University of Southern California Dissertations and Theses

The effects of a reading fluency intervention on the reading outcomes of middle school students with severe reading disabilities

(USC Thesis Other)

The effects of a reading fluency intervention on the reading outcomes of middle school students with severe reading disabilities

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

THE EFFECTS OF A FLUENCY INTERVENTION PROGRAM ON THE
READING OUTCOMES OF MIDDLE SCHOOL STUDENTS
WITH SEVERE READING DISABILITIES

by

Sally Atwood Spencer

A Dissertation Presented to the
FACULTY OF THE ROSSIER SCHOOL OF EDUCATION
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF EDUCATION

December 2008

Copyright 2008 Sally Atwood Spencer

ii

DEDICATION

When you spend four years of your life working on a new venture, you end up
with a lot of people to thank—people who have supported you, encouraged you, and
shared the ride. The ones mentioned here were the most prominent, but the rest of you
know who you are. If I listed everyone it would take me another 160 pages. And nobody
wants that…
First, of course, I must thank my wonderful dissertation committee. My co-chair,
Dr. Frank Manis, guided me through all the ups and downs of running a complicated,
school-based study. His expertise and positive attitude made a difficult undertaking feel
like an adventure. Frank, I learned so much from working with you, and hope that my
research career someday lives up to the great example you set. My other co-chair, Dr.
Dennis Hocevar, was a source of ongoing encouragement and statistical expertise, and
just an overall great person to work with. Thank you, Dennis, for your time, your positive
attitude and your excellent know-how.
The third member of my committee, Dr. Sue Sears, has been a mentor and a
friend for many years, as well as a role-model and guide through my entire teaching
career. Sue, I look to you as my example of what it means to be a special education
professor. You always hold high standards for yourself, and are never satisfied with the
easy solution; you continually strive to be a global thinker, to be compassionate, and to
always keep the best interests of the students in mind. You are a special colleague and
friend, and I can’t thank you enough for all your time and guidance.

iii

I also want to thank my family, who were really proud and excited for me,
even after four years of “I’m too busy to get together”. Very, very special thanks go to
my sister Ginny, who spent two entire weeks helping me score, organize and input my
data into the computer. I never would have been done on time without your help, Zee—
you are an excellent big sister. You truly earned your Assistant-Doctorate degree.
The members of Dr. Manis’ lab played a huge role in this process; this large study
would have been impossible without their help. Very special thanks go to Alison
Zumberge, Dr. Jennifer Bruno, Jason Batten, Rachel Beattie and Jason Goldman for
training testers, testing tons of kids, overseeing the scoring, and generally always being
there when I needed someone to bounce ideas off of. They all are so talented, and I feel
so lucky to have worked with them.
One amazing friend, Dr. Rachel Friedman-Narr, has contributed more to my
doctoral process than almost anyone. She was never too busy or too tired to listen to me
ramble on about my research, share ideas with me, read my drafts, and generally be my
sounding board through four long years. Rach, I don’t know what I would have done
without your input and enthusiasm. You won the award for friend-of-the-decade before I
was even halfway through my program, and I will never be able to repay you for the
listening, the feedback, the inspiration, the encouragement, and… well… the true
friendship you’ve given me.
Finally, I need to express my love and gratitude to my amazing, miraculous
husband Layne. How did I get lucky enough to find someone who never stops believing
in me, who spurs me on to try harder and do better, and who loves me through the

iv

process? Where do you find a partner who will sit patiently while you talk about
outliers and statistical analysis and not only listen, but understand? This man is brilliant
and funny and sweet, and he has suffered through four years of bad moods and
exhaustion with barely a complaint. Layne, you are the love of my life, and I adore you.
Thank you for helping me through this. I owe it all to you.

v

TABLE OF CONTENTS
Dedication ii
List of Tables viii
Abstract ix
Chapter 1: Introduction 1
Overview of the Problem 1
The Importance of Reading Fluency 3
The Efficacy of Fluency Intervention 5
Limitations of the Current Research Base 6
Purpose of the Current Study 8
Study Design 10
Limitations and Delimitations of the Study 11
Definition of Terms 12

Chapter 2: Review of the Literature 15
Models of Reading and Reading Fluency 15
Defining Fluency 30
Correlating Factors to Fluent Reading 33
Fluency and Comprehension 40
The 2001 Great Leaps Study 67
Summary and Hypotheses 71

Chapter 3: Methods 73
Settings 73
Participants 74
Assessments 81
Study Procedures 93
Data Analysis 96

Chapter 4: Results 98
Summary of the Research Methods 98
Evaluation of Data Characteristics 100
Examination of Influences on Study Implementation 105
Reading Outcome Data 107
The Relationship Between Student Characteristics and Student Outcomes 115
Summary 119

vi

Chapter 5: Discussion
The Intervention’s Impact on Reading Fluency 122
The Intervention’s Impact on Reading Comprehension 125
The Relationship Between Phonemic Awareness Skill and Reading Growth 128
The Relationship Between RAN Skill and Reading Growth 130
Implications for Practice 130
Limitations of the Study 134
Future Directions 135

References 139

Appendices
Appendix A: Sample of Great Leaps Phonics/Words Page 149
Appendix B: Sample of Great Leaps Phrases Page 150
Appendix C: Sample of Great Leaps Story 151
Appendix D: Great Leaps Equal Ratio Chart 152
Appendix E: Example of Skills for School Success Workbook Page 153
Appendix F: Skills for School Success Record Page 154
Appendix G: Comparisons of Pre and Posttest Mean Reading Scores Within
Groups, GORT Measures, Outliers Included 155
Appendix H: Comparisons of Pre and Posttest Mean Grade-Level
Equivalencies Within Groups, GORT, Outliers Included 156
Appendix I: Comparison Of GORT Mean Residual Gain Scores by Group,
Outliers Included 157
Appendix J: Comparisons of Pre and Posttest Mean Scores For GORT
Comprehension Within Groups, Outliers Included 158
Appendix K: Comparison Of GORT Mean Comprehension Scores by
Group, Outlier Included 159
Appendix L: Open-Ended Comprehension Questions for GORT Assessment,
Form A 160

vii

LIST OF TABLES
Table 1: Demographic Information by Group 81
Table 2: Assessment Batteries in the Order Administered at the Pretest 86

Table 3: Comparison of Pretest Means for Experimental and Control Groups 104
Table 4: Comparisons of Pre And Posttest Mean Reading Standard Scores Within
Groups, TOWRE and Woodcock Measure 108

Table 5: Between-Group Comparisons Of Mean Residual Gain Scores 110

Table 6: Comparisons of Pre And Posttest Mean Reading Standard Scores Within
Groups, GORT 111

Table 7: Comparisons of Pre and Posttest Mean Grade-Level Equivalencies Within
Groups, GORT 111

Table 8: Comparison Of GORT Mean Residual Gain Scores by Group 112

Table 9: Comparisons of Pre and Posttest Mean Scores For GORT Comprehension
Within Groups 114

Table 10: Comparison Of GORT Mean Comprehension RGS Scores by Group 114

Table 11: Correlations between reading gains and underlying related factors 117

viii

ABSTRACT
Despite recent advances in the science of teaching reading, there still exists a
small percentage of students who fail to make the expected progress in reading-related
skills. As they get older, these students are at great risk for dropout, behavior problems,
and learned helplessness. Even if these struggling readers learn to decode adequately,
fluency still remains a problem for many, and little is known about the effectiveness of
fluency interventions for older students with severe reading delays. This study used a
randomized experimental design to test the efficacy of a fluency intervention program on
the reading outcomes of 60 students with severe reading disabilities in grades six through
eight. Students in the experimental group received ten minutes a day of one-on-one
fluency intervention with a trained paraprofessional, utilizing the Great Leaps Reading
Program (Campbell, 1999); students in the control group participated in ten minutes a day
of study skills instruction, also from trained paraprofessionals. Students were assessed
pre and post intervention on a variety of fluency and comprehension measures, as well as
measures of reading-related factors such as phonemic awareness and Rapid Automatic
Naming (RAN). Results showed that the implementation of one-on-one programs such as
this in typical urban middle schools may be problematic, due to a variety of conflicting
obligations for students and paraprofessionals. Nevertheless, students in the experimental
group made significant progress in fluency as compared to students in the control group
in reading fluency. No significant gains were seen in reading comprehension. Students
with higher phonemic awareness skills seemed to make the most progress, but no
relationship was found between fluency gains and RAN.

1

CHAPTER I: INTRODUCTION

Overview of the Problem
The body of research on reading instruction is vast. For decades, researchers have
been trying to discover exactly what elements make up the most effective reading
instruction. During the past 15 years, reading researchers have succeeded in isolating and
identifying several components that seem to be common to effective reading programs,
and a consensus now exists in most quarters about what constitutes an effective reading
program for students in early grades: instruction in phonemic awareness, phonemic word
study, fluency, vocabulary and comprehension. In addition, it is believed that the
instruction must be systematic, individualized, and explicit, using a variety of language-
rich materials and ongoing assessment (California Reading Task Force, 1995; National
Reading Panel, 2000).
However, even with the highest quality reading instruction, there are inevitably
some students who do not achieve grade-level expectations. These students, despite
intensive intervention based on best-practice in reading instruction, do not make adequate
progress in reading, and are often the ones who end up identified as having reading
disabilities. Many subsequently receive special education services due to delayed reading
acquisition, and these students may continue to work on early literacy skills well into
middle and high school. Various researchers have given these students labels, such as
“treatment resistors” (Torgesen, 1999), or “nonresponders” (Fuchs, Fuchs, McMaster,
Yen, & Svenson, 2004); however, no matter what they are called, the answer to how best

2

to help these low-achieving students make progress towards grade-level reading skills
remains elusive.
For secondary students with reading disabilities, reading problems may have
serious consequences. Many secondary students with learning disabilities have higher
rates of absenteeism and dropout than typical students, and fewer than a quarter of them
go on to pursue higher education (Blackorby & Wagner, 1996). Students with disabilities
frequently fall further and further behind their typically developing peers, until by the
time they reach middle and high school they are operating at a significant deficit (Deshler
et al., 2001).
Nationally, approximately 65% of eighth graders with learning disabilities are
reading below the 20
th
percentile (National Center for Educational Statistics, 2007), and
only 7% are rated as proficient. In urban districts these numbers tend to be even more
extreme: in 2006 in Los Angeles County, 8% of middle school students identified with
disabilities were at or above the proficient level in reading, and in the Los Angeles
Unified School District (LAUSD) this figure dropped to 4% (California Department of
Education, n.d.). In New York City, results of state testing for 2006/2007 show only 4%
of the eighth grade students in special education read at or above the proficient level
(New York City Department of Education, n.d.) Clearly, the instructional methods best
suited for teaching reading to students with pervasive reading disabilities are not yet well-
identified or well-established in our nation’s urban schools.

3

The Importance of Reading Fluency
In recent years, some theorists have speculated that the ability to read at a
reasonable and efficient rate may be among the most significant characteristics of
proficient readers (Adams, 1990), yet it is very common for students with the most
pervasive reading difficulties to have trouble achieving automaticity in word and passage
reading (Chard, Vaughn & Tyler, 2002; Lyon & Moats, 1997; Nikolopoulour,
Goulandris, & Snowling, 2003). For secondary students in particular, fluency is identified
as one of the critical variables to successful reading (Archer, Gleason & Vachon, 2003).
As a result, the attention of the reading community has turned to fluency as one potential
key to remediation for struggling secondary readers.
Yet the acquisition of reading fluency remains problematic for many secondary
students with reading disabilities. Although researchers over the past ten years have
identified that intensive instruction in phonemic awareness and phonemic decoding will
increase the word identification skills of most impaired readers (Foorman et al., 1997;
Torgesen, Wagner & Rashotte, 1997; Vellutino, Fletcher, Snowling & Scanlon, 2004),
these same interventions have not proven effective in improving students’ fluency (Lyon
& Moats, 1997; Torgesen & Hudson, 2006). Significantly, some researchers have found a
close relationship between reading fluency and reading comprehension, and conjectured
that students who have trouble learning to read fluently may also ultimately have trouble
comprehending text. Meyer and Felton (1999), in a review of fluency research and
practice, put it even more definitively, stating, “A major reason for focusing on the
development of fluent reading is the theoretical relationship between fluency and

4

comprehension… Given the escalating demands for reading skills in our technological
society, it is critical that researchers and practitioners focus on fluency as an important
component of reading instruction” (p. 284).
The theoretical foundations for fluency instruction are, indeed, strong. Proponents
of the information processing theories of reading have suggested a link between a
student’s ability to read fluently and the amount of cognitive resources available for
comprehension (LaBerge & Samuels, 1974; Perfetti, 1985). These theorists advocate that
readers have a limited amount of cognitive attention that must be allocated among all the
conscious processes taking place in the brain. They propose that dysfluent readers have to
allocate an inordinate amount of their attention to decoding, and so have few resources
left for comprehension of text.
Other reading theorists, including connectivists such as Adams (1990) and
Stanovich (1980), hypothesize that reading is facilitated by an interconnected series of
processes that ultimately support a student’s ability to comprehend text. The models
proposed by these researchers involve simultaneous interactions among a variety of
reading processes, any of which may compensate for weaknesses in another. Both Adams
and Stanovich discuss potential limitations caused by dysfluent decoding, and suggest
that lack of fluency is likely to impair comprehension. It seems, then, that despite slightly
different paradigms, many respected reading theorists believe that fluent, automatic
decoding of words can lead the reader to be able to comprehend text more easily (Adams,
1990; LaBerge & Samuels, 1984; Perfetti, 1985; Stanovich, 1980).

5

The Efficacy of Fluency Intervention
Despite the belief by many in the theoretical community that fluency is a critical
component in the reading process, there is currently no consensus about the effectiveness
of fluency intervention on students’ reading comprehension. While some recent meta-
analyses of fluency intervention studies have found a positive relationship between
fluency and comprehension (National Reading Panel, 2000; Therrien, 2004), the
researchers are quick to point out the dearth of experimental studies to substantiate these
contentions. Other studies have tried to examine that link explicitly and have found little
or no evidence supporting the idea that increasing fluency will lead to stronger reading
comprehension (Fleisher, Jenkins & Pany, 1979; Irausquin, Drent, & Verhoeven, 2005;
Martin-Chang & Levy, 2005; Therrien, Wickstrom, & Jones, 2006;). It is clear that the
field is still in need of studies that investigate the link between fluency and
comprehension explicitly through the use of experimental research design.
Nevertheless, some recent studies point to a strong relationship between reading
fluency and middle school students’ outcomes on standardized reading assessments.
Torgesen (2005) correlated the results of tests of fluency, verbal ability, non-verbal
ability and memory with seventh graders’ results on the Florida Comprehensive
Assessment Test (FCAT) and found that 43% of the variability on the test could be
correlated with the students’ reading fluency. The only stronger indicator of success was
verbal ability, at 51% (Torgesen, 2005).
In addition, Torgesen (2005) found that students who performed at the lowest
levels on the FCAT were significantly more impaired in reading fluency than in decoding

6

and verbal knowledge. He identified the plight of students who were poor readers as
they moved up through middle school and high school, stating that as students get older
the text gets more complex, necessitating higher levels of verbal knowledge and
reasoning skills. However, students who do not read fluently read less efficiently and less
often, and as a result are unlikely to maintain the levels of cognitive growth and increased
verbal knowledge that they need to succeed in more complex text. In these cases, the
influence of poor fluency becomes more and more profound as students get older
(Torgesen, 2005).
It may be stated, then, that studies have confirmed the importance of reading
fluency on secondary students’ ability to successfully compete in grade level curriculum
and testing. It can further be stated, however, that the link between fluency practice and
increased reading comprehension is still tenuous. Since the ability to effectively
comprehend text must be the goal of any reading program, it should be noted that the
current state of the research leaves many questions that need to be investigated.
Limitations of the Current Research Base
Despite several meta-analyses on the efficacy of fluency intervention as well as
several studies that investigate the link between fluency training and reading
comprehension, many questions remain. When the National Reading Panel (NRP, 2000)
did its exhaustive search of the research base on fluency, they found only 16 fluency
studies conducted since 1970 that had pre- and post-test measures of reading ability and
an experimental design, and only some of these measured the effects on reading
comprehension. Additionally, although many studies exist that examine the outcomes of

7

fluency interventions in terms of reading efficiency, very few used an experimental
design that can imply causation; clearly, more experimental studies are needed that can
distinguish whether fluency intervention has a positive impact on both reading decoding
and reading comprehension. Recent meta-analyses of fluency interventions made the
following recommendations for future studies: studies should include a relatively long
implementation period, with the intervention administered by adults rather than peers.
They should provide regular corrective feedback to students and include predetermined
performance criteria, adjustment of the reading levels of the materials according to
student progress, and the use of a control group (Chard, Vaughn & Tyler, 2002; NRP,
2000; Therrien, 2004).
There is also a need for more research investigating the factors that contribute to a
student’s ability to develop reading fluency. Currently, research in this area has mostly
focused on two factors: students’ aptitude in phonemic awareness, and students’ overall
processing speed. Research into the former has been conducted over several decades, and
there is now a high degree of consensus that phonemic awareness ability is highly
correlated to reading ability and fluency (see for example Bradley & Bryant, 1978;
Manis, Custodio & Szeszulski, 1993; Wagner et al., 1993).
As a result of this consensus, in recent years researchers have begun to focus
more of their investigation into the influence of processing speed on students’ reading
aptitude. Processing speed is often measured by a student’s ability to quickly recall letters
or numbers, and is known as rapid automatic naming, or RAN. Many researchers (for
example, Manis, Doi & Bhadha, 2000; Manis, Seidenberg & Doi, 1999) have stated that

8

a student’s ability to recall these stored codes is highly related to his or her ability to
read fluently. Many of these studies have found that RAN is uniquely and strongly
related to reading ability, and some research concludes that RAN ability may have a
particular influence on the outcomes of older struggling readers (Allor, 2002). However,
it is still unclear the extent to which low RAN correlates with impaired fluency
development, and the influence RAN ability may have on a student’s ability to benefit
from fluency intervention techniques.
Recently, another line of research has investigated the differential effects of RAN
and phonemic awareness skills on the outcomes of struggling readers (Wolf & Bowers,
1999). These researchers propose a double-deficit model, suggesting that there are three
distinct groups of poor readers: those with poor phonemic awareness skills, those with
poor RAN skills, and those with deficits in both areas. Wolf and Bowers label these
students as “double-deficit” readers. The existence of these three distinctive groups of
disabled readers, as well the discrete influence of each of these factors, is still under
investigation, and remains in question.
Purpose of the Current Study
The purpose of this study is to investigate the effect of an explicit fluency
intervention program on the reading skills of students in two Los Angeles middle schools.
The students in this study are all identified with mild to moderate disabilities, and were
chosen from the bottom five percent of students in their respective schools. All were
reading well below grade level before participating in the intervention, despite receiving
special education services.

9

This study investigates the effectiveness of an individually administered
fluency intervention program on the reading skills of these students, examining the
outcomes in terms of decoding ability and fluency rates, as well as potential changes in
reading comprehension. In addition, this study seeks to examine the relationship between
the students’ phonemic awareness skill and naming ability on their reading outcomes, and
to determine the practicality of this type of one-on-one intervention for typical urban
middle schools. The research questions are the following:
(1) Is there a difference in reading rate and accuracy (together defined as fluency)
between students receiving a daily fluency intervention and the control group?
(2) Is there a difference in reading comprehension between students receiving a daily
fluency intervention and the control group?
(3) Is there a difference between the fluency and comprehension of students with
poor phonemic awareness skills and those with stronger phonemic skills as
measured before the intervention?
(4) Is there a difference between the fluency and comprehension of students with
poor RAN skills and those with stronger RAN skills as measured before the
intervention?
(5) Can a one-on-one daily fluency intervention program be implemented
successfully in two typical urban middle schools?

10

Study Design
Following the recommendations of the National Reading Panel (2000), this study
uses an experimental, randomized control group design. Also as recommended by
previous research, the study uses adults to administer the interventions, using a
commercially available fluency program that builds in error feedback, documentation of
student progress and regularly increasing reading levels. In this way, the study seeks to
remediate some of the identified limitations in the current research base (Chard, Vaughn
& Tyler, 2002; NRP, 2000; Therrien, 2004).
Participants were pre-tested on a variety of standardized measures including
measures of verbal language ability, RAN, phonemic awareness, phonemic decoding,
word recognition, reading fluency and reading comprehension. Only students who were
reading below the fourth grade level and who had average to above average ability in
terms of verbal language were included in the study sample.
Study subjects were randomly assigned to one of two groups: the experimental
group participated in a fluency intervention program called Great Leaps (Campbell,
1998), and the control group received an equal amount of study skills intervention. Study
skills was chosen as the control activity because of its intrinsic value to this population of
students with disabilities—it was important that the control group’s participation in the
project be productive and well spent. In addition, although the study skills curriculum
consists of reading and writing activities, the study skills program does not include any
explicit reading instruction. Both groups received their intervention programs 3 to 4 times
a week for ten minutes, working one-on-one with a trained paraprofessional.

11

Pre and posttest measures were analyzed using a variety of statistical measures,
including t-tests to compare the means of the two groups, correlation matrices, and
analyses of variance (ANOVA). The results were examined to look for reading outcomes
attributable to the intervention, as well as patterns of differential effects related to RAN
ability and phonemic awareness skill.
Limitations and Delimitations of the Study
This study has several identifiable limitations. Since most paraprofessionals only
worked with one of the treatment groups, there are most likely differential effects
attributable to the individual skills of the adults involved. In addition, the current cultures
and climates at the two middle schools were quite different, and the study effects may
also be influenced by this difference. Issues related to implementation of the programs
and assessments are also problematic. All of these potential limitations are discussed in
detail in subsequent sections.
Since the body of research on reading and reading fluency is quite extensive, the
delimitations of this study include a focus on fluency research related to severely delayed
readers in the middle school grades. An effort has been made to synthesize factors
believed to be particularly relevant to this population, and the findings may not be
assumed to apply to students outside that specific population. Furthermore, the students
in this study are all identified as having mild to moderate disabilities, and are from urban
environments. Again, the findings may not be generalizable to students who do not have
similar disabilities, or who come from very divergent populations.

12

In the following chapters, the literature base concerning fluency as an integral
component of reading will be examined, and recent studies on fluency interventions for
middle school students with reading delays will be discussed. Subsequently, the methods
of the study will be described in detail, including the statistical analysis of the data. The
results of the data analysis will be presented in chapter four, and finally those results will
be discussed in relationship to the existing literature base. Recommendations for future
research will also be presented.
Definition of Terms
Automaticity
Fast and accurate reading of words and passages, with minimal attention devoted
to decoding.
Fluency
A developmental reading process that is the product of accuracy and rate in
decoding. Once fully developed, fluency implies accurate and effortless decoding and
appropriate expression when reading (Wolf & Katzir-Cohen, 2001).
Fluency Intervention
A supplemental instructional activity specifically focused on increasing students’
reading fluency levels.
Learning Disability
“A disorder in one or more of the basic psychological processes involved in
understanding or in using spoken or written language, which may manifest itself in an
imperfect ability to listen, think, speak, read, write, spell or to do mathematical

13

calculations” (Individuals with Disabilities Education Act, 2004). This disorder should
not be the result of a visual, hearing or motor disability, a general developmental delay,
emotional, cultural or environmental factors, limited English proficiency or lack of
education.
Phonemic Awareness
The ability to hear and orally manipulate the individual sounds within words.
Phonemic awareness is an oral skill, and does not involve the manipulation or association
of letters. It is believed to be one of the most strongly correlated skills to reading success
(Kame’enui et. al, 1997).
Phonemic Decoding
The ability to associate sounds with letters, and to use this knowledge to read
words.
Phonological Processing
Similar to phonemic awareness, this is the ability to orally distinguish the smallest
units of sound within words, known as phonemes, and to subsequently identify and make
sense of the words they form.
Prosody
Appropriate rate, intonation, and expression in reading.
Rapid Automatic Naming (RAN)
The ability of an individual to quickly and effortlessly say the names of colors,
letters, numbers or other common features. RAN is believed to be correlated with reading
acquisition.

14

Reading Disability
A type of learning disability that manifests itself in the impaired ability to read as
compared to typical students of the same age. May include impairments in decoding
ability, fluency, and/or comprehension of text.
Reading Models
Theoretical representations of the processes believed to be involved in reading.
Treatment Resistors/Nonresponders
Both terms are used to signify students who have received well designed,
research-based reading interventions, but who have failed to develop age-appropriate
reading skills (Fuchs, Fuchs, McMaster, Yen, & Svenson, 2004; Torgesen, 1999).

15

CHAPTER II: REVIEW OF THE LITERATURE
This literature review will focus first on some of the historically significant
models of reading and reading fluency, and the interactions between the various
components of reading outlined in each of these theories. The purpose of this discussion
is to frame the context within which fluency research is now being conducted, and to
create a definition of fluency that will guide the current study. Although many more
reading models exist, these have been chosen because they each specifically address the
issue of automatic word recognition in some significant manner. Each reading theory will
be summarized and critically analyzed in order to create a clear understanding of reading
fluency and its component parts.
Next, these underlying components of reading fluency will be examined, and
correlational studies will be examined to determine what, if any, core cognitive factors
may influence a student’s ability to become a fluent reader. Finally, intervention studies
that examine the relative importance of each of the components of fluency will be
reviewed. These studies will be analyzed to see what can be learned about the
significance of each of these components in building reading fluency, and the findings
will be discussed in the context of the current intervention.
Models of Reading and Reading Fluency
Historically, reading fluency first came to the attention of the education
establishment in 1908 with the work of Edmond Huey (1908/1968). Huey, whose efforts
are now considered to be visionary for his time, was the first to discuss the importance of
repeated practice in reading in order to develop fluency and speed, likening the automatic

16

processes needed for fluent reading to the automatic swinging of a tennis racquet. He
proposed that systematic rehearsal was key to accumulating the skills and sub-skills
needed to read more quickly and accurately, stating that “repetition progressively frees
the mind from attention to details, makes facile the total act, shortens the time, and
reduces the extent to which consciousness must concern itself with the process” (Huey,
1968, p. 65). Huey’s seminal investigations, however, were not further developed for
many years. It wasn’t until 1974 that LaBerge and Samuels began to reexamine Huey’s
writings and think about the automatic processes involved in fluent reading. The model
they developed, as well as many subsequent models, was rooted in his work, and
presented reading as a hierarchical progression of component skills. These early
hierarchical models will be considered here first.
Hierarchical Models of Reading Fluency
The attention allocation model. Building on Huey’s idea of automaticity, LaBerge
and Samuels proposed a model of reading fluency based on the concept of attention
allocation. The authors established their model on the assumption that “we can only
attend to one thing at a time, but we may be able to process many things at a time as long
as not more than one requires attention” (LaBerge & Samuels, 1974, p. 294, emphasis
added). For example, when a person drives a car most of the processes are automatic, and
he or she may carry on a conversation with ease while driving. However, as soon as
something happens that requires attention, such as a child chasing a ball into the street,
the driver will need to stop talking in order to concentrate attention on the situation
(Purcell-Gates, 1997).

17

LaBerge and Samuels suggest that reading develops through a serious of
stages, and that each stage requires automaticity in order to move efficiently into the next
one; before a stage becomes automatic, however, readers have to allocate attention to the
processes embedded within it. If a reader is asked to move on to a new stage before he or
she has sufficiently automated the previous one, it may overburden the attentional
resources and hinder acquisition of the new skills (LaBerge & Samuels, 1974). (It should
be noted that the authors’ criteria for automaticity is that the skill can be performed
without conscious attention; a person will perform it naturally, even without meaning to.
For example, skilled readers automatically read billboards as they pass, even if they don’t
intend to; it takes conscious effort for skilled readers not to decode the text they see.)
LaBerge and Samuels’ hierarchical reading model assumes that the first thing that
happens when a person views text is that the symbols on the page are processed and the
identifying features are noted. For instance, when looking at the letters t and h, the
identifying features would be the crossing line on the t and the half loop on the h. Their
theory proposes that, as the identification of individual letters becomes automatic, less
attention is paid to these features; the graphical features then become “chunked” in long-
term memory as the letter itself, and the features no longer require significant attention.
At that point the student would move forward into the next stage: the formation of a code
from the letters, or word recognition. According to LaBerge and Samuels (1974), word
recognition requires automaticity of letter recognition (the previous stage) or the word
recognition process will be slow and laborious.

18

The authors go on to discuss the various levels of automaticity that must occur
as a student moves on toward more proficient reading, and the allocation of attention that
is characterized in each of the subsequent stages. For example, a more proficient reader
may recognize a common word and activate its meaning automatically, with no
purposeful allocation of attention. The reader might also recognize and make meaning
from a group of words, again without allocating attention. However, some allocation of
attention might be needed to identify familiar word parts or individual letter sounds in a
very complex word or group of words. In that case, the automaticity and speed of reading
is likely to decrease as the reader splits his attentional resources between decoding and
meaning (LaBerge & Samuels, 1974).
Gradually, the model proposes, students begin to move from letter-by-letter and
word-by-word decoding into the chunking of words, which are labeled “unit codes.” The
authors posit this chunking as an interaction between visual memory, phonological
memory and episodic memory. At this level of interaction, the reader is beginning to
develop automaticity, and LaBerge and Samuels hypothesize that at this point little
attention is devoted to the visual system, as the phonological code is automatically
triggered.
In the final step of the model, the authors discuss the high levels of attention
needed to perform more complex comprehensional tasks, and they speculate that this type
of higher-level comprehension will be difficult for readers who cannot perform the
underlying processes automatically. Thus, if a student becomes more automatic at the
subcomponents of decoding, it can be assumed that their ability to comprehend will be

19

improved. They also emphasize the importance, at this level, of the reader having a
good deal of semantic vocabulary information stored in long term memory.
The attention allocation model of reading fluency proposed by LaBerge and
Samuels was highly influential in the reading community, as it proposed a new way of
looking at the resources involved in skillful reading. Perfetti (1985) used their model as
the basis for his own, which focused more on the interaction between the amount of effort
involved in focusing attentional resources and the successful or unsuccessful outcome of
reading. That model will be examined next.
The verbal efficiency model. Perfetti’s verbal efficiency model (1985) is closely
related to LaBerge & Samuels’ attention allocation theory, and although it is presented as
a linear/hierarchical set of skills, the author does discuss the possibility of a cyclical
relationship that could allow one process to compensate to some extent for another.
Similar to LaBerge and Samuels, it is built on the assumption that a limited amount of
attentional and memory resources are available for reading, and as with their model,
Perfetti assumes that the rapid, effortless decoding of words is a critical predecessor to
efficient and productive comprehension of text. However, the Perfetti model proposes an
interaction between the amount of resources needed for any part of the reading process
(called “cost”) and the efficiency of the ultimate outcome of the reading, namely reading
comprehension (called “product”). As the cost of any process goes up due to the need for
increased attention or memory resources, the product (comprehension) is likely to go
down (Perfetti, 1985).

20

Perfetti bases his theory on the assumption that reading requires resource
allocation from both active working memory and attention, and that the outcome is
limited by the efficiency of each of the processes that draws upon those allocations. He
proposes that there are three main processing elements, all of which may vary in their
levels of efficiency: lexical access, or the ability to decode the words; propositional
encoding, the process of chunking text into small related units of meaning while reading;
and schema activation, which involves relating text to previous knowledge to make
meaning from the whole. A weakness in any one of these processes can require too much
“cost” from attention or memory, and can affect the reader’s ability to make meaning
from the text, the “product.” In a skilled reader, the processes of lexical access and
schema activation should be relatively automatic, and should require little in the way of
resources. However, in a delayed reader any or all of these processes may require a
significant allocation of resources, thereby robbing the reader of the ability to focus
resources on the process of comprehending the text (Perfetti, 1985).
Over time, researchers began to see some limitations inherent in the hierarchical
theories of reading. These linear, bottom-up models were unable to account for many of
the occurrences seen in laboratory and school-based studies (Stanovich, 2000).
Phenomena observed both anecdotally and in research studies, such as the effects of
syntax and context, were not explained by the models that proposed a strict, bottom-up
sequence of skills. As a result, researchers began examining the idea of an interaction
between the various component skills involved in reading. Some of those interactive
models will be examined next.

21

Nonhierarchical Models of Reading
The interactive compensatory model. In 1977, Rumelhart created one of the best
known of the interactive processes models. Although this model is not commonly
assumed to specifically address reading fluency, Stanovich used it as the basis of his own
interactive compensatory model of reading, which does address the issues of automaticity
and allocation of resources (Stanovich, 1980). In Stanovich’s interactive compensatory
model, readers who are struggling in any of the processes involved in reading may use
other processes to compensate for the weakness. According to Stanovich, this
compensation technique is frequently observed in students who have trouble with
decoding skills, and who use higher-level skills such as context to help them make sense
of the text. He argues that by adopting this interactive compensatory model of reading,
we agree that reading is not a linear process, and that it is not necessary that “the
initiation of a higher level process must await the completion of all the lower ones”
(Stanovich 2000, p. 23).
Stanovich’s model also utilizes the two-process theory of expectancy developed
by Posner and Snyder (1975), which proposed that there are two different processes by
which a reader may use context to improve word recognition. The first of these,
automatic-activation, occurs when previously stored information is activated in the
memory without conscious effort. Posner and Snyder theorize that in automatic-
activation, semantically related locations in the brain are activated, and that all of these
processes are done automatically, thus leaving most of the attentional capacity unused. In

22

the case of automatic-activation, attention is free to be concentrated on higher-level
processes related to comprehension.
The second type of cognitive system that Posner and Snyder discuss is called the
conscious-attention mechanism. Conscious-attention is activated more slowly, and will
be called into play when there is no automatic activation of previously stored
information. The conscious-attention mechanism requires a high allocation of attention,
and thus may rob cognitive resources from the higher level processes such as
comprehension. Building on this theory for his interactive compensatory model,
Stanovich argues that the interactive nature of all the reading processes allows higher
level processes to support deficits in lower level skills as necessary; thus, if a student is
unable to recognize a word automatically, the conscious-attention system may be
activated to help the student break down and decode the word, or to use context as a clue
to its identity. As a result, however, the student may have less processing capacity left for
higher-level reading processes (Stanovich, 2000).
It seems clear that the interactive compensatory model extends the work of
LaBerge and Samuels by allowing feedback from higher level processes to lower, and yet
it’s important to note that they share a common paradigm: that rapid, automatic
recognition of words is key to fluent reading, and that without automaticity, students will
lack the resources to comprehend effectively. The next model discusses in even greater
detail the processes that make automaticity possible.
The interactive processing model. Adams (1990) proposes a model of fluent
reading that is highly related to the Stanovich model (1980) in that all the elements of the

23

model are interactive, and each compensates for weaknesses in the other. Her model is
composed of four components: the orthographic processor, the phonological processor,
the meaning processor, and the context processor (Adams, 1990).
In the interactive processing model, the orthographic processor is the only
element of the model that interacts directly with text; Adams theorizes that all text is
primarily processed visually, and that the function of the orthographic processor is to
recognize and interpret the symbols on the page, and to associate them together into small
chunks of recognizable text similar to syllables. She proposes that the orthographic
processor is able to recognize common letter patterns, such as those found in word
families or affixes, and uses those letter patterns to quickly and efficiently decode even
the most difficult text.
Simultaneously, the orthographic processor is sharing the incoming visual
information with the meaning and phonological processors. Each serves as a kind of
meaningful feedback loop that interacts with the information interpreted by the
orthographic system, checks it for accuracy, and adds to it through its own distinctive
process. In the case of the phonological processor, it serves several functions. First is to
provide a “back up system” to the orthographic processor, checking the spelling patterns
for familiarity against its bank of phonologically stored patterns. Secondly, this element
helps the primary system to decode unfamiliar words, which in the case of a struggling
reader may be most of the text. This phonological back-up system accounts for the
sound-by-sound phonetic processing that we see in many beginning readers and which is
not evident in more accomplished readers (Adams, 1990). According to Adams,

24

accomplished readers have learned to use the orthographic processor to recognize
most of the letter patterns by sight, and thus rely less on the phonemic expertise of the
phonological processor.
The third important function of the phonological processor is to retain the chunks
of language as they are processed and decoded by the orthographic processor. It does this,
Adams theorizes, through the process of subvocalization: holding the oral image of the
words in short term memory just long enough for the meaning and context processors to
make use of them. However, Adams notes that if a struggling reader is using too much of
the phonological system to decode each word, it is unlikely that there will be resources
left for this type of retention of text, and comprehension will most likely suffer.
The meaning processor, as its name implies, is primarily concerned with creating
meaning from the words that are decoded and checked by the orthographic and
phonological systems. It is dependent on a reader’s vocabulary knowledge and world
knowledge, as well as the speed at which a reader can process the information through
the lower two systems. If the processing of text through the phonological and
orthographic systems is too slow, it may be difficult for the meaning processor to hold on
to the relationship between the parts of the text, and comprehension may suffer. The
meaning processor receives constant feedback from the context processor, which will
enrich its routine and guide it in making logical choices based on the context of the text
(Adams, 1990).
It is important to note that in this model, as in Stanovich’s, all the components
work simultaneously, and all serve as back-up compensatory systems as needed. For

25

those students who decode too slowly for the meaning processor to make sense of the
text, the context processor will kick in and provide additional support. When the
orthographic system runs in to an unfamiliar word, it receives both context and decoding
support through the other systems. The system is interactive and compensatory, much as
Stanovich’s model, and lends itself well to instructional application. For example, Adams
recommends that the complexity of text should be manipulated depending on the purpose
for reading: if the purpose is to gain new information, the complexity of both the words
and the sentence structure should probably be simplified so the orthographic,
phonological and context processors can support the work of the meaning systems. If the
purpose of the reading is to facilitate word recognition, novel words might be added to
the text, but the overall complexity of the sentence structure should be simplified in order
to allow the meaning processor to provide more support to the orthographic and
phonological systems (Adams, 1990).
Summary. All of the models examined above hypothesize about the processes
underlying reading in typical, fluent readers, and use that to discuss what may happen
when readers are not able to read fluently. However, other models have been proposed
that focus primarily on atypical readers, and attempt to explain the component processes
of dysfluent reading from that perspective. These models are not overall models of
reading processes, but are specifically centered on the process of fluency. Two of these
models will be investigated next.

26

Atypical Fluency Models
The systems analysis approach. Berninger, Abbott, Billingsley and Nagy (2001)
created an interactive model of fluency based on the concept that dysfluent readers may
have a variety of timing deficits rather than one global deficit that manifests itself as
“poor fluency.” The researchers studied 102 first through sixth graders identified with
reading and/or writing deficits. Students were tested on reading rates for graded passages,
real words out of context, and lists of pseudo words. They found that some subjects were
slow on all three measures, but others were slow on only one or two of the measures
(Berninger, Abbott, Billingsley & Nagy, 2001).
Based on these results, the authors classified their dysfluent readers according to
three different subtypes: low efficiency, low automaticity and poor executive functioning
(Berninger, et al., 2001). The error patterns for the three subtypes varied considerably.
Low efficiency readers were very accurate but very slow, whereas low automaticity
readers were both inaccurate and slow, with a pattern of errors that included lots of
repetitions and pauses, and many self-corrections based on meaning. Conversely, those
readers typed as having poor executive functions also made lots of errors, but did not use
meaning to correct themselves; these readers seemed to have a more random pattern of
errors that could be related to attention as well as faulty underlying processes such as
phonological skills or automaticity. The authors reflected that a single measure of oral
reading rate might not be enough to differentiate between the contributory factors
described here, and suggested that error analysis might be key to identifying the specific
subtype of dysfluency present.

27

As a result of their research, Berninger et al. (2001) proposed a systems model
to define the fluency process and to suggest treatment possibilities. Their model
represents reading as layers of interaction between three core components: orthography,
phonology, and morphology. Each of the three components interacts directly with the
others, but all are also affected by a syntactic layer, which guides the processing of
context. The authors theorize that each layer works on both a lexical and sub-lexical
level, and that each component may work alone or in concert to contribute to the reading
process.
They go on to discuss the instructional implications of their model, stating that,
“fluency is a multi-dimensional concept which may operate differently at the word,
sentence, and text levels” (Berninger et al., 2001, p. 396). According to their model of
fluency, practitioners need to assess a student to discover the specific type of dysfluency
and create interventions accordingly; for example, students in the poor executive
functioning group may need instruction in the reading processes as well as in
metacognitive strategies such as self-monitoring. On the other hand, students in the low-
efficiency group will probably benefit from simple repeated readings to increase
automaticity, without needing instruction in the component skills such as phonological
processing (Berninger et al., 2001).
Although the Berninger et al. model examines the interactive cognitive processes
fundamental to reading fluency, it is also very focused on classroom application. The
goal of their model is to clarify the assessment and treatment of dysfluent readers. In the
next model, concern about application is also at the forefront of the researchers’ thinking.

28

This model, similar to Berninger’s, proposes that there may be more than one type of
dysfunction at work in readers with poor fluency.
Double-deficit model. In 1999, Wolf and Bowers first published their
comprehensive theory of double-deficits in dyslexic readers. With an extensive review of
existing literature, they hypothesized that existent studies indicated support for the idea of
three discrete types of developmental dyslexia: those with deficits in phonological
decoding, those with deficits in naming-speed, and those with deficits in both—the
double-deficit readers (Wolf & Bowers, 1999).
The authors cite decades of work that has built a consensus about the existence of
phonological impairments associated with diminished reading ability in many disabled
readers. However, they argue that naming speed predicts a separate and distinctive part of
the reading process, defining it as the reader’s ability to rapidly process visual
orthographic information, link that information to stored orthographic representations of
letters, integrate those representations with sublexical linguistic and phonological
representations, and articulate the outcome (Wolf & Bowers, 1999, p. 418).
The double-deficit model does not attempt to define the hierarchy or linear order
of processes that occur as a person reads, but rather emphasizes the importance of speed
within each of the underlying processes. According to their model, readers have high
processing-speed requirements in eight discrete procedures as they deal with text:
attention, visual, mental representation, sensory information, integration, lexical,
affective and motoric (Wolf & Bowers, 1999, p. 417). The authors cite many studies that
have examined these naming processes and their relationship to both reading outcomes

29

and phonological deficits, and conclude that, although there is not 100% agreement in
the literature, most studies confirm that there are strong and discrete relationships of
naming speed to word and text reading fluency that differ from the relationships found
between phonological processing and word attack or decoding skills (see for example
Manis, Doi & Bhadha, 2000). They also summarize a body of research that mostly
concurs that the influence of naming speed occurs primarily in the first few years of
reading development, and diminishes significantly in typical readers after second or third
grade; however, in impaired readers that influence may last much longer (McBride-
Chang & Manis, 1996).
Wolf and Bowers are adamant that the main focus of their work, however, is less
theoretical, and more linked to the appropriate diagnosis and treatment of dysfluent
readers (Wolf & Bowers, 1999). They state that, “if we are better able to understand the
additional dimensions represented by deficits in naming speed in impaired readers… we
may be better prepared to design better diagnostic batteries and treatments that
correspond to the readers’ needs” (p. 430). They call for more research into fluency
interventions that address automaticity at the level of connected text, but also in
sublexical processes (Wolf & Bowers, 1999).
In a later article, Wolf and Katzir-Cohen (2001) expand on this discussion by
using several of the reading models outlined above to highlight the importance of
automaticity of lower level skills as being critical to fluency development (Adams, 1990;
LaBerge & Samuels, 1974; Perfetti, 1985). They discuss fluency as involving the
development of proficiency in every underlying process of reading, including

30

orthography, phonology, semantics, morphology, syntax, attention, auditory
perception, memory, lexical access, and prosody. According to the double-deficit model,
impairment in any of these processes can produce dysfluent reading (Wolf & Katzir-
Cohen, 2001).
Defining Fluency
Although each of the models outlined above examines reading fluency from a
slightly different perspective, each of them acknowledges that there are almost certainly
multiple systems at work when students learn to read fluently. It seems to be generally
agreed that these processes must include automatic recognition of words (orthographic
processing), both at the single word and the passage level, as well as phonemic skills
(phonological processing) and use of context (semantic processing) to support the
recognition of unfamiliar words. All of the models also emphasize the importance of
automaticity in sub-lexical skills such as letter and phoneme recognition.
Equally as important, each of those models acknowledges that comprehension is
the ultimate goal of reading fluency, and each model seeks to explore how fluency, or the
lack of it, can affect a reader’s ability to understand and interact with text in a meaningful
way. Huey (1908/1968), LaBerge and Samuels (1974), and Perfetti (1985) all view
reading fluency as the key to making cognitive resources available for comprehension;
Stanovich, on the other hand, extends that concept with the notion that comprehension
skills can act as a compensating factor for struggling decoders (2000). In Adams’ (1990)
model the processes of fluent decoding and comprehension support each other, and the
interaction between them shifts as a result of developmental reading growth. Conversely,

31

the final two models seek to identify the underlying causes of dysfluent reading, and
define the process of fluency development in terms of potential barriers (Berninger et al.,
2001, Wolf & Bowers, 1999). For these two models, the development of comprehension
may begin by identifying and remediating the specific areas of fluency deficit.
However, as a result of the differences in these models, each of them leads us to
define fluency in a somewhat different manner. Through the eyes of the early hierarchical
models fluency becomes the outcome of practice in automatic decoding and word
recognition; it might be defined as a simple interaction between reading rate and
accuracy, which then leads to the ability to comprehend. For Adams and Stanovich,
fluency is one piece in an interactive process; it can facilitate comprehension or be
facilitated by it, but the development of fluency is not necessarily a linear evolution.
Nevertheless, if it can be assumed that the ultimate outcome of any reading remediation
program is to improve comprehension—and the theories outlined above seem to have
consensus on that point—then it would appear that a definition of fluency must include
comprehension of text as part of that definition.
Wolf and Katzir-Cohen (2001) used much of the theoretical research summarized
above to create a definition of fluency that seems to incorporate many of the critical
components identified in these models, including the essential outcome of improved
reading comprehension. This definition recognizes a developmental curve to reading
fluency, gradually moving the focus from lower-level processes to higher-level
comprehension of text:

32

In its beginnings, reading fluency is the product of the initial development of
accuracy and the subsequent development of automaticity in underlying
sublexical processes, lexical processes, and their integration in single-
word reading and connected text. These include perceptual, phonological,
orthographic, and morphological processes at the letter, letter-pattern, and
word levels, as well as semantic and syntactic processes at the word level
and connected-text level. After it is fully developed, reading fluency refers
to a level of accuracy and rate where decoding is relatively effortless;
where oral reading is smooth and accurate with correct prosody; and
where attention can be allocated to comprehension (Wolf & Katzir-Cohen,
2001, p. 219).

It is this definition of fluency, founded on the theoretical models previously
discussed, that I will use in subsequent sections of this paper. In addition, it is these
components of fluency, as defined by Wolf and Katzir-Cohen (2001), that will be
examined in relation to the existent literature base: sublexical processes, lexical
processes, and the integration of both into meaningful comprehension of text.
However, first it seems important to review the literature on factors related to
fluency and dysfluency. A body of work exists that attempts to use correlational research
to verify the underlying characteristics proposed by many of the reading models outlined
above, and to link those characteristics to a student’s ability to read fluently. Since the
current study seeks to remediate the fluency levels of disabled readers, is seems critical to
identify the factors that may impede or contribute to their ultimate success. This body of
literature will be examined in the subsequent section.
Correlating Factors to Fluent Reading
For the last several decades, researchers have been exploring the link between a
child’s ability to process sounds and his or her reading aptitude; there is now a high
degree of consensus that a child’s phonemic awareness skill (PA) is highly correlated

33

with his or her ultimate capacity to read fluently and expertly (see Bradley & Bryant,
1978; Manis, Custodio & Szeszulski, 1993; Wagner et al., 1997, for example). In
addition, many researchers view phonemic skill as a stable characteristic similar to other
cognitive traits such as IQ (Wagner et al., 1997), and hypothesize that it is established
early in a child’s development, and remains relatively constant over time, despite growth
in reading ability (Manis, Custodio & Szeszulski, 1993).
In recent years, then, research has focused less on reconfirming the impact of
phonemic awareness on reading ability, and more on differentiating the impact of
phonemic awareness from other processing deficits that seem to provide an impediment
to reading (Schatschneider, Carlson, Francis, Foorman, & Fletcher, 2002). One area
currently under active investigation is naming speed.
The Contributions of Phonemic Awareness vs. Naming Speed
Naming speed, also called rapid automatic naming (RAN), is one measure of a
person’s ability to quickly retrieve phonological codes, such as the names of objects,
colors, letters or numbers, from long-term memory (Wagner et al., 1997). As outlined by
Wolf and Bower (1999) in the double-deficit model, some researchers believe that a
child’s ability to recall these stored codes efficiently and effortlessly has an enormous
impact on their ability to read fluently. This impact can be seen on a variety of levels; for
example, a child needs to be able to recall the letter sounds in order to decode words, and
also to recall the word itself once the sounds have been decoded (Wolf & Bower, 1999).
As could be discerned in the discussion of the double-deficit theory above, some
researchers believe that this ability to process and recall code quickly has a discrete and

34

powerful impact on children’s reading fluency, and many are now working to confirm
the truth of that assertion.
However, it is clear upon examining the research that there is not yet a consensus
about the discrete impact of processing speed on students’ reading ability, nor on the
value of RAN as a measure for predicting that impact. In a 2003 meta-analysis of
research on rapid naming, Swanson, Trainin, Necoechea and Hammill identified three
hypotheses about PA and RAN underlying much of the current research: (1) That PA and
RAN operate independently, (2) that they vary as a function of maturity with average
readers, and (3) that RAN is a complex process involving many underlying components
that contribute to reading separately from PA (Swanson et al., 2003). From the 35 articles
they included in their analysis, these authors construed that the evidence does not support
the contention that either PA or RAN are strong predictors of real word reading when
compared to other factors such as vocabulary or IQ, that they aren’t robustly correlated to
reading comprehension, and that evidence only weakly supports the double-deficit model.
However, they acknowledge that their findings may have been confounded because they
grouped together typical readers and students with reading disabilities (Swanson et al.,
2003).
Another synthesis of RAN research came up with quite different results (Allor,
2002). In this review, which synthesized and discussed the results of 32 articles published
between 1984 and 1998, Allor found that study results clearly imply a unique variance to
reading ability for PA that is unrelated to RAN, stating that all but one of the studies she
reviewed were in consensus on this point. In addition, although she acknowledged that

35

results concerning RAN were less definitive, the evidence seems to imply that it
counts for some unique variance in developing readers, including those who are older but
do not yet have fluent word reading skills (Allor, 2002). She is careful to point out,
however, that it is still unclear what the relationship is between the two factors, and how
much unique variance they each supply.
Since the publication of these two syntheses, more studies have been conducted
seeking to clarify the discrete contributions of PA and RAN, and to investigate the
usefulness of RAN as a predictive measure. Many researchers have now reported finding
that there is a separate variance accounted for by RAN when compared to PA (Katzir,
Kim, Wolf, Kennedy, Lovett & Morris, 2006; Katzir et al., 2006; Manis, Doi & Bhadha,
2000; Manis, Seidenberg & Doi, 1999; Meyer, Wood, Hart & Felton, 1998; Savage &
Frederickson, 2005; Stage, Abbott, Jenkins & Berninger, 2003; Sunseth & Bowers, 2002;
Wood, Hill, Meyer & Flowers, 2005). In addition, it seems that many studies are finding
a strong relationship between PA and phonologically-based reading skills such as
decoding and nonsense word reading (Katzir, Kim, Wolf, Kennedy, Lovett & Morris,
2006; Katzir et al., 2006; Manis, Doi & Bhadha, 2000; Meyer, Wood, Hart & Felton,
1998; Savage & Frederickson, 2005; Sunseth & Bowers, 2002), and a separate strong
correlation between RAN and orthographic skills such as automatic word recognition and
fluency (Katzir et al., 2006; Manis, Doi &Bhadha, 2000; Savage & Frederickson, 2005;
Schatschneider et al., 2002; Sunseth & Bowers, 2002). In addition, many studies are
finding that the influence of RAN ability in typical readers is strongest in the
developmental years of first and second grade (Manis, Seidenberg & Doi, 1999;

36

Torgesen, Wagner, Rashotte, Burgess & Hecht, 1997), but that it can persist in
students with reading disabilities well into eighth grade and beyond (Meyer, Wood, Hart
& Felton, 1998).
Many of these same researchers are finding support for the double-deficit theory
in their findings (Manis, Doi &Bhadha, 2000; Savage & Frederickson, 2005; Stage,
Abbot, Jenkins & Berninger, 2003). Manis, Doi and Bhadha found that a group of
students with double-deficits was qualitatively different from another group with only PA
deficits, and that the readers with the least reading impairment in their study were those
with single deficits in naming (2000). The most impaired readers in their sample were
those who had double-deficits that included the slowest naming times. Stage, Abbot,
Jenkins and Berninger, on the other hand, found evidence of the possibility of two types
of double-deficit in students with dyslexia: according to their results, students could
conceivably have a double-deficit including RAN and PA deficits, or one made up of
RAN and orthographic deficits (Stage, Abbot, Jenkins & Berninger, 2003).
As can be seen, the inquiries in this area are ongoing, but more and more
investigators seem to be finding evidence of discrete deficits in a variety of areas. Many
are confirming the theory of Wolf and Bowers (1999) that proposes two discrete areas of
deficiency for dyslexic readers: PA and RAN. In addition, some researchers have tried to
identify the impact of RAN on reading comprehension, but the results there are extremely
inconsistent. Manis, Seidenberg and Doi (1999) found that both RAN and PA correlated
separately with comprehension, while Meyer, Wood, Hart and Felton (1998) failed to
find any correlation between RAN and comprehension in their study. However, in a later

37

study Wood, Hill, Meyer and Flowers found RAN to add predictive variance to eighth
grade comprehension in severely impaired readers (2005). Savage and Frederickson
found only phonetic skills to predict comprehension (2005), while Katzir et al. (2006)
found that RAN, PA and orthographic skills were all correlated to reading
comprehension. Clearly, more work needs to be done in this area to confirm the
predictive and correlational relationship between RAN abilities and the development of
reading comprehension skills.
Instructional interventions and RAN. A few researchers have applied the double-
deficit model to instructional interventions, seeking to find out if discrete groups of
students, identified by their specific areas of deficit, would react differently to a range of
interventions; sadly, neither of these studies provide results in terms of reading
comprehension. Lovett, Steinbach and Fritjers (2000) worked with a sample of 166
students with severe reading disabilities aged seven to thirteen, giving them three
different types of instruction. One group received a program focused on phonological
analysis and blending at the letter-sound level, another got a program that taught a
systematic strategy for breaking down and decoding multiple-syllable words, and a third
group functioned as a control, getting an equivalent amount of study skills instruction
(Lovett, Steinbach & Fritjers, 2000).
Their results did show a differential effect from the interventions according to
deficit type, but all three groups of students, PA-deficit, RAN-deficit and double-deficit
(DD) benefited greatly from both interventions as compared to the control group. The
group with DD did the poorest, but still benefited significantly from both types of

38

instruction. In their discussion of the results, the authors stated that students with DD
were the most impaired readers. In addition, in alignment with the earlier findings of
Manis and colleagues (2000), they found that the RAN-deficit students were overall less
impaired than those with PA deficits. In fact, in their study the students with only RAN
deficits had higher verbal IQ than the students with PA only or DD. It’s interesting to
note that the RAN times of the low-RAN students did not increase as a result of the
interventions. It may be that RAN is also a stable characteristic in the population of
students with RAN deficits and significant reading delays (Lovett, Steinbach & Fritjers,
2000).
Levy, Bourassa and Horn (1999) examined three types of word instruction on
low-RAN and high-RAN poor readers. Their sample of 128 second-graders were taught
words through three different conditions: a whole-word method, a rime-family method,
or through the application of letter-by-letter phonemic decoding. The students learned
words in each of the conditions in separate trainings, and were assessed on the outcomes
after each training period was over (Levy, Bourassa & Horn, 1999).
Results showed that the low-RAN group generally learned words more slowly
than the others, but eventually achieved the same level of acquisition for words learned in
rime families or phonetically. However, words learned as whole units were never fully
mastered by the low-RAN group.
When tested for retention, words learned in the phonemic and whole-word
conditions were retained better than the words in the rime group by all learners. The
authors theorized that because rime family words were learned more quickly, they

39

provided the students with less practice and less opportunity for overlearning.
(However, it is significant that the low-RAN group learned fewer words with the whole-
word condition, and so retained fewer words from that condition overall.)
Interestingly, all three training conditions generalized to new words reliably,
except the whole-word condition for the low-RAN group. Generalization to words that
shared traits to trained words was highest for the phonemic condition and lowest for
whole word (Levy, Bourassa & Horn, 1999).
The authors discussed their findings in light of Wolf and Bowers’ assertion that
low-RAN readers have difficulty linking letters together into larger orthographical units,
thus making retention of new words difficult (1999). However, since low-RAN learners
in this study did well with rime-family instruction, which involves recognition of larger
orthographic patterns, they proposed that their results only partially support the Wolf and
Bowers model (1999). They also reported that rime-family learning may not be an
effective technique for the most disabled learners, as it may not allow enough rehearsal
for overlearning to occur (Levy, Bourassa & Horn, 1999).
It can be seen from this discussion that many researchers are beginning to
acknowledge the possibility of discrete influences of RAN and PA on reading
acquisition, and some are even beginning to look at the link between RAN and reading
comprehension. It might be hypothesized from these findings that a fluency intervention
with impaired middle-school readers may evidence differential effects according to the
underlying deficits responsible for the students’ reading difficulties. However, it’s

40

encouraging to note that Lovett, Steinbach and Fritjers (2000) found that low-RAN,
low-PA and DD readers all benefited from well-constructed reading remediation.
In the following section, the literature on fluency intervention programs will be
examined; in particular I will focus on interventions aimed at increasing the reading
comprehension of middle-school students by building their reading fluency. These
studies will be considered in relationship to Wolf and Katzir-Cohen’s definition of
fluency (2001) as outlined above, recognizing the importance of sublexical and lexical
processes in a student’s ability to fluently decode and comprehend text.
Fluency and Comprehension
Many researchers have sought to investigate and expand upon the reading theories
outlined above through empirical studies. As previously discussed, numerous
correlational studies have explored the unique characteristics of readers in relationship to
reading fluency, and these studies are ongoing today. Other researchers have created
longitudinal studies that examine the development of reading proficiency over time, and
which seek to identify why some students become fluent and others do not.
There is also a large body of work that is focused on intervention techniques
designed to help students increase their fluency, and which explores the most efficacious
ways to do so. That literature will be examined in this section. In accordance with the
Wolf and Katzir-Cohen definition of fluency, I will specifically focus on studies that
examine the link between fluency and comprehension. In addition, I will expressly
concentrate on interventions designed to increase the reading fluency of secondary

41

students with reading disabilities, many of whom struggle to increase their proficiency
at reading and comprehending unfamiliar text.
To begin, it seems the most efficient way to summarize the multitude of studies
done since the 1970s may be to discuss in detail several meta-analyses that have been
conducted during the last ten years. These syntheses of research will be examined first.
Syntheses of Fluency Intervention Research
In the past several years, as the focus on fluency has intensified, several reviews
and meta-analyses of the existent literature base have been conducted. Each of these had
a unique frame of reference that affected the conclusions that were drawn by the
researchers, however all the authors profess to have found a relatively consistent positive
relationship between fluency intervention and reading comprehension.
The largest and most publicized of these studies was conducted by the National
Reading Panel (NRP, 2000). Their study looked at experimental research that
encompassed a variety of fluency-building methods such as repeated reading,
neurological impress, peer tutoring, assisted reading, and a host of others, both at the
lexical and connected-text levels. They included studies with children from kindergarten
through twelfth grade, with both typical learners and students with LD (NRP, 2000). The
panel concluded:
Such procedures had a consistent and positive impact on word recognition,
fluency, and comprehension as measured by a variety of test instruments
and at a range of grade levels… Analysis indicated that repeated reading
procedures have a clear impact on the reading ability of non-impaired
readers through at least grade 4, as well as on students with various kinds
of reading problems throughout high school (p. 3-3, Executive Summary).

42

The average effect size for comprehension in their meta-analysis was small but
significant at .35 (NRP, 2000).
It is additionally noteworthy that the NRP identified fluency practice as an
effective technique for typical readers only through fourth grade. This finding has been
duplicated by other researchers (see Fuchs, Fuchs, Hamlett, Walz, & Germann, 1993)
who have found that for average students, the greatest amount of fluency growth occurs
during the years of early reading development in the primary grades, and tapers off
through junior high school. More research is needed to confirm whether the benefits of
fluency intervention will consistently transfer to older students reading below grade level
(Fuchs, Fuchs, Hamlett, Walz, & Germann, 1993).
Therrien (2004) did another meta-analysis using the same population guidelines
as the National Reading Panel, but with a specific focus on repeated reading in connected
text. However, he differentiated his findings, looking specifically at the impact of cued
reading (where the student is cued to concentrate on speed or comprehension), corrective
feedback, peer vs. adult support, and preset performance criteria. He also discriminated
between transfer vs. non-transfer studies, a transfer study being one that generalized the
effects of the intervention to the reading of novel text.
Therrien’s meta-analysis, like the NRP, found positive correlations between
increased fluency and improved reading comprehension. He also found consistently
higher outcomes for adult-supported than for peer-supported interventions, with a strong
effect size of .71 for comprehension when an adult led the intervention. Interestingly, he
found that the transfer effects of fluency instruction were higher for students with LD

43

than for typical readers, particularly when combined with corrective feedback and
specific performance criteria (Therrien, 2004). These results correlate strongly with the
findings of Chard, Vaughn and Tyler (2002), who synthesized fluency studies that
included students with LD in first through fifth grade. Their results confirmed the impact
of adult feedback and monitoring, and also identified the importance of moving students
into increasingly harder text during the course of the intervention (Chard, Vaughn &
Tyler, 2002).
Two other syntheses in recent years have examined a wide variety of fluency
studies to identify patterns, one through single-case meta-analysis (Morgan & Sideridis,
2006) and one through the process of vote counting (Kuhn & Stahl, 2003), in which no
meta-analysis was done. Both found consistent reading growth to be a result of the
fluency interventions, although Morgan and Sideridis did not examine the effects on
comprehension.
The studies reviewed above included a wide variety of fluency building activities,
as diverse as repeated word practice, peer tutoring, the neurological impress method, and
repeated reading. From these meta-analyses and syntheses it seems that the majority of
fluency intervention studies conducted in the last forty years have found a positive
correlation between increased fluency and reading comprehension, particularly for
students with LD. However, it should be noted that most of these researchers mentioned a
serious deficiency of experimental research in this area; without a control group, it is
impossible to conjecture whether fluency intervention creates better comprehension, or
whether some other confounding factor is at work. Many of the authors called for more

44

studies using control groups, and identified several other unresolved issues related to
fluency instruction; common among these were the challenge of identifying the most
effective level of text to use during interventions, the optimum length for fluency
training, the long-term sustainability of fluency increases, and the feasibility of chunking
text into phrases or prosodic segments for fluency practice (Chard, Vaughn & Tyler,
2002; Kuhn & Stahl, 2003)
In the past several years, since these meta-analyses were conducted, more studies
have been performed that closely examine the link between fluency intervention and
reading comprehension. These studies can be divided into two main categories: studies in
which students practice fluent reading at the word or lexical level, and studies that
examine students’ reading in connected text. These two categories of research will be
explored next.
Lexical Processes and Reading Fluency
The first study to examine the use of isolated word practice and its effects on
reading comprehension was published in 1979 (Fleisher, Jenkins & Pany, 1979). Ever
since, researchers have been disputing the results and seeking to replicate or disprove
them. Dozens of studies have been done over the years that sought to identify whether
isolated word training could increase students’ ability to decode, read fluently, and
comprehend text. Even now, almost thirty years later, the answers to these questions are
still in dispute.
The seminal study by Fleisher, Jenkins and Pany (1979) was the first training
study to try to discover the effects of single-word training on passage fluency and

45

comprehension, and the results have been debated and retested by numerous
researchers ever since. The authors set themselves the task of trying to confirm Perfetti’s
“bottleneck” hypothesis, which states that poor readers use most of their processing
ability in decoding and thus have very little left over for comprehension of text.
According to the bottleneck theory, if students become automatic in their word
recognition they will free up processing space that can be devoted to comprehension
(Perfetti & Hogaboam, 1975). Theoretically, then, if students’ fluency levels increase,
their comprehension should automatically improve as well.
Fleisher and her colleagues worked with 36 fourth and fifth graders who were
identified as either good or bad readers; the good readers had an average reading level of
7.6, while the bad readers averaged a 2.8 reading level. The researchers identified two
short reading passages, written at the 7.1 and 6.3 grade levels, respectively; the poor
readers were then trained on a list of words that included all the words in one of these two
passages. The students were drilled repeatedly with flash cards until they could read all
the words from that passage within approximately one second of exposure, and then they
were given the passage to read. As they read, they were assessed on accuracy, speed, and
ability to answer comprehension questions about the passages.
As a control, two other conditions were created. The good readers were asked to
read the word list and one of the passages without any training, and the poor readers were
also given the same “no training” procedure as the good readers, using the passage on
which they had not been trained. In essence, the poor readers were “serving as their own
control group” for the untrained passage (Fleisher, Jenkins & Pany, 1979, p. 35).

46

Results of the passage reading showed that the poor readers improved their
fluency rates significantly on the trained passages, reading an average of 91 words per
minute (wpm) as compared to 61 wpm on the untrained passages. They also significantly
reduced their errors on the trained passages. However, the results did not generalize to
comprehension; despite their increased fluency levels, there were no effects of training
found on the poor readers’ ability to answer comprehension questions about the reading.
As might be expected, the poor readers did not match the good readers on any measure
(Fleisher, Jenkins & Pany, 1979).
In order to confirm their results, the authors decided to reproduce the study with
some methodological changes: in the second study they trained the poor readers even
more stringently until the poor readers matched or exceeded the fluency levels of the
good readers on the word lists. In addition, in order to avoid the possibility of students
worrying too much about speed over comprehension, students were not overtly timed as
they read the passages (Fleisher, Jenkins & Pany, 1979).
In the second study, the poor readers in the word-training group increased their
fluency significantly, however they were still unable to match the speed of the good
readers when reading the passages. Although the poor readers were able to exceed the
wpm of the good readers when reading words in isolation (95 wpm for the poor readers
vs. 90 wpm for the good readers), in passage reading the poor readers averaged 107 wpm,
as compared with 151 wpm for the good readers. Comprehension for the poor readers
was significantly less than for the good readers, and not significantly higher than on the
untrained passages.

47

One of the primary criticisms of this study is the use of above grade-level
passages as the medium for fluency assessment (Levy, Abello & Lysynchuk, 1997).
According to Stanovich (1986), poor readers do less reading than good readers, and as a
result they get systematically further behind in terms of vocabulary, concept development
and reading comprehension strategies. This theory, called the Mathew Effect, is generally
accepted by the educational establishment, and poses a serious question about the use of
above grade-level passages in the Fleisher et al. study; even though students were trained
in the recognition of the words from the passages, they were not trained in the meaning of
the words, or in the concepts covered in the passages. It is very possible that although
their fluency increased to near normal levels, the poor readers did not possess the
experience with text needed to successfully comprehend what they were decoding at
these advanced levels.
Nevertheless, Fleisher and her colleagues maintained that their studies disproved
Perfetti’s bottleneck theory, and asserted that isolated word practice is an inadequate
strategy for increasing students’ reading skills in the classroom (Fleisher, Jenkins &
Pany, 1979). Many researchers in subsequent years have replicated and reproduced
portions of their studies, and many have criticized their findings, yet the debate on the
ability of single-word training to improve overall reading comprehension remains
unresolved. As I review several recent studies that investigate the effectiveness of word-
level interventions and their ability to improve reading comprehension, it will be seen
that questions about the bottleneck theory are still being posed, and portions of Fleisher’s
study design are still being replicated and adjusted.

48

Two recent publications were found that studied the effects of word-level
training on reading comprehension; both of these examined the efficacy of practicing
single words in context vs. out of context (Irausquin, Drent, & Verhoeven, 2005; Martin-
Chang & Levy, 2006). The authors of the first study worked with 28 children, ages seven
to ten, divided into two groups: a “speed” group who worked on a variety of computer-
based activities designed to help the students learn to recognize consonant-vowel-
consonant (C-V-C) words quickly and accurately, and a “context” or control group, who
worked on computer activities using C-V-C words in a series of semantic-based activities
(Irausquin, Drent, & Verhoeven, 2005). In the speed group, the computer automatically
adjusted the speed of the activities in order to ensure accuracy, as well as to push the
students to higher levels of fluency as quickly as possible.
Students were assessed pre- and post-intervention on their ability to read C-V-C
words as well as untrained monosyllabic words containing consonant clusters. They also
were tested on their ability to read connected text fluently and for comprehension. The
results of the post-tests showed significant advantages for the students in the speed group.
They had higher gains in ability to read C-V-C words, both in terms of speed and
accuracy, but perhaps more importantly, they also showed significantly higher
achievement on the untrained consonant cluster words (Irausquin, Drent, & Verhoeven,
2005). This would seem to imply that they were able to generalize the effects of the
intervention to more difficult words.
The speed group also did better on the connected text: more students in the speed
group were reading higher grade-levels of text at the end of the intervention. It is also

49

significant, however, that the control group also had many students increase their
reading levels. The authors attribute this to a possible increase in syntactic awareness
resulting from their participation in the semantic activities, leading to increased ability to
decode (Stanovich, Cunningham & Freeman, 1984, as cited in Irausquin, Drent, &
Verhoeven, 2005).
In the second study (Martin-Chang and Levy, 2006), the subjects were 48 students
in the fourth grade; 22 were good readers, with an average tested reading level of seventh
grade, and 24 were poor readers, with an average reading level of second grade. Two
different experiments were conducted with each group. In the out-of-context experiment,
students were asked to practice a list of 85 words on the computer over the period of
several days; at the end of the training, the students were asked to read the list of words
for accuracy and speed, and then to read the words within the context of a third-grade
level training passage. In the in-context experiment, students received the same number
of exposures to the training words, but within the context of a grade-level story. Students
sat with a facilitator, who read the story to them. The students were required to read all
the highlighted training words as quickly as they could, however they were only required
to read the training words, as the rest of the story was read to them by the facilitator. The
students were then assessed in the same manner as they were for the out-of-context
experiment: reading the word list and two passages (Martin-Chang and Levy, 2006).
As might be expected, good readers had higher outcomes for both accuracy and
fluency across all conditions. However, both good and poor readers made significant
gains in fluency and accuracy in both experiments, with poor readers making

50

proportionately greater gains in both areas (Martin-Chang and Levy, 2006). When
comparing the two interventions, both good and poor readers made statistically greater
gains in fluency from the in-context training, but both groups learned more words from
the out-of-context training, and there were no significant differences in accuracy.
Significantly, none of the training groups in either of these two studies made a
significant increase in reading comprehension as a result of the fluency intervention
(Irausquin, Drent, & Verhoeven, 2005; Martin-Chang & Levy, 2006). It would seem that
based on these recent studies, Perfetti’s bottleneck theory may, in fact, be called into
question. In addition, the interactive reading models discussed in the previous section
assume that as word recognition becomes more automatic, cognitive processes are freed
and become available for comprehension (Stanovich, 2000; Adams, 1990); the results of
these studies seem to belie those theories. However, one other study was found,
admittedly a bit older than the others being presented here, that seems to provide an
explanation for these findings that may be consistent with the fluency theories previously
discussed. That study will be examined next.
In 1997, a study was published that looked at the effects of isolated word training
on students’ ability to read those same words within the context of passages. The authors
were interested in whether students’ ability to read passages that contained the trained
words would be superior to their ability to read untrained passages, and whether there
would be corresponding increases in comprehension (Levy, Abello & Lysynchuk, 1997).
The subjects were 28 fourth graders selected as “poor readers” due to low scores
on word reading assessments. In the first experiment the students were trained in 72

51

content words, taken from a third-grade level story containing a total of approximately
91 content words. For four days in a row, students were trained by computer on the list of
72 words, with six repetitions each day for a total of 24 training trials. On the fifth day
the students were asked to read two passages—one that included the trained words, one
that didn’t. Students read each passage three times, answering four discrete
comprehension questions after each reading. Passage readings were assessed for fluency,
accuracy, and accuracy of comprehension (Levy, Abello & Lysynchuk, 1997). The
results showed that although the students all read more quickly and more accurately in
the passages that contained the trained words, there was no difference in comprehension
levels between the trained and untrained passages.
Interested in further investigating the lack of comprehension gains, the authors
hypothesized that perhaps the increase in reading fluency evidenced by the students in the
first experiment was not enough to influence their comprehension. Citing the seminal
study by Fleisher, Jenkins and Pany (1979) in which the students were trained on all the
words in the passages until they reached a fluency level of approximately one word per
second, the researchers designed a follow-up study to further investigate their research
questions.
In the second experiment the procedures were very similar to the first study, with
a few important modifications. The participants were fourth graders designated as poor
readers who had not participated in the first experiment. This time, during the training
segment, students were encouraged to read the words more quickly by cutting down the
length of time the words were presented on the computer to 1.5 seconds. In addition, the

52

authors reduced the number of words in the passages, and trained the students in all of
the content words found in the passage, as in the procedures used by Fleisher et al. All
students reached a word recognition level of approximately one word per second during
the training portion of the second experiment.
Similar to the first experiment, the results of the second one showed that all
students made significant progress in their ability to read the trained words fluently and
accurately both out-of-context and in the passages. Again as in the first experiment,
students read the passages with the trained words much more quickly and accurately than
those that contained other words. However, the results of the second experiment
evidenced one noteworthy difference from the first. On the second experiment the
students scored significantly higher on comprehension of the trained passages vs. the
untrained ones. The authors contend that their studies show that training on words in
isolation helps increase the fluent and accurate reading of those words in context. In
addition, they assert that these gains must be “well consolidated” in order for them to
generalize to comprehension; it seems that until words are automatized to around a word
per second, comprehension may not be affected (Martin-Chang & Levy, 2006).
A review of the work of Breznitz and Berman (2003) can inform these results a
little more. In a series of related studies, Breznitz and colleagues found that by pushing
students to read at their maximal rate, they made fewer reading errors and had higher
comprehension than when allowed to read at their typical rate. Several years later, the
author replicated these findings with a group of dyslexic readers, finding a 35% error
reduction and a 20% increase in comprehension in this population when pushed to

53

increase their reading speeds. The authors assert that reading acceleration influences
comprehension, attention span, enhances short term and working memory and increases
word retrieval (Breznitz & Berman, 2003). It seems possible that this acceleration effect
may have played a strong role in the comprehension outcomes in some of the previous
studies.
Fluency intervention at the sublexical level. Very few studies have undertaken
interventions aimed at creating fluency in sublexical skills, however one study was found
linking sublexical processes to reading comprehension in novel reading material. The
authors cite the verbal efficiency theory of Perfetti (1985) as the theoretical framework
for their study, stating that the strong version of Perfetti’s model implies that increasing a
student’s ability to decode quickly and efficiently should bring about an automatic
increase in comprehension (van der Bosh, van Bon & Schreuder, 1995).
Students in this study were given training in reading four- or five-letter nonsense
words in either a timed or untimed condition; the authors hypothesized that the timed
condition would facilitate automaticity of decoding, and thus increase reading
comprehension. Although the results showed that the students in the timed group did
improve their ability to decode real words quickly and accurately as compared to the
untimed group, neither group improved their comprehension as a result of this brief, 16-
session intervention (van der Bosh, van Bon & Schreuder, 1995).
It doesn’t seem surprising that a discrete focus on sublexical skills would be only
tenuously linked to reading comprehension. In all of the theoretical models previously
reviewed there are multiple other components believed to supplement or support an

54

individual’s ability to use phonemic decoding skills to facilitate comprehension.
However, in multiple other studies that didn’t report on reading comprehension, training
in these sublexical components was found to have positive effects on students’ ability to
read untrained words both in and out of context (Berends & Reitsma, 2007; Kairaluoma,
Ahonen, Aro & Holopainen, 2007; Thaler, Ebner, Wimmer, & Landerl, 2004; Tressoldi,
Vio & Iozzino, 2007), so it may be conjectured that sublexical-level fluency training
could be a valuable piece of a more extensive fluency training program.
From the studies reviewed above, it is obvious that researchers are still working to
develop a consensus about the effectiveness of isolated word training and its ability to
positively impact fluency and reading comprehension. The attention allocation theories of
Perfetti (1985), used as a foundation by many researchers, have yet to be definitively
confirmed in this context. In the next section research studies that have investigated the
efficacy of building comprehension through fluency interventions in connected text will
be presented; as this is a widely used classroom practice, it may be conjectured that more
consensus exists around the value of this practice.
Fluency Interventions in Connected Text
In the years since the meta-analyses outlined above were published, quite a few
studies have been conducted to investigate the efficacy of using various types of repeated
reading interventions with middle and high school students with disabilities. These
studies will be discussed below. However, first it seems important to review the original
publication about repeated reading, which has influenced thousands of classrooms and
inspired dozens of research studies with its findings. This study, initially published in

55

1979, was republished in 1997 in recognition of its tremendous impact on reading
instruction in American classrooms.
The naissance of repeated reading. In 1997 The Reading Teacher journal
reprinted the seminal article by S.J. Samuels (1979) that encouraged the educational
establishment to consider the technique of repeated reading to increase fluency. In his
article, Samuels discussed the importance of developing automaticity in reading skills,
and suggested repeated reading as one technique for doing so. He described the process
of repeated reading as “rereading a short meaningful passage several times until a
satisfactory level of fluency is reached” (Samuels, 1979/1997, p.377), and he cited a
study in which one student increased her reading rate from a starting point of 30 wpm to
almost 90 wpm after several readings of a text. More significantly, perhaps, this student’s
initial rate got higher on each passage she practiced, so that by the fifth passage her
starting rate was already almost 80 wpm before practicing. Samuels claimed that this
increase in initial reading speed demonstrated that the technique of repeated reading was
helping this student generalize her training into increased reading skills across texts.
Samuels recommended putting an emphasis on speed rather than accuracy, as he
observed that most students increase their accuracy as their speed improves. He cautions
against “scaring” students by putting too much focus on accuracy, stating, “… the student
becomes fearful of making a mistake, and consequently the pace of reading slows down.
In fact, if we overemphasize accuracy, we tend to impede fluency” (Samuels, 1979/1997,
p. 377).

56

Samuels used the theoretical framework of LaBerge and Samuels (1974) to
justify the practice of repeated reading, and stated that it would have a positive effect on
comprehension since “with each additional rereading, the student is better able to
comprehend because the decoding barrier to comprehension is gradually overcome” (p.
378). The article goes on to compare the practice of repeated reading to practice playing
the piano or tennis, noting that in both music and athletics, beginners are given small
units of activity to rehearse over and over again until they are able to perform
automatically and with mastery. He recommends giving beginning readers the same type
of repeated exercise to build their skills (Samuels, 1979/1997).
Although this article offered little in the way of empirical evidence to support the
procedure of repeated reading, it had a huge impact on the educational establishment. In
fact, in the introduction to the republishing of the article, the editor asserts that Samuels’
work initiated a new line of research into the use of repeated reading to build fluency and
comprehension, which has generated strong evidence that repeated reading is effective
for increasing students’ reading abilities (Dowhower, S. 1997). However, as will be seen
in the review of recent research on repeated reading, it is still far from conclusive that this
technique helps all students increase their ability to comprehend novel text.
Using repeated reading with secondary students with disabilities. Since the
publication of the meta-analyses of fluency interventions by Chard, Vaughn and Tyler
(2002), the National Reading Panel (2000), and Therrien (2004), a number of studies
have been conducted that specifically examine the use of fluency intervention to increase
the reading comprehension of students with reading delays in middle and high school.

57

However, only one piece of research was found that incorporates the recommendations
of these meta-analyses: a longer implementation period, implementation by an adult,
regular provision of corrective feedback, a predetermined performance criterion (such as
a target wpm), adjustment of passage level according to student performance, and the use
of a control group (Chard, Vaughn and Tyler 2002; National Reading Panel, 2000;
Therrien, 2004). Not surprisingly, the primary author of this study was also the author of
one of the meta-analyses. Because of its strong methodological foundations, this study
will be examined here first.
Therrien, Wickstrom, and Jones (2006), worked with 30 students, grades four
through eight, all of whom were identified with LD or at risk for LD. All the participants
were reading at least two grades below the level of their class placement. The 30
participants were grouped via stratified random assignment, whereby the reading levels
of the participants were evenly distributed between a control group and a treatment
group, but students were randomly assigned within those reading-level limitations.
Pretest measures showed no significant differences between the groups.
Each of the 15 students in the intervention group worked one-on-one with a
trained adult tutor to read short passages and answer comprehension questions. The
students were first taught to use a series of prompts to help them generate questions about
the story structure and plot, then they read and reread each passage until reaching a
predetermined wpm, up to a maximum of four times per passage. The students generated
questions based on the standardized prompts pre-taught to them by their tutors, and then
answered passage-specific questions. The reading levels of the passages were adjusted

58

according to how proficiently the student could reach the target wmp in each passage.
Students worked one-on-one with the tutors in 10 to 15 minute sessions, reading a total of
50 passages each over a four-month period. The published study did not specify whether
the intervention group participated in any activities other than the typical instruction
occurring in their classrooms.
The reading fluency of each group was assessed pre- and post-intervention using
the oral reading component of the Dynamic Indicators of Basic Early Literacy Skills
(DIBELS, University of Oregon, 2005), and their overall reading achievement was
measured using the Broad Reading Scale of the Woodcock-Johnson Achievement Test III
(WJRIII, Woodcock, Woodcock, McGrew & Mather, 2001). Results showed that the
students in the intervention group averaged a 2.07 grade level increase in fluency passage
level over the course of the intervention, and these students also significantly increased
the number of comprehension questions they answered correctly during the intervention
sessions. In addition, significant increases in fluency were found for the treatment group
as compared to the control group as measured on the DIBELS. It is important to note,
however, that although the scores of the treatment group on the WJR III averaged higher
than the control group, no overall statistically significant differences were found between
groups in students’ reading comprehension scores (Therrien, Wickstrom, & Jones, 2006).
In the discussion of the results, the authors cite their students’ ability to generalize
their fluency gains to novel text as a significant success of this study. They theorize that
the reason for this success is their application of the principles identified by the meta-
analyses of fluency intervention research: a longer implementation period, predetermined

59

performance criteria, and passage levels adjusted according to student success (Chard,
Vaughn and Tyler 2002; National Reading Panel, 2000; Therrien, 2004). They further
state that although the findings in terms of reading comprehension were not statistically
significant, studies with larger populations and longer duration need to be conducted in
order to fully determine the effect of the intervention on overall reading ability. In
addition, they discuss the possibility that there might be differences in outcomes between
students identified with LD and those who are not, but their sample size was too small to
investigate that possibility. (Therrien, Wickstrom, & Jones, 2006). It is also highly
possible that other correlated factors, such as students’ naming abilities or phonemic
awareness skills, may have had an impact, but these factors were not assessed in this
study.
Significantly, none of the other recent studies published with secondary students
use a control group, and the majority uses a multiple baseline design to administer
fluency interventions to very small groups of students. However, three studies were found
that worked with larger groups of secondary students with LD. Two of these studies,
Manset-Williamson & Nelson (2005) and Marchand-Martella, Martella, Orlob & Ebey
(2000), used one-on-one interventions to work on decoding and fluency skills with
secondary students, and both found increases in comprehension as a result.
The Manset-Williamson & Nelson (2005) study was conducted with 21 volunteer
participants ages 9 - 14, nominated by their schools as low-performers in reading; an
unspecified number of students were identified with LD. All participants’ average
reading levels were at least two years below grade level, with a mean grade equivalent

60

score of 2.5 on the WJR III Broad Reading Cluster (Woodcock, McGrew & Mather,
2001).
The students in this study were randomly assigned to one of two treatment types
each of which included a significant component of repeated reading, but one group also
received explicit instruction in comprehension skills. Both groups received some
additional instruction in decoding. Students participated in six weeks of one-on-one
tutoring with a trained college student, four days a week for one hour per day, for a total
of 20 hours of intervention over the course of a summer (Manset-Williamson & Nelson,
2005).
Results showed that both groups made significant progress in fluency on all
measures, both transfer and non-transfer, with a grade equivalent gain of 6 months for
both groups over the course of the five-week intervention. Both groups also made
significant progress on the training measure of oral retell comprehension, but not on the
transfer measure of comprehension, the WJR III. The authors theorize that the
interventions worked on the most immediate measure of comprehension, the oral retell,
which is all that can probably be expected during such a short intervention (Manset-
Williamson & Nelson, 2005). It might also be questioned whether the WJR III is
sensitive enough to pick up changes in comprehension occurring over only five weeks of
intervention.
The Marchand-Martella et al. study (2000) used peer tutors rather than adults to
work with 22 ninth-grade students, both with and without LD, and once again the
intervention involved repeated readings as well as decoding instruction. This study did

61

not specifically target the teaching of any comprehension strategies, however
comprehension was computed as a dependent measure (Marchand-Martella et al., 2000).
The students were pre- and post-tested on the Gates MacGinitie reading test (MacGinitie
& MacGinitie, 1992), and were placed in intervention materials according to the pre-test
results. The instruction in this study lasted a little bit longer than the previous one, taking
place four days a week in 50-minute sessions over the course of 80 school days.
Results showed significant improvement in fluency and comprehension for the
majority of the students, with an average increase of 1.5 grade levels in comprehension
over the course of the 80-day intervention. Intriguingly, the students in the lowest level of
materials did not evidence significant comprehension gains, and students in the highest
level of materials did not demonstrate any fluency gains. It might be conjectured that the
lowest students, whose decoding skills were still at a very basic level, did not achieve the
automaticity needed to improve comprehension, while the highest students, who read an
average of 135 wpm on the pre-intervention assessment, might have experienced a
ceiling effect for fluency.
The results of these three studies, as well as the meta-analyses previously
discussed, seem to point out some important facts about fluency interventions for
secondary students: in order for students to make gains in reading comprehension, the
intervention must be of significant duration that the fluency rates are appreciably
increased and stabilized. It also seems possible to infer that the more severe the reading
delay, the longer the intervention needs to be in order to evidence comprehension gains.
In addition, all of these studies implemented programs that included other reading

62

components besides just repeated reading. It seems difficult to tease-out the efficacy of
repeated reading as an intervention technique without studies that use repeated reading as
the sole intervention, and which compare it to a control group of untreated students.
One other recent study implemented a fluency intervention with a significant
number of secondary students, however, this study did so within the context of general
education instruction in middle-school content classes (Bryant, Vaughn, Linan-
Thompson, Ugel, Hamff, & Hougen, 2000). Students were not given any individualized
interventions, but participated in classrooms where the teachers were trained to provide
fluency strategies to all students.
In this study, ten sixth-grade teachers were trained in a variety of reading
strategies, including techniques for decoding multiple syllable words, repeated reading in
expository text for fluency, and a specific comprehension strategy. The teachers were
then encouraged to apply the strategies across all the content areas in which they taught
for four months, with the goal of increasing the reading skills of all the students in their
classes. The student participants consisted of 60 sixth-graders, including14 students
identified with LD, who were identified as typical learners, low achievers, or LD. The
results of the assessments were analyzed in terms of these group differences.
All three groups increased their fluency scores significantly from pre- to post-test,
however, the students with LD were not able to significantly increase their reading
comprehension. The authors speculate that students with LD will need more time to
internalize comprehension strategies, and that many students with LD may need
individualized or small-group instruction in order to evidence significant overall reading

63

gains (Bryant et al., 2000). It seems clear from these studies that fluency interventions
are very successful at raising the fluency levels of secondary students with LD, but
individualized, long-term intervention may be needed in order for those gains to have an
effect on reading comprehension.
Small-scale studies of fluency intervention. The remainder of the recent secondary
studies used ten or fewer students in multiple-baseline interventions, and they all worked
with students who not only had reading delays, but who were also identified with either
behavioral or emotional problems. These studies seem particularly important since there
is a high prevalence of behavior problems associated with students with LD in secondary
school, and interventions designed for secondary struggling readers are likely to be
impacted by behavioral issues in some students. All of these studies involve students
working one-on-one with an adult.
In the longest of these interventions the students received instruction three times a
week for ten weeks. The study was completed with four high-school aged students with
severe reading disabilities in a residential school; all students also had behavioral
challenges. Four typical readers (also students with behavior challenges) were chosen for
comparison (Valleley & Shriver, 2003).
The intervention consisted of repeated readings of fourth and fifth grade level
passages from the Timed Reading Series. The students read the same passage until they
demonstrated at least 1 wpm increase three times in a row, then they moved on; passages
were read a maximum of ten times and a minimum of four. It should be noted that these
students were quite resistant to the intervention, and only three of them finished the

64

whole ten weeks. In addition, their progress was very slow: on many of the passages
the students were not able to achieve a three word per minute increase in fluency within
ten repetitions of the story. It is difficult to know if this was because of behavior
problems or very pervasive reading disabilities, however no comprehension gains were
seen from the study (Valleley & Shriver, 2003).
In a similar study, Alber-Morgan, Ramp, Anderson and Martin (2007) worked
with four students aged 12 to 15, all of whom were identified with behavior problems;
two of these were identified with LD and two with emotional disabilities. Their pre-
intervention reading levels ranged from grade 2 to 6.
These students worked one-on-one with an adult in ten-minute increments, daily
for six weeks. Interventions included timed reading with comprehension questions
(baseline), repeated reading, and repeated reading with prediction.
Results of informal assessment using the training passages found strong increases
in fluency and comprehension for the repeated reading, but no significant changes when
the prediction phase was added. Students were not assessed on transfer passages, so
generalization to novel text cannot be inferred (Alber-Morgan, Ramp, Anderson &
Martin, 2007).
If one conclusion can be drawn from the fluency intervention research reviewed
to this point, it is that the methodology of the research limits the ability to make
generalizations about the findings. Most studies lack control groups, the size of the
samples ranges from two or three students to large groups, and the intervention periods
vary widely from several days to several months. In addition, the concurrence of other

65

reading interventions confound the findings noticeably, and make it difficult to
attribute gains specifically to fluency programs. Clearly, the ability of fluency
interventions to have a positive impact on secondary students’ reading comprehension
can still be disputed at this point.
However, it seems that the studies that are most effective at increasing students’
reading comprehension have a few elements in common, and those elements seem to be
in concordance with the findings of the recent meta-analyses of the research (Chard,
Vaughn and Tyler 2002; National Reading Panel, 2000; Therrien, 2004): interventions
need a sizeable implementation period, they should be administered by an adult, they
should include predetermined performance criteria, and the passage levels should be
adjusted according to student progress. Although the existence of these characteristics do
not seem to guarantee the success of a study, research that is missing one or more of these
criteria consistently seems to have trouble finding a link between fluency gains and
reading comprehension. In addition, the work of Breznitz and Berman (2003) also needs
to be considered: fluency interventions must push students to their maximal levels in
order to observe an effect on reading comprehension.
One more study must be examined, as this study provides the impetus for the
current research project. In addition, the fluency intervention program used in this study,
the Great Leaps reading program (Campbell, 1995), has the potential of addressing many
of the methodological considerations raised in the previous discussions: it is designed to
be implemented one-on-one by an adult, it incorporates systematic feedback and

66

predetermined performance criteria, and students are systematically moved into more
difficult material as they experience success (Campbell, 1995).
The 2001 Great Leaps Study

In 2001, a study was published that summarized the outcomes from using a
systematic fluency intervention program called Great Leaps (Campbell, 1995) with
struggling middle-school readers (Mercer, Campbell, Miller, Mercer & Lane, 2000). The
subjects were 49 middle-school students identified with learning disabilities, and
although the specific demographics of the subjects were not given, the school was 30-
40% African American and less than 10% Latino, with the remainder of the students
presumed to be of Caucasian background. Approximately 80% of the school’s students
qualified for free or reduced-price lunch.
All students in the study were tested both pre- and post-intervention for fluency
and reading levels using curriculum-based measurement (CBM). CBM is an assessment
procedure that uses school curriculum as the tool for assessment; students were observed
and timed reading graded passages selected from the school district’s basil reading series.
Results were recorded in terms of reading speed and accuracy (Mercer et al., 2000).
This study was conducted over a period of three years, and began with a single
group of 11 sixth graders with an average reading level of lower than first grade (.63).
These 11 students participated in the intervention for the first year, and a year later 19
more were added (average reading level 1.9). During the third year another group of 19
students was added, with the average reading level for this final group being a little above
first grade level (1.53). Group one received an average of 24 months of intervention,

67

group two’s average was 15.5 months, and group three had an average of 7.2 months
in the intervention program (Mercer et al., 2000).
The students participated in three elements of fluency practice in each
intervention session: part one consisted of sublexical and lexical fluency practice
involving speeded reading of individual letters, phonics patterns such as igh or ite, and
individual word practice. Part two, also practiced daily, consisted of timed reading of
short two to four word sight phrases such as “he was” or “it could only be”. Part three of
the daily practice was oral repeated reading in connected text.
Each day, students were asked to read one page of text in parts one, two and three,
with the goal of completing each page in one minute, with less than three errors. Once the
student mastered a page, he or she "leapt" forward to a harder page for the next session.
Subsequent pages in each part got more difficult, with increased demands for both
decoding and reading speed. Adequate fluency rates at each level were determined by the
program’s designer, according to the grade level of the passage and in accordance with
researched standards (Campbell, 1995). Students’ progress on all measures was kept via a
graph, used to motivate students and document ongoing progress (Mercer et al., 2000).
The results of this study were highly encouraging, and seemed to point to the
effectiveness of this intervention for improving reading and fluency levels of middle-
school non-responding readers. Pre- and post-test CBM grade levels were compared to
assess the effectiveness of the interventions, and fluency rates on these passages were
calculated; dependent t-tests were used to analyze the results. The intervention results
were found to be statistically significant for every group for both fluency and reading

68

level change. Group one showed an average growth of 3.14 grade levels over the three
years of intervention (p=.0001, effect size 13.43), group two improved an average of 3.08
grade levels during two years (p=.0001, effect size 2.67), and group three increased 1.82
grade levels in a single year of intervention (p=.0001, effect size 2.01). Students’ fluency
rates on CBM increased 40.36 wpm for group one (p=.0001, effect size 2.42), 29.16
WPM for group two (p=.0002, effect size 1.52) and 31.95 WPM for group three
(p=.0001, effect size 1.55) (Mercer et al., 2000). It should be noted that the final fluency
levels were calculated on harder passages at students’ improved reading levels, making
these results even more notable.
In light of the theoretical foundations discussed previously, it doesn’t seem
surprising that this study evidenced substantial outcomes for the participants. Many of the
components identified in Wolf and Katzir-Cohen’s definition of fluency (2001) are
addressed in this program, including “phonological, orthographic, and morphological
processes at the letter, letter-pattern, and word levels, as well as… syntactic processes at
the word level and connected-text level” (Wolf & Katzir-Cohen, 2001, p. 219). The
phonological, orthographic and morphological components are addressed by the
sublexical and lexical elements of the program (part one), and syntactic processes are
addressed through both the sight phrases (part two) and the connected text (part three); it
seems clear that this program was designed to incorporate much of what is currently
understood about the processes underlying fluent reading. However, it should be pointed
out that one of the significant limitations of this study is that it did not assess the
students’ reading comprehension. Although the students seemed to make significant

69

reading progress, particularly in light of their excessively low initial reading rates in
middle school, it cannot be determined from this study whether they were able to
comprehend more effectively because of the increase in fluency they experienced.
Nevertheless, the results of this study plainly imply that middle-school students with the
lowest reading levels can benefit from a systematic and well-designed fluency
intervention program.
Several weaknesses to this study were pointed out by the authors, and will be
discussed here. Firstly, since the study did not incorporate a control group, it is possible
that the effects were due primarily to the one-on-one attention these students received
daily during the intervention, and weren’t caused by the intervention materials or
procedures. Secondly, the authors discuss lack of comprehension assessment as a
significant limitation of this study. Obviously, if the students’ comprehension was not
improved, the gains in fluency and reading levels may not generalize to overall improved
reading ability. Thirdly, although the vast majority of students in this study made
significant progress during the intervention, a few did not, and the authors recommend an
investigation of the specific characteristics of those students who did not make progress
(Mercer et al., 2000).
It is the goal of the current study, then, to build on the results of the Mercer et al.
study (2000) by using the same intervention program and population as their study while
remediating some of the identified weaknesses. The current study seeks to identify
whether the Great Leaps fluency intervention program, when used with seriously delayed
middle-school readers with disabilities, will increase their reading fluency and

70

comprehension. Accordingly, this study will utilize a randomized experimental design,
as well as standardized measures of reading comprehension in the pre- and post-test
assessments. In addition, this study will examine the practicality of using a one-on-one
intervention program such as this in typical urban schools.
Summary and Hypotheses
While it is clear that fluency interventions are effective in increasing the reading
rates of secondary students with learning disabilities, the results in terms of
comprehension are anything but conclusive. Repeated reading interventions and word-
level training increased students’ ability to read more quickly and accurately, but the
effects didn’t always transfer to novel text, and there wasn’t always an associated
increase in comprehension. Additionally, confounding factors such as students’ naming
ability or phonemic awareness skills may have an impact on their ability to benefit from
fluency interventions.
However, the research is very clear about one thing: more studies are needed that
utilize experimental procedures to isolate the effects of fluency interventions with large
groups of students. In addition, most researchers concur that these studies should be of a
significant length, should be implemented by adults in a one-on-one or small group
setting, and should incorporate feedback and preset performance criteria for the students.
It is the purpose of the current project to add to the current literature base by creating just
such a study. Middle-school students with learning disabilities will be randomly divided
into a treatment or a control group, and will receive either regular, daily fluency
interventions for the duration of a school year, or will receive an alternate daily lesson in

71

the same format. The results will be analyzed to determine the effects of the fluency
intervention on the students’ reading fluency and comprehension, as well as to determine
what underlying concomitant factors exist that may have impacted the students’ success.
The aim of this study, then, is to test the following set of null hypotheses:
(1) There will be no significant difference between students receiving a daily
fluency intervention and a control group on reading rates and accuracy
(defined as fluency).
(2) There will be no significant difference between students receiving a daily
fluency intervention and a control group on reading comprehension.
(3) There will be no significant difference between the fluency and
comprehension outcomes of students with poor phonemic awareness skills
and those with stronger phonemic skills.
(4) There will be no difference between the fluency and comprehension outcomes
of students with poor rapid naming skills and those with stronger rapid
naming skills.
In the following chapter, the methods designed to address these hypotheses will
be outlined, and the proposed statistical analysis will be discussed.

72

CHAPTER III: METHODS

The purpose of this chapter is to articulate the methods that were used to conduct
this study. First, the settings will be described and the participants will be defined,
including selection procedures, characteristics and demographics. Then, the assessments
will be identified, and the rationale for the selection of the assessments will be presented.
Finally, study procedures and materials will be described and the data analysis
approaches will be discussed.
Settings
The intervention study took place at two schools located on the outskirts of a large
urban region on the west coast. Both schools are part of this city’s school district, which
serves over 700,000 students, 266,000 of whom are English Learners, and 75,000 of
whom are qualified to receive services for Special Education. The transiency rate in this
district is approximately 28%. More than 33,000 certificated teachers are employed by
the district, 8,000 of whom have less than five years of teaching experience.
School one is a Title 1 school that serves grades six through eight. Enrollment for
the 2006/2007 school year (the most recent year for which information is available) was
1819, with 295 students designated as English Learners. The transiency rate was 16.5
percent, with 94.38 percent actual attendance. The school had 66 certificated teachers in
2006/2007, 19 of whom have less than five years of teaching experience, 25 of whom
have more than ten years of teaching experience.
School two is also a Title 1 school serving grades six through eight. School two
had 1526 students in 2006/2007, with 408 designated as English Learners, 92 of whom

73

were reclassified as English proficient before the end of the school year. The
transiency rate for school two was 19.92 percent, with 95.16 percent actual attendance. In
2006/2007 the school had 60 certificated teachers, only seven of whom have been
teaching for five years or less, 25 of whom have been teaching for more than ten years.
Both schools use a traditional ten-month school calendar.
Both schools have well-maintained campuses and good facilities. Although school
two is located in a neighborhood characterized by a lower overall socio-economic
condition than school one, the school is in good condition, and is well supplied with a
variety of academic resources.
Participants
Characteristics
Participants in the study were sixth, seventh and eighth graders who were
enrolled in Special Day Programs (SDPs) at these two middle school sites. SDPs are
self-contained special education classrooms created by the school district to serve
students who are believed to be unable to benefit from instruction in general education
classes due to the impact of a disability. The disabling conditions represented in these
classrooms (which are identified as being for students with mild to moderate disabilities)
include students identified with specific learning disabilities (SLD), mild mental
retardation (MR) autism (AUT), and a variety of impairments that fall under the label of
other health impairment (OHI). Students with OHI eligibilities may have Attention
Deficit Disorder (ADD), Attention Deficit Hyperactivity Disorder (ADHD), or a range of
other learning or health factors that affect the students’ ability to be successful in school.

74

Students in the SDPs were chosen for this study for a number of reasons.
Firstly, these students are usually at the very bottom of the school population in terms of
academic achievement. As a result, there was a high likelihood that the majority of these
students would qualify for participation in this study due to below average fluency rates
and poor overall reading ability. Since the earlier study on which this one was based
(Mercer, Campbell, Miller, Mercer & Lane, 2000) used students who averaged a mid
first-grade reading level in middle school, it was assumed that the students in the SDPs
were the most likely to have similarly low levels of reading.
There was an additional, more pragmatic reason for choosing these students. All
SDP classes in this school district are assigned paraprofessionals to assist in the
classrooms. Since the remediation program for this study was designed to be
administered by trained paraprofessionals under the supervision of a credentialed teacher,
using students in the SDPs required less reallocation of resources by the middle schools,
and facilitated the creation of a control group activity in the designated classrooms.
However, it is important to note that due to the large numbers of students that participated
at each school, the schools ended up assigning additional paraprofessionals to take part in
the study. At school one, six SDP paraprofessionals and 3 paraprofessionals assigned to
Resource classrooms participated in the study. Each paraprofessional worked with 3
students in either the experimental group, the control group, or both. At school two, a
variety of additional paraprofessionals were assigned to participate in the program,
including those assigned as one-on-one assistants to particular students; at this school a
total of nine paraprofessionals also participated, each working with from 3 to 5 students,

75

depending on the classroom assignment. Only one paraprofessional at school two
worked with students from both the experimental and control groups. All total, there were
19 paraprofessionals who implemented the study at the two sites.
It must be emphasized that the Mercer et al. study (2000) did not use students
with disabling conditions such as ADD, ADHD, autism or mental retardation. However,
the reality in most public schools today is that students who are performing at the low end
of the academic index and who are eligible for special education services due to reading
problems, often have other confounding issues affecting their ability to succeed in school.
In order to find out if the reading intervention is practical for real school use, it seemed
reasonable to include students with the typical range of disabling factors found in the
special education population in ordinary schools. As a result, all students in the SDP
programs identified by their teachers as potential candidates were assessed for eligibility.
Recruitment of Participants
The principals at the school sites gave permission for this study to occur at their
schools. The university’s Institutional Review Board, as well as the faculty and
administration at both schools, considered the intervention program to be standard of care
in reading instruction, and as a result, the letters that went home describing the study
were in the form of passive consent: parents were given the option to opt out of the study.
If they did not return the letter, it was assumed that they consented to have their child
participate in the fluency intervention. This method of passive consent is frequently used
in educational studies when the intervention consists of normal educational practices, as
it does in this case (Range, Embry, & Macleod, 2001). Three parents opted to withdraw

76

their children from participation in the study. In addition, an assent form was read and
discussed with each student prior to assessment; students were given the option of opting
out of the study at that time. Three additional students chose not to participate at that time
and were not assessed.
Eligibility and Selection Procedures
In order to be identified as eligible for the study, students needed English reading
fluency delays of at least three academic years. At school two all students in the SDP
classrooms who assented to participate were tested for eligibility. At school one,
however, due to unexpected time constraints, teachers made recommendations about
students who they believed would qualify as fluency delayed, and only those students
were pre-tested.
Eligibility to participate was determined by examining students’ scores on two of
the assessments: the Word Identification (WID) subtest of the Woodcock Reading
Mastery Test and the Grey Oral Reading Test (GORT). If a student scored below a raw
score of 63 (fourth-grade equivalency) on the WID, he or she was given all assessments
and was automatically included in the intervention. A score below 63 implied that the
student was reading at least two grades below grade level and would benefit from the
intervention. If a student scored above a raw score of 78 on the WID (sixth-grade
equivalency), testing was discontinued and the student was not included in the
intervention; students scoring above the sixth grade level on word identification were
considered too proficient to benefit from the fluency intervention. If students scored
between the raw scores of 63 and 78 on WID, their performance on the GORT was

77

analyzed to determine eligibility; students who fell into this group and who were able
to read the GORT passages fluently at or above grade five were excluded from
participation in the study. It was determined that the fluency intervention program was
unlikely to improve the reading of this group of students. If students in this “bubble”
group were unable to read GORT passages fluently at or above the fifth grade level they
were considered eligible for the study.
In addition, students were omitted from the study if they were unable to fulfill the
minimum requirements of any of the assessment tools used. For example, if a student
could not understand the directions, could not attend to the tasks long enough to complete
them, or could not achieve the basil for the Woodcock tests, he or she was omitted from
the study. This was done for two reasons: all students needed to be able to fulfill at least
the minimum requirements on all the assessments in order to ensure accurate assessment
and comparison of the pretest and posttest conditions. In addition, it was decided that
students with very severe receptive or expressive language problems were unlikely to
benefit from this fluency intervention, and it would not be a good use of their
instructional time. Several students were screened out of the sample through this process.
A minimum sample size of 22 was determined using Cohen's Power Tables
(1988), for a power of .80 and an estimated effect size of 1.0. The effect size of 1.0 was
chosen based on the previous study (Mercer, Campbell, Miller, Mercer & Lane, 2000),
which found an effect size of 2.01 (p=.0001) for a similar intervention with 49 students.
However, the actual sample in the current study was larger.

78

At school two, 37 students were found eligible for the study and participated in
the intervention project. Part way through the second semester, one student was
transferred to another school due to persistent behavior problems and was unable to
complete the intervention, and another student left the school. In total, 35 students at
school two completed the intervention sessions and participated in both pre and
posttesting.
At school one, 36 students were originally found eligible for the study and began
the intervention. However, after one week of implementation the paraprofessionals
complained about the additional workload, and at that point the intervention was
postponed and a new schedule was developed. Based on feedback from the
paraprofessionals, the teachers and the administration, it was decided to reduce the total
number of students involved in the intervention so that no single paraprofessional worked
with more than 3 students. Since most of the initial assessments had not been scored at
that point, they could not be used to determine which students would be removed from
the study. Instead, English teachers at each grade level were asked to look at the list of
qualified students and nominate students to be removed from the intervention. The
teachers elected to remove students who they felt were the most proficient readers, or
whose behavior and/or attendance might make it difficult for them to participate in a
meaningful way. After the reorganization, 26 students remained in the intervention at
school one, all of whom completed the intervention sessions. Overall, a total of 61
students at both sites completed the study: 34 students were in the fluency program, and
27 were in the control activity.

79

Demographics
The participants in this study were 18 girls and 43 boys in grades six through
eight, with ages ranging from 10 years 11 months to fifteen years old. 36 of the students
were Hispanic, 11 were Caucasian, 10 were African American, 3 Armenian and 1 of
Filipino heritage. 36 of the students were identified by their schools as speaking English
as their primary language, while 21 of them had Spanish listed as their home language,
and 3 spoke Armenian at home. All the participants were identified with disabilities: 53
students had special education eligibility as students with learning disabilities (SLD), two
had eligibilities of mental retardation (MR), 3 were identified as Other Health Impaired
(OHI), two students had eligibilities of Specific Language Impairment (SLI), and one
student was identified with autism (AUT). The researcher was unable to determine the
percentage of students who had comorbid conditions of ADD or ADHD as this condition
is rarely officially diagnosed in this school district. However, since the reported
comorbidity rates of LD and ADD/ADHD range from 10% to as high as 80% (Jakobson
and Kikas, 2007), and since the students in this study are those most impaired by learning
problems at their respective schools, it may be assumed that a generous proportion of the
study population had comorbid conditions of ADD or ADHD.
The distribution of the students in the two groups in terms of gender and
disabilities was relatively equal; the experimental group had 10 girls and 24 boys, while
the control group had 8 girls and 19 boys. Each group had one student identified with MR
and one with SLI, and there were two students with OHI in the experimental group and
one in the control group. However, despite random assignment, the two groups did not

80

turn out to be totally equivalent in terms of ethnicity. 41% of the students in the
experimental group were Hispanic, 24% Caucasian, 26% African American, with two
Armenian students and one Filipino. In the control group 74% of the students were
Hispanic, 19% Caucasian, with one Armenian and one African American student (see
Table 1).

Table 1
Demographic Information by Group
Instructional Condition
Exp Cont
N 34 27
Mean age 12.6 12.6
Age range 10.1 - 15.0 11.1 - 14.5
Gender ratio 10F/24M 8F/19M
Hispanic 16 20
Caucasian 7 5
African American 9 1
Armenian 1 1
Filipino 1 0
SLD 29 24
OHI 2 1
MR 1 1
SLI 1 1
AUT 1 0

Assessments
All participants were assessed before and after the study on a battery of tests that
are non-invasive and that are routinely employed in school practice to measure academic
progress of individual students. Initial assessments involved approximately 95 - 120
minutes of work per student, and were broken into two batteries which were spread out

81

over two to three test days to minimize disruption to the academic program and to
lower student anxiety. Post-intervention assessments took about 60 minutes per student,
and were completed in one sitting. Skills tested included word identification,
phonological decoding, reading comprehension, reading fluency at the word and
paragraph level, phonological awareness, verbal skills and naming speed. Assessments
were deliberately arranged in order to intersperse high-stress and lower-stress
assessments, and to vary the type of tasks within the battery. The order of assessments
was the same for all students. See Table 2 for a complete list of the assessments in the
order they were administered at the pretest.
Rationale for Choice of Assessments
The assessments for this study were chosen to measure the students’ reading
fluency and comprehension, as well as many of the underlying processes that are believed
to be highly correlated to reading success. Reading fluency was assessed using leveled
passages and standardized measures of automatic word recognition. Standardized
measures of rapid naming, phonemic decoding, verbal ability, and phonological
processing were given in order to assess students’ levels in underlying related skills
before the intervention. The measures were analyzed to evaluate changes in reading
levels after the intervention, to identify correlating factors that may relate to
responsiveness to the intervention, and for the purpose of comparing the sample to other
studies.

82

Types of Assessments
Assessments of phonological processing. Phonological processing ability has been
shown to be highly correlated to reading ability (National Reading Panel, 2000).
Assessments of phonological processing were included to measure these underlying skills
related to reading success, and were analyzed for a possible correlation to responsiveness
to the intervention.
Phonological processing was assessed using the Comprehensive Test of
Phonological Processing (CTOPP) (Wagner, Torgesen & Rashotte, 1999a). The CTOPP
assesses three kinds of phonological processing skills: phonological awareness, (the
student’s ability to distinguish and manipulate sounds within words,) phonological
memory, (the student’s ability to store and access phonological information in short-term
memory,) and rapid naming (the student’s ability to quickly and efficiently access
information from long-term memory). Based on the review of the literature, two of these
constructs—phonological awareness and rapid naming— were chosen as skills believed
to be correlated with reading fluency, and were assessed using the following measures.
The first construct, phonological awareness, was measured using the Phoneme
Elision sub-test of the CTOPP. This test measures a student’s ability to manipulate
phonemes in isolation from text. In the Phoneme Elision test, students are asked to repeat
a word, and then say the word again with one phoneme deleted. For example, students are
asked to say mike, then mike without the /k/.
The second construct, rapid naming, was measured using the Rapid Letter
Naming sub-test of the CTOPP. In this test, students are asked to say the names of a

83

series of lower-case letters as quickly as they can. It is scored for both speed and
accuracy.
The CTOPP has been shown to have a high degree of reliability for test/retest,
interscorer and content sampling: 56 out of 69 samples had reliability over .76, with no
sample under .67. It has also evidenced high construct validity; when constructs were
measured using a Comparative Fit Index, the model earned a .99 out of a possible 1.0,
supporting the construct validity of the test.
Assessments of word recognition. Two tests were used to examine students’
ability to read real single words automatically and efficiently: the Sight Word Efficiency
subtest of the Test of Word Reading Efficiency (TOWRE) (Wagner, Torgesen &
Rashotte, 1999b) and the Word Identification sub-test of the Woodcock Reading Mastery
Test—Revised/Normative Update (Woodcock, 1996). The TOWRE was used to assess
students’ ability to rapidly read words under timed conditions, while the Word
Identification test was used to assess students’ ability to recognize common lists of words
in an untimed format.
The TOWRE is a nationally normed test of word reading accuracy and fluency. It
was normed on 1,507 people in 30 states, approximately ten percent of whom had
identified disabilities. The TOWRE subtests and alternate forms have test/retest
reliability of between .83 and .92 for students between the ages of 10 and 18, an
interscorer reliability of .99, and content reliability higher than .93 for all subtests. When
correlated to the Woodcock Reading Mastery Test (Revised) correlations were between
.89 and .94, indicating strong concurrent validity.

84

The Sight Word Efficiency subtest of the TOWRE asks students to read a list
of real words of increasing difficulty, timing them to see how many words they can read
in 45 seconds. Words range in difficulty from go and dog to selection and grandiose. The
subtest has two forms, each of which has 104 items. As recommended by the authors of
the test, students were given both forms of the assessment at each administration, and the
scores were averaged to give a more reliable measure of ability.
The second measure of word reading was the Word Identification subtest of the
Woodcock Reading Mastery Test—Revised/Normative Update (WRMT-R/NU). The
WRMT-R/NU is a battery of standardized reading tasks that assess many of the functions
related to reading success. Sub-tests of the WRMT-R/NU are commonly used in schools
and in research studies to measure decoding, automatic word recognition and
comprehension abilities. The WRMT-R/NU was renormed in 1995/1996 with a
population of approximately 3,700 people. The sample included a representative
proportion of students with disabilities as specified by 1994 census data. Internal
consistency ranges from .68 to .98, with a median of .91 for all subtests. Split-half
consistency ranges from .86 to .99. Concurrent validity was established with the
Woodcock-Johnson Psycho-Educational Battery, the Iowa Test of Basic Skills, the
Peabody Individual Achievement Test, and the Wide Range Achievement Test.

85

Table 2
Assessment Batteries in the Order Administered at the Pretest
Battery One
Assessment Reading Construct
Woodcock Reading Mastery Tests—
Revised/Normative Update (WRMT-R/NU)
Word Identification subtest*

WRMT-R/NU
Word Attack subtest*

Gray Oral Reading Test (GORT)*

WISC IV
Similarities
Vocabulary
Word recognition

Phonemic decoding

Reading fluency

Verbal ability
Battery Two

WRMT-R/NU
Passage Comprehension subtest*

Test of Word Reading Efficiency (TOWRE)
Sight Word Efficiency subtest*
Phonemic Decoding Efficiency subtest*

Comprehensive Test of Phonological
Processing (CTOPP)
Rapid Letter Naming subtest
Phoneme Elision subtest

Reading Comprehension

Word recognition fluency
Phonemic decoding fluency

Rapid automatic naming
Phoneme manipulation
* Assessment repeated after the intervention

86

The WRMT-R/NU Word Identification subtest was used to provide another
measure of students’ ability to read common words, a skill highly correlated to reading
fluency. It requires students to read aloud from a list of everyday words of increasing
complexity, ranging in difficulty from words such as stop and play to naïve and
carnivorous. The WRMT-R/NU Word Identification Test is not timed, however, if a
student does not respond within five seconds he is given a zero and prompted to read the
next word.
Assessments of phonemic decoding. Two tests of phonemic decoding were
administered. The first was the Phonemic Decoding Efficiency subtest of the TOWRE,
which asks students to quickly decode a list of nonwords of increasing difficulty.
Students are given 45 seconds to decode as many nonwords as possible from the list of 63
items. Items range from nonwords such as ik and pog to stremfick and revignuf. There are
two forms of the subtest available, and the test-retest reliability for students aged 10 to 18
is .89. As with the Sight Word Efficiency subtest, both forms of the test were given at
each administration, and the scores were averaged.
The second decoding assessment used was the Word Attack subtest of the
WRMT-R/NU. The tasks in this test are similar to those of the TOWRE Decoding
Efficiency test, but the tasks are untimed. Nonwords range from simple items such as dee
to more challenging items such as gnouthe. Split-half reliability of the Word Attack
subtest for students in 8
th
grade is .90. Both of these assessments are widely used in
research studies as reliable and sensitive measures of students’ abilities to use phonetic
decoding skills.

87

Assessments of reading comprehension. Since the ultimate goal of any reading
intervention must be for students to understand what they are reading, it is critical to
assess participants’ reading comprehension. For this purpose, the students in this study
were asked to take the Passage Comprehension subtest of the WRMT-R/NU. This subtest
uses a modified cloze procedure, requiring students to read a sentence or short passage
and to fill in a missing word appropriately. An example of a test item on the Passage
Comprehension test is “A map shows where places are. It helps you know how to get to
those __________.” Students are asked to say a word that logically fits in the blank
space. The manual of the WRMT-R/NU states that the tasks were designed to “avoid
confusing general passage comprehension skills with specialized vocabulary skills”
(Woodcock, 1996, p. 8), and they provide a quick and standardized measure of students’
ability to read and comprehend short, self-contained passages. Split-half reliability for the
Passage Comprehension subtest is .92 for grade eight, and it has been used in a large
number of studies as a reliable measure of reading comprehension.
The Gray Oral Reading Test (GORT, Wiederholt & Bryant, 2001) also includes a
standardized measure of reading comprehension, however, a recent study has called into
question the validity of this type of multiple-choice measure for passage comprehension
(Keenan & Betjemann, 2006). The authors of this study found that a large percentage of
students could use their background knowledge to correctly guess the answers to many of
the multiple-choice questions on the GORT without reading the passages. As a result, for
this study it was decided to rewrite the questions on the GORT passages as open-ended
questions, rather than use the multiple-choice questions provided by the authors of the

88

test. Every effort was made to ensure that the open-ended questions could not be
answered through the use of background knowledge, and several adults were asked to
look at them to verify that the answers couldn’t be intuited from commonly known
information. A variety of types of questions were created, including literal/factual
questions, questions that required the students to use inference, and questions that
required students to summarize or categorize information. A complete list of the
questions used is included in Appendix L.
In order to get a clear comparison of students’ change on these rewritten
questions, the same version of the GORT was given at both the pre- and post-tests, and
the students’ answers were compared for improvement on the number of correct answers
before and after the intervention. Ratios of number of questions answered correctly per
passage were calculated. If significantly more students in the experimental group
improved their ability to answer the open-ended questions on the post-test, it might be
inferred that the improvement was due to the influence of the intervention.
Since these open-ended comprehension questions had not been validated previous
to this study, correlations were conducted looking for relationships between the GORT
comprehension growth scores and the growth scores on the other reading measures. A
small but statistically significant positive correlation was found between the GORT
comprehension scores (as measured by the ratio of questions answered correctly per
passage) and growth on Woodcock Comprehension subtest (.264, p=.042). This suggests
that the comprehension questions as rewritten may be measuring the same construct as
the Comprehension subtest of the Woodcock assessment.

89

Assessments of reading fluency. Reading fluency was assessed in a variety of
ways. The Rapid Letter Naming Test discussed above is an assessment of students’ letter
naming fluency, and the TOWRE Phonemic Decoding Efficiency test measures fluency
in nonsense word decoding. The TOWRE Sight Word Efficiency test is not only a test of
word recognition, but it also tests fluency at the word level. All of these tests have been
discussed above.
For passage level fluency, the students were assessed using the Gray Oral
Reading Test (GORT). The format of the GORT is very similar to typical classroom
tasks. The GORT asks students to read aloud from one of 14 developmentally sequenced
passages. Students continue on to sequentially harder passages until a ceiling is reached
for accuracy and fluency. In order to obtain a standardized fluency score, students are
timed as they read, and a fluency score is calculated from the number of words read
correctly per minute on each passage.
For this study, all students started reading on the first passage of the GORT, and
continued on until they reached a ceiling for the passage score (accuracy plus rate).
Comprehension was not used to help determine the ceiling.
The GORT was normed on more than 1,600 individuals, including a
representative sample of people with disabilities. Reliability coefficients are reported at
.90 or higher, and the authors also report high test-retest, alternate forms and inter-scorer
reliability (Wiederholt & Bryant, 2001).
Assessments of language ability. General language assessments were used to
identify students’ underlying strengths and deficits in these reading-related skills, as well

90

as possible correlations with responsiveness to this intervention. Two subtests of the
Wechsler Intelligence Scale for Children (WISC IV, Wechsler, 2003) were administered
to assess students’ overall language ability: the Vocabulary subtest and the Similarities
subtest. These subtests are designed to assess a student’s oral language skills, which are
highly correlated to reading success (Prifitera, Saklofske, & Weiss, 2005). The
Vocabulary subtest asks a student to supply a verbal definition of a word or concept. For
example, students were asked “What does brave mean?” Responses were scored 0-1-2
according to quality. The Similarities subtest asks students to explain how two different
objects (e.g. horse and cow) or concepts (e.g. hope and fear) are alike. Scoring is 0-1-2
according to the quality of the response. Split-half reliability for the subtests of the WISC
IV is as follows. Vocabulary: .87; Similarities: .81. These reliability figures have been
tested with both typical and special populations, and the WISC IV is a commonly used
measure of students’ verbal and nonverbal ability in a variety of research and educational
studies.
Retesting Procedures
Typically, researchers use alternative forms of assessments (if available) when
conducting post-testing during intervention studies. However, in the present study the
choice was made to use the same form of the GORT in order to compare the number of
correct answers on the open-ended questions we created. In order to maintain
consistency, then, the same form of each assessment was used at both the pre- and post-
testing.

91

A recent study supported the efficacy of this practice in reading assessments for
students with reading disabilities. Cirino et al. (2002) examined the practice effects of
reassessment on the same form of reading tests with students with reading disabilities,
and found that:
…the reading performances of students with RD do not change
significantly over time, at least when they are not attending school; there
is no indication that practice effects operated as a whole… These results
indicate that reading and reading related performance do not improve
based solely on the passage of time, or with repeated exposure to reading
lists or related stimuli. Therefore, if a significant and substantively
meaningful change were to occur over time on these measures, this change
could likely be attributed to the effectiveness of the instruction received by
the children (Cirino et al., 2002, p. 536).

Furthermore, since the current study uses a randomized experimental design, it may
be assumed that any differential finding between the assessment results of the
experimental and the control group is attributable to the effects of the intervention,
despite the form of assessment used. In discussing the consequences of test-retest effect
on measurement of treatment outcomes, McArthur (2007) proposed that “…a randomized
control trial is the best tool we have to test whether a treatment has a real effect on a
disease or disorder over and above confounds such as placebo effects, Hawthorn effects,
test–retest effects, and regression-to-mean effects” (p. 251).
Study Procedures
Students from the SDP classrooms were assessed as outlined above at the
beginning of the study, and those meeting the eligibility criteria were chosen to
participate in the study. Participants at each school were randomly assigned to one of two
groups, experimental or control, and the students were then clustered and assigned to a

92

paraprofessional for the intervention. The assignment of the paraprofessionals was
done according to the schedules of the students and the classrooms.
The experimental group participated in the Great Leaps fluency intervention
program (outlined below) for ten minutes a day (Campbell, 1999). The control group
participated in a similar 10-minute intervention, utilizing the Skills for School Success
study skills program (described below) (Archer & Gleason, 2002). Students at school two
participated in interventions from October through April, and at school one the
interventions went from November to May. Although the study was designed for all
students to complete six months of intervention, (allowing for holidays and breaks,) and
to participate in their assigned program four times a week, in reality the implementation
was inconsistent. That will be discussed in more detail in the subsequent chapter.
Intervention Activities
The experimental group. The fluency intervention program that was used was
Great Leaps Reading, developed by Kenneth U. Campbell (1999). It consists of fluency
activities at three levels: sounds or individual words (Appendix A), short sight phrases
(Appendix B), and connected text (Appendix C). Every student read one page at each
level every day. Students were placed in the program through a simple assessment
process that was conducted at the first few sessions, and which is outlined in the program
guidelines. At each subsequent 10-minute session, students worked with a trained
paraprofessional, being timed on repeated readings of text at each level. Students were
timed on each fluency sheet only once per day, but they were allowed to practice the

93

sheets after their session as recommended by the program. In actuality, very few
students chose to practice the interventions outside the required 10-minute sessions.
The Great Leaps Reading Program is based on research that emphasizes the
importance of practice in the skills underlying fluent reading, such as phonetic decoding,
rapid recognition of common sight words, and repeated reading to build automaticity
(Mercer, Campbell, Miller, Mercer & Lane, 2000). All subskills were practiced in a timed
format, and students graphed their progress on an equal-ratio chart that shows
incremental units of progress (Appendix D). This type of charting has been shown to
increase the academic progress and attentional behaviors for students with learning
disabilities (Shimabukuro, Prater, & Jenkins, 1999). As students experienced success on a
worksheet, they “leapt forward” to a higher level, thus tracking their own progress as they
moved into progressively more difficult material. Great Leaps passage content is chosen
to be motivating to students, and passages are based on common interpersonal language
to make them accessible to students with reading and vocabulary delays (Campbell,
1999). As the students advanced through the program they were moved into
progressively harder material written at higher grade-levels, and expected to gain fluency
at each developmental level.
The control group. Students in the control group also received ten minutes per day
of individual, one-on-one intervention with a paraprofessional. However, these students
were working on developing their general classroom and study skills by using the Skills
for School Success program (Archer & Gleason, 2002).

94

The Skills for School Success program is a research-based program designed
to increase students’ abilities to use their textbooks, take tests, keep track of their
assignments and study (Archer & Gleason, 2002). Students who were in the control
group worked through this program, reading items in a workbook and filling in answers
with the support of the paraprofessional. A sample workbook page can be seen in
Appendix E. Paraprofessionals were asked to keep track of their intervention sessions
using a daily record sheet (Appendix F).
Based on the interest levels of the materials as well as the reading levels of the
students, the fourth grade edition of the Skills for School Success program was originally
chosen for the control group. However, as the study progressed and the students got
further into the workbook, it was found that the reading level of the materials was much
too difficult for some of the students, and the paraprofessionals were given the option of
moving into third grade materials instead. They were encouraged to use their judgment to
decide which level of materials to use with which students.
Training of the Paraprofessionals
Paraprofessionals were trained on both programs at the beginning of the school
year by the primary researcher. Additionally, supplementary trainings were held at each
school right before the interventions began, and the researcher was also available to
answer questions and give guidance for the first few weeks of the program. Several
paraprofessionals chose to get additional one-on-one support at the beginning of the
study. After a few weeks had passed, each paraprofessional was observed by the

95

researcher, and feedback on implementation was given to ensure consistency of
implementation.
Paraprofessionals were asked to use their skills to solve problems as they
occurred, but were also given the phone number of the researcher in case of problems that
couldn’t easily be resolved. Support was given over the course of the intervention on a
variety of issues, including lack of time in the classroom to implement the program, and
problems with individual students. Most problems with students were resolved quickly,
however one student consistently caused trouble during the intervention sessions.
Consideration was being given to dropping him from the study, but he was expelled from
the school before the decision was made.
Data Analysis
Data was analyzed using correlational statistical procedures, ANOVAs and t-tests
for comparison of group means. First, independent group t-tests were used to determine
the experimental and control groups’ equivalencies on all measures. Next, paired-sample
t-tests were used to examine reading growth from pre to post-intervention within each
separate group.
Residualized gain scores (RGS) were calculated for each student to gauge reading
change, and independent group t-tests were conducted to see if there were statistical
differences between the mean RGS of each group on reading outcomes. Correlations
compared pretest levels of phonological processing ability, verbal ability, and letter
naming speed to the reading RGS of the students in order to identify potential
relationships between these theoretically important factors and the reading outcomes.

96

Finally, individual groups of students were examined to determine if any particular
group of skills or attributes contributed to students’ responsiveness to the intervention.
In the following chapter the results of these analyses will be presented.
Subsequently, in the final chapter these results will be discussed in detail, and analyzed in
relationship to the research questions and the existent literature base.

97

CHAPTER IV: RESULTS
The purpose of this chapter is to report the results of the data analysis as described
in the previous chapter. First, a summary of the research methods will be presented, then
descriptive statistics will be imparted, and will be examined to look for patterns and
potential irregularities. Next, results of group pretest comparisons will be studied for
group equivalencies, and data on the reliability of the implementation will be examined to
look for relationships between the way the interventions were conducted and the student
outcomes.
Comparisons of group means will be introduced in two parts: data from the
TOWRE and Woodcock assessments will be presented first, followed by data from the
GORT measures. Both sets of data will be examined in two ways—first through within
group, pre/posttest analyses, and then through comparisons of group means. Following
the group comparisons, descriptive analyses of GORT comprehension data will be
reported, and correlational data will be explored to investigate potential relationships
between student characteristics and student outcomes. Finally, investigations related to
phonological processing and RAN will be presented. The chapter concludes with a
summarization of the findings in terms of the research questions.
Summary of the Research Methods
This study used a randomized experimental design to explore the effects of a
fluency intervention program on the reading outcomes of students with significant
reading disabilities. Students were pretested on a number of measures of reading skill that
included two subtests of the Test Of Word Reading Efficiency or TOWRE (Phonemic

98

Decoding Efficiency and Sight Word Efficiency), three subtests of the Woodcock
Reading Mastery Test (Comprehension, Word Identification and Word Attack), and the
Gray Oral Reading Test (GORT). The GORT produces three sets of scores: Rate,
Accuracy and Passage.
Additionally, students were pretested on two measures of general language ability
using two subtests of the Wechsler Intelligence Scale for Children (Vocabulary and
Similarities), and two tests of phonological processing using the Comprehensive Test of
Phonological Processing (CTOPP). The two subtests of the CTOPP given were Phoneme
Elision and Rapid Letter Naming.
After pretesting, eligible students were randomly assigned to two groups. The
experimental group received daily ten-minute sessions in a fluency intervention program
entitled Great Leaps (Campbell, 1999). The program was administered by trained
paraprofessionals who worked one-on-one with three to five students daily. Students in
the control group also worked one-on-one with a trained paraprofessional daily for ten
minutes, using a study skills program entitled Skills for School Success (Archer &
Gleason, 2002).
At the end of the intervention period, which was approximately 6 months,
students were reassessed on the GORT, Woodcock and TOWRE measures, and results
were analyzed to look for changes in reading proficiency.

99

Evaluation of Data Characteristics
Missing Data Points
Prior to any analyses, the data were examined to look for missing values. In terms
of the reading and reading-related measures, only three scores out of 793 were missing.
One student was found to be missing two data points from the pretesting, Comprehension
and Sight Word Efficiency, and another student was missing the pretest administration of
Phonemic Decoding Efficiency. These three scores were omitted during analyses using a
pairwise deletion; since no multiple regressions were used in the analysis, and these data
points are missing completely at random (Howell, 2007), using a pairwise deletion on
these three points is unlikely to have an effect on the statistical outcomes of the study.
More significantly, however, multiple subjects were found to be missing data
related to study implementation. Due to poor record keeping by some of the
paraprofessionals, five subjects were missing data for Sessions Per Week, and four more
subjects were missing data on both Number Of Sessions and Sessions Per Week. In all,
13 data points out of 122 related to study implementation are missing. To compensate for
the missing data, these implementation factors were also analyzed using pairwise
deletion, but analyses concerning the effects of implementation obviously must be
interpreted with caution.
Examination of the Woodcock and TOWRE Reading Data
Histograms and boxplots were generated of all the pretest and posttest reading
scores to examine the data for normality and outliers. The data from the two TOWRE
measures (Phonemic Decoding Efficiency and Sight Word Efficiency) and from the three

100

Woodcock reading tests (Comprehension, Word Attack and Word Identification) were
for the most part normally distributed, with some tendency for mild skewness. The
control group showed a tendency for more irregularity than the experimental group on
most measures.
Boxplots for these five measures, Phonemic Decoding Efficiency (PDE), Sight
Word Efficiency (SWE), Comprehension (COMP), Word Attack (WA) and Word
Identification (WI), revealed a few outliers, but no extreme outliers. Outliers are defined
as data points lying between 1.5 and 3 times the interquartile range outside the box
(National Institute of Standards and Technology, n.d.). Boxplot data points falling outside
the box by more than three times the interquartile range are labeled as extreme outliers.
None of these five tests showed abnormal distribution or extreme outliers on either the
pre or posttest data that indicated any cause for concern about the reliability of the data.
Examination of the GORT Reading Data
Conversely, both pre and post test data from the GORT assessment showed
extreme skewness in all three measures, with multiple outliers. The data for the
experimental group was skewed and had several outliers, but none of the outliers were
flagged as “extreme”. However, the control group data was even more skewed, and
generated seven extreme outliers in the pre and post measures.
When data is affected by extreme outliers, statistical measures that rely heavily on
an assumption of normality, such as the t-test or ANOVA, may produce a bias of
estimates and/or p-values (Cohen & Cohen, 1975; High, 2000). Randomly discarding
outliers is not recommended, though, as it may distort the outcomes and misrepresent the

101

true nature of the data. Obtaining a true statistical picture of the data seemed to be a
concern with the GORT results because of the large number of extreme outliers generated
in the control group.
In order to learn more about these unusual cases, all of the extreme outliers were
carefully examined to look for irregularities in scoring or distinguishing characteristics
that may have caused these scores to fall so far outside the norm (High, 2000; Howell,
2004). In four of the six subjects no atypical factors were identified, and these four
subjects’ scores remained in the sample. Although the scores fell outside the norm for the
group, they seemed to be accurate, and to fairly represent the range of students in the
control group.
However, two students’ scores repeatedly showed up as extreme outliers in both
the pre and post test data, and irregularities were found with these two students that
caused concern. One student had been flagged during pretesting as a student whose
reading might be too high to be included in the study, but the follow up evaluation that
should have occurred during pretesting was somehow overlooked. While investigating
the outliers this student’s initial test scores were reexamined, and it was determined that
she shouldn’t have qualified for inclusion in the study. After careful consideration, this
student’s scores were removed from the data; it is believed that she is not representative
of the study population, and the inclusion of her scores would most likely cause
misinterpretation of the data. For the remainder of this chapter, all results reported will be
for the remaining 60 subjects, with this student’s scores omitted.

102

The second student of concern was also flagged repeatedly as an extreme
outlier in the posttest data. When his assessments were closely examined, it was
discovered that his pretest GORT scores were minimized due to the repetition of an entire
sentence during the assessment. (Scoring rules of the GORT require repeated words to be
counted as accuracy errors—the repetition of an entire sentence can radically diminish a
student’s scores, but may not be representative of his or her true reading potential.) In this
study, all GORT tests were scored twice; once by the tester during the assessment, and
again by the researcher to guarantee consistency of scoring. When this student’s GORT
assessments were rescored, his standard scores were lowered from 5 for Rate (RA), 8 for
Accuracy (ACC) and 7 for Passage (PAS) in the initial scoring to 4, 3, and 3 because of
the repetition of the sentence. This caused the student’s posttest scores to seem highly
inflated by comparison, and the student was flagged as an extreme outlier.
In order to remediate the effects of this second exceptional case, the GORT data
was analyzed both with and without this students’ extreme scores, and will be reported
both ways. Tables reporting GORT scores within the body of this chapter will be
presented with the outlier removed, as this is believed to be more representative of the
true nature of the data. Tables giving results with the extreme outlier included are
presented in the Appendices.
Pretest Equivalencies
Mean scores on pretest assessments were compared across groups to check for
group equivalency. Groups were found to be equivalent on all pretest measures with

103

equal variances between groups (p>.05, see Table 3). Students were also compared by
school, and were found to be equivalent on all pretest measures (p>.05).

Table 3
Comparison of Pretest Means for Experimental and Control Groups
Measure Exp.
Mean

SD
Cont.
Mean

SD
Sig.
(2-
tailed)
1. Elision 5.26 1.9 5.12 2.4 .791
2. Rapid Letter Naming 6.47 0.4 6.73 0.4 .647
3. Similarities 6.44 3.5 5.50 2.8 .262
4. Vocabulary 5.12 0.4 4.15 0.5 .132
5. Comprehension 72.09 1.4 71.23 1.8 .708
6. Sight Word Efficiency 77.36 1.3 77.08 1.6 .891
7. Phonemic Decoding Efficiency 72.00 1.2 71.72 1.6 .887
8. Word Identification 75.32 1.4 72.58 1.7 .221
9. Word Attack 84.09 1.5 79.62 1.9 .066
10. GORT Rate 1.50 0.2 1.58 0.2 .778
11. GORT Accuracy 1.38 0.1 1.38 0.2 .993
12. GORT Passage 1.59 0.2 1.58 0.2 .963
*Significant, p<0.5, equal variances not assumed
Measures 1 - 5 use a scaled score with a mean of 10 and a standard deviation of 3.
Measures 6 - 9 are standard scores, with a mean of 100 and a standard deviation of 15.
Measures 10 – 12 are standard scores with a mean of 10 and a standard deviation of 3.

Next, ANOVAs were used to explore the effects of gender, ethnicity, home
language, school and type of disability on the students’ scores. No significant main
effects were found on the students’ reading outcomes for any of these characteristics
(p>.05).

104

Examination of Influences on Study Implementation
Three conditions were identified as factors related to study implementation:
Number of Sessions, Sessions per Week, and Paraprofessional. These three factors were
examined to both assess the consistency of the intervention, and to look for possible
relationships between these implementation factors and student outcomes.
Implementation of the intervention programs varied considerably from the
original design. Although the paraprofessionals were expected to work with the students
four days a week, the realities of school life, including student absences, paraprofessional
absences, assemblies, classroom tests and a myriad of other factors, prevented this from
taking place. While some paraprofessionals were making every effort to complete three
or four sessions with each student each week, others had weeks where only one session
per student was completed. Over the course of the school year it became apparent that
there was a wide discrepancy in the number of intervention sessions being completed for
each student, so a goal was set for the participants: paraprofessionals were asked to try to
complete 60 sessions with their students before the scheduled posttest dates.
The actual number of sessions completed by all but one student ranged from 33 to
65; remarkably, one student completed 99 sessions. The average number of sessions
completed was 53.2 across all students. Approximately 47 out of the 60 students
completed more than 50 sessions. The average number of sessions per week ranged from
2.1 to 4.1, with the overall group average being 2.7 sessions per week; sessions were
spread out over 17 to 24 weeks. Notably, the record keeping system for the control group
was not as explicit as that of the experimental group, and as a result accurate numbers of

105

hours and sessions were not always available for this group. In addition, two
paraprofessionals lost their records for some or all of the year, resulting in the missing
data points discussed above.
Because of this large discrepancy in implementation, correlations were run to look
at the relationship between the total number of sessions each student completed, the
average number of sessions they completed per week, and their outcomes on the reading
measures. In the experimental group, a significant correlation was found on only one
reading measure: Sessions Per Week was positively correlated with PAS outcomes
(p=.025). In the control group no significant correlations were found, however, as
mentioned above, it is important to interpret the control group findings with caution due
to the poor record keeping of some of the paraprofessionals in this group. No significant
differences were found between the experimental and the control group on either of these
implementation measures.
In addition to the correlations mentioned above, an ANOVA was run to look at
the effect of Paraprofessional on the reading outcomes. No significant main effects were
found for Paraprofessional on any reading outcome measures. To explore this in more
depth, means plots comparing each reading measure by Paraprofessional were generated.
Examination of these means plots revealed widely varying mean outcomes for the various
paraprofessionals on each measure, with no systematic patterns related to any individual
paraprofessional.

106

Reading Outcome Data
As mentioned above, the reading outcomes will be presented in two parts. First,
results of intra-group and inter-group comparisons for the TOWRE and Woodcock
measures will be examined. Subsequently the GORT measures will be similarly
examined and discussed. Effect sizes for all findings were calculated using Cohen’s-d
procedures (Cohen & Cohen, 1975).
Results of TOWRE and Woodcock Measures
Table 4 presents the pre and posttest reading standard scores for both groups of
students on the TOWRE and Woodcock assessments. Paired sample t-tests were
conducted to compare the pre and posttest scores within each group. As shown in Table
4, the experimental group made statistically significant growth on SWE (p=.002) and
PDE (p=.002) from the pre to the post assessment with small effect sizes of 0.29 and
0.30. This finding is of particular interest as these are measures of fluency at the phoneme
and word level. The other three measures, COMP, WID and WA, which are not measures
of fluency, showed no significant mean improvement for the students in the experimental
group (p>.05).
The control group also made statistically significant progress on SWE (p=.031)
from pretest to posttest, although the effect size is small. The control group showed no
significant outcomes on any of the other measures (p>.05).

107

Group
Experimental Control
Measure
Pre Post t ES Sig.
(2-tailed)
Pre Post t ES Sig.
(2-
tailed)
Comprehension
(COMP)
72.09 72.50 .263 0.03 .476 71.23 71.69 .454 0.05 .384
SD
8.2 8.0 9.3 9.7
Sight Word
Efficiency
(SWE)
77.36 79.82 3.42 0.29 **.002 77.08 78.69 2.28 0.20 *.031
SD
7.7 7.9 8.2 8.1
Phonemic
Decoding
Efficiency
(PDE)
72.00 74.26 3.43 0.30 **.002 71.72 71.04 -.624 0.08 .538
SD
7.2 7.8 7.8 8.88.1 10.1
Word
Identification
(WID)
75.32 74.47 -1.17 -0.11 .794 72.58 71.69 -.994 -0.16 .654
SD
8.5 7.2 8.7 9.7
Word Attack
(WA)
84.09 83.29 -0.72 -0.09 .252 79.62 78.62 -.886 -0.10 .330
SD
8.8 8.5 9.6 10.3
Table 4

Comparisons of Pre And Posttest Mean Reading Standard Scores Within Groups, TOWRE and Woodcock Measures
* Significant p<.05 **Significant p<.01

108

After considering the change within each group, means of the experimental and
control groups were compared to look for differences between the two groups. This was
done by calculating residualized gain scores from the pre and posttest scores on each
measure using a linear regression model; these scores differ from regular gain scores in
that they measure each subject’s distance from the group regression line rather than from
his or her own initial score. Residualized gain scores reduce the possibility of multiple
errors when using data collected at different time points, and mediate the effects of
regression toward the mean generally seen in intervention data (Hand & Taylor, 1987).
Using the residualized gain scores (RGS), independent group t-tests were
conducted to examine the difference in the mean RGS for each group. Table 5 shows the
results of these t-tests on the TOWRE and Woodcock measures.
As can be seen in Table 5, when comparing means of the RGS between the
experimental and control groups, the experimental group made significantly more
progress on the PDE measure than the control group (p=.025) with a moderate effect size
of 0.41. The rest of the scores of the experimental group are not statistically different
from those of the control group on these five measures.

109

Table 5
Between-Group Comparisons Of Mean Residual Gain Scores
Measure Group N Mean SD t ES Sig.
(2-
tailed)

PDE Experimental 34 .273 0.80 2.325 0.41 *.025
Control 26 -.091 0.98

SWE Experimental 33 .085 1.02 .670 0.18 .506
Control 26 -.091 0.98

WA Experimental 34 .058 1.01 .659 0.17 .512
Control 26 -.113 .97

WID Experimental 34 .088 .73 .928 0.23 .391
Control 26 -.152 1.26

COMP Experimental 33 -.055 .93 -.069 0.02 .946
Control 26 -.038 .95
*Significant p<0.05, equal variance not assumed

Results of GORT Measures
Tables 6 and 7 show the mean pre and posttest standard scores and grade level
equivalencies for the GORT measures. In Table 6 it can be seen that the experimental
group had statistically significant within-group increases on the standard scores of all
three GORT measures (p<.001), with moderate to large effect sizes. The control group
also had increases on all measures, but not to the same level of statistical significance.
Comparing within-group pre and post test scores on grade equivalencies (Table
7), the experimental group again had statistically significant increases on all the measures
(p<.001), with moderate effect sizes. The control group showed no significant gains in
grade level equivalency scores on any measures. It should be noted again that all results

110

Table 6
Comparisons of Pre And Posttest Mean Reading Standard Scores Within Groups, GORT

*Significant, p<.001 **Significant, p<.01 ***Significant, p<.05 (2-tailed)
All GORT measures use a scaled score with a mean of 10 and a standard deviation of 3.

Table 7
Comparisons of Pre and Posttest Mean Grade-Level Equivalencies Within Groups,
GORT
*Significant, p<.01 (2-tailed)

Group
Experimental Control
Measure Pre Post t ES Sig. Pre Post t ES Sig.
Rate 2.09 2.87 4.73 0.59 *.000 2.16 2.75 4.20 0.56 **.001
SD 1.1 1.5 1.0 1.1
Accuracy 1.62 2.47 4.80 0.70 *.000 1.52 1.78 1.72 0.21 .148
SD 1.0 1.4 1.4 1.1
Passage 1.79 2.69 5.68 0.71 *.000 1.63 2.02 2.46 0.31 ***.040
SD 1.0 1.5 1.1 1.4
Group
Experimental Control
Measure Pre Post t ES Sig. Pre Post t ES Sig.
Rate 1.50 2.29 4.73 0.64 *.000 1.48 1.80 1.62 0.27 .119
SD 0.9 1.5 1.1 1.3
Accuracy 1.38 2.06 3.44 0.57 *.002 1.32 1.48 1.28 0.16 .212
SD 0.8 1.5 1.0 1.0
Passage 1.59 2.26 3.92 0.57 *.000 1.52 1.72 1.31 0.19 .203
SD 0.9 1.4 1.0 1.1

111

reported in this section are analyzed with one outlier removed, as discussed above.
Results of GORT analyses with the outlier replaced can be seen in Appendices G and H.
Residualized gain scores were again used to compare gains made in the
experimental group to those made in the control group. Table 8 presents the results of
these comparisons for the GORT data. It can be seen from this table that with one
extreme outlier removed, the experimental group made statistically significant gains in all
three GORT measures when compared to the control group with moderate effects sizes of
0.59 to 0.62. Results with the outlier replaced can be seen in Appendix I.

Table 8
Comparison Of GORT Mean Residual Gain Scores by Group
Measure Group N Mean SD t ES Sig.
(2-tailed)
RA Experimental 34 .155 .94 2.22 0.59 *.030
Control 25 -.373 .85

ACC Experimental 34 .223 1.11 2.41 0.62 **.019
Control 25 -.353 .72

PAS Experimental 34 .150 .87 2.34 0.61 *.023
Control 25 -.389 .89

*Significant p<.05, equal variance assumed
**Significant p<.05, equal variance not assumed

Descriptive comprehension data from the GORT assessment. In addition to the
Woodcock Comprehension test, students’ reading comprehension was also measured
using the GORT passages. The GORT passages are designed to be administered with
multiple-choice questions that produce a standard score for comprehension along with the

112

scores for rate, accuracy and overall passage fluency. However, a study published in
2006 called into question the validity of many of the GORT comprehension questions,
finding that most of them could be answered by children who haven’t read the passages
at rates well above chance (Keenan & Betjemann, 2006). As a result, the present study
used a set of open-ended comprehension questions designed by the researcher.
Comprehension growth on these new questions was calculated by giving the same
version of the GORT at both test administrations, then comparing the actual number of
questions answered correctly between administrations. Paired sample t-tests were
conducted to check for significant change from pre to posttest scores.
Table 9 shows the results of these computations. The increase in the number of
passages read (PR) and the number of questions answered correctly (QC) on the GORT
measure before and after the intervention was highly significant for both groups, with
moderate effect sizes. The final columns on Table 9 shows the ratio scores (RS) for
GORT comprehension. This score was created by dividing the number of questions read
correctly by the number of passages read, both for pre and posttests. As illustrated on the
table, no significant increase in ratio score was seen between pre and posttest in either
group.
Independent group t-tests were conducted to compare the means of the RGS of
the two groups on these GORT comprehension measures (See Table 10). No significant
between-group differences were found on any of the measures (p>.05). Results of the
GORT comprehension data with the outlier included can be seen in Appendices J and K.

113

Table 9
Comparisons of Pre and Posttest Mean Scores For GORT Comprehension Within
Groups

Group PR:
pre
PR:
post
ES Sig.

QC:
pre
QC:
post
ES Sig.

RS:
pre
RS:
post
ES Sig.

Exp
N=34
2.53 3.76 0.63 *.000 9.53 13.85 0.65 *.000 3.39 3.94 0.01 .955
Cont
N=25
2.52 3.28 0.43 **.013 9.56 12.20 0.46 **.015 3.94 3.97 0.04 .866
* Significant, p<.001 **Significant, p<.01
PR: Passages read; QC: Questions correct; RS: Ratio score.

Table 10
Comparison Of GORT Mean Comprehension RGS Scores by Group

Measure Group N Mean SD t ES Sig.
(2-tailed)
PR Experimental 34 .099 0.99 1.23 0.32 .223
Control 25 -.213 0.93

QC Experimental 34 .027 1.01 -.142 -0.04 .888
Control 25 .064 0.94

RS Experimental 34 -.016 1.03 -.142 -0.04 .887
Control 25 .022 0.96

114

The Relationship Between Student Characteristics and Student Outcomes

In order to explore the impact of some of the underlying factors thought to effect
reading proficiency, such as rapid automatic naming, phonemic awareness and general
verbal skills, correlations were run examining the interaction between the reading gain
scores and the pretest scores in Elision, Rapid Letter Naming, Similarities, and
Vocabulary (hereafter referred to as reading-related factors). Table 11 shows the results
of those correlations.
From the correlational table it can be seen that Elision (a measure of phonemic
awareness) was positively correlated with two reading gain scores: SWE (p=.037) and
COMP (p=.043). COMP was also correlated with Similarities (p=.047), a measure of
general verbal ability.
On further examination, one other interesting finding emerges from the
correlational data. There was no significant correlation between the growth scores in
COMP and any of the growth scores on the fluency measures—PDE, SWE, RA, ACC or
PASS (p>.05). From these data it seems that students’ growth in comprehension, where it
occurred, was not related to their growth in fluency.
The relationship between two reading-related factors, Elision (a measure of
phonological awareness) and RLN (a measure of rapid automatic naming) were of
particular interest in this study. In the following sections those factors are explored in
more depth.

115

The Relationship of Phonological Processing to Reading Gains
In order to investigate the effectiveness of the intervention on students with
stronger phonological processing skills, students were ranked according to their scores on
the Elision measure, then divided into a high and a low group. The 35 students in the high
group scored between 5 and 12 on this measure. The 25 students in the low group had
scores ranging from 4 to 1—more than two standard deviations below normal. (Due to
the large number of students who received a score of 5 on this measure, the groups were
not evenly distributed.)
When the RGS of these two groups were compared using independent group t-
tests, the high Elision group significantly outperformed the low group on RA (p=.013),
with a moderate effect size of 0.67. Results approached significance on PAS (p=.055, ES
0.50). No other significant comparisons were found (p>.05).
In an attempt to explore this more fully, the students were re-ranked, with only the
students who had Elision scores in the average range (seven or above) included in the
high group. With this grouping, the 12 students with average skills in phonemic
awareness scored significantly higher than the other students on measures of Sight Word
Efficiency (p=.008, ES 0.85) and Word Attack (p=.029, ES0.71) with strong effect sizes
for both conditions. It seems that students with average phonemic awareness skills
outperformed others on two measures of word reading, one of which, SWE, measures
word reading fluency.

116

Correlations
1 .206 -.104 -.098 -.050 .112 .061 .213 .231 .041
.115 .430 .457 .710 .397 .648 .105 .079 .757
60 60 60 60 59 59 59 59 59 59
.206 1 .267* .109 .273* .117 .225 .010 .131 .265*
.115 .039 .409 .037 .379 .086 .941 .322 .043
60 60 60 60 59 59 59 59 59 59
-.104 .267* 1 .456** -.069 .008 -.047 .060 -.013 .259*
.430 .039 .000 .605 .952 .723 .652 .921 .047
60 60 60 60 59 59 59 59 59 59
-.098 .109 .456** 1 -.065 -.035 -.122 .125 -.099 .060
.457 .409 .000 .625 .790 .358 .346 .454 .651
60 60 60 60 59 59 59 59 59 59
-.050 .273* -.069 -.065 1 .198 .051 -.039 -.029 .071
.710 .037 .605 .625 .137 .705 .771 .830 .591
59 59 59 59 59 58 58 58 58 59
.112 .117 .008 -.035 .198 1 .163 .174 .263* -.080
.397 .379 .952 .790 .137 .223 .192 .046 .550
59 59 59 59 58 59 58 58 58 58
.061 .225 -.047 -.122 .051 .163 1 .436 ** .767** .130
.648 .086 .723 .358 .705 .223 .001 .000 .329
59 59 59 59 58 58 59 59 59 58
.213 .010 .060 .125 -.039 .174 .436 ** 1 .744** -.012
.105 .941 .652 .346 .771 .192 .001 .000 .926
59 59 59 59 58 58 59 59 59 58
.231 .131 -.013 -.099 -.029 .263* .767 ** .744 ** 1 .095
.079 .322 .921 .454 .830 .046 .000 .000 .478
59 59 59 59 58 58 59 59 59 58
.041 .265* .259* .060 .071 -.080 .130 -.012 .095 1
.757 .043 .047 .651 .591 .550 .329 .926 .478
59 59 59 59 59 58 58 58 58 59
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Pearson Correlation
Sig. (2-tailed)
N
Rapid Letter
Naming: SS
Elision: SS
Similarities: ss
Vocabulary: SS
SWE rgs
PDE rgs
RATE rgs
ACC rgs
PAS rgs
COMP rgs
Rapid Letter
Naming: SS Elision: SS
Similariti
es: ss
Vocabul
ary: SS SWE rgs PDE rgs RATE rgs ACC rgs PAS rgs COMP rgs
Correlation is significant at the 0.05 level (2-tailed). *.
Correlation is significant at the 0.01 level (2-tailed). **.
Table 11

Correlations between reading gains and underlying related factors

117

The Relationship of RAN to Reading Gains
Similarly to above, students were also ranked according to their skills in rapid
automatic naming by using their scores on the Rapid Letter Naming assessment. The
range of scores on this assessment was from 1 to 12, with 27 students having a score of 7
or above. These 27 students were designated as “high RAN” students. Coincidentally,
approximately half of the students in this group (27 out of 60) had scores of seven or
above, which puts them into the average range for RAN ability, so this comparison was
only run once.
Independent group t-tests were conducted to compare the RGS of the high RAN
group, whose scores were in the average range, to the rest of the students. No significant
differences were found between the two groups on any of the reading gain scores.
Students Who Gained In Fluency And Comprehension
One more interesting group emerged from the data: there were 12 students who
made gains in both fluency and comprehension over the course of the intervention period.
Seven of these students were in the experimental group and five were in the control
group. In order to learn more about these 12 students, separate correlations were run with
implementation factors and reading-related factors.
For the 12 students who made gains in fluency and comprehension, their PAS
scores were significantly correlated with Number of Sessions (p=.006) and Sessions per
Week (p=.001), and their PDE scores were correlated with Number of Sessions (p=.047).
No other reading scores were correlated with implementation factors. When looking at

118

reading-related factors, COMP was correlated with Similarities (p=.038), and SWE
was negatively correlated with RAN (p=.034).
To explore the make-up of this group in more detail, independent group t-tests
were run to compare these students’ mean scores to the mean scores of the rest of the
students on the reading-related factors, implementation factors, and initial fluency and
comprehension scores. Results found no significant difference between the two groups on
any of the implementation factors, or on initial comprehension and fluency scores.
However, as a group, the high achieving students were found to have significantly higher
mean scores on one reading-related measure: Elision (p=.026), a measure of phonemic
awareness skill. This difference had a strong effect size of 0.70.
Summary
In this summary the results of the data analysis will be reviewed in accordance
with the research questions proposed in Chapter One. These results will be discussed and
analyzed in depth in the following chapter.
Is there a difference in fluency growth between the two groups? When comparing
the group means of the experimental and control groups on the eight reading measures
with one extreme outlier removed, the experimental group made significantly more
progress on four measures: TOWRE Phonemic Decoding Efficiency, and GORT Rate,
Accuracy and Passage. All four of these measures are fluency related measures. It seems
that the experimental group made significantly more improvement in reading fluency
than the control group during the intervention period.

119

Is there a difference in reading comprehension growth between the two
groups? Neither group made significant progress in reading comprehension as measured
by the Woodcock Comprehension assessment, and there was no statistically significant
difference between the groups in terms of comprehension, either on the Woodcock
measure, or on the open-ended GORT questions. Although students in both groups
answered more questions correctly on the GORT post test, there was no difference in the
ratio of questions correct per passage from pre to post test. No correlation was found
between reading fluency gains and comprehension gains.
Is there a difference in outcomes between those students with poor phonemic
awareness skills and those with stronger skills? Students with stronger phonemic
awareness skills, as measured by their scores on the Elision assessment, had significantly
higher gain scores than the rest of the students on one reading measure: RA (p=.013), and
approached significance on PAS (p=.055). Students whose Elision scores were in the
average range scored higher than the rest on SWE (p=.008) and WA (p=.020). In a
related finding, the 12 students who made gains in both fluency and comprehension had
significantly higher scores on Elision than the rest of the students (p=.026).
Is there a difference in outcomes between those students with poor RAN skills and
those with stronger skills?
No correlations were found between RAN skills and any of the reading gains. In
addition, no significant differences were found in gain scores between students with high
RAN scores and those with low.

120

Can a one-on-one fluency intervention program be implemented successfully in
two urban middle schools? The implementation of the program varied considerably
across the 18 paraprofessionals who participated, so the ability to implement the program
with fidelity may be called into question. However, all paraprofessionals remained
engaged in the program throughout the school year, and the results of the intervention
showed significant mean outcomes in fluency for the students involved. It seems that
overall the implementation was successful.
In the following chapter, these results will be discussed in detail and analyzed in
light of the existing literature base. In addition, implications for practice will be
presented, and recommendations for future research will be made.

121

CHAPTER V: DISCUSSION

For years teachers have been told that students in the primary grades are “learning
to read”, and that after third grade they should be “reading to learn”. Unfortunately, for
some students this is a fallacy. A small percentage of young people move through the
grade levels without the requisite reading skills to access and participate in most of the
academic materials presented to them. It seems critical that these struggling students
should keep working on “learning to read,” even into their high school years if needed,
but the methods best suited to helping them achieve success remain elusive.
The current study sought to investigate this problem by exploring the use of a
fluency intervention program in middle school special education classrooms. The
following research questions were proposed:
(1) Is there a difference in reading rate and accuracy (together defined as fluency)
between students receiving a daily fluency intervention and the control group?
(2) Is there a difference in reading comprehension between students receiving a daily
fluency intervention and the control group?
(3) Is there a difference between the fluency and comprehension of students with
poor phonemic awareness and those with stronger phonemic skills as measured
before the intervention?
(4) Is there a difference between the fluency and comprehension of students with
poor RAN skills and those with stronger RAN skills as measured before the
intervention?

122

(5) Can a one-on-one daily fluency intervention program be implemented
successfully in two typical urban middle schools?
The results of the study will be discussed in this chapter, addressing the findings
for each research question in the order presented and in relationship to the current
literature base. Next, implications for practice will be discussed, seeking to integrate the
findings from this study with typical middle school classroom practices. Subsequently,
limitations and weaknesses of the current study will be examined, and finally, the chapter
will conclude with recommendations for the direction of future research.
The Intervention’s Impact on Reading Fluency
Although the statistical analyses were somewhat confounded by the presence of
two extreme outliers, once the outliers were removed it seems clear that the intervention
was successful for the remaining 59 students. Students in the experimental group did
make significantly more progress in fluency than the control group, as measured by their
performance on four measures: TOWRE Phonemic Decoding Efficiency, and GORT
Rate, Accuracy and Passage. This difference in performance between the two groups
occurred despite uneven implementation of the program, and despite a significant
decrease in the intensity of the intervention from the original study design.
The reasons for this success are likely related to the design of the Great Leaps
fluency program (Campbell, 1999), which closely replicates the recommendations of the
majority of the literature on effective fluency instruction. This program builds in specific
and clear performance criteria for the students, and systematically moves them into
harder material as they master each level. Equally as important, adults implemented the

123

intervention, and it incorporated regular error correction and feedback. These
components are exactly those that were recommended by Chard, Vaughn and Tyler
(2002), Therrien (2004) and the National Reading Panel (2000) in their meta-analyses of
fluency research.
In addition, the program gives fluency practice on a variety of levels, focusing on
sounds, single words, short phrases and whole passages. As defined by Wolf and Katzir-
Cohen (2001), fluency includes “…development of automaticity in underlying sublexical
processes, lexical processes, and their integration into single-word reading and connected
text” (p. 219). Great Leaps, unlike most other fluency intervention programs,
systematically addresses the development of fluency in all of these processes, and it may
be this methodical fluency practice on a variety of levels that accounts for much of the
success of the program in the current study.
It seems important to discuss these results more explicitly in comparison to the
original Great Leaps intervention study (Mercer et al., 2000). That study, as published,
doesn’t reveal the exact number of intervention sessions the students received, so direct
comparisons are difficult to make. In addition, Mercer and colleagues used curriculum-
based assessments that are not directly equivalent to the standardized measures used in
the current study. However, when examining the grade level changes in the students who
completed six months of intervention in the previous study (n=7), the average change for
those students was +1.57, with a range of 0.5 to 2.5. In the current study, the actual mean
change in grade level scores on the GORT Passage assessment for the experimental
group was +.90 (n=33), with a range of -1.3 to 2.8. Although the assessments used in

124

these two studies are difficult to equate, it seems likely that if the actual number of
sessions completed could be compared, the levels of growth in the two studies might be
similar.
Another important point must be made, however. When comparing fluency gains
on the GORT measures for each group separately, both groups made significant progress
from pre to posttesting. In addition, both groups showed progress on at least one of the
TOWRE fluency measures from pre to posttest. It seems that the interventions that were
received by both groups had a positive effect on their fluency growth. Although there
isn’t a third group of no-treatment students with which to compare, the improvement
made by the control group leads one to believe that the opportunity to work intensively
for ten minutes a day with a paraprofessional on any reading-related task may create
fluency benefits for the students involved. This seems particularly noteworthy when you
consider the low initial rates of reading fluency that these students had obtained in more
than six years of schooling. While the statistics clearly imply that the Great Leaps
intervention had the greatest impact on increasing students’ reading rate and accuracy,
the increase in rate and accuracy for the control group is also an important finding, and
one worthy of attention.
The Intervention’s Impact on Reading Comprehension
Sadly, however, the fluency program as implemented in this study did not seem to
have any measurable impact on reading comprehension. Although many students made
gains in the number of passages they read on the GORT assessment and thus were able to
answer more comprehension questions, the average number of correct questions per

125

passage did not significantly change. Furthermore, no mean growth was seen on the
Woodcock Passage Comprehension test, and no positive correlations were found between
gains in comprehension and gains on any of the fluency related reading measures.
Although the research is far from conclusive, many studies have found a positive
relationship between fluency and comprehension. In fact, the National Reading Panel
(2000), Kuhn and Stahl (2003) and Therrien (2004), after examining multiple studies
with a variety of different types of students, all concluded that there was a strong and
positive relationship between increased fluency and increased reading comprehension.
That relationship was not seen in the current study.
The reasons for this may be numerous, but first and foremost the extreme nature
of the present population must be considered. Most fluency studies with reading-delayed
students draw subjects from the bottom 15 – 30% of the reading population (see, for
example, Manset-Williamson & Nelson, 2005; Marchand-Martella, Martella, Orlob, &
Ebey, 2000; Therrien, Wickstrom, & Jones, 2006). In the present study, the students’
reading delays were much more profound. In fact, 48 out of 60 of the students in the
present study scored at or below the tenth percentile on the pretest Word Identification
assessment. More than half scored below the fifth percentile. Clearly, these middle school
students represent the most significantly impaired readers, and if a history of good
reading instruction can be assumed, they epitomize Torgesen’s definition of “treatment
resistors”: students who, despite receiving high-quality reading instruction, fail to
develop word reading skills commensurate with their levels of general verbal ability
(Torgesen, 2000).

126

Importantly, Torgesen also mentions that many of this population of treatment-
resisting students are delayed not only in word reading, but in vocabulary and language
skills as well. This assertion is especially applicable to struggling readers in middle
school, who have missed years of reading experiences that would help them develop age-
appropriate vocabulary, as well as syntactical and conceptual knowledge that can lead to
strong reading comprehension (Stanovich, 1986). So for these students, even if they can
acquire strong word-reading skills, they are still likely to have problems with reading
comprehension related to deficits in general language proficiency (Torgesen, 2000, p.
57). That condition seems to be in place in the current study.
This conjecture can be confirmed to some extent by looking at the measures of
language ability given to the students in this study. Assuming average performance to be
one standard deviation on either side of the mean, any score on the WISC scale below
seven would be considered below-average performance. Forty-one out of 60 of the
students in this study scored seven or below on the Similarities measure, and 57 out of 60
scored seven or below on the Vocabulary assessment. Without a doubt, the majority of
these students have deficits in aspects of language proficiency that might impair their
ability to benefit globally from an increase in reading fluency.
Nevertheless, gain scores on Comprehension were found to be mildly positively
correlated with students’ scores on Similarities in the total population (.259, p=.047).
These findings imply that, despite an overall pattern of low language proficiency, those
students with more aptitude in this area were more likely to show improved reading
comprehension at posttest. It’s also interesting to note that among the 12 students who

127

made progress in both fluency and reading comprehension (five of whom were in the
control group), 9 of these 12 students had scores of three or below on one or both of the
verbal measures. Although this group of 12 students also had a positive correlation
between Comprehension and Similarities (.602, p=.038), it seems clear that overall verbal
ability, at least as measured by these two WISC assessments, is not the only contributor
to these students’ capacity to gain reading comprehension benefit from fluency increases.
The ambiguity of these findings, both in the general study population and in the 12 high
achieving students, invites further exploration into this relationship.
The Relationship Between Phonemic Awareness Skill and Reading Growth
When correlations were analyzed for all the students, two of the residual gain
scores on reading measures showed significant positive correlations with phonemic
awareness (Elision): Sight Word Efficiency and Comprehension. These results suggested
that as students’ phonemic awareness increased, their likelihood for reading speed and
comprehension gains increased as well.
In order to explore these effects in more detail, participants were ranked
according to their score on the Elision measure, then divided into a high and a low group.
This process was done twice—once by simply dividing the students in half according to
their scores on the measure, and once by putting only the students who scored in the
average range (8 or above) on Elision in the high group.
When the two halves (high and low) were compared, there was a significant
difference in their residual gain scores on the GORT Rate assessment—a test that
measures reading speed without accounting for accuracy. The comparison also

128

approached significance for the GORT Passage assessment. It seems that the students
ranked in the top half for phonemic awareness skill outperformed the others in terms of
reading fluency. However, when the students with phonemic awareness skills in the
average range were compared to the rest of the students (whose scores were below
average,) a slightly different picture emerged. The nine students with average skills in
phonemic awareness scored significantly higher than the other students on measures of
Sight Word Efficiency and Word Attack, with strong effect sizes for both conditions.
Although both analyses imply that phonemic awareness skill was related to increased
reading outcomes, the specifics varied according to where the cut-off point was placed.
Nevertheless, research has confirmed time and time again the association between
reading proficiency and phonemic awareness skill, particularly in the population of
students with learning disabilities (see for example Fletcher et al., 1994; Manis, Custodio
& Szeszulski, 1993; Share & Stanovich, 1995). It does not seem surprising that students
with stronger phonemic awareness skills would evidence greater reading growth in an
intervention, and in this case, it seems that students with higher phonemic awareness
outperformed others on a variety of measures. However, the artificiality of this cut-off
point must be pointed out; these findings give the suggestion of a pattern, but must be
interpreted carefully.
One other interesting finding related to phonemic awareness emerged from the
data. For the twelve students who achieved gains in both fluency and reading
comprehension, their mean score in Elision was significantly higher than the mean of the
other students, with a strong effect size of 0.70. From that finding it might be concluded

129

that, overall, the students who benefited most from both interventions were those who
had the highest ability in phonemic awareness. Again, this seems to confirm the link
between phonological ability and reading that has been found in numerous earlier studies.
The Relationship Between RAN Skill and Reading Growth
When students were ranked according to their RAN scores and divided in half
into groups of high and low, there were no significant differences in the reading
outcomes of the two groups. Neither were any correlations seen between RAN and gains
on any of the reading measures.
These findings would seem to support the work of researchers such as Manis,
Seidenberg and Doi (1999) and Torgesen, Wagner, Rashotte, Burgess, and Hecht (1997)
who found the strongest influence of RAN in first and second graders. Although Meyer,
Wood, Hart and Felton (1998) argued that the influence of naming speed can have a
persistent effect on reading ability for older students with reading disabilities, the results
of the current study do not support that contention.
Implications for Practice
Can a One-On-One Fluency Intervention Program Be Implemented Successfully in Two
Typical Urban Middle Schools?
The final research question addresses the implications of this study for typical
classroom practices in urban middle schools. Were these two schools able to implement
the program with fidelity, given their limited resources and the high demands on teacher
and student time?

130

In the original Great Leaps study (Mercer et al., 2000) the school dedicated
specific paraprofessionals to implementing the fluency program; in other words, the
paraprofessionals were reassigned from their regular responsibilities. This reallocation of
resources was not possible at the two schools that participated in the present study.
Instead, paraprofessionals were asked to take on the intervention in addition to their
regular workload. This potentially could have had a considerable impact on the teachers
to whom they were assigned, as well as the students in their classrooms.
Clearly, from the data on the number of sessions per week and the total number of
sessions completed, the paraprofessionals were not able to give the students regular daily
treatments as originally designed. If the interventions had been done four days per week,
students would have completed more than 90 sessions in the six months of the
intervention. Most completed less than 60. Other obligations to their classrooms and
students frequently interrupted the intervention sessions, and several of the
paraprofessionals complained about feeling burdened by the additional workload. A few
teachers also mentioned that the intervention sessions disrupted their regular classroom
routines.
Nevertheless, both schools were able to complete six months of the intervention
with very little attrition from the students and none from the paraprofessionals.
Furthermore, the fluency increases experienced by the experimental group show potential
for the program as a viable tool for struggling readers. From this perspective, it seems
that the fidelity of implementation, although not as expected, was a qualified success;
with some creative use of personnel and space, this program could most likely be

131

implemented in a typical urban middle school to the benefit of the students. All the
same, it must be observed that there was a strong correlation between the number of
sessions students received per week and their gain in fluency. Obviously, if the
implementation had been more intensive, the outcomes for the students would most likely
have increased.
Notably these students, all of whom were in the bottom five to ten percentile in
reading, were receiving no other reading intervention at their schools. The pressure to
achieve high scores on standardized assessment has pushed aside remedial programs in
many schools and forced a focus on grade-level content, even when the students aren’t
able to access it by reading it for themselves. The value of a 10-minute a day
intervention, then, can be appreciated in light of the time constraints and excessive
demands experienced by many secondary special education teachers. The results of the
current study clearly support the potential of a quick, one-on-one remediation program to
increase the reading fluency of severely impaired students. The fact that it can be
implemented by paraprofessionals, and doesn’t take up the limited time of the teachers,
increases the value to impacted schools.
Several other implications for school practice emerged from the findings. First, it
is clear from these results that although a fluency program can increase the decoding
skills of this population, that alone is not enough to improve their overall reading ability.
Indeed, if reading is defined as the ability to understand what is read, then the success of
this study was limited at best. Although it can be agreed that students need to learn to
decode well enough to access text, clearly this population needs additional instruction in

132

subjects such as vocabulary, reading comprehension strategies, and other general
language skills in order to gain meaning from the text they decode (Torgesen, 2000).
Another important implication for school practice concerns the utilization of
paraprofessionals. In special education programs around the country it is typical to find
paraprofessionals assigned to assist in the classrooms. Yet studies have repeatedly shown
that teachers feel unprepared to serve as supervisors or managers for their
paraprofessionals (Carroll, 2001; French, 2003), and are unclear as to whether the
paraprofessional’s primary responsibility should be to assist them or assist the students
(French, 1998). In light of those findings, the positive effect that both these interventions
had on the reading outcomes of the students seems important. Undoubtedly, engaging
paraprofessionals in one-on-one instruction with students using easy to implement
materials such as Great Leaps (Campbell, 1999) and Skills for School Success (Archer &
Gleason, 2002) can be a good use of their time, and can be of considerable benefit to the
children.
In addition, research has repeatedly supported the effectiveness of small group
and one-on-one instruction with this population (Elbaum, Vaughn, Hughes & Moody,
1999; Torgesen, 2004; Vaughn, 2001), and the results of this study support that assertion.
It seems that self-contained, self-directed materials focused on deficit skills may be a
good way to increase the effectiveness of paraprofessionals in special education classes
through one-on-one tutoring experiences.

133

Limitations of the Study
This study has several limitations that must be addressed. The creation of a
control group was a strength of this study, and allowed the results to be attributed to the
effectiveness of the fluency program. Nevertheless, it is clear from the findings that
having a third group, who received no intervention, would have benefited the analysis.
Since the control group evidenced unexpected gains in fluency, having a no-treatment
third group might have informed the results significantly.
Another limitation of this study was the possible effect of the different skill levels
of the paraprofessionals. Although correlational analyses didn’t reveal any group effects
accorded to the paraprofessionals, it is quite possible that some of them had greater skills
in working with the students than others. Structuring the study so that all the
paraprofessionals worked equally with both the experimental and control groups would
have reduced the possibility of differential effects due to the skills of the personnel
involved.
Another significant shortcoming was the inconsistency of the program
implementation. The number of sessions per student and the number of sessions per week
varied widely, and much of this could be attributed to each paraprofessional’s individual
commitment (or lack of commitment) to the program. Although significant results were
found, they were not as robust as might be anticipated. It seems likely that had there been
more treatment fidelity, the results for the students would have been greater.
The use of the same form of the tests for pre and posttesting was also a potential
limitation. While some research has shown that using the same form of an assessment

134

with students with reading disabilities is unlikely to produce test-retest effects (Cirino
et al., 2002), it’s still possible that some consequence of test repetition was present in the
results. This seems most likely in the GORT, where the narratives might be memorable to
the students. In particular, this test-retest effect could perhaps account for some of the
increase in fluency evidenced by the control group on the GORT assessment.
However, it is important to discuss two factors that mitigate this observation.
First, since the increases in fluency by the experimental group were statistically
significant when compared to the control group, and both groups used the same form of
the GORT, the between-group difference can rightfully be attributed to the intervention.
As cited in chapter three, “…a randomized control group is the best tool we have to test
whether a treatment has a real effect… over and above confounds such as placebo effects,
Hawthorn effects, test-retest effects, and regression-to-mean effects” (McArthur, 2007).
In addition, the other assessment that was most likely to show test-retest effect
would have been the Woodcock Passage Comprehension, which also uses short
narratives in the assessment. Notably, no mean change was seen in achievement on that
test for either group. This seems to imply that test-retest effect, at least for the Woodcock
assessment, was not a factor for the majority of the students.
Future Directions
This study raises many questions that might guide future research. Primarily,
further investigation into the characteristics of the students who made fluency progress
would be useful. Many studies, including this one, have shown that fluency interventions
can increase the reading decoding of students with significant reading delays, but there

135

were some students who did not respond to the intervention. Identifying the specific
characteristics of responders versus non-responders might inform the use of such
instructional programs, and help maximize the time of all concerned. In particular,
assessments that measure students’ listening comprehension ability, such as the Listening
Comprehension and Understanding Directions subtests of the Woodcock-Johnson III
(Woodcock, McGrew & Mather, 2001) might be a good addition to the current
assessment battery, and might prove to be more indicative of the language skills needed
for comprehension in this particular population of students.
If specific characteristics can be determined for the students who do not respond
to interventions such as this, it’s possible that other types of fluency instruction might be
more useful for those students. The Great Leaps program focuses mainly on increasing
students’ decoding ability and reading speed through repeated-reading, however, other
fluency researchers have pointed out the importance of including additional elements in
fluency instruction. Rasinsky et al. (2005) discussed the need for fluency activities that
emphasize meaning and prosody. These researchers believe that a focus on speed alone
can actually inhibit reading growth in some students; future research should explore the
use of different types of fluency interventions for specific sub-groups of reading-delayed
students.
Similarly, Mokhtari and Thompson (2006) proposed that students’ ability to
understand syntax and to chunk language into meaningful phrases may have as great an
impact on fluency as phonemic processing. Although Great Leaps does include some
emphasis on phrasing, there is no discussion of how these phrases relate to text in terms

136

of syntax or meaning. Exploring the addition of syntax and prosody to a program of
repeated-reading could produce significant findings that would enrich the use of fluency
interventions in schools.
More investigation also needs to be done on the relationship between fluency and
reading comprehension in this particular population of students with the most severe
reading disabilities. Recent work by Torgesen (2005) has shown that 40% of the variance
in standardized test scores of seventh graders can be attributed to fluency. However, it is
unclear if this percentage applies across the population, and particularly to the students in
the bottom five percent of readers. For this group of profoundly delayed readers, many
questions remain about how to best use our limited time to increase their reading skills so
they can be self-sufficient in school and life. More studies with this population are
needed to inform our choices.
Finally, more research on the use of paraprofessionals to provide remediation
seems necessary. Are there other programs available that could be implemented easily
and effectively by paraprofessionals assigned to special education classes? How can we
best use their skills and affinities without increasing the burden on the paraprofessionals
and their special education teachers? The findings from this study, as well as the original
Great Leaps study (Mercer et al., 2000), clearly lead to the conclusion that
paraprofessionals can provide a great benefit to schools when used for intervention with
struggling students, but finding the ideal materials and circumstances to do so warrants
further investigation. Identifying the best resources and conditions for intervention

137

programs such as this could reduce the inconsistency of implementation so evident in
the current study.
Conclusion
Although much has been learned about effective reading instruction in the past ten
years, many questions remain about how to help the small percentage of students with the
most pervasive reading problems. Exploring the efficacy of fluency intervention is one
way that researchers are seeking to help all students become readers.
More research in this area is needed to find the best combination of instructional
methods that will work for students identified with profound reading disabilities. This
study provides more information about how to build fluency in these learners, and its
impact on reading comprehension for this troubling population. In order to open the door
for students of all abilities to become fluent, successful readers, more investigation of
fluency and its impact on the reading comprehension of treatment resistors is essential.

138

REFERENCES

Adams, M.J. (1990). Beginning to read: Thinking and learning about print. Cambridge,
MA: MIT Press.

Alber-Morgan, S.R., Ramp, E.M., Anderson, L.L., & Martin, C.M. (2007). Effects of
repeated readings, error correction, and performance feedback on the fluency and
comprehension of middle school students with behavior problems. The Journal of
Special Education, 41(1), 17-30.

Allor, J.H. (2002). The relationship of phonemic awareness and rapid naming to reading
development. Learning Disability Quarterly, 25(1), 47-58.

Archer, A.L., & Gleason, M.M. (2002). Skills for school success. North Billerica, MA:
Curriculum Associates.

Archer, A.L., Gleason, M.M., & Vachon, V.L. (2003). Decoding and fluency:
Foundational skills for struggling older readers. Learning Disability Quarterly,
26(2), 89-102.

Berends, I.E. & Reitsma, P. (2007). Orthographic analysis of words read during fluency
training promotes reading of similar new words. Journal of Research in Reading,
30(2), 129-139.

Berninger, V.W., Abbott, R.D., Billingsley, F., & Nagy, W. (2001). Processes underlying
timing and fluency of reading: Efficiency, automaticity, coordination, and
morphological awareness. In M. Wolf (Ed.) Dyslexia, fluency and the brain.
Timonium, MD: York Press.

Blackorby, J., & Wagner, M. (1996). Longitudinal postschool outcomes of youth with
disabilities: Findings from the National Longitudinal Transition Study.
Exceptional Children, 62, 399-414.

Bradley, L., & Bryant, P. (1978). Difficulties in auditory organization as a possible cause
of reading backwardness. Nature, 27, 746-747.

Breznitz, Z., & Berman, L. (2003). The underlying factors in word reading. Educational
Psychology Review, 15(3), 247-265.

Bryant, D.P., Vaughn, S., Linan-Thompson, S., Ugel, N., Hamff, A., & Hougen, M.
(2000). Reading outcomes for students with and without reading disabilities in
general education middle-school content area classes. Learning Disability
Quarterly, 23(4), 238-252.

139

California Department of Education (n.d.). Testing and accountability data. Retrieved
February 24, 2008 from http://star.cde.ca.gov/star2007/index.asp

California Reading Task Force (1995). Every child a reader. Sacramento, CA:
Department of Education.

Campbell, K.U. (1999). Great leaps reading (4
TH
ed.). Gainesville, FL: Diarmuid, Inc.

Carroll, D. (2001). Considering paraeducator training, roles and responsibilities.
Teaching Exceptional Children, 34(2), 60-64.

Chard, D.J., Vaughn, S., & Tyler, B.J. (2002). A synthesis of research on effective
interventions for building reading fluency with elementary school students with
learning disabilities. Journal of Learning Disabilities, 35, 386-407.

Cirino, P.T., Rashid, F.L., Sevcik, R.A., Lovett, M.W., Frijters, J.C., Wolf, M., & Morris,
R.D. (2002). Psychometric stability of nationally normed and experimental
decoding and related measures in children with reading disability. Journal of
Learning Disabilities, 35, 525-538.

Deshler, D.D., Schumaker, J.B., Lenz, B.K., Bulgren, J.A., Hock, M.F., Knight, J., &
Ehren, B.J. (2001). Ensuring content-area learning by secondary students with
learning disabilities. Learning Disability Research and Practice, 16, 96-108.

Dowhower, S. (1997). Introduction to S.J. Samuels, “The method of repeated readings.”
The Reading Teacher, 50, 376.

Elbaum, B., Vaughn, S., Hughes, M., & Moody, S. W. (1999). Grouping practices and
reading outcomes for students with disabilities. Exceptional Children, 65, 399–
415.

Fleisher, L.S., Jenkins, J.R., & Pany, D. (1979). Effects on poor readers’ comprehension
of training in rapid decoding. Reading Research Quarterly, 15(1), 30-48.

Fletcher, J.M., Shaywitz, S.E., Shankweiler, D.P., Katz, L., Liberman, I.Y., Stuebing,
K.K., Francis, D.J., Fowler, A.E., & Shaywitz, B.A. (1994). Cognitive profiles of
reading disability: Comparisons of discrepancy and low achievement definitions.
Journal of Educational Psychology, 86, 6-23.

Foorman, B.R., Francis, D.J., Winikates, D., Mehta, P., Schatschneider, C., & Fletcher,
J.M. (1997). Early interventions for children with reading disabilities. Scientific
Studies of Reading, 1(3), 255-277.

140

French, N.K. (1998). Working together: Resource teachers and paraeducators. Remedial
and Special Education, 19, 357-368.

French, N.K. (2003). Paraeducators in special education programs. Focus on Exceptional
Children, 36(2), 1-16.

Fuchs, L.S., Fuchs, D., Hamlett, C.L., Walz, L., & Germann, G. (1993). Formative
evaluation of academic progress: How much growth can we expect? School
Psychology Review, 22(1), 27-48.

Fuchs, D., Fuchs, L.S., McMaster, K.L., Yen, L., & Svenson, E. (2004). Nonresponders:
How to find them? How to help them? What do they mean for special education?
Teaching Exceptional Children, 37(1), 72-78.

Hand, D.J., & Taylor, C.C. (1987). Multivariate analysis of variance and repeated
measures: A practical approach for behavioural scientists. Boca Raton, FL:
Chapman and Hall.

High, R. (2000). Dealing with outliers: How to maintain your data’s integrity. Retrieved
June 8
th
, 2000 from the University of Oregon website:
http://cc.uoregon.edu/cnews/spring2000/outliers/html

Howell, D.C. (2007). Treatment of missing data. Retrieved June 10, 2008 from the
University of Vermont web site:
http://www.uvm.edu/~dhowell/StatPages/More_Stuff/Missing_Data/Missing.html

Howell, D.C. (2004). Fundamental statistics for the behavioral sciences (5
th
ed.).
Belmont, CA: Thomson Wadsworth.

Huey, S.E. (1908/1968). The psychology and pedagogy of reading. Cambridge, MA: MIT
Press. (Original work published 1908).

Irausquin, R.S., Drent, J., & Verhoeven, L. (2005). Benefits of computer-presented speed
training for poor readers. Annals of Dyslexia, 55(2), 246-265.

Jakobson, A., & Kikas, E. (2007). Cognitive functioning in children with and without
attention-deficit/hyperactivity disorder with and without comorbid learning
disabilities. Journal of Learning Disabilities, 40(3), 194-202.

Kairaluoma, L., Ahonen, T., Aro, M., & Holopainen, L. (2007). Boosting reading
fluency: An intervention case study at subword level. Scandinavian Journal of
Educational Research, 51(3), 253-274.

141

Kame'enui, E. J., Simmons, D. C., Baker, S., Chard, D. J., Dickson, S. V., Gunn, B.,
Smith, S. B., Sprick, M., & Lin, S. J. (1997). Effective strategies for teaching
beginning reading. In E. J. Kame'enui, & D. W. Carnine (Eds.), Effective teaching
strategies that accommodate diverse learners. Columbus, OH: Merrill.

Katzir, T., Kim, Y., Wolf, M., Kennedy, B., Lovett, M., & Morris, R. (2006). The
relationship of spelling recognition, RAN, and phonological awareness to reading
skills in older poor readers and younger reading-matched controls. Reading and
Writing, 19, 845-872.

Katzir, T., Kim, Y., Wolf, M., O’Brien, B., Kennedy, B., Lovett, M., et al. (2006).
Reading fluency: The whole is more than the parts. Annals of Dyslexia, 56(1), 51
– 82.

Keenan, J.M. & Betjemann, R.S. (2006). Comprehending the Gray Oral Reading Test
without reading it: Why comprehension tests should not include passage-
independent items. Scientific Studies in Reading, 10(4), 363-380.

Kuhn, M.R., & Stahl, S.A. (2003). Fluency: A review of developmental and remedial
practices. Journal of Educational Psychology, 95(1), 3 – 21.

LaBerge, D., & Samuels, S.J. (1974). Toward a theory of automatic information
processing in reading. Cognitive Psychology, 6, 293-323.

Levy, B.A., Abello, B., & Lysynchuk, L. (1997). Transfer from word training to reading
in context: Gains in reading fluency and comprehension. Learning Disability
Quarterly, 20(3), 173-188.

Levy, B.A., Bourassa, D.C., & Horn, C. (1999). Fast and slow namers: Benefits of
segmentation and whole word training. Journal of Experimental Child
Psychology, 73, 115-138.

Lovett, M.W., Steinbach, K.A., & Fritjers, J.C. (2000). Remediating the core deficits of
developmental reading disability: A double-deficit perspective. Journal of
Learning Disabilities, 33, 334-358.

Lyon, G.R., & Moats, L.C. (1997). Critical conceptual and methodological considerations
in reading intervention research. Journal of Learning Disabilities, 30, 578-589.

MacGinitie, W.H., & MacGinitie, R.K. (1992). Gates-MacGinitie Reading Tests (3rd
ed.). Chicago: Riverside.

142

Manis, F. R., Custodio, R., & Szeszulski, P.A. (1993). Development of phonological
and orthographic skill: A 2-year longitudinal study of dyslexic children. Journal
of Experimental Child Psychology, 56, 64-86.

Manis, F.R., Doi, L.M., & Bhadha, B. (2000). Naming speed, phonological awareness,
and orthographic knowledge in second graders. Journal of Learning Disabilities,
33, 325-333, 374.

Manis, F.R., Seidenberg, M.S., & Doi, L.M. (1999). See Dick run: Rapid naming and the
longitudinal prediction of reading subskills in first and second graders. Scientific
Studies of Reading, 3(2), 129-157.

Manset-Williamson, G., & Nelson, J.M. (2005). Balanced, strategic reading instruction
for upper-elementary and middle school students with reading disabilities: A
comparative study of two approaches. Learning Disabilities Quarterly, 28(4), 59-
74.

Marchand-Martella, N., Martella, R.C., Orlob, M., & Ebey, T. (2000). Conducting action
research in a rural high school setting using peers as corrective reading instructors
for students with disabilities. Rural Special Education Quarterly, 19(2).

Martin-Chang, S.L., Levy, B.A. (2006). Word reading fluency: A transfer appropriate
processing account of fluency transfer. Reading and Writing, 19, 517-542.

McArthur, G. (2007). Test-retest effects in treatment studies of reading disability: The
devil is in the details. Dyslexia, 13, 240-252.

McBride-Chang, C., & Manis, F.R. (1996). Structural invariance in the associations of
naming speed, phonological awareness, and verbal reasoning in good and poor
readers: A test of the double deficit hypothesis. Reading and Writing: An
Interdisciplinary Journal, 8, 323-339.

Mercer, C.D., Campbell, K.U., Miller, M.D., Mercer, K.D., & Lane, H.B. (2000). The
effects of a reading fluency intervention for middle-schoolers with specific
learning disabilities. Learning Disabilities Research and Practice, 15(4), 179-189.

Meyer, M.S., & Felton, R.H. (1999). Repeated reading to enhance fluency: Old
approaches and new directions. Annals of Dyslexia, 49, 283-306.

Meyer, M.S., Wood, F.B., Hart, L.A., & Felton, R.H. (1998). Selective predictive value
of rapid automatized naming in poor readers. Journal of Learning Disabilities,
31(2), 106-117.

143

Mokhtari, K., & Thompson, H.B. (2006). How problems of reading fluency and
comprehension are related to difficulties in syntactic awareness skills among fifth
graders. Reading Research and Instruction, 46(1), 73-94.

Morgan, P.L., & Sideridis, G.D. (2006). Contrasting the effectiveness of fluency
interventions for students with or at risk for learning disabilities: A multilevel
random coefficient modeling meta-analysis. Learning Disabilities Research and
Practice, 21(4), 191-210.

National Institute of Standards and Technology (n.d.). What are outliers in the data?
Retrieved June 10, 2008 from
http://www.itl.nist.gov/div898/handbook/prc/section1/prc16.htm

National Reading Panel. (2000). Report of National Reading Panel: Teaching children to
read. Bethesda, MD: National Institute of Child Health and Human Development.

National Center for Educational Statistics (2007). NAEP data explorer. Retrieved
February 24, 2008 from http://nces.ed.gov/nationsreportcard/nde/viewresults.asp

New York City Department of Education (n.d.). Assessment and accountability data.
Retrieved June 27, 2008 from http://schools.nyc.gov/daa/2007ela/default.asp

Nikolopoulour, D, Goulandris, N., & Snowling, M. J. (2003). Developmental dyslexia
in Greek. In N. Goulandris, (Ed.), Dyslexia in different languages (pp. 1-14).
London, UK: Whurr Publishers.

Perfetti, C.A. (1985). Reading ability. Oxford, England: Oxford University Press.

Perfetti, C.A., & Hogaboam, T. (1975). Relationship between single word decoding and
reading comprehension. Journal of Educational Psychology, 67, 461-469.

Posner, M.I., & Snyder, C.R.R. (1975). Facilitation and inhibition in the processing of
signals. In P.M.A. Rabbitt & S. Dornic (Eds.), Attention and performance
(Volume 5, pp. 669-682). New York: Academic Press.

Prifitera, A., Saklofske, D.H., & Weiss, L.G. (2005). WISC-IV clinical use and
interpretation: Scientist-practitioner perspectives. San Diego, CA: Elsevier
Academic Press.

Purcell-Gates, V. (1997, May). There’s reading… and then there’s reading: Process
models and instruction. Focus on Basics: Connecting Research and Practice, 2,
Article A. Retrieved July 14, 2007, from http://www.ncsall.net/?id=460

144

Range, L., Embry, T., & MacLeod, T. (2001). Active and passive consent: A
comparison of actual research with children. Ethical Human Sciences and
Services, 3(1), 23-31.

Rasinski, T.V., Padak, N.D., McKeon, C.A., Wilfong, L.G., Friedauer, J.A., & Heim, P.
(2005). Is reading fluency a key for successful high school reading? Journal of
Adolescent and Adult Literacy, 48(1), 22-27.

Rumelhart, D. (1978). Toward and interactive model of reading. In S. Dormic (Ed.),
Attention and performance (Vol. 1). Hillsdale, NJ: Erlbaum.

Samuels, S.J. (1979/1997). The method of repeated readings. The Reading Teacher, 50,
376-381.

Savage, R., & Frederickson, N. (2005). Evidence of a highly specific relationship
between rapid automatic naming of digits and text-reading speed. Brain and
Language, 93, 152-159.

Schatschneider, C., Carlson, C.D., Francis, D.J., Foorman, B.R., & Fletcher, J.M. (2002).
Relationship of rapid automatized naming and phonological awareness in early
reading development: Implications for the double-deficit hypothesis. Journal of
Learning Disabilities, 35, 245-256.

Share, D.L., & Stanovich, K.E. (1995). Cognitive processes in early reading
development: Accommodating individual differences into a model of acquisition.
Issues in Education: Contributions from Educational Psychology, 1, 1-57.

Shimabukuro, S.M., Prater, M.A., Jenkins, A., & Edelen-Smith, P. (1999). The effects of
self-monitoring of academic performance on students with learning disabilities
and ADD/ADHD. Education and Treatment of Children, 22(4), 397-414.

Stage, S.A., Abbott, R.D., Jenkins, J.R., & Berninger, V.W. (2003). Predicting response
to early reading intervention from verbal IQ, reading-related language abilities,
attention ratings, and verbal IQ-word reading discrepancy: Failure to validate
discrepancy method. Journal of Learning Disabilities, 36, 24-33.

Stanovich, K.E. (1986). Matthew effects in reading: Some consequences of individual
differences in the acquisition of literacy. Reading Research Quarterly, 21, 360-
407.

Stanovich, K.E. (2000). Progress in understanding reading: Scientific foundations and
new frontiers. New York, NY: The Guilford Press.

145

Stanovich, K.E. (1980). Toward an interactive-compensatory model of individual
differences in the development of reading fluency. Reading Research Quarterly,
16, 32-71.Taylor, S.E., Frackenpohl, H., & Pattee, J.L. (1960). Grade level norms
for the components of the fundamental reading skill. Bulletin #3, New York:
Huntington, Educational Development Laboratories.

Stanovich, K. E., Cunningham, A. E., & Freeman, D. J. (1984). Intelligence, cognitive
skills and early reading progress. Reading Research Quarterly, 19, 278–303.

Sunseth, K., & Bowers, P.G. (2002). Rapid naming and phonemic awareness:
Contributions to reading, spelling and orthographic knowledge. Scientific Studies
of Reading, 6(4), 401-429.

Swanson, H.L., Trainin, G., Necoechea, D.M., & Hammill, D.D. (2003). Rapid naming,
phonological awareness and reading: A meta-analysis of the correlational
evidence. Review of Educational Research, 73, 407-440.

Thaler, V., Ebner, E.M., Wimmer, H., & Landerl, K. (2004). Training reading fluency in
dysfluent readers with high reading accuracy: Word specific effects but low
transfer to untrained words. Annals of Dyslexia, 54(1), 89 – 113.

Therrien, W.J. (2004). Fluency and comprehension gains as a result of repeated reading.
Remedial and Special Education, 25, 252-261.

Therrien, W.J., Wickstrom, K., & Jones, K. (2006). Effect of a combined repeated
reading and question generation intervention on reading achievement. Learning
Disabilities Research and Practice, 21(2), 89-97.

Torgesen, J. (2005). Effective interventions for older students with reading disabilities:
Lessons from research. Paper presented at the annual meeting of the Council for
Exceptional Children, Baltimore, MD.

Torgesen, J. (2004). Avoiding the devastating downward spiral: The evidence that early
intervention prevents reading failure. American Educator, 28(3), 6-47.

Torgesen, J. (1999). Individual differences in response to early interventions in reading:
The lingering problem of treatment resisters. Learning Disabilities Research and
Practice 15, 55–64.

Torgesen, J.K., & Hudson, R.F. (2006). Reading fluency: Critical issues for struggling
readers. In S.J. Samuels (Ed.) What research has to say about reading fluency.
Newark, DE: International Reading Association.

146

Torgesen, J.K., Wagner, R.K., & Rashotte, C.A. (1997). Prevention and remediation
of severe reading disabilities: Keeping the end in mind. Scientific Studies of
Reading, 1(3), 217-235.

Torgesen, J.K., Wagner, R.K., Rashotte, C.A., Burgess, S., & Hecht, S. (1997).
Contributions of phonological awareness and rapid automatic naming ability to
the growth of word-reading skills in second-to-fifth-grade children. Scientific
Studies of Reading, 1(2), 161-185.

Tressoldi, P.E., Vio, C., & Iozzino, R. (2007). Efficacy of an intervention to improve
fluency in children with developmental dyslexia in a regular orthography. Journal
of Learning Disabilities, 40(3), 203-209.

University of Oregon. (2005). Dynamic Indicators of Basic Early Literacy Skills
(DIBELS): Oral Reading Fluency, http://dibels.uoregon.edu/measures/orf.php.
2005.

Valleley, R.J., & Shriver, M.D. (2003). An examination of the effects of repeated
readings with secondary students. Journal of Behavioral Education, 12(1), 55-76.

van den Bosch, K., van Bon, W.H.J., & Schreuder, R. (1995). Poor readers’ decoding
skills: Effects of training with limited exposure duration. Reading Research
Quarterly, 30(1), 110-125.

Vaughn, S. (2001). Instructional grouping for reading with students with LD.
Intervention in School and Clinic, 36, 131-138.

Vellutino, F.R., Fletcher, J.M., Snowling, M.J., & Scanlon, D.M. (2004). Specific reading
disability (dyslexia): What have we learned in the past four decades? Journal of
Child Psychiatry and Psychiatry, 45(1), 2 – 40.

Wagner, R.K., Torgesen, J.K., & Rashotte, C.A. (1999a). Comprehensive test of
phonological processing. Austin, TX: Pro-Ed.

Wagner, R.K., Torgesen, J.K., & Rashotte, C.A. (1999b). Test of word reading efficiency.
Austin, TX: Pro-Ed.

Wagner, R.K., Torgesen, J.K., Rashotte, C.A., Hecht, S.A., Barker, T.A., Burgess, S.R.,
et al. (1997). Changing relations between phonological processing abilities and
word-level reading as children develop from beginning to skilled readers: A 5-
year longitudinal study. Developmental Psychology, 33, 468-479.

147

Wechsler, D. (2003). Wechsler Intelligence Scale for Children-Fourth Edition
Administration and Scoring Manual. San Antonio, TX: Harcourt Assessment.

Wiederholt, J.L., & Bryant, B.R. (2001). Gray Oral Reading Tests-4 (4
th
ed.) Austin, TX:
Pro-ed.

Wolf, M., & Bowers, P. (1999). The “double-deficit hypothesis” for the developmental
dyslexias. Journal of Educational Psychology, 91, 1-24.

Wolf, M. & Katzir-Cohen, T. (2001). Reading fluency and its intervention. Scientific
Studies of Reading, 5(3), 211-239.

Wood, F.B., Hill, D.F., Meyer, M.S., & Flowers, D.L. (2005). Predictive assessment of
reading. Annals of Dyslexia, 55(2), 193-216.

Woodcock, R.W. (1996). Woodcock reading mastery test—revised/normative update.
Circle Pines, MN: American Guidance Service.

Woodcock, R.W., McGrew, K.S., & Mather, N. (2001). Woodcock-Johnson III Tests of
Cognitive Abilities. Itasca, IL: Riverside Publishing.

148

Appendix A
Sample of Great Leaps Phonics/Words Page

149

Appendix B
Sample of Great Leaps Phrases Page

150

Appendix C
Sample of Great Leaps Story

151

Appendix D
Great Leaps Equal-Ratio Chart

152

Appendix E
Example of Skills for School Success Workbook Page

153

Appendix F
Skills for School Success Record Page

154

Appendix G
Comparisons of Pre and Posttest Mean Reading Scores Within Groups, GORT Measures,
Outlier Included
*Significant p<.0001 **Significant, p<.05 (2-tailed)

Group
Experimental Control
Measure Pre Post t ES Sig. Pre Post t ES Sig.
Rate 2.09 2.87 4.73 0.59 *.000 1.70 2.22 2.27 -0.32 **.032
SD 1.1 1.5 1.3 1.9
Accuracy 1.62 2.47 4.80 0.70 *.000 1.52 1.74 1.80 -0.16 .083
SD 1.0 1.4 1.3 1.4
Passage 1.79 2.69 5.68 0.71 *.000 1.70 2.07 1.99 -0.29 .057
SD 1.0 1.5 1.1 1.4

155

Appendix H
Comparisons of Pre and Posttest Mean Grade-Level Equivalencies Within Groups,
GORT, Outlier Included
* Significant, p<.01 **Significant, p<.05 (2-tailed)

Group
Experimental Control
Measure Pre Post t ES Sig. Pre Post t ES Sig.
Rate 1.50 2.29 4.73 0.64 *.000 2.30 3.01 4.44 0.56 *.000
SD 0.9 1.5 1.1 1.4
Accuracy 1.38 2.06 3.44 0.57 *.002 1.68 2.02 1.98 0.34 .059
SD 0.8 1.5 1.0 1.0
Passage 1.59 2.26 3.92 0.57 *.000 1.80 2.30 2.72 0.475 **.012
SD 0.9 1.4 1.0 1.1

156

Appendix I
Comparison Of GORT Mean Residual Gain Scores by Group, Outlier Included

Measure Group N Mean SD t ES Sig.
(2-tailed)
RA Experimental 34 .155 .94 1.38 0.35 .173
Control 27 -.195 1.04

ACC Experimental 34 .223 1.11 2.20 0.53 *.048
Control 27 -.280 .74

PAS Experimental 34 .150 .87 1.34 0.34 .186
Control 27 -.189 1.12

*Significant, p<.05

157

Appendix J
Comparisons of Pre and Posttest Mean Scores For GORT Comprehension Within
Groups, Outlier Included
Group PR ES Sig.

QC ES Sig.

RS ES Sig.

pre 2.53 9.53 3.39 Exp
N=34
post 3.76

0.63

*.000
13.85

0.65

*.000
3.94

0.01

.955
pre 2.70 10.26 3.93 Cont
N=27
post 3.59

0.45

**.005
12.89

0.43

***.010
3.89

0.05

.849
PR: Passages read; QC: Questions correct; RS: Ratio score
* Significant, p<.001 **Significant, p<.01 ***Significant, p<.05 (2-tailed)

158

Appendix K
Comparison Of GORT Mean Comprehension Scores by Group, Outlier Included

Measure Group N Mean SD t ES Sig.
(2-
tailed)
PR Experimental 34 .100 0.99 .882 0.23 .381
Control 27 -.126 .99

QC Experimental 34 .137 1.04 1.22 0.32 .228
Control 27 -.173 .91

RS Experimental 34 .027 1.01 .238 0.06 .813
Control 27 -.034 .99

159

Appendix L

Open-Ended Comprehension Questions for GORT Assessment, Form A

A-1
1. What was the man holding?
A pretty box; a present; a box; a surprise

2. Who ran to meet the man?
A little girl; his daughter

3. What did the man plan to do with the box?
Give it to the little girl
Give it as a present: QUERY—“Who is he giving it to?”

4. Why do you think Father wanted to give his little girl a surprise?
To make her feel good; to make her happy; because she was good

5. How do you think the girl felt when she saw the surprise?
Happy; good; excited; other similar adjectives

A-2
1. When does the little girl ride the bike slowly?
When a car is coming; when she sees a car
NOT: When she stops at a red light; when she sees a red light

2. Where do you think the girl is riding her bike?
In the street; in the road
In the city or town: QUERY—“Where in the city (town)?”

3. What is the main idea of this story?
Something along the lines of:
Riding a Bike; The New Bike; The Girl and Her Bike; Safe Bike Riding; The
Good Bike Rider

4. How would you describe this girl’s bike riding?
Safe; careful; cautious; safe; similar adjectives
Good: QUERY—“What makes her a good rider?”

5. How would you describe the bike?
New; yellow; has white stars
NOT: fast; slow; red; green

160

A-3

1. Who were the people who were working in the empty lot?
Neighbors; people from the neighborhood; people who lived nearby
Friends; boys; girls; parents: QUERY—“Can you be more specific?”

2. Who built the fence?
The parents; a group of parents

3. What did the boys take out of the lot?
Any of the following: old boards; trash; dry branches

4. Why did the people work so hard on the lot?
To make it safe; so the children would have a safe place to play; because the
lot was messy; because there was no place for the children to play

5. What do you think will happen next?
The people will go home; children will play in the new park

A-4
1. What did the girls do every morning at the farm?
Gather eggs
“Wake up”; “eat breakfast”: QUERY: What else did they do?
NOT: Fish, milk the cow, any other farm activity

2. What’s one thing the farmer taught the girls?
Any of the following: How to milk a cow; how to fish; how to gather eggs;
how to pick corn

3. What’s another thing the farmer taught the girls?
Any of the following not answered on the previous question: How to milk a
cow; how to fish; how to gather eggs; how to pick corn

4. Why did the girls have to wait to drink the milk?
It needed to cool off/to get cold
NOT: They had to milk the cow first

5. How do you know the farm was near a body of water?
They caught fish; they had ducks

161

A-5
1. Why was the jay thirsty?
She had just flown a long way; from flying
NOT: There was no water

2. Why couldn’t the jay drink the water?
Because it was too low in the jar; there wasn’t enough; the jar was too empty

3. What does the story make you think about the personality of this bird?
She was smart; she was clever; she was tricky; she was persistent
She was thirsty: QUERY—“What else? What kind of personality did he
have?”

4. Why were stones important in this story?
The jay used them to get the water; She dropped them in the jar to raise the
water level; some similar answer

5. How did the jay feel at the end of the story?
Proud; pleased with herself
Tired or happy: QUERY—“How else do you think she felt?”

A-6
1. In the era of the cowboy, what did the skilled cowboys do?
Round up the herds (cows); take the herds (cows) to market; take the cows on
a trail drive
NOT: Ride horses; take care of cows

2. What was the purpose of the trail drive?
To get the cows to the market; to sell the cows
NOT: To get the cows into the ranches; to round up the cows

3. What’s one way the cattle business changed?
Barbed wire made the land more manageable; cattle were shipped by train;
barbed wire made the cows easier to round up

4. What’s another way the cattle business changed?
Any of the above answers not given previously.

5. The cowboys’ skill at herding was essential for what activity?
Long trail drives
“Getting cows to market”: QUERY: How did they use the herding to get
the cows to the market?

162

A-7
1. Why do you think Cesar’s family had no permanent home?
They were poor; they moved around a lot picking crops
NOT: Because he went to lots of different schools

2. Why did Cesar go to so many different schools?
Because his family traveled from farm to farm picking crops; because his
parents were farm workers

3. Why was Cesar concerned about the farm workers?
They were poor; they were suffering; they needed health care; they needed
safer housing

4. How did the workers get better conditions?
They went on strike; they held peaceful protests; they held marches; they
organized a union
They joined together: QUERY—“Then what did they do?”

5. What kinds of protests did Cesar lead?
Peaceful ones; nonviolent protests; protest marches

A-8
1. Why was Mark nervous?
He had a tough assignment; he had to pick up a scary prisoner
He had his first assignment: QUERY—“Why did that make him nervous?”

2. How did the prisoner surprise Mark?
He looked peaceful; he was polite; he didn’t look like a criminal

3. What was the prisoner probably being tried for?
Assault; ruthless attacks on people; attacking people
“Hurting people”: QUERY: Can you be more specific?

4. What was Mark’s occupation?
Sheriff’s deputy
“Cop”; “policeman”: QUERY: Can you be more specific?

5. How does this saying apply to this story: “You can’t judge a book by its cover”?
Mark couldn’t judge the prisoner’s character by the way he looked/acted

Abstract (if available)

Abstract Despite recent advances in the science of teaching reading, there still exists a small percentage of students who fail to make the expected progress in reading-related skills. As they get older, these students are at great risk for dropout, behavior problems, and learned helplessness. Even if these struggling readers learn to decode adequately, fluency still remains a problem for many, and little is known about the effectiveness of fluency interventions for older students with severe reading delays. This study used a randomized experimental design to test the efficacy of a fluency intervention program on the reading outcomes of 60 students with severe reading disabilities in grades six through eight. Students in the experimental group received ten minutes a day of one-on-one fluency intervention with a trained paraprofessional, utilizing the Great Leaps Reading Program (Campbell, 1999)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Effects of guided reading (teacher supported) on reading comprehension scores for primary students who are learning disabled English language learners

PDF

What about the children left-behind? An evaluation of a reading intervention program

PDF

The effectiveness of the literacy for success intervention at Wilson Middle School

PDF

Examining the impact of LETRS professional learning on student literacy outcomes: a quantitative analysis

PDF

The effects of a multi-linguistic diagnostic spelling intervention on the writing achievement and writing self-perception beliefs of secondary students: phonology, orthography, and morphology

PDF

The effect of site support teams on student achievement in seven northern California schools

PDF

"Thinking in the middle" what instructional approaches are employed by urban middle school teachers to effect changes in African American students' learning outcomes and performances?

PDF

The knowledge, motivation and organizational influences that shape a special education teacher’s ability to provide effective specialized academic instruction

PDF

Improving foundational reading skills growth in middle school: a promising practices study

PDF

Reading prereferral intervention: developing, implementing and monitoring

PDF

The effect of a language arts intervention on the academic achievement of English language learners

PDF

The comparison of hybrid intervention and traditional intervention in increasing student achievement in middle school mathematics

PDF

Examining the effectiveness of the intervention programs for English learners at MFC intermediate school

PDF

An evaluation of the impact of a standards-based intervention on the academic achievement of English language learners

PDF

Raising student achievement at Eberman Elementary School with effective teaching strategies

PDF

Perceptions of inclusion: high school students diagnosed with learning disabilities and their level of self-efficacy

PDF

Globalization, curricular elements, organizational practices and perceived student outcomes in California schools

PDF

Impact of positive student–staff relationships and the social–emotional outcomes of Black high school students classified with an emotional disability

PDF

The effect of reading self-efficacy, expectancy-value, and metacognitive self-regulation on the achievement and persistence of community college students enrolled in basic skills reading courses

PDF

A social skills curriculum intended to increase the employment success of secondary transition aged students with moderate to severe intellectual disabilities

Asset Metadata

Creator Spencer, Sally Atwood (author)

Core Title The effects of a reading fluency intervention on the reading outcomes of middle school students with severe reading disabilities

School Rossier School of Education

Degree Doctor of Education

Degree Program Education

Publication Date 09/04/2008

Defense Date 07/14/2008

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag at risk students,fluency,learning disabilities,OAI-PMH Harvest,Reading,reading disabilities,reading interventions,Special Education

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Hocevar, Dennis (committee chair), Manis, Franklin R. (committee chair), Sears, Sue (committee member)

Creator Email saspence@usc.edu,sspencer@csun.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1584

Unique identifier UC1315345

Identifier etd-Spencer-2336 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-107478 (legacy record id),usctheses-m1584 (legacy record id)

Legacy Identifier etd-Spencer-2336.pdf

Dmrecord 107478

Document Type Dissertation

Rights Spencer, Sally Atwood

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu