are posted as sup- 2015; Gonzalez-Ledo et al., 2015; Unzueta &
plemental material: Table 2 describes student Barbetta, 2012) and technology-based graphic
demographics, independent variables, and organizers with strategy instruction and self-
dependent variables; and Table 3 shows meth- regulation components (Ahn, 2017; Evmenova
odological quality indicators. et al., 2016; Good, 2019). One study used
explicit instruction to teach multiple text-writ-
ing skills within sentences and paragraphs
Student Demographics (Viel-Ruma et al., 2010).
The average age of students was 12 years old There were a total of 15 fluency studies
(SD = 3.4) with a range of 6 years to 19 years (36%). In most fluency studies, instruction—
old. There were 144 males (70%) and 61 either SRSD or explicit instruction—was
females (30%). A total of 93 students had paired with timed practice and performance
high-incidence disabilities (45%), 72 had feedback. In several studies, SRSD instruc-
low-incidence disabilities (35%), and 40 had tion was first delivered, followed by timed
no disability (20%). For specific disability practice composing 10-min persuasive essays
types, 68 students had an emotional-behav- and performance feedback on genre elements
ioral disorder (EBD; 33%), 62 had LD (30%), (e.g., persuasive essay parts) or holistic qual-
40 had no disability (20%), 35 had OHI ity (Cerar, 2012; Garwood et al., 2019; Hoover
(17%), 31 had autism (15%), 10 had S/LI et al., 2012; Mason et al., 2010, 2013; Mong
(5%), and five had intellectual disability (2%). Cramer & Mason, 2014). In some SRSD stud-
Students with comorbid disabilities were ies, goal setting (e.g., number of essay parts or
coded in multiple categories. Ninety-four stu- quality score) was used to indicate when stu-
dents were White (46%), 49 were African dents attained mastery or fluency (Cerar,
American (24%), 29 were not specified 2012; Garwood et al., 2019; Mason et al.,
(14%), 23 were Hispanic (11%), seven were 2010, 2013; Mong Cramer & Mason, 2014).
multiracial (3%), two were Asian (1%), and Besides SRSD, a combination of explicit
one was Native American (<1%). instruction, timed practice (i.e., 3-min or
12-min practice trials), performance feed-
back, and goal setting was used in two studies
Independent Variables (Hough et al., 2012; Rodgers, 2019). In a
There were a total of 27 acquisition studies smaller number of fluency studies, timed
(64%). Most acquisition studies featured a practice was provided without additional
SRSD strategy to teach planning, writing, and instruction. In those studies, students engaged
revising an essay or narrative (Asaro-Saddler, in timed practice writing 3-min narratives
2014; Asaro-Saddler & Bak, 2012; Asaro-Sad- (Bauernschmidt, 1991; Geisler et al., 2009;
152
Table 1. Top Regression Model: Description of Parameters.

Parameter Name Description Estimate SE t p


Default estimates: Parameters based on brief writing tasks completed in acquisition studies
γ000 Intercept Level of TWWpM when other variables 7.06 0.79 8.89 <.01
are 0 during baseline
γ100 Session Trend of TWWpM during baseline −0.01 0.02 −0.29 .77
γ200 Intervention iteration Trend of TWWpM during intervention 0.16 0.02 6.36 <.01
γ300 Intervention Level of TWWpM during intervention 0.97 0.16 5.99 <.01
γ400 Postintervention Level of TWWpM during postintervention 1.34 0.18 7.29 <.01
Added estimates: Parameters added to model if present in studies
γ500 Extended tasks Level of TWWpM on extended writing −3.57 0.55 −6.52 <.01
tasks across phases
γ600 Fluency studies Level of TWWpM in fluency studies across 2.15 1.15 1.87 .07
phases
γ700 Extended Tasks × Intervention Iteration Trend of TWWpM on extended writing −0.07 0.02 −3.37 <.01
tasks during intervention
γ800 Fluency Studies × Intervention Iteration Trend of TWWpM during fluency −0.06 0.02 −3.14 <.01
intervention sessions
γ900 Age Level of TWWpM above the mean age of 0.27 0.13 2.13 <.05
12 years old
Random effects assigned to levels
u00 s Level 2 random effects of intercept Random changes in level at student level
r0 ps Level 3 random effects of intercept Random changes in level at study level

Note. TWWpM = total words written per minute. Level is mean across phase(s), and trend is rate of growth per session. Brief writing tasks were ≤10 min, and extended writing
tasks were >10 min.
Datchuk et al. 153

Lavik, 2014; Stotz et al., 2008; Taylor, 2010; Handwritten responses were most com-
Wolfe et al., 2000) and 15-min journal narra- mon, in 35 studies (83%), but students typed
tives (Regan et al., 2005). Following timed their responses in seven studies, or 17% of
practice, students received performance feed- total studies (Ahn, 2017; Dunn & Miller,
back on an aspect of their writing: TWW 2016; Evmenova et al., 2016; Good, 2019;
(Bauernschmidt, 1991; Geisler et al., 2009; Hashey, 2015; Nordness et al., 2019; Unzueta
Stotz et al., 2008; Taylor, 2010; Wolfe et al., & Barbetta, 2012). In 18 studies (43%), hand-
2000), CWS (Lavik, 2014), and journal com- written student responses were later typed by
position (Regan et al., 2005). In addition, researchers to ease data analysis and adjust
some of those studies also featured goal set- for potential presentation effects (e.g., hand-
ting (Lavik, 2014; Rodgers, 2019; Wolfe writing legibility) associated with scoring
et al., 2000) and graphing (Bauernschmidt, (Asaro-Saddler, 2014; Asaro-Saddler & Bak,
1991; Geisler et al., 2009; Lavik, 2014; Stotz 2012; Asaro-Saddler & Saddler, 2010; Caroll,
et al., 2008; Taylor, 2010; Wolfe et al., 2000). 2018; Cerar, 2012; Mong Cramer & Mason,
2014; Hoover et al., 2012; Hough et al., 2012;
MacArthur & Philippakos, 2010; Mason
Dependent Variables et al., 2010, 2013; McKeown et al., 2015;
Writing tasks with an extended time limit Ray et al., 2019; Regan et al., 2005; Saddler,
(i.e., >10 min) were administered in the 2006; Shen & Troia, 2018; Southall, 2011;
majority of studies: They were administered Werunga, 2018).
in 24 studies (57%) with an average time
limit of 26.2 min (SD = 10.9) and a range of
Methodology
12 to 50 min (Ahn, 2017; Asaro-Saddler,
2014; Asaro-Saddler & Bak, 2012; Asaro- Multiple-baseline or multiple-probe designs
Saddler & Saddler, 2010; Bishop et al., 2015; were used in all but three studies (Bauern-
Caroll, 2018; Dunn, 2012, 2015; Dunn & schmidt, 1991; Lavik, 2014; Wolfe et al.,
Miller, 2016; Evmenova et al., 2016; Gonza- 2000). Following quality recommendations
lez-Ledo et al., 2015; Good, 2019; Hashey, (WWC, 2020), all studies except for two
2015; Hough et al., 2012; MacArthur & (Lavik, 2014; Lewandowski, 2011) met stan-
Philippakos, 2010; McKeown et al, 2015; dards—with or without reservations—for
Mourgkasi & Mavropoulou, 2018; Nordness minimum number of data points per phase and
et al., 2019; Ray et al., 2019; Regan et al., minimum number of phases. In most studies,
2005; Saddler, 2006; Shen & Troia, 2018; 24 studies (57%), data collection was inter-
Unzueta & Barbetta, 2012; Werunga, 2018). rupted (i.e., no data were collected) during
In two additional studies (5%), extended time intervention, and baseline performance was
limits were provided of 30 min (Miller & compared with a postintervention phase (Ahn,
Little, 2018) and 90 min (Southall, 2011); 2017; Asaro-Saddler, 2014; Asaro-Saddler &
however, writing duration (i.e., time students Bak, 2012; Asaro-Saddler & Saddler, 2010;
spent engaged in writing) was only a portion Bishop et al., 2015; Cerar, 2012; Dunn, 2012,
of the time limit, ranging from 2 min to 76 2015; Dunn & Miller, 2016; Evmenova et al.,
min. Writing tasks with brief time limits (i.e., 2016; Garwood et al., 2019; Good, 2019;
≤10 min) were administered in 16 studies Hashey, 2015; Hoover et al., 2012; Hough
(38%), and tasks were either 3 min or 10 min et al., 2012; MacArthur & Philippakos, 2010;
(Bauernschmidt, 1991; Cerar, 2012; Gar- McKeown et al., 2015; Mong Cramer &
wood et al., 2019; Geisler et al., 2009; Hoover Mason, 2014; Mourgkasi & Mavropoulou,
et al., 2012; Lavik, 2014; Levy, 2018; Lewan- 2018; Nordness et al., 2019; Ray et al., 2019;
dowski, 2011; Mason et al., 2010, 2013; Saddler, 2006; Shen & Troia, 2018; Werunga,
Mong Cramer & Mason, 2014; Rodgers, 2018). All studies except for one (Bauern-
2019; Stotz et al., 2008; Taylor, 2010; Viel- schmidt, 1991) reported treatment fidelity. A
Ruma et al., 2010; Wolfe et al., 2000). total of 16 studies (38%) did not report IOA:
154 Exceptional Children 88(2)

Two studies reported IOA for other writing second model accounted for approximately
fluency metrics (Dunn, 2012; Lavik, 2014), 14% of the AICc weight, and each remaining
and the remaining 14 studies used computer model accounted for approximately 5% or
software (e.g., Microsoft Word) to determine below of the remaining AICc weight. With a
TWW (Asaro-Saddler, 2014; Asaro-Saddler clear separation in AICc weight, there was a
& Bak, 2012; Asaro-Saddler & Saddler, 2010; high probability that the top model accounted
Dunn & Miller, 2016; Hashey, 2015; MacAr- for a sufficient amount of variance in the cri-
thur & Philippakos, 2010; Mason et al., 2010, terion variable (Burnham et al., 2011; Long,
2013; Mong Cramer & Mason, 2014; Nord- 2012). In addition, the evidence ratio of the
ness et al., 2019; Ray et al., 2019; Regan et al., top two models (i.e., dividing the AICc weight
2005; Shen & Troia, 2018; Southall, 2011). of the top model by the second model) indi-
cated the top model was approximately 5
times more likely than the second model
Data Analysis of TWWpM (Burnham & Anderson, 2002). The top model,
We calculated and compared a total of 35 dif- shown in Equation (1), has 13 parameter esti-
ferent regression models. The top model mates (see Table 1), an AICc value of
accounted for 74% of the corrected Akaike 14186.54, an AICc weight of 0.74, and a log
information criterion (AICc) weight. The likelihood of −7080.21.

Y
ips = γ 000 + ( γ100 ) ( Session ) + ( γ 200 ) ( Intervention Iteration ) + ( γ 300 ) ( Intervention )

+ ( γ 400 ) ( Post Intervention ) + ( γ 500 ) ( Extended Tasks ) + ( γ 600 ) ( Fluency Studies )


+ ( γ 700 ) ( Extended Tasks x Intervention Iteration ) (1)

+ ( γ 800 ) ( Fluency Studies x Intervention Iteration )


+ ( γ 900 ) ( Age ) + u00 s + r0 ps

The top-performing model includes sev- Extended Tasks × Intervention Iteration and
eral factors thought a priori to explain vari- Fluency Studies × Intervention Iteration) are
ance in TWWpM: age (i.e., above or below added to the model for studies that included
the mean age of 12 years old), time limit of these variables. For example, students tended
writing task (i.e., brief or extended), interven- to have higher rates of TWWpM on brief writ-
tion type (i.e., acquisition or fluency), and ing tasks than on extended writing tasks: Dur-
when postbaseline data were collected (i.e., ing baseline, students averaged 7.06 TWWpM
during intervention or postintervention). Not on brief writing tasks but only 3.49 TWWpM
present in the top model are some factors on extended writing tasks. That is, estimates
thought a priori to explain variance, such as show that the level of TWWpM on extended
several demographic variables (i.e., disability writing tasks had, on average, 3.57 fewer
status, gender, and race), independent vari- TWWpM than on brief writing tasks (i.e.,
ables (i.e., SRSD and non-SRSD), and depen- 7.06 – 3.57 = 3.49 TWWpM). The contribu-
dent variables (i.e., transcription modes of tion of individual parameters to the top model
handwriting and typing). ranges from small to large. Specifically, the
As shown in Table 1, the default parameter parameters of session and fluency studies
estimates are the TWWpM produced on brief have relatively small t values, large standard
writing tasks by students (i.e., average age of errors, and nonsignificant p values, meaning
12 years old) within acquisition studies. Three these estimates are likely to change given the
parameter estimates (i.e., age, extended tasks, inclusion of additional data from future stud-
and fluency studies) and two interactions (i.e., ies (Burnham & Anderson, 2002).
Datchuk et al. 155

Discussion (Ciullo & Reutebuch, 2013; Graham, 2006;


Graham et al., 2013; Rogers & Graham, 2008).
Writing interventions had a positive effect on A variety of acquisition interventions were
the level and trend of TWWpM across studies. used, with SRSD and graphic organizer instruc-
During baseline, students tended to show little tion being the most frequent. Both instructional
to no growth across time. Intervention pro- approaches (i.e., SRSD and graphic organizer
duced an increase in level of 0.97 TWWpM instruction) were primarily used to improve
and a gradual increase in trend of 0.16 aspects of writing other than TWW, such as
TWWpM each intervention session or approx- genre elements (e.g., essay parts) and holistic
imately 1.6 TWWpM every 10 sessions. In writing quality, meaning gains in TWW largely
addition, the postintervention level was 1.34 resulted as an auxiliary benefit from focusing
TWWpM. This is impressive change given on other aspects of writing.
that on average, students tend to improve by Similar to prior studies (e.g., Hier & Eckert,
0.17 TWW to 0.47 TWW per 3 min each 2016; Mason et al., 2013), we found positive
week, which is prorated to 0.06 TWWpM to effects for fluency interventions that featured
0.16 TWWpM each week (AIMSweb, 2016). timed practice paired with one or more compo-
nents of initial instruction (e.g., explicit
Writing interventions had a positive instruction or SRSD), performance feedback,
effect on the level and trend of goal setting, and graphing of performance. Of
TWWpM across studies. note, we categorized all interventions that pro-
vided explicit support for speed or automatic-
Following recommendations to categorize ity (i.e., timed practice) as fluency
writing interventions by intended learning out- interventions, even though some interventions
come (Codding et al., 2019), we found level featured a combination of initial instruction for
and trend effects differed by acquisition or flu- acquisition and timed practice for fluency
ency. Acquisition intervention resulted in a (e.g., Hoover et al., 2012). We did this to
slightly steeper trend: Acquisition interven- reflect the intended learning outcome of stud-
tions had an average gain of 0.16 TWWpM per ies (i.e., fluency and not just acquisition) and
session, and fluency interventions had an aver- to reflect limitations with the TWW metric:
age gain of 0.10 TWWpM per session (i.e., Because TWW does not account for accuracy
0.06 TWWpM fewer per session). This differ- (i.e., words without regard to spelling, syntax,
ential effect in trend may partially stem from or semantics), the distinction between when
differences in level. Cautious interpretation is initial instruction to improve accuracy stops
warranted because of its relatively minor con- and when fluency practice to improve speed
tribution to the top regression model (see the begins is blurred. In future studies, researchers
fluency studies parameter in Table 1); however, may want to investigate more nuanced distinc-
fluency studies had higher level of performance tions between intervention types (e.g., combi-
across experimental phases (i.e., baseline, flu- nation acquisition and fluency).
ency intervention, and post–fluency interven- When compared with a similar meta-analy-
tion), an average of 2.15 TWWpM more than sis that estimated effects of writing interven-
acquisition studies. With a higher level of per- tions on level and trend of CWS per minute
formance, it is possible students in fluency (CWSpM; Datchuk, Wagner, et al., 2020), our
studies were approaching their performance results suggest that effects may differ if writ-
ceiling and had little additional room for gains ing fluency is measured by TWW or CWS. In
in trend. Conversely, students in acquisition the present meta-analysis, intervention resulted
studies had lower levels of performance but in a larger change in TWW level (i.e., 0.97
more room to make incremental gains in trend. TWWpM) than trend (i.e., 0.16 TWWpM). In
We found acquisition interventions improved the prior meta-analysis (Datchuk, Wagner,
writing fluency, extending prior reviews that et al., 2020), the opposite was true for CWS
have typically reported effects on writing length level and trend: Intervention resulted in a
or count of TWW on untimed writing tasks smaller change in CWS level (i.e., –0.19
156 Exceptional Children 88(2)

CWSpM) than trend (i.e., 0.39 CWSpM). schools: brief writing tasks for screening and
This difference may partially stem from what progress monitoring (Romig et al., 2020) and
TWW and CWS measure. Namely, TWW extended writing tasks for instruction and
accounts for increases in any kind of written assessment in core instruction (Ray et al.,
production (e.g., misspelled or spelled cor- 2016). Compared with brief writing tasks
rectly), whereas CWS has more stringent cri- (i.e., see default parameter estimates in Table
teria: It accounts for multiple aspects of 1), we found students on extended writing
production, such as use of correct spelling and tasks had lower levels of performance by 3.57
syntax (Ritchey et al., 2016). As such, TWW TWWpM across experimental phases (i.e.,
may show more immediate change in level baseline, intervention, and postintervention)
than CWS; however, CWS may show greater and less steep trends during intervention by
gains in trend as more aspects of production 0.07 TWWpM per session. This finding cor-
(e.g., rules of syntax) are addressed by inter- responds to theories of writing development
vention. Caution must be exercised when and writing fluency (Berninger & Amtmann,
comparing results between meta-analyses. 2003; Ritchey et al., 2016). Specifically, on
Differences between TWW and CWS may brief writing tasks, students have time to
primarily reflect variance across students and engage in sustained transcription and text
studies: The present TWW meta-analysis and generation with little opportunity to stop and
the prior CWS meta-analysis (Datchuk, Wag- engage aspects of self-regulation. Extended
ner, et al., 2020) overlapped by only four pri- writing tasks by definition have longer time
mary studies (Bishop et al., 2015; Lavik, limits and provide more opportunity for stu-
2014; Lewandowski, 2011; Stotz et al., 2008). dents to engage in self-regulatory aspects of
Future writing intervention research is ulti- writing, such as pausing to consider plans and
mately needed to understand differential revisions, thereby lowering writing speed.
effects of writing interventions on writing flu-
ency metrics, including TWW and CWS.
Methodological Quality
Similar to prior meta-analytic findings, we
found writing fluency varied by age but did Studies typically had high methodological
not substantially vary by other student demo- quality with two important caveats. First, a
graphics of disability status, gender, and race sizable number of studies, 16 studies (38%),
(Datchuk, Wagner, et al., 2020). Specifically, did not report IOA, and many of these studies
older students (i.e., above the mean age of 12 used automated software to calculate TWW.
years old) tended to show higher levels of Although automated software likely increases
TWWpM than younger students. It is impor- replicability of procedures, it raises specific
tant to note that the student sample was fairly concerns for studies in which researchers
homogenous: The sample comprised mostly typed handwritten student responses for anal-
White males with EBD or LD (i.e., 47% of stu- ysis. Because scoring TWW depends on judg-
dents were White and 65% EBD or LD), and ments of space and legibility, it is important
this lack of diversity may have contributed to that studies not only report IOA but also detail
our findings. Given low prevalence of certain the process by which handwritten student
racial groups and disability types (e.g., His- responses are transferred to typed responses.
panic and autism), we collapsed categories of Second, most studies (i.e., 57% of total
race (i.e., White and non-White) and disability studies and 70% of acquisition studies) inter-
type by prevalence (i.e., high and low inci- rupted data collection during intervention and
dence); however, these did not prove to sub- continued data collection postintervention. For
stantially contribute to variance in TWWpM. those studies, we estimated postintervention
Results also extend research on measure- level; however, if data were collected during
ment of writing fluency. We included studies intervention, it is logical to assume that effect
that administered brief (i.e., ≤10 min) or estimates (i.e., level and trend) may decrease
extended (i.e., >10 min) writing tasks because given the gradual nature of skill acquisition. In
both types of writing tasks are used within addition, interrupting data collection creates a
Datchuk et al. 157

semiopaque picture of intervention effects dur- interventions include timed practice writing to
ing intervention and makes it difficult to esti- prompts (e.g., 10-min persuasive essays) paired
mate when changes in level and trend may with initial instruction (e.g., SRSD) and addi-
occur. This may delay data-based decision tional practice components of performance
making central to intervention delivery, such as feedback, goal setting, and graphing. Although
judging performance in relation to set criteria the primary goal of acquisition or fluency inter-
and altering intervention procedures as needed vention may be to improve other aspects of
(Riley-Tillman et al., 2020). In future studies, writing, such as the number of genre elements,
researchers should consider continuing data the rate of TWWpM may increase as a second-
collection during intervention to provide more ary benefit. As a result of intervention, students
precise estimates of intervention effects. may show an increase in level (i.e., older stu-
dents showing higher levels than younger stu-
dents) and a gradual increase in trend of
Limitations and Future Directions TWWpM. Students may show steeper upward
There are three limitations and future directions. trends from acquisition intervention than from
First, although no student demographics were fluency intervention. Changes in both level and
present in the top model, the student sample was trend may be more apparent on brief writing
largely homogenous, comprising mostly White tasks used for progress monitoring, such as
males with EBD or LD. Future intervention CBM-WE, and less apparent on longer writing
research should include more diverse samples of tasks in which students are expected to engage
students. Second, level and trend estimates were in planning and revising.
calculated only for parameters in the top model.
Some common aspects of intervention and mea- To improve the writing fluency . . .
surement, such as SRSD and transcription mode of students with and without
(i.e., handwritten or typed), were not part of the disabilities, practitioners should
top model and did not have accompanying effect
consider using several acquisition
estimates. In future meta-analyses, researchers
may want to consider use of other meta-analytic and fluency interventions.
techniques to estimate additional parameters of
interest. Third, we did not formally assess for References
influence of publication bias. We did include References marked with an asterisk indicate stud-
gray literature (i.e., unpublished doctoral disser- ies included in the meta-analysis.
tations), but publication bias is still likely given *Ahn, S. Y. (2017). Using a mobile-based graphic
the relatively small sample size. Future research organizer with embedded procedural facilita-
tion for idea generation to support persuasive
should investigate ways of assessing for publica-
essay writing for middle school students
tion bias because current techniques, such as with emotional and behavioral disorders
Egger’s test for funnel plot symmetry (Egger (Publication No. 10685724) [Doctoral disser-
et al., 1997), do not meet assumptions of multi- tation, George Mason University]. ProQuest
ple regression parameters used in the present Dissertations and Theses Global.
meta-analysis. AIMSweb®. (2016). AIMSweb national norms
table: Written expression—Total words writ-
ten. Pearson.
Implications for Practice *Asaro-Saddler, K. (2014). Self-regulated strategy
There are several implications for practice. To development: Effects on writers with autism
improve the writing fluency (i.e., rate of TWW spectrum disorders. Education and Training in
Autism and Developmental Disabilities, 49(1),
on timed writing tasks) of students with and
78–91.
without disabilities, practitioners should *Asaro-Saddler, K., & Bak, N. (2012). Teaching
consider using several acquisition and flu- children with high-functioning autism spectrum
ency interventions. Acquisition interventions disorders to write persuasive essays. Topics in
include SRSD and graphic organizer instruc- Language Disorders, 32(4), 361–378. https://
tion to teach essay or narrative writing. Fluency doi.org/10.1097/TLD.0b013e318271813f
158 Exceptional Children 88(2)

*Asaro-Saddler, K., & Saddler, B. (2010). Planning Ciullo, S., & Reutebuch, C. (2013). Computer-
instruction and self-regulation training: Effects based graphic organizers for students with LD:
on writers with autism spectrum disorder. A systematic review of literature. Learning
Exceptional Children, 77(1), 107–124. https:// Disabilities Research & Practice, 28(4), 196–
doi.org/10.1177/001440291007700105 210. https://doi.org/10.1111/ldrp.12017
Bates, D., Machler, M., Bolker, B. M., & Walker, Codding, R. S., Kromminga, K. R., & Running, K.
S. C. (2015). Fitting linear mixed-effects mod- (2019). Behavioral interventions for academic
els using lme4. Journal of Statistical Software, performance: A summary of the literature. In S.
67(1), 1–48. https://doi.org/10.18637/jss.v067. G. Little, & A. Akin-Little (Eds.), Behavioral
i01 interventions in the schools: Evidence-based
*Bauernschmidt, M. C. (1991). Repeated writing positive strategies (2nd ed., pp. 143–169).
to improve writing fluency for students with American Psychology Association.
mild handicaps (Publication No. 9210158) Cutler, L., & Graham, S. (2008). Primary grade
[Doctoral dissertation, University of Wisconsin– writing instruction: A national survey. Journal
Milwaukee]. ProQuest Dissertations and Theses of Educational Psychology, 100(4), 907–919.
Global. http://doi.org/10.1037/a0012656
Berninger, V. W., & Amtmann, D. (2003). Preventing Datchuk, S. M., & Kubina, R. M. (2013). A review
written expression disabilities through early of teaching sentence-level writing skills to
and continuing assessment and intervention for students with writing difficulties and learning
handwriting and/or spelling problems: Research disabilities. Remedial and Special Education,
into practice. In H. L. Swanson, K. R. Harris, 34(3), 180–192. https://doi.org/10.1177/07419
S. Graham, H. L. Swanson, K. R. Harris, & S. 32512448254
Graham (Eds.), Handbook of learning disabili- Datchuk, S. M., Rodgers, D. B., Wagner, K., Hier,
ties (pp. 345–363). Guilford Press. B. O., & Moore, C. T. (2020). Effects of writ-
*Bishop, A. E., Sawyer, M., Alber-Morgan, S. ing interventions on the level and trend of total
R., & Boggs, M. (2015). Effects of a graphic words written: A meta-analysis [Data set].
organizer training package on the persua- https://osf.io/e2fqb/?view_only=d0df68af8db
sive writing of middle school students with 7450badbc2907e3e2eb7c
autism. Education and Training in Autism and Datchuk, S. M., Wagner, K., & Hier, B. O. (2020).
Developmental Disabilities, 50(3), 290–302. Level and trend of writing sequences: A review
Burnham, K. P., & Anderson, D. R. (2002). Model and meta-analysis of writing interventions for
selection and multimodel inference: A practi- students with disabilities. Exceptional Children,
cal information-theoretic approach (2nd ed.). 86(2), 174–192. https://doi.org/10.1177/00144
Springer. 02919873311
Burnham, K. P., Anderson, D. R., & Huyvaert, Deno, S. L., Marston, D., & Mirkin, P. (1982).
K. P. (2011). AIC model selection and multi- Valid measurement procedures for continuous
model inference in behavioral ecology: Some evaluation of written expression. Exceptional
background, observations, and comparison. Children, 48(4), 368–371. https://doi.org/10
Behavioral Ecology and Sociobiology, 65, 23– .1177/001440298204800417
35. https://doi.org/10.1007/s00265-010-1029-6 Drevon, D., Fursa, S. R., & Malcolm, A. L.
*Caroll, M. L. (2018). The impact of using self-reg- (2017). Intercoder reliability and validity
ulated strategy development to increase expos- of WebPlotDigitizer in extracting graphed
itory writing outcomes in students at-risk for data. Behavior Modification, 41(2), 323–339.
emotional and behavioral disorders [Doctoral https://doi.org/10.1177/0145445516673998
dissertation, University of Texas–Austin]. *Dunn, M. (2012). Illustrating story plans: Does
University of Texas–Austin Scholar Works. a mnemonic strategy including art media ren-
https://hdl.handle.net/2152/68132 der more elaborate text? Journal of Visual
*Cerar, N. I. (2012). Teaching students with emo- Literacy, 31, 93–123. https://doi.org/10.1080/
tional and behavior disorder how to write 23796529.2012.11674696
persuasive essays fluently (Publication No. *Dunn, M. (2015). Students at-risk of having dys-
3506246) [Doctoral dissertation, George graphia: Applying assistive technology tools
Mason University]. ProQuest Dissertations and to help with pre writing. Educational Policies
Theses Global. and Current Practices, 1, 99–111.
Datchuk et al. 159

*Dunn, M., & Miller, D. (2016). Improving story *Gonzalez-Ledo, M., Barbetta, P. M., & Unzueta,
writing: Integrating the story mnemonic C. H. (2015). The effects of computer graphic
strategy with iPad apps for art and keyboard- organizers on the narrative writing of elementary
ing. International Journal for Research in school students with specific learning disabili-
Learning Disabilities, 3(1), 11–28. ties. Journal of Special Education Technology,
Egger, M., Smith, G. D., Schneider, M., & Minder, 30(1), 29–42. https://doi.org/10.1177/016264
C. (1997). Bias in meta-analysis detected by a 341503000103
simple, graphical test. British Medical Journal, *Good, K. E. (2019). The pen or the cursor: A
316, 629–634. https://doi.org/10.1136/bmj.315 single-subject comparison of a paper-based
.7109.629 graphic organizer and a computer-based
*Evmenova, A. S., Regan, K., Booykin, A., graphic organizer to support the persuasive
Good, K., Hughes, M., MacVittie, N., Sacco, writing of students with emotional and behav-
D., Ahn, S. Y., & Chirinos, D. (2016). ioral disorders or mild autism (Publication
Emphasizing planning for essay writing No. 13864282) [Doctoral dissertation, George
with a computer-based graphic organizer. Mason University]. ProQuest Dissertations
Exceptional Children, 82(2), 170–191. https:// and Theses Global.
doi.org/10.1177/0014402915591697 Graham, S. (2006). Strategy instruction and the
Fearrington, J. Y., Parker, P. D., Kidder-Ashley, teaching of writing: A meta-analysis. In C.
P., Gagnon, S. G., McCane-Bowling, S., & MacArthur, S. Graham, & J. Fitzgerald (Eds.),
Sorrell, C. A. (2014). Gender differences in Handbook of writing research (pp. 187–207).
written expression curriculum-based mea- Guilford.
surement in third- and through eighth-grade Graham, S. (2018). A revised writer(s)-within-
students. Psychology in the Schools, 51(1), community model of writing. Educational
85–96. https://doi.org/10.1002/pits.21733 Psychologist, 53(4), 258–279. https://doi.org/
Furey, W. M., Marcotte, A. M., Hintze, J. M., & 10.1080/00461520.2018.1481406
Shackett, C. M. (2016). Concurrent validity Graham, S., Harris, K. R., & McKeown, D. (2013).
and classification accuracy of curriculum- The writing of students with learning disabili-
based measurement for written expression. ties, meta-analysis of self-regulated strategy
School Psychology Quarterly, 31(3), 369–382. development writing intervention studies, and
https://doi.org/10.1037/spq0000138 future directions: Redux. In L. Swanson, K.
Gansle, K. A., Noell, G. H., Vanderheyden, A. M., R. Harris, & S. Graham (Eds.), Handbook of
Slider, N. J., Hoffpauir, L. D., Whitmarsh, learning disabilities (2nd ed., pp. 405–438).
E. L., & Naquin, G. M. (2004). An examina- Guilford.
tion of the criterion validity and sensitivity *Hashey, A. I. (2015). A technology-enhanced
to brief intervention of alternate curriculum- approach to self-regulated strategy develop-
based measures of writing skill. Psychology ment: Engaging adolescents with emotional and
in the Schools, 41(3), 291–300. https://doi. behavioral disorders in argumentative writing
org/10.1002/pits.10166 (Publication No. 3714605) [Doctoral disserta-
*Garwood, J. D., Werts, M. G., Mason, L. H., tion, University at Buffalo–State University of
Harris, B., Austin, M. B., Ciullo, S., Magner, New York). ProQuest Dissertations and Theses
K., Koppenhaver, D. A., & Shin, M. (2019). Global.
Improving persuasive science writing for sec- Hier, B. O., & Eckert, T. L. (2016). Programming
ondary students with emotional and behavioral generality into a performance feedback writing
disorders education in residential treatment facil- intervention: A randomized controlled trial.
ities. Behavioral Disorders, 44(4), 227–240. Journal of School Psychology, 56, 111–131.
https://doi.org/10.1177/0198742918809341 https://doi.org/10.1016/j.jsp.2016.03.003
*Geisler, J. L., Hessler, T., Gardner, R., III, & *Hoover, T. M., Kubina, R. M., & Mason, L.
Lovelace, T. S. (2009). Differentiated writing H. (2012). Effects of self-regulated strat-
interventions for high-achieving urban African egy development for POW + TREE on high
American elementary students. Journal of school students with learning disabilities.
Advanced Academics, 20(2), 214–247. https:// Exceptionality, 20(1), 20–38. https://doi.org/1
doi.org/10.1177/1932202X0902000202 0.1080/09362835.2012.640903
160 Exceptional Children 88(2)

*Hough, T. M., Hixson, M. D., Decker, D., & A. (2020). Intervening with multiplication
Bradley-Johnson, S. (2012). The effec- fact difficulties: Examining the utility of the
tiveness of an explicit instruction writ- instructional hierarchy to target interventions.
ing program for second graders. Journal of Journal of Behavioral Education. Advance
Behavioral Education, 21, 163–174. https:// online publication. https://doi.org/10.1007/
doi.org/10.1007/s10864-012-9146-0 s10864-020-09388-0
Keller-Margulis, M. A., Mercer, S. H., Payan, *Mason, L. H., Kubina, R. M., & Hoover, T.
A., & McGee, W. (2015). Measuring annual (2013). Effects of quick writing instruction
growth using written expression curricu- for high school students with emotional distur-
lum-based measurement: An examination bances. Journal of Emotional and Behavioral
of seasonal and gender differences. School Disorders, 21(3), 163–175. https://doi.
Psychology Quarterly, 30(2), 276–288. https:// org/10.1177/1063426611410429
doi.org/10.1037/spq0000086 *Mason, L. H., Kubina, R. M., Valasa, L. L., &
Kim, Y. S. G., Gatlin, B., Al Otaiba, S., & Wanzek, Mong Cramer, A. (2010). Evaluating effec-
J. (2018). Theorization and an empirical inves- tive writing instruction for adolescent students
tigation of the component-based and develop- in an emotional and behavior support setting.
mental text writing fluency construct. Journal Behavioral Disorders, 35(2), 140–156. https://
of Learning Disabilities, 51(4), 320–335. doi.org/10.1177/019874291003500205
https://doi.org/10.1177/0022219417712016 Mazerolle, M. J. (2017). AICcmodavg (Version
Koenig, E. A., Eckert, T. L., & Hier, B. O. (2016). 2.1-1) [Computer software]. https://cran.r-
Using performance feedback and goal setting project.org/web/packages/AICcmodavg/index.
to improve elementary students’ writing flu- html
ency: A randomized controlled trial. School McFarland, J., Hussar, B., Zhang, J., Wang, X.,
Psychology Review, 45(3), 275–295. https:// Wang, K., Hein, S., Diliberti, M., Forrest
doi.org/10.17105/SPR45-3.275-295 Cataldi, E., Bullock Mann, F., & Barmer,
*Lavik, K. B. (2014). The effectiveness of a goal A. (2019). The condition of education 2019.
setting intervention that incorporates perfor- National Center for Education Statistics.
mance feedback, self-graphing, and reinforce- https://nces.ed.gov/pubsearch/pubsinfo.
ment on improving the writing skills of high asp?pubid=2019144
school students (Publication No. 3671433) *McKeown, D., Kimball, K., & Ledford, J. (2015).
[Doctoral dissertation, Kent State University]. Effects of asynchronous audio feedback on
ProQuest Dissertations and Theses Global. the story revision practices of students with
*Levy, R. A. (2018). Effectiveness of a modi- emotional/behavioral disorders. Education
fied approach to small-group SRSD for sec- and Treatment of Children, 38(4), 541–564.
ond grade students with writing difficulties https://doi.org/10.1353/etc.2015.0020
(Publication No. 10969829) [Doctoral dis- McMaster, K. L., Lembke, E. S., Shin, J., Poch,
sertation, North Carolina State University]. A. L., Smith, R. A., Jung, P.-G., Allen, A.
ProQuest Dissertations and Theses Global. A., & Wagner, K. (2020). Supporting teach-
*Lewandowski, S. C. (2011). The effects of a modi- ers’ use of data-based instruction to improve
fied version of self-regulated startegy develop- students’ early writing skills. Journal of
ment and self-talk internalization strategy on Educational Psychology, 112(1), 1–21. https://
writing and self-talk for elementary students doi.org/10.1037/edu0000358
with attention difficulties (Publication No. *Miller, K. M., & Little, M. E. (2018). Examining
3468986) [Doctoral dissertation, Michigan the effects of SRSD in combination with
State University]. Proquest Dissertations and video self-modeling on writing by third
Theses Global. grade students with learning disabilities.
Long, J. D. (2012). Longitudinal data analysis for Exceptionality, 26(2), 81–105. https://doi.org/
the behavioral sciences using R. Sage. 10.1080/09362835.2017.1283622
*MacArthur, C. A., & Philippakos, Z. (2010). *Mong Cramer, A., & Mason, L. H. (2014). The
Instruction in a strategy for compare-contrast effects of strategy instruction for writing and
writing. Exceptional Children, 76(4), 438–456. revising persuasive quick writes for middle
https://doi.org/10.1177/001440291007600404 school students with emotional and behavioral
Maki, K. E., Zaslofksy, A. F., Knight, S., disorders. Behavioral Disorders, 40(1), 37–51.
Ebbesmeyer, A. M., & Chelmo-Boatman, https://doi.org/10.17988/0198-7429-40.1.37
Datchuk et al. 161

*Mourgkasi, V., & Mavropoulou, S. (2018). Story & Ortiz, M. (2016). Indicators of fluent writ-
composition, mental state language and self- ing in beginning writers. In. K. D. Cummings,
regulation strategy instruction for writers & Y. Petscher (Eds.), The fluency construct
with autism spectrum conditions. Journal of (pp. 21–66). Springer.
Research in Special Educational Needs, 18(1), *Rodgers, D. B. (2019). A paragraph text-writ-
36–49. ing intervention for secondary students with
*Nordness, P. D., Hagaman, J. L., Herskovitz, R., intellectual and developmental disabilities:
& Leader-Janssen, E. (2019). POWER UP: A A single-case design study (Publication No.
persuasive writing strategy for secondary stu- 13811948) [Doctoral dissertation, University
dents with emotional and behavioral disorders. of Iowa]. ProQuest Dissertations and Theses
Journal of Education and Learning, 8, 32–42. Global.
https://doi.org/10.5539/jel.v8n4p32 Rohatgi, A. (2018). WebPlotDigitzer (Version 4.1)
Parker, D. C., & Burns, M. K. (2014). Using the [Computer software]. https://arohatgi.info/
instructional level as a criterion to target WebPlotDigitzer.
reading interventions. Reading & Writing Rogers, L. A., & Graham, S. (2008). A meta-analy-
Quarterly, 30(1), 79–94. https://doi.org/10.10 sis of single subject design writing intervention
80/10573569.2012.702047 research. Journal of Educational Psychology,
Parker, D. C., McMaster, K. L., & Burns, M. K. 100(4), 879–906. https://doi.org/10.1037/0022
(2011). Determining an instructional level -0663.100.4.879
for early writing skills. School Psychology Romig, J. E., Miller, A. A., Therrien, W. J., &
Review, 40(1), 158–167. https://doi.org/10.10 Lloyd, J. W. (2020). Meta-analysis of prompt
80/02796015.2011.12087735 and duration for curriculum-based measure-
R Core Team. (2020). R: A language and envi- ment of written language. Exceptionality.
ronment for statistical computing (Version Advance online publication. https://doi.org/10
4.0) [Computer software]. R Foundation for .1080/09362835.2020.1743706
Statistical Computing. https://www.R-project. Romig, J. E., Therrien, W. J., & Lloyd, J. W.
org (2017). Meta-analysis of criterion valid-
Raudenbush, S. W., & Bryk, A. S. (2002). ity for curriculum-based measurement in
Hierarchical linear models: Applications and written language. The Journal of Special
data analysis methods (2nd ed.). Sage Education, 51(2), 72–82. https://doi.org/10.11
Ray, A. B., Graham, S., Houston, J. D., & Harris, 77%2F0022466916670637
R. (2016). Teachers’ use of writing to sup- *Saddler, B. (2006). Increasing story-writing ability
port students’ learning in middle school: A through self-regulated strategy development:
national survey in the United States. Reading Effects on young writers with learning dis-
and Writing, 29, 1039–1068. https://doi. abilities. Learning Disability Quarterly, 29(4),
org/10.1007/s11145-015-9602-z 291–305. https://doi.org/10.2307/30035555
*Ray, A. B., Graham, S., & Liu, X. (2019). Effects Shapiro, E. S. (2011). Academic skill problems:
of SRSD college entrance exam instruction for Direct assessment and intervention (4th ed.).
high school students with disabilities or at-risk Guilford.
for writing difficulties. Reading and Writing, *Shen, M., & Troia, G. A. (2018). Teaching chil-
32, 1507–1529. https://doi.org/10.1007/s111 dren with language-learning disabilities to plan
45-018-9900-3 and revise compare-contrast texts. Learning
*Regan, K. S., Mastropieri, M. A., & Scruggs, Disability Quarterly, 41(1), 44–61. https://doi.
T. E. (2005). Promoting expressive writing org/10.1177/0731948717701260
among students with emotional and behav- *Southall, C. (2011). Self-regulated strategy devel-
ioral disturbance via dialogue journals. opment writing instruction for adolescents
Behavioral Disorders, 31, 33–50. https://doi. with autism spectrum disorder [Doctoral dis-
org/10.1177/019874290503100107 sertation, University of Georgia]. University
Riley-Tillman, T. C., Burns, M. K., & Kilgus, S. P. of Georgia-Athenaeum. http://hdl.handle.net/
(2020). Evaluating educational interventions: 10724/27602
Single-case design for measuring response to *Stotz, K. E., Itoi, M., Konrad, M., & Alber-
intervention (2nd ed.). Guilford. Morgan, S. R. (2008). Effects of self-graphing
Ritchey, K. D., McMaster, K. L., Al Otaiba, S., on written expression of fourth grade students
Puranik, C. S., Grace Kim, Y. S., Parker, D. C., with high-incidence disabilities. Journal of
162 Exceptional Children 88(2)

Behavioral Education, 17, 172–186. http://doi. Education Evaluation and Regional Assistance.
org/10.1007/s10864-007-9055-9 https://ies.ed.gov/ncee/wwc/Handbooks
*Taylor, C. M. (2010). The effects of repeated Wolery, M., Busick, M., Reichow, B., &
writing on secondary students’ writing flu- Barton, E. E. (2010). Comparison of over-
ency (Publication No. 3425372) [Doctoral lap methods for quantitatively synthesiz-
dissertation, Ohio State University]. ProQuest ing single-subject data. The Journal of
Dissertations and Theses Global. Special Education, 44(1), 18–28. http://doi.
*Unzueta, C. H., & Barbetta, P. M. (2012). The org/10.1177/0022466908328009
effects of computer organizers on the per- *Wolfe, L. H., Heron, T. E., & Goddard, Y. L.
suasive writing of Hispanic middle school (2000). Effects of self-monitoring on the on-
students with specific learning disabilities. task behavior and written language perfor-
Journal of Special Education Technology, mance of elementary students with learning
27(3), 15–30. http://doi.org/10.1177/016264 disabilities. Journal of Behavioral Education,
341202700302 10(1), 49–73. https://doi.org/10.1023/A:10166
U.S. Department of Education Institute of 95806663
Education Sciences, National Center for
Educational Statistics. (2011). NAEP data Authors’ Note
explorer [Data set and software]. http://nation-
Study data and materials are shared online
sreportcard.gov/ndecore/xplore/NDE
(Datchuk, Rodgers, et al., 2020).
*Viel-Ruma, K., Houchins, D. E., Jolivette, K.,
Fredrick, L. D., & Gama, R. (2010). Direct
instruction in written expression: The effects Open Science Practices
on English speakers and English language
learners with disabilities. Learning Disabilities
Research & Practice, 25(2), 97–108. http:// For publishing data and materials, this article
doi.org/10.1111/j.1540-5826.2010.00307.x earned badges for Open Data and Open Materials.
*Werunga, R. N. (2018). Effects of self-regulated The public content for the data, code, and materials
strategy development on the writing skills and may be retrieved from https://osf.io/e2fqb/?view_
problem behaviors of students with emotional and only=d0df68af8db7450badbc2907e3e2eb7c.
behavioral disorders (Publication No. 10846952)
[Doctoral dissertation, University of North
Carolina–Charlotte]. ProQuest Dissertations and
Supplemental Material
Theses Global. Supplemental material for this article is available
What Works Clearinghouse. (2020). What Works online.
Clearinghouse standards handbook: Version
4.1. U.S. Department of Education, Institute Manuscript received November 2020; accepted
of Education Sciences, National Center for June 2021.
Copyright of Exceptional Children is the property of Sage Publications Inc. 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.

You might also like