711223

research-article2017
LDXXXX10.1177/0022219417711223Journal of Learning DisabilitiesKoen et al.

Article
Journal of Learning Disabilities

The Location and Effects of Visual

1­–17
© Hammill Institute on Disabilities 2017
Reprints and permissions:
Hemisphere-Specific Stimulation on sagepub.com/journalsPermissions.nav
DOI: 10.1177/0022219417711223
https://doi.org/10.1177/0022219417711223

Reading Fluency in Children With journaloflearningdisabilities.sagepub.com

the Characteristics of Dyslexia

Bobbie Jean Koen, PhD1, Jacqueline Hawkins, EdD1,

Xi Zhu, PhD1, Ben Jansen, PhD1, Weihua Fan, PhD1,
and Sharon Johnson, PhD1

Abstract
Fluency is used as an indicator of reading proficiency. Many students with reading disabilities are unable to benefit from
typical interventions. This study is designed to replicate Lorusso, Facoetti, Paganoni, Pezzani, and Molteni’s (2006) work
using FlashWord, a computer program that tachistoscopically presents words in the right or left visual hemi-field in
English and locates through fMRI imaging the processing areas involved in fluency development. Our participants were
15 students who were ages 8 to 19 years and had reading disabilities randomly assigned to Intervention (n = 9) and
Delayed Intervention (n = 6) groups. Functional imaging studies focused on analyzing activations in the left hemisphere (LH)
superior temporal gyrus, the inferior frontal gyrus, and the LH inferior occipito-temporal/fusiform area (visual-word form
area [VWFA]). Analysis of intervention data showed that 6 of the 9 Intervention group participants (67%) achieved levels
of automatic processing and increased their reading rate by an average of 20 words per minute after participating in the
FlashWord intervention. Analyses of fMRI group activation maps and mean activation levels in regions of interest document
processing changes in VWFA activations that could be related to the increase in reading speed and confirm these locations
as essential to developing fluency.

Keywords
brain imaging, neuropsychology, phonological processing, response to intervention

As a persistent component of reading disability, the behav-
ioral and neurobiological mechanisms that influence flu-
ency are not well understood. The neurobiological origins
of fluency can be seen in the early work of physiologist
Donald Hebb, who proposed the concept of unitization in
1949. LaBerge and Samuels (1974) went on to apply this
idea to more complex visual levels (familiar letter patterns)
and in other modalities such as phonological representa-
tions, focusing on the automaticity of processing that relates
to the neural efficiency hypothesis (Mrazik & Dombrowski,
2010). Educators have long used fluency as a measure of
reading performance and a sign of superior comprehension
but have not been able to prescribe instructional practices
that improve reading speed for all children, especially those
with specific reading disabilities. Students are expected to
read fluently as a function of instruction and maturity, and
the common preoccupation with measuring fluency as the
rate and accuracy of oral reading ignores the multiple other
dimensions of fluency, particularly the contributions of
lower level subskills: graphological features of letters,
orthographic regularities of letter combinations, the seman-
tic features of words, and the semantic-syntactic constraints
of word sequences, investigated first by Doehring (1976).
Kame’enui, Simmons, Good, and Harn (2000) proposed
a developmental conceptualization of fluency that included
the building of proficiency in underlying component skills
of reading, such as phoneme awareness, effectively merg-
ing the influences of skill development with processing
speed and accuracy into a continuum of reading proficiency.
It is this continuum that Wolf and Katzir-Cohen (2001) refer
to in their comprehensive definition of fluency that illus-
trates the developing of automaticity including perceptual,
phonological, orthographic, and morphological processes at
the letter, letter-pattern, and word levels as well as semantic
1
University of Houston, TX, USA
Corresponding Author:
Bobbie Jean Koen, Policy and Planning, University of Houston, 320K
Student Service Center 1, Houston, TX 77204-5029, USA.
Email: bjkoen@central.uh.edu
2 Journal of Learning Disabilities 00(0)
and syntactic processes at the word level and the connected-
text level. Since the development of reading fluency
depends on every process and skill used in reading,
Kame’enui (2007) proposes that it also requires an increase
in accuracy and proficiency in every underlying compo-
nent. It would follow that failure to acquire these processes
and skills could result in serious and persistent reading
dysfunctions.
Although it is limited, scientific investigation of fast pro-
cessing includes not only behavioral data but results from
many new technologies. The component-based definition
of fluency provides the theoretical framework for investi-
gating how the relative contributions of letter-sound asso-
ciation, phonological awareness, orthographic pattern
recognition, comprehension, and rapid letter naming affect
fluent word and connected-text reading in children with the
characteristics of dyslexia. Using multivariant analysis of
the results of a battery of reading skills measures of 123
second- and third-grade student with dyslexia, Katzir et al.
(2006) found that rapid naming, orthographic pattern recog-
nition, and word reading fluency moderately predicted dif-
ferent dimensions of connected-text reading (i.e., rate,
accuracy, and comprehension) whereas phonological
awareness contributed only to the comprehension dimen-
sion of connected-text reading when controlling for the
children’s gender, age, socioeconomic status, and IQ. The
unexpected finding that rapid naming was more related to
reading speed than was phonological awareness under-
scores the importance of these subskills.
Misra, Katzir, Wolf, and Poldrack (2004) used fMRI to
investigate this rapid naming phenomenon more closely by
looking at the activation patterns elicited by serial letter
rapid automatized naming (RAN) tasks and object RAN
tasks. Results from both letter and object naming scans,
when compared to fixation, indicated significant activations
within all three systems of the reading network: the frontal
areas, bilaterally along the ventral visual pathway, and in
the left hemisphere (LH) dorsal posterior regions. However,
areas that were differentially activated were more active
during the Letter Naming task, especially the angular gyrus
(important for the interpretation of orthographic symbols)
and the superior parietal lobule. This finding underscores
the conclusion that RAN of letters activates many of the
same regions and pathways as used when reading words
and is, therefore, of greater predictive ability to reading flu-
ency than are object RAN tasks.
Unexpected challenges arise from the documentation of
natural hemispheric and regional subsystem shifts in read-
ing behavior that must be considered in evaluating neuro-
biological data. Licht, Bakker, Kok, and Bouma (1988)
used event-related potentials related to word naming during
a 4-year longitudinal study and found that most children
shift the processing of words from the right hemisphere
(RH) to the left by the end of Grade 1, beginning of Grade
2 (ages 6–7 years). This was the first evidence of a long-
suspected major change in brain regions for the develop-
mental processing of written language.
Shaywitz et al. (2002) found another naturally occurring
developmental shift in activation of the three cortical read-
ing subsystems that occurs slightly later, around Grade 4
(age 10.5 years). They observed that younger nonimpaired
readers showed stronger engagement of the dorsal temporo-
parietal system—including the angular gyrus, supramarginal
gyrus in the inferior parietal lobule, and posterior aspect of
the superior temporal gyrus (STG or Wernicke’s area)—and
the anterior system—posterior aspect of the inferior frontal
gyrus (IFG)—but limited use of the ventral system—LH
inferior occipito-temporal/fusiform area, extending anteri-
orly into the middle temporal gyrus (MTG) and inferior tem-
poral gyrus. In contrast, nonimpaired readers older than 10.5
years of age showed increased engagement of the ventral
system, which is associated with increasingly skilled read-
ing, that is, positively correlated with higher reading scores.
These results would seem to support the suggestion offered
by Pugh et al. (2001) and Sandak, Mencl, Frost, and Pugh
(2004) that this ventral system fails to develop in students
with the characteristics of dyslexia, not because of impair-
ment but as a result of lack of proper stimulation. This
exploratory study intends to look for evidence of fluent pro-
cessing that could include both of these naturally occurring
processing shifts in reading development.
Problem
As the field of cognitive neuroscience becomes more pre-
cise in the identification of the cortical subsystems that sup-
port the development of reading, a central question involves
the behavioral and neurobiological evidence in various sys-
tems to brain-based intervention procedures. Shaywitz et al.
(2004) found increasing LH activation in students with dys-
lexia using fMRI as a result of explicit alphabetic principle
and phonological awareness training. Also, studies using
visual hemisphere-specific stimulation (VHSS) have dem-
onstrated surprising increases in fluency for reading-dis-
abled readers.
Bakker, Bouma, and Gardien (1990) identified children
with dyslexia in light of the known hemispheric subservi-
ence in learning to read as L-dyslexics or P-dyslexics, based
on error analysis, distribution of brain responses, and
behavioral measures. They suggest that L-dyslexics pre-
dominately generate LH strategies from the very onset of
learning to read and therefore are relatively insensitive to
the perceptual features of the text. L-dyslexics manifest a
hurried and inaccurate style of reading with many substan-
tive errors. P-dyslexics are children who began the learn-
ing-to-read process in the RH but never progress from there
and so are overly sensitive to perceptual features of the text
and read slowly with a fragmented style.
Koen et al. 3
Bakker et al. (1990) theorized that since L-type dyslex-
ics had difficulty using right-hemispheric strategies dur-
ing initial reading, they might benefit from specific
stimulation of the RH and that the opposite is true for
P-dyslexics: they had not shifted to LH processing and so
would profit from specific stimulation of the LH. In gen-
eral, specific stimulation of a hemisphere can be accom-
plished tactilely or visually, by the lateral presentation of
reading material in the left visual field or to the fingers of
the left hand in L-dyslexics and in the right visual field or
to the fingers of the right hand in P-dyslexics. This study
actually treated the children using a wooden tactile train-
ing box and plastic letters that are manipulated out of
sight. L-types were given easy-to-visualize concrete words
to form and trace with the left hand to stimulate the RH,
and P-types were given difficult-to-visualize abstract
words to form and trace with the right hand to stimulate
the LH. While P-dyslexics showed a decrease in fragmen-
tation errors on both word and text reading after interven-
tion, L-dyslexics decreased substantive errors only on text
reading. The authors identified several limitations in their
methodology and intervention that may have contributed
to the somewhat mixed results, but the positive effects of
even motor stimulation to the less activated hemisphere on
reading performance are encouraging. Further, these
results suggest that the dyslexia subtyping procedures
appear to be valid techniques for matching interventions
to processing systems.
Lorusso, Facoetti, Paganoni, Pezzani, and Molteni
(2006) employed computer technology to achieve much
stronger results in an Italian population of impaired readers
based on the strength of these theoretical and neurobiologi-
cal foundations. These researchers adopted the subtyping
of students with the characteristics of dyslexia proposed by
Bakker et al. (1990) and added M-type dyslexia: a mixed
type demonstrating both slow and inaccurate reading. They
extended the theory to propose that M-dyslexia children
would benefit from stimulation of both hemispheres, alter-
nately. Their new technology included a modified version
of a computerized system for VHSS, “FlashWord” (Masutto
& Fabbro, 1995). After 1,440 min (24 hours) of interven-
tion in Italian, these researchers used only behavioral mea-
sures and found that all students with the characteristics of
dyslexia, regardless of subtype, not only improved in accu-
racy and fluency compared to nonimpaired controls but
also showed gains in spelling, memory, and general pro-
cessing speed. Additionally, the students with the charac-
teristics of dyslexia gained 0.33 syllables per second more
in reading speed during the same period of time than their
nonimpaired controls. These remarkable results suggest
that putting pressure on the system by requiring very fast
processing of the presented stimuli may produce a greater
degree of automatization of the component processes. It is
this automatization of the underlying lexical and sublexical
processes that Wolf and Katzir-Cohen (2001) validated as
critical influences on fluent reading of connected text in
their comprehensive definition of fluency.
Hypothesis
The main goal of the present analysis is twofold. First, we
begin with the reconceptualizations of the definition of flu-
ency and used fMRI technology to localize brain activity in
regions of interest (ROIs) before and after VHSS training in
students with the characteristics of developmental dyslexia.
Using postintervention activation maps, we expect to iden-
tify brain regions involved in the training of fast processing
in reading and determine the effects in these brain regions
that signify the development of fluency. Comparing the
activation maps of the students with the characteristics of
dyslexia who participated in the FlashWord Intervention
condition with those of the students in the Delayed
Intervention condition should reveal the specific effects of
the VHSS training on the processing activities in the read-
ing core subsystems. Based on the three core reading sub-
systems first documented by Shaywitz and Shaywitz
(1999)—the anterior processing subsystem located in the
IFG of mostly the LH, which facilitates phonological
encoding; the temporal-parietal subsystem found in the
inferior MTG and STG in both hemispheres, which pro-
vides rule-based analysis and learning; and the occipital-
temporal subsystem found in extra striate areas located
posterior to V1, which applies orthographic encoding, and
the visual-word form area (VWFA), which supplies sight
words—statistical analyses should produce the relative
influence of these regions on the development of fast
processing.
Second, we expected to quantify the extent to which
VHSS training increases fluency scores in students with the
characteristics of dyslexia. We tested the hypothesis that
subtyping students with the characteristics of dyslexia and
administering VHSS intervention based on those subtypes
(FlashWord-modified and in English) would improve flu-
ency performance across dyslexia subtypes more effec-
tively than other currently used reading fluency programs.
By comparing the fluency scores from the beginning of the
semester and the end of the semester of the children with the
characteristics of dyslexia who completed the VHSS inter-
vention with those who did not, the difference was expected
to quantify the greater increases in fluency achieved by the
students with dyslexia who completed their assigned inter-
vention. This should demonstrate the efficacy of this inter-
vention with this special population and within all dyslexia
subtypes: L, P, and M.
We investigated the following research questions:
1. Regarding postintervention activations, what effects
in the ROIs signify the development of fluency? We
4 Journal of Learning Disabilities 00(0)

Table 1.  Study Participant Demographics.

Participant code Gender Age (years) Race Reading instructional grade level
a
Delayed Intervention group
CB F 8 African American 1
CC F 10 Anglo 4
PA F 11 African American 4
MG M 14 Anglo 5
MJ M 19 Anglo 8
ML F 16 Anglo 2
Group summary M=2 Range = 8–19 Anglo: 4 (67%) Range = 1–8
F=4 Mode: NA African American: 2 (33%) Mode: 4
M = 13 M=4
Intervention groupb
PC M 8 Anglo 2
PE M 10 African American 4
MA M 16 Anglo 7
MC M 16 Anglo 4
MD F 16 Anglo Primer-3
ME M 17 Anglo 3
MF M 15 Anglo 8
MI M 16 Anglo 6
MN M 19 Anglo 10
Group summary M=8 Range = 8–19 Anglo: 8 (89%) Range = Primary 3–10
F=1 Mode: 16 African American: 1 (11%) Mode: 4
M = 15 M=4
All participants summaryc M = 10 Range = 8–19 Anglo: 12 (80%) Range = Primary 3–10
F=5 Mode: 16 African American: 3 (20%) Mode: 4
M = 14 M=5

Note. F = female; M = male; NA = not available.

a
n = 6. bn = 9. cN = 15.

are especially interested in tracking the hemispheric Methodology

and systematic shift as a function of participation in
the intervention.
2. To what extent does VHSS training increase flu-
ency scores in students with the characteristics of
dyslexia? Is it more effective than other commonly
used fluency instructional techniques?
Evidence suggests that there could also be a reduction
in activation in the ROIs as a result of the intervention.
Haier and Benbow (1995) found that the people with the
highest scores on a cognitive task used less brain energy
and concluded that intelligence is associated with a more
efficient brain rather than a harder working brain (Haier
et al., 1988). This “neural efficiency hypothesis” implies
that more efficient brains require less energy and hence
demonstrate less overall cortical activation (Mrazik &
Dombrowski, 2010). The exploratory nature of this study
must anticipate both higher and possibly lower activation
levels in these participants as a result of the intervention
depending on the relative reading development of the
participants.
Sample
A total of 20 students, 5 females and 15 males, ranging in
age from 8 to 19 years, were recruited from three private
schools in a large urban setting. All of the families self-
selected their students’ participation in the study by return-
ing a Release of Confidentiality and were interviewed by
the principal investigator to discuss in detail MRI safety
information and answer any questions about the research or
intervention. As a result of these visits, 2 students were
found to be unsuitable for fMRI imaging (English as a sec-
ond language and orthodontia), and they agreed to with-
draw from the study (see Table 1).
Analysis of standardized test results showed that all of
the rest of the students, except for 2, qualified as students
with the characteristics of developmental dyslexia according
to the International Statistical Classification of Diseases
and Related Health Problems, 10th revision (World Health
Organization, 1992) criterion, which was used by Lorusso
et al. (2006) to identify their participants. This criterion was
interpreted for this study as performance on a standardized
Koen et al. 5

Table 2.  Standardized Reading Testing and Intervention Program Assignment Justification.

Standardized test, grade level, Analysis of errors;

Participant year, subtest = scale score or Reading level, justification for program Dyslexia Recommended
code percentile ranking wpm, errors assignment subtype program
CB Stanford, 1, 2010, Primer, 50, no Slow processing; good Mixed LH/RH
Word Reading = 481 errors comprehension
CC Stanford, 4, 2010, 4, 128, no errors Reports using meaning to L-type RH
Reading Comprehension = 627 decode unknown words
PA Stanford, 4, 2010, Total Reading 2, 92, 2 errors Both errors were meaning L-type RH
= 581 based
PC Out of state, 2, 2010, Visual 3, 78, no errors General skill level and L-type RH
Processing/Reading (no personality type
scores available)
PE Stanford, 4, 2010, Total Reading 1, 40, 6 errors 5 errors were phonics based P-type LH
= 508
MA Stanford, 7, 2010, Reading 7, 53, 4 errors All errors were meaning Mixed RH/LH
Comprehension = 688 based; slow processing
MC Stanford, 4, 2010, Word Study 4, 69, 3 errors All errors were phonics P-type LH
= 578 based
MD Stanford, Primer-3, 2010, 3, 87, 8 errors 5 errors were phonics based; Mixed LH/RH
Reading Vocabulary = 45th 3 errors were meaning
percentile based
ME Stanford, 3, 2010, Word Study 2, 54, 6 errors All errors were meaning L-type RH
= 544 based
MF Stanford, 8, 2010, Reading 8, 102, no errors Slow processing P-type LH
Comprehension = 662
MG Stanford, 5, 2010, Spelling = 4, 48, 5 errors All errors were meaning L-type RH
<12th percentile based
MI Stanford, 6, 2010, Word Study 5, 106, 4 errors 3 errors were meaning based; Mixed RH/LH
= 605 1 error was phonics based
MJ Stanford, 8, 2010, Reading 8, 98, 3 errors Errors were different; L-type RH
Comprehension = <8th personality type
percentile
ML Stanford, 2, 2010, Word Study 1, 70, 5 errors 4 errors were meaning based L-type RH
= <6th percentile
MN Stanford, 10, 2010, Vocabulary 8, 115, 5 errors All errors were phonics P-type LH
= 703 based

Note. wpm = words per minute; LH = left hemisphere; RH = right hemisphere.

assessment of text reading that is reflected in a Stanford
Achievement Test scale score of 600 or less in at least one of
the tests for speed and accuracy in reading and/or spelling,
despite an average IQ, as ensured by each school administra-
tor (“Beyond the Numbers,” 2012). When scale scores were
not available, percentile measurements of 50 or less were
used to show lack of reading proficiency. These scores,
taken with the reading level/age discrepancy of the student,
were determined to qualify the remaining 16 students (see
Table 2).
The participating students represent a wide range of ages
and reading grade levels. Both randomly chosen treatment
groups also show this range of functioning, but the
Intervention group has a higher average age and a greater
reading grade level range than the Delayed Intervention
group. There were also more girls than boys in the Delayed
Intervention group, with the opposite in the Intervention
group. One student refused to participate in the interven-
tion, so his data were excluded, bringing the total to 15
participants.
Additionally, preintervention fluency assessment scores
were used to subtype the students with the characteristics of
dyslexia. All participants were classified as having P-, L-,
or M-dyslexia on the basis of their reading speed and read-
ing errors (“time-consuming errors” such as fragmentations
and repetitions or “substantive errors” such as substitutions
and omissions) according to the following formula based on
Lorusso et al. (2006):
1. P-type if reading speed is at least 1 SD below age
mean (i.e., z < –1) and the proportion of time-con-
suming errors over total errors is ≥ 60%.
6 Journal of Learning Disabilities 00(0)
2. L-type if reading speed is no more than 1 SD below
age mean (i.e., z ≥ –1) and the proportion of substan-
tive errors over total errors is ≥ 60%.
3. M-type in all other cases (prevalence of time-con-
suming errors but reading speed above −1 SD; prev-
alence of substantive errors but speed below −1 SD;
presence of an equivalent amount of both kinds of
errors).
The process of using these subtypes to match the opti-
mal hemisphere-specific stimulation to each student for
remediation purposes required some modifications to
accommodate different kinds of reading errors found in
English readers (see Table 2). In addition to the “substan-
tive” errors described by Bakker et al. (1990), some readers
of English tend to create meaning based on their experi-
ences and vocabulary. These fabrications tend to have little
to do with what is printed on the page but make perfect
sense when lifted from the printed text. Students who com-
mitted these kinds of errors were considered L-type since
they were using the context of the sentence to produce
meaningful, though wrong, substitutions and were given
the RH intervention program. All students provided base-
line fluency scores and end-of-study fluency scores through
their regular classroom assessments when possible.
Experimental Design
This fMRI experiment used a mixed design in that the
events of interest were randomized with perceptual controls
to provide robust event-related activation maps and esti-
mates of hemodynamic response, modeled after one used
by Richards et al. (2006). The word pairs (phonological
analysis) and letter match (perceptual control) stimuli in
this study were randomly presented every 12 and 18 s,
respectively, within each run. Analysis of the brain activa-
tion provided four contrasts—Word Pairs task, Letter Match
task, Word Pairs greater than Letter Match, and Letter
Match greater than Word Pairs—and should isolate the pho-
neme-mapping processes and the associated brain regions
uniquely involved in phonological analysis apart from letter
processing alone.
The Letter Match task required the child to decide
whether two letter strings (e.g., szpy and sxpy) printed in
all black letters and presented simultaneously one above
the other matched exactly. The length of the letter strings
was comparable to the length of the pseudowords used in
the phonological analysis task. This control task required
attention to all letter positions but did not involve assign-
ing speech sounds to letters. The child was directed to
press a “Yes” button if the letter strings matched or to
press a “No” button if the letter strings did not match. For
the phonological task condition, Word Pairs, the word
pairs were decodable nonwords printed in black, each
Figure 1.  Sample Visual Stimuli.
containing a letter or group of letters printed in a contrast-
ing color, also presented visually, one above the other
(see Figure 1). The child was directed to press the “Yes”
button if the contrasting-color letter(s) in the top word
could stand for the same sound as the contrasting-color
letter(s) in the bottom word and to press a “No” button if
the target letters stood for different sounds. All responses
had to occur within the 12- to 18-s duration of a slide in
order to be recorded.
During the fMRI scanning, 116 total slides were ran-
domly presented to each participant, 60% (69) showing
word pairs and 40% (47) showing letter matches. Slides
with the letter match stimulus appeared for 12 s, and slides
showing word pairs were presented for 18 s, for a total func-
tional scanning time of 27 min per participant. One half of
each of the task condition slides were yes. All words and
letters were presented at the center of the screen printed in
60-pica black (or pink) Calibri lower-case letters.
Procedures
Scans were conducted at the Baylor College of Medicine
Human Neuroimaging Lab. Before each scan, the students
were thoroughly trained on the Word Pairs and Letter
Matching tasks. Following the preintervention scanning
session, the students in the Intervention condition partici-
pated in the computerized VHSS program, FlashWord,
modified according to their dyslexia subtype at their
schools. Each school adopted its own schedule for interven-
tion, and volunteers generally administered the program
after training from the principal investigator. The principal
investigator maintained weekly communication with the
volunteers and students, updating and scoring each stu-
dent’s Intervention binder and meeting with volunteers to
make sure that minimum competencies were being met as
the processing time was being decreased. Finally, all of the
students with the characteristics of dyslexia returned for the
postintervention scanning session using the exact same
design, task, and procedures as before. Care was also taken
to ensure that postintervention, the students were scanned
Koen et al. 7
using the same machine as they had been for the preinter-
vention scan.
Intervention. FlashWord, version 2.2, written by Franco
Fabbro and Cristina Masutto (copyright 1995–2004 by Edi-
trice TecnoScuola) and used for this research by permis-
sion, is a computer program that uses a game format to
present words or phrases in the right or left visual hemi-
field at increasingly rapid rates. The ability to process words
and phrases at rates of 250 to 100 milliseconds is generally
considered to reflect “emerging fluency” (Bakker et al.,
1990). The student sees the words or phrases projected on
either the right or left side of the computer screen, stimulat-
ing either the right or left visual field and the opposite brain
hemisphere, according to his or her dyslexia subtype. Ocu-
lar fixation is monitored by asking the child to follow a
luminous dot oscillating up and down on the screen at an
adjustable speed. A word is flashed only if the child clicks
on the mouse at the exact moment the dot is crossing the
central target. The student’s task is to read the words as they
are flashed on the screen.
The word lists created for the English version of
FlashWord were modeled on the Italian word lists following
general guidelines from the authors. The Left Hemisphere
Program word lists were structured to mirror traditional
reading instruction sequences: short vowel patterns, long
vowel silent “e” patterns, regular vowel and consonant pat-
terns, suffixes, irregular vowel and consonant patterns,
vowel-r patterns, diphthongs, final stable syllables, and pre-
fixes. The words themselves were generated from lists of
words for teachers like The Yellow Pages for Students and
Teachers (The Kid’s Stuff People, 1980) and Cypress-
Fairbank’s Dyslexia Handbook (an internally developed
resource for teachers in the Region IV Education Service
Center in Texas). The phrases were found in Fry Instant
Phrases (Fry, n.d.-a) and the Dyslexia Handbook. To create
the RH program, nonsense words and phrases were formed
following the phonetic structure of the LH lists. Additional
lists of related words were original to this study, while high-
image-value words (words that tend to be memorized
instead of phonologically decoded) were selected from
Fry’s 600 (Fry, n.d.-b; see Appendix A).
Before the beginning of each lesson with the student, cri-
teria for word presentation (target speed and tolerance)
were set and recorded, and presentation times were recorded
in the student’s intervention binder. Students using the RH
program also chose the colors of the words and background
and the font displayed. The longest presentation times
(1,000 milliseconds for words and 2,000 milliseconds for
phrases) were used in the first sessions, initially so that the
child was able to read the word list with at least 70% accu-
racy, and later when more complex stimuli were presented
for the first time. As the child’s reading performance
improved, presentation times were shortened in the
following sessions so as to keep pressure on the system of
word recognition. The lists of 25 to 80 words/phrases
became increasingly difficult in terms of word length and
complexity so that final lists also included short sentences
selected according to familiarity and predictability. To fur-
ther stimulate anticipation in P-types, there was progressive
shortening of presentation times within the same session.
Nine of the students in the Intervention condition com-
pleted 1,440 min (24 hours) of the recommended version of
the FlashWord computer program during one semester of
school. Six students in the Delayed Intervention group par-
ticipated in the usual reading fluency program at their
schools and used the FlashWord computer program only
after the second scan. All students participated in the two
fMRI scans and statistical analysis, so the Delayed
Intervention group provides a true comparison of the VHSS
treatment effects on this target population.
Acquisition of MRI and fMRI scans.  Structural and functional
MRIs were performed on one of two Siemens head-only 3T
Allegra Magnetic Resonance scanners. Scanning included a
192 transversial slice, high-resolution set of anatomical
images in plane with functional data (TR/TE 1200/2.93 mil-
liseconds; fast spoiled gradient echo pulse sequence; 0.96 ×
0.96 × 0.89 millimeters; 256 × 208 matrix). This anatomical
series was followed by four fMRI series using two-dimen-
sional gradient echo planar pulse sequence (TR/TE 2000/50
milliseconds, 26 transversial slices; at 4 millimeters with
0% overlap, 64 × 64 matrix). Children viewed the stimulus
on a rear-projection video display (NEC GT2150) using a
mirrored lens attached to the head coil. If a student required
glasses to see the stimulus clearly, magnetic resonance–
compatible frames with insertable polycarbonate lenses
(Solo Bambini) were created. Participants indicated their
responses on two 2-button optical response pads (Current
Designs, Inc.), one held in each hand. Most of the students
allowed the principal investigator to mark “Y” or “N” on
their thumbs to help them remember. They indicated a
“Yes” or “No” response by pressing a certain button within
the 12- to 18-s stimulus presentation in order to be counted
as correct. These responses were monitored during scan-
ning to ensure that the participant was awake and on task.
Total MRI scans lasted approximately 32 min, 10 s.
fMRI image and data analysis.  Initially the fMRI data were
visually analyzed to assess the amount of movement and
its effect on the quality of the activation maps. Preprocess-
ing of the data included slice timing correction to mathe-
matically reconcile imaging differences, co-registration to
a standard brain to align the images, and correction of
severe head motion to improve activation detection. Nor-
malization was used to create group data, spatial smooth-
ing was done using a 3-millimeter blurring filter, and
temporal filtering was conducted. Due to a large amount
8 Journal of Learning Disabilities 00(0)

Table 3.  Fluency Summary.

Preintervention: Program/ Reached fluency? Postintervention:

Participant wpm/grade lowest speed: (100–250 milliseconds) wpm/grade Difference
code reading level words-phrases or in DI group (Yes/No) reading level (wpm)
CB 24/2 LH DI 50/2.0 26
CC 128/4 RH DI 120/5.0 −8
PA 92/2 RH DI 116/3.0 14
PC 78/3 RH/75–125 Yes 120/4.0 42
PE 40/1 LH/150–200 Yes for words 51/1.5 11
MA 53/7 RH/250–700 No, yes 63/8.0 10
LH/80–150
MC 69/4 LH/80–80 Yes 95/4.5 26
MD 87/3 LH/500–700 No, no 75/3.5 −12
RH/300–400
ME 54/2 RH/400–1000 No 55/3.0 1
MF 102/8 LH/70–80 Yes 118/8.0 16
MG 48/4 RH DI 52/4.0 6
MI 106/5 RH/200–500 No, yes for words 103/6.0 −3
LH/70–500
MJ 98/8 RH DI 103/8.0 5
ML 70/1 RH DI 71/1.0 1
MN 115/8 LH/40–80 Yes 131/8.0 16

Note. wpm = words per minute; DI = Delayed Intervention; LH = left hemisphere; RH = right hemisphere.

of movement during some of the scans, it was also neces-
sary to remove from estimation those images that were
produced while parameters exceeded the acceptable range.
In our final data analysis, there were no participants with
movement greater than 3 millimeters of translation or 3°
of rotation. There were also no significant differences
between the root mean square of absolute movement
(movement from this time point to a reference time point)
or relative movement (movement from this time point to
next time point) across x, y, or z axes of translation or
three rotations between intervention group and delayed
group (p > .5).
At the first level of analysis, the preprocessed fMRI
images, the stimulus onset times for each condition, and the
six movement parameters were correlated to create a design
matrix of the data. This model was estimated, and four con-
trasts were created: the Letter Match condition (showing
activation, a correlation coefficient at each voxel of the
brain, for only the letter match stimuli), the Word Pairs con-
dition (showing activation from only the word pairs stim-
uli), Letter Match Greater than Word Pairs condition
(showing activation from the letter match stimuli that sur-
passed the activation from the word pairs stimuli), and
Word Pairs Greater than Letter Match condition (showing
activation from the word pairs stimuli that surpassed the
activation from the letter match stimuli).
The average activation of magnetic resonance signal for
each ROI was determined using cluster-size thresholding, a
technique wherein data are analyzed using a relatively
liberal alpha value (e.g., p < .1) for voxel-wise comparisons.
For this study, we used a minimum of five voxels to define a
cluster. These clusters were identified within the three core
reading subsystems: the anterior processing system in the
IFG, the temporal-parietal system in the STG and MTG, and
the occipital-temporal system in the VWFA. These regions
have already been identified as collaborative areas that pro-
duce particular reading behaviors.
This experiment evaluated the influence of the core read-
ing subsystems (the occipital-temporal, temporal-parietal,
and anterior activation sites based on previous research)
and other sources of variability (nuisance factors or error)
on fluency and accuracy using multiple linear regression.
Our analysis should have revealed the relative influence of
each of the core reading subsystems on the development of
fluency.
Results
Behavioral Data
Preintervention fluency scores collected in January and pos-
tintervention fluency scores gathered in May or June were
examined by participant and by group to determine ranges of
net gain and mean gain (see Table 3). Participants in the
Intervention group (n = 9) produced preintervention scores
ranging from 40 to 115 words per minute (wpm) (M = 78
wpm) and postintervention scores ranging from 51 to 131
wpm (M = 90 wpm). Participants in the Delayed Intervention
group (n = 6) had preintervention scores ranging from 24 to
Koen et al. 9
128 wpm (M = 77 wpm) and postintervention scores ranging
from 50 to 120 wpm (M = 85 wpm). The net gain during this
6-month period demonstrated by the Intervention group was
11.9 wpm and by the Delayed Intervention group was 7.3
wpm. A total of 2 participants in the Intervention group and
1 in the Delayed Intervention group actually produced
slower postintervention scores compared to their preinter-
vention assessment, so the averages mask substantial gains
made by some students in both groups.
A paired-samples t test compared preintervention flu-
ency scores with postintervention scores for participants in
the Intervention and Delayed Intervention groups. The cor-
relation (.879) of these scores was significant (p < .001),
indicating a strong relationship between the scores. The t
test for the independent samples was also significant, t(14)
= −2.81, p < .05, indicating that there were differences
between the reading fluency scores of the students in both
groups. A very high significant correlation (.94, p < .01)
was found between those participants in the Delayed
Intervention group and their pre- and postintervention flu-
ency scores, but no significant difference was found
between preintervention fluency (M = 76.67, SD = 37.32)
and postintervention fluency (M = 85.33, SD = 31.68).
Examining the relationship of the Intervention group par-
ticipants and their pre- and postintervention fluency scores
revealed another high and significant correlation (.85, p <
.01); again, no significant difference was found between
preintervention fluency (M = 78.22, SD = 26.31) and pos-
tintervention fluency (M = 90.11, SD = 30.12).
To determine whether there was any interaction effect
attributable to the intervention, pre- and postintervention
fluency scores were analyzed by means of a two-way
mixed-design ANOVA having two levels of fluency scores,
the Intervention and the Delayed Intervention groups as the
between-subjects factor and the two assessment points—
preintervention and postintervention—as the within-subject
factor. The interaction effect of Fluency Score × Group was
found not to be statistically significant, F(1, 13) = .04, p >
.05, indicating comparability between the two groups’
scores (Intervention group: M = 84.17, SE = 9.98, and
Delayed Intervention group: M = 81.0, SE = 12.0). However,
the within-subject main effect of pre- and postintervention
fluency scores was found to be statistically significant, F(1,
13) = 6.68, p < .05, partial η2 = .31. Results showed that the
pre- and postintervention fluency scores for the Intervention
and Delayed Intervention groups differed significantly from
each other. These data suggest that there was change in the
students’ reading speed over time, but this was not necessar-
ily attributed to the effects of the intervention.
Intervention Data
To further examine the effects of fluent processing on
reading scores, the results of 1,440 min of intervention
(quantified in milliseconds and representing a change in
speed of processing) was used as a measure of achieved
fluency for the Intervention group only. This evidence of
processing change was analyzed by means of a two-way
mixed-design ANOVA having two levels of reading flu-
ency scores (pre- and postintervention) as a within-sub-
ject factor and two levels of fluency: those students (n =
6) who reached levels of emerging fluency, 100 millisec-
onds or less, and those (n = 3) who did not, as a between-
subjects factor. The between-subjects main effect of the
fluency rate achieved during intervention was significant,
F(1, 8) = 5.38, p = .05, indicating fluency differences
between the students who achieved fast processing as
measured through the FlashWord intervention and those
who did not.
Additionally, a postintervention correlational statis-
tical analysis was conducted to investigate relationships
between the achievement of fluent processing and spe-
cific ROI activation. We created scatterplots to illus-
trate active variations in individual student levels of
activation in the selected ROIs and the student’s
achieved level of fluency in the FlashWord program.
We used data from the Intervention group only and
included those participants who did not produce mea-
sureable activations for continuity. We analyzed the
data separately for each ROI so comparisons would be
possible (see Figure 3).
Data from fMRI Analysis
These group activation maps represent the condition of
brain activation resulting from the phonological analysis of
word pairs over the activation resulting from a perceptual
control, the visual matching of strings of nonpronounceable
letters, in a sample of participants. The calibration maps
varied somewhat from scan to scan depending on the
robustness of the activation detected. The comparisons
included all data from the Delayed Intervention group (pre–
Delayed Intervention and post–Delayed Intervention), and
all data from the Intervention group (pre-Intervention and
post-Intervention) and were analyzed separately. The p
value for all equaled .01.
The activation in the pre–Delayed Intervention group
(Word Pair > Letter Match contrast) maps showed more
RH activation than LH (LH = 9, RH = 10; see Figure 2).
The largest area of activation was found in the Lingual
Gyrus_R (t score = 15.68) but included only 112 voxels at
a peak coordinate of 20 –84 –12. This area includes activa-
tions in the lingual gyrus, fusiform gyrus, and Brodmann
area 37, evidence of activity in word recognition and auto-
matic processing in multiple sections of the VWFA. The
second largest cluster, while only 63 voxels, involved the
same area in the LH with very high peak intensities
(10.8266 for the LH cluster).
10 Journal of Learning Disabilities 00(0)

Figure 2.  Selected Group Activation Maps—Delayed Intervention Group.

Note. Activation maps are from random effects analysis for phonological processing of word pairs greater than perceptual control of letter match
condition for subjects in the Delayed Intervention group only at (A) preintervention and (B) postintervention. Activation slice A is a saggital view, slice
B is a coronal view, and slice C is the transverse view. The “glass brain” feature shows internal activations for each slice. The calibration bar indicates t
values for comparison at each area presented on the slice views.

Similarly, the activation in the post–Delayed Intervention
group (Word Pair > Letter Match contrast) maps continued
to show more RH activation than LH (LH = 4 cluster, RH =
6 clusters, interhemisphere = 2 clusters) but fewer total
areas of activation than the pre–Delayed Intervention map.
The peak intensity (t scores) for the post–Delayed
Intervention group (3.3649) was less than the peak intensity
for the pre-Intervention condition (3.7469). The largest area
of activation was found in the Fusiform Gyrus_R (60 vox-
els, peak coordinate = 44 −64 −20, t score = 7.2379) and the
second largest in the Fusiform Gyrus_L (38 voxels, peak
coordinate −40 −56 −20, t score = 12.3386). Again, the lin-
gual gyrus, fusiform gyrus, and related Brodmann areas 18,
37, and 20 showed activation and relate to automatic pro-
cessing of phonological features in the VWFA, high-level
visual processing and recognition memory.
The activation in the pre-Intervention group (Word Pair
> Letter Match contrast) maps showed more LH activation
than RGH (LH = 4, RH = 1), and only five clusters met the
five-voxel minimum requirement (see Figure 4). The
Koen et al. 11
Figure 3.  Fluency Correlations With Region of Interest
Activations.
largest cluster was found in the posterior cingulate with a
peak coordinate of 8 –60 12 (t score = 4.1414) and included
cuneus, lingual gyrus, and Brodmann area 30. These fea-
tures are responsible for basic visual processing, word rec-
ognition, and a default mode network. The second largest
area of activation was found in the Fusiform Gyrus_R and
included only 20 voxels at a peak coordinate of 40 –68 –20
(peak intensity = 3.6417). This cluster included the fusi-
form gyrus, Brodmann area 19, lingual gyrus, and Brodmann
area 18, which showed evidence of automatic processing
and word recognition, with visual cortex activation in areas
of feature extraction and attention.
The activation in the post-Intervention group (Word Pair
> Letter Match contrast) maps showed slightly more RH
activation than LH (LH = 6, RH = 8) and more clusters of
activation in the post-Intervention scans than the pre-Inter-
vention scans. The peak intensity (t scores) for both condi-
tions was nearly equal (2.998 for pre- and 2.8965 for
postintervention). However, there was much higher peak
intensity in comparable features in the post condition in the
Fusiform Gyrus_L and _R. The Fusiform Gyrus_L was 132
voxels around a peak coordinate of −40 −64 −20 (peak
intensity = 11.8261) and included middle occipital gyrus,
lingual gyrus, fusiform gyrus, and Brodmann area 18. These
areas are thought to involve change detection, word recog-
nition, automatic processing of phonological features, and
visual processing.
Further, the second largest cluster in this condition was
also located in the Fusiform Gyrus_R and also had a higher
intensity (9.137) than its preintervention counterpart
(3.6417). It encompassed 129 voxels at a peak coordinate of
40 –72 –20 and included middle occipital gyrus, fusiform
gyrus, Brodmann areas 19 and 37, lingual gyrus, and infe-
rior temporal gyrus. These areas represent change detec-
tion, automatic processing of phonological features, feature
extraction and attention, word recognition, and representa-
tion of complex features.
Statistical Analysis of fMRI data
The statistical parametric mapping program (SPM8) was
used to analyze the level of activation present measured by
t scores in the ROIs for each participant, using the Word
Pairs over Letter Match condition, which was expected to
focus on phonological analysis processing in these areas.
The creation of these ROIs was based initially on Talaraich
coordinates for the IFG, STG, and VWFA from relevant lit-
erature (Shaywitz & Shaywitz, 1999). Individual differ-
ences were accommodated by using the highest value from
with the identified area. These values were calculated from
the postintervention scans for the Intervention and Delayed
Intervention groups only (see Table 4).
An independent-samples t test compared the mean
activation level captured in t scores from voxels in the
IFG located generally at Talairach coordinates: −46, 35,
12 for the Word Pairs over the Letter Match condition in
12 Journal of Learning Disabilities 00(0)

Figure 4.  Selected Group Activation Maps—Intervention Group.

Note. Activation maps are from random effects analysis for phonological processing of word pairs greater than perceptual control of letter match
condition for participants in the Delayed Intervention group only at (A) preintervention and (B) postintervention. Activation slice A is a saggital view,
slice B is a coronal view, and slice C is the transverse view. The “glass brain” feature shows internal activations for each slice. The calibration bar
indicates t values for comparison at each area presented on the slice views.

the Intervention (M = 1.43, SD = 1.40) and Delayed
Intervention (M = 1.29, SD = 1.41) groups. This compari-
son was not statistically significant, indicating that the
level of activation in this ROI was not different between
the two groups of students. Identical analysis of the STG
and the VWFA yielded similar results. In the STG, gener-
ally located at Talairach coordinates −59, −21, 12 for the
Word Pairs over Letter Match condition, no difference
was found for activation levels, for the Intervention group
(M = 1.42, SD = 1.73) and Delayed Intervention group
(M = 1.20, SD = 1.31). In the VWFA, generally located at
Talairach coordinates −42, −57, −6 for the same condition
no difference was found for the Intervention group
(M = 1.37, SD = 1.30) and Delayed Intervention group (M
= 1.70, SD = 1.32).
The measured levels of activation in the ROIs—IFG,
STG, and VWFA, located as previously noted—were ana-
lyzed by means of a two-way mixed design ANOVA having
two levels of achieved fluency (fluent and not fluent). Neither
the main effect comparing the means of the brain regions nor
Koen et al. 13

Table 4.  Measured Region of Interest Activation Postintervention.

Participant Intervention group/ LH inferior frontal LH superior temporal Visual-word form

code achieved fluency? gyrus (t score) gyrus (t score) area (t score)
CB DI/NA 0.00 0.00 2.37
CC DI/NA 2.56 2.40 2.46
PA DI/NA 0.00 0.00 0.00
PC I/Yes 0.00 0.00 0.00
PE I/Yes 2.39 2.39 2.39
MA I/Yes 2.48 3.36 2.57
MC I/Yes 2.34 3.90 2.33
MD I/No 2.34 0.00 0.00
ME I/No 0.00 0.00 0.00
MF I/Yes 0.00 0.00 2.50
MG DI/NA 0.00 0.00 2.65
MI I/No 3.38 3.10 2.52
MJ DI/NA 2.51 2.38 2.73
ML DI/NA 2.67 2.36 0.00
MN I/Yes 0.00 0.00 2.35

Note. LH = left hemisphere; DI = Delayed Intervention group; NA = not available; I = Intervention group.

the interaction effect of the participant’s fluency level on
brain activation in these areas was significant (F < 1.0).
Using a clustering threshold of five voxels, a sample of
the activation locations was found postintervention in the
condition of Word Pairs over Letter Match for a fluent par-
ticipant. Table 5 contains a partial list of LH activation sites,
noting the location, relative size, and maximum recorded t
score for this individual.
These data show some expected activation areas with
substantial groups of voxels involved and some surprising
lack of activation within the ROIs. The largest activated
cluster in the IFG ROI was the IFG (1.52), but activation in
the STG (3.10) as well as Brodmann areas 41 (3.17) and 42
(3.94) was much stronger.
The largest activated cluster in the STG ROI was the
STG (2.56), but again, other areas showed stronger levels of
activation. The postcentral gyrus activation (3.87) was
unusual in that this area is the primary somatosensory cor-
tex receiving all sensory input, especially touch, and there
was no variation in the motor demands of the task that
would explain activation in this area. The activation found
in Brodmann areas 13 (3.08) and 40 (3.16) reflects refine-
ment of sound/symbol associations and connective tissue
between lateral and medial layers.
The largest activation in the VWFA ROI was found in the
smallest clusters identified. Brodmann areas 21 (2.05) and
35 (2.01) appeared to support automatic processing through
their connection to the MTG, thought to access word mean-
ing, and the perirhinal cortex, important to memory. The left
aspect of the fusiform gyrus showed the strongest activation
(3.06), as would be expected if automatic retrieval of letter
patterns was being used.
The hemodynamic response is the change in magnetic
resonance signal on T2* images following local neuronal
activity. Table 6 shows the peak activation levels that were
processed from data taken from scans of fluent (A) and non-
fluent (B) participants focusing on the ROIs: IFG, STG, and
VWFA. The levels of normalized flow signal from voxels in
the IFG, STG, and VWFA ROIs of a sample fluent partici-
pant showed different levels of activation. Analyses of the
levels of activation indicated that the STG ROI activation
was the strongest (0.15), the IFG activation was next robust
(0.135), and the VWFA was the least strong (0.083).
As might be expected, the same activation assessments
in a sample nonfluent participant were lower in amplitude
compared to the activation in the fluent sample. However,
the order of magnitude was different for the nonfluent sam-
ple: the STG ROI activation was the strongest (0.046), the
VWFA activation was next robust (0.033), and the IFG was
the least strong (0.025).
The results from the scatterplot comparison of fluency and
selected ROI activation (see Figure 3) further support impor-
tant variations in processing that promote automaticity in
written word processing. The results from the comparison of
fluency and IFG activation showed a weak correlation. Two
students did not achieve fluent processing even though they
demonstrated high levels of activation. Since the IFG is asso-
ciated with phonological encoding and sound/symbol asso-
ciations, these activations may indicate more effort’s being
applied to the decoding of words, interfering with fluent pro-
cessing. The results from the comparison of fluency and STG
activation showed a stronger correlation as well as higher
levels of activation. However, this area of the brain is associ-
ated with semantic encoding and word meanings, so this
14 Journal of Learning Disabilities 00(0)

Table 5.  Locations of Significant Activation in a Fluent activation in the VWFA. This was an expected outcome in
Participant. that even the level of activation here was less, overall, than
Cluster Maximum the activation levels in the IFG or STG.
Structure x y z size t score
ROI-IFG Discussion
  LH inferior frontal −48 24 12 523 1.52
gyrus This research was designed to test the hypothesis that sub-
  LH superior temporal −60 −28 12 352 3.10 typing students with the characteristics of dyslexia and
gyrus administering VHSS intervention based on those subtypes
  LH Brodmann area 41 −56 −20 12 147 3.71 (FlashWord-modified), would improve fluency perfor-
  LH Brodmann area 22 −64 −8 4 129 1.45 mance across dyslexia subtypes more effectively than other
  LH insula −36 −16 12 119 1.97 currently used reading fluency programs. It was expected
  LH Brodmann area 42 −60 −20 −12 114 3.94 that VHSS training, based on hemispheric processing dif-
  LH Brodmann area 13 −40 −16 12 73 1.93 ferences demonstrated through reading behaviors, would
  LH precentral gyrus −56 −8 12 67 1.66 increase fluency scores in students with the characteristics
ROI-STG of dyslexia who participated in the intervention signifi-
  LH superior temporal −40 −40 16 233 2.56 cantly more than those with dyslexia who did not partici-
gyrus pate. The evidence available to date indicates that this type
  LH angular gyrus −52 −64 36 86 2.46 of VHSS training does increase fluency in reading-disabled
  LH insula −42 −16 16 68 2.21 Italian students with reading disability (Lorusso et al., 2006;
  LH postcentral gyrus −52 −31 52 33 3.87
Tressoldi, Lorusso, Brenbati, & Donini, 2007). Further, it
  LH Brodmann area 13 −44 −16 16 29 3.08
was hypothesized that regarding postintervention activa-
  LH inferior parietal −52 −36 28 26 2.74
lobule
tion, brain regions that are critical to the training of fast pro-
  LH Brodmann area 40 −52 −24 16 11 3.16 cessing in reading would be identified, and the effects in the
  LH Brodmann area 39 −52 −68 28 10 2.42 brain that signify the development of fluency would be
  LH sub-gyral −48 −8 16 8 1.73 documented. One pattern of processing that has been previ-
ROI-VWFA ously identified shows the shifting of reading processes
  LH sub-gyral −36 −4 −32 30 1.54 from the anterior system located bilaterally to the LH ven-
  LH middle temporal −40 0 −32 19 1.42 tral system. Identification of the brain regions activated
gyrus postintervention was also expected to provide evidence for
  LH Brodmann area 20 −44 −8 −32 7 1.80 the shift from right to left hemispheric processing within the
  LH Brodmann area 21 −40 −4 −32 5 2.05 temporo-parietal system of developing readers and greater
  LH Brodmann area 35 −24 −16 −32 5 2.01 activation of the LH occipito-temporal/fusiform area sig-
  LH fusiform (aal) −28 −24 −32 5 3.06 naling the use of automatic word identification in the VWFA
Note. ROI = region of interest; IFG = inferior frontal gyrus; LH = left
(Licht et al., 1988; Shaywitz et al., 2002).
hemisphere; STG = superior temporal gyrus; VWFA = visual-word form Emerging automaticity was defined for this study as pro-
area. cessing in the 100 to 250 milliseconds range, and the results
from the Intervention group data set included data from par-
ticipants who achieved a fast level of word and phrase pro-
Table 6.  Hemodynamic Activation in Regions of Interest.
cessing through their interaction with FlashWord and data
Inferior Superior Visual-word from participants who, despite the same length of treatment,
Participant frontal gyrus temporal gyrus form area did not reach levels of automaticity. Of the 9 students in the
Fluent 0.135 0.150 0.083 Intervention group, 6 reached word fluency, and 5 of the 9
Nonfluent 0.025 0.046 0.033 also achieved phrase processing that reached this standard.
Those students who displayed only phonics-based errors in
reading connected text and worked for the entire intervention
activation without fluent processing might indicate a reliance time in the LH program made the most substantial increases
on word meanings to aid decoding. in both processing and reading speed (mean processing speed
The results from the comparison of fluency and VWFA = 80 milliseconds and mean reading speed = +14 wpm). Only
activation showed a clear relationship between strength of 1 student demonstrated only meaning-based errors, used
activation and automaticity in reading. This scatterplot the RH program exclusively, and showed the highest net
displayed all students who achieved very fast rates of pro- gain in fluency overall (processing speed = 75 milliseconds
cessing and also demonstrated relatively high rates of and reading speed = +42 wpm). Of the 3 students who
Koen et al. 15
demonstrated mixed processing deficits, 2 were assigned to
both the RH and the LH programs and reached automaticity
in the LH program but not the RH program. The students
who split their time between programs made the least amount
of progress. These students showed mixed results: 1 demon-
strated a net gain of 10 wpm and the other a net loss of 3 wpm
in fluency. It is suggested that continued work with the inter-
vention program could achieve the desired level of automa-
ticity and that strengthening processing in the RH is inherently
more difficult than strengthening in the LH. This is perhaps
because older RH readers have established these inefficient
connections over time, and these must be abandoned (extin-
guished) in order to create more effective ones.
We also found differences in the pre- and postinterven-
tion fluency scores, which varied by group, in spite of a
small sample size and unequal samples. The average differ-
ence between preintervention and postintervention fluency
scores for the Intervention group was 11.9 wpm, and the
average difference for the Delayed Intervention group was
7.3 wpm. As the Delayed Intervention participants demon-
strated little or no change in their reading speed during the
study, Intervention participants showed change approach-
ing statistical significance. Data showing that the pre- and
postscores for both groups differed from each other signifi-
cantly from the beginning of the study to the end could be
attributed to the effects of the intervention. However, many
of the Delayed Intervention students also made gains with-
out it. It would be difficult to assess the exact cause of such
reading increases without understanding a myriad of poten-
tial factors affecting each participant.
The group activation maps offer interesting evidence for
hemispheric processing changes in this population. The results
from the pre– and post–Delayed Intervention conditions con-
firm that, even without the intervention, many of these stu-
dents were using LH features to process phonological
information and later showed enhanced LH processing of
phonological features in areas known for automaticity, per-
haps as a result of age and reading instruction as described by
Shaywitz et al. (2002) in nonimpaired older readers. Even
though these students were already processing phonological
information in these areas of automaticity, they did not achieve
enhanced processing in terms of activation areas and intensity,
suggesting a static condition from the pre- to the postinterven-
tion scan for these students. It is interesting to note that 5 of
the 6 students in the Delayed Intervention group were evalu-
ated to have RH weaknesses based on their reading behaviors
that were not addressed during the course of the study. Clearly,
there were changes in processing in the RH that these students
experienced without the benefit of the intervention, although
increases in functioning were more diffuse and less intense
than in the pre- to post-Intervention group data.
The activation data from the Intervention group, pre
and post, show a different outcome. The peak intensity
scores suggest that before the intervention more effort was
focused on trying to recognize the nonwords and compar-
ing the letters accurately than on automatic retrieval of
phonological features in the VWFA. However, postinter-
vention, their activation patterns suggest a high level of
automatic processing in both hemispheres: in the cerebel-
lum posterior lobe and the lingual gyrus itself. The stu-
dents in this group were initially evaluated to have a mix
of processing weaknesses, based on reading behaviors,
with 2 students demonstrating RH weaknesses, 4 showing
LH weaknesses, and 3 exhibiting weaknesses in both
hemispheres. These students could be considered more
disabled in reading; yet as a result of the intervention, they
achieved processing levels that exceeded the Delayed
Intervention group in number of active clusters (8 for the
post–Delayed Intervention group and 14 for the post-
Intervention group) and intensity of activation (3.3649 for
the post–Delayed Intervention and 11.8261 for the post-
Intervention groups). This evidence supports the develop-
ment of the ventral processing system over the anterior
processing system as shown in literature.
We also found evidence indicative of the expected
changes in neural processing from the effects of the inter-
vention in the comparison of data from selected ROIs.
There were consistently higher amplitudes of activation in
all explored ROIs in the fluent group (Intervention sub-
group) compared to the amplitude of activation seen in the
nonfluent group (all of the Delayed Intervention group).
Higher levels of activation in the STG (sound/symbol asso-
ciations), IFG (encoding phonological features), and
VWFA (automatic word retrieval) for the Intervention
group could be the direct result of the selective hemispheric
stimulation provided by the intervention and therefore rep-
resented most definitively the expected differences in pro-
cessing. Higher levels of activation of the STG could
indicate that most of the processing in the IFG ROI
involved sound/symbol associations with support in the
primary and auditory association cortex. In other words,
less effort was required to accomplish the phonological
analysis task by this participant, which may provide evi-
dence for the neural efficiency hypothesis. The postcentral
activation found in the STG ROI could be evidence of
compensatory systems being used for phonological analy-
sis in immature processing systems. Further, the correla-
tion of fluent processing with decreasing levels of activation
in the IFG, STG, and VWFA may indicate that higher acti-
vations in these ROIs contributed to faster processing and
offer evidence for the “neural efficiency hypothesis.” Thus,
taken together, the activation locations identified for this
fluent participant generally followed activation patterns
found in the literature that are thought to support fluent
processing (Goswami, 2008; Pugh et al., 2001; Simos,
Breier, Fletcher, Bergman, & Papanicolaou, 2005).
16 Journal of Learning Disabilities 00(0)
Most striking is the lack of activation in the ROIs in the
nonfluent students, when compared to the average t scores
of students who achieved fluency. These extremely low lev-
els of measured activation reflect virtually no significant
activation and make this a potentially important finding.
The current literature on amplitude of signals in fMRI has
revealed that these measurements are often the focus of
resting state versus focused activation studies and are not
commonly used in educational research (Steffener, Tabert,
Reuben, & Stern, 2009). These results would seem, how-
ever, to provide further evidence of hypoactivation of key
brain regions required for reading by nonfluent participants.
These findings underscore the effects of the forced process-
ing from the FlashWord intervention program on the devel-
opment of automaticity.
The results of this study show that students do not
achieve at the same rate using FlashWord; individual differ-
ences in processing strengths and weaknesses seemed to
influence student proficiency within the given intervention
time constraint. However, the high rate of success for those
in the Intervention group, both in achieving a level of
automatized processing and in increasing reading speed of
connected text, provides validation of the benefits of accu-
rate subtyping and VHSS. All but 2 participants in the
Intervention group achieved emerging fluency in either
word or phrase processing components of their assigned
program. Even though statistical analysis was not possible
due to the extremely small samples, it is clear that the suc-
cess of the intervention depends on accurate assessment of
reading errors and assignment to the correct FlashWord pro-
gram. Regardless, the data showed that general activation
patterns identified in literature increased RH or bilateral
activation in brain features involved in phonological pro-
cessing and hypoactivation in LH features and the VWFA
(Goswami, 2008; Pugh et al., 2001; Shaywitz et al., 2002;
Shaywitz et al., 2004; Simos et al., 2005).
Study Limitations
There were several limitations to the effectiveness of this
research. The sample was too small in size relative to
unequal comparison groups and did not allow for reliable
comparison of the dyslexia subtypes as they related to each
other and to the intervention program they used. The small
sample and the participants’ youthfulness—and often their
disability as well—also increased the challenge of obtain-
ing useable data. Movement artifacts decreased the number
of functional images, and the high rate of “squeeze bulbing”
displayed during the postintervention scanning served to
limit the quantity of postintervention images that were
available for analysis. Being limited to the analysis of
selected ROIs from the literature may have reduced the sig-
nificant findings based on the demands of the scanner task.
More extensive analysis of areas involved in phonological
processing only, or the addition of a third scanner task
requiring the automatic recall of sight words, could improve
the statistical analysis outcomes.
It remains to be seen whether the human brain reaches
some level of development that decreases the plasticity of
neural processes to respond to neurobiological training.
Similar to this study’s results, Tressoldi et al. (2007) found
robust increases in teenage Italian students’ reading speed
after the intervention comparable to those achieved by their
younger colleagues. Until that level of diminishing plastic-
ity is determined, the potential for VHSS to effect positive
change for nonfluent readers of every age is unknown but
extremely promising.
Authors’ Note
This study was approved by the University of Houston Institutional
Research Board and consent/assent was given through the
University of Houston Informed Consent form.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) received no financial support for the research,
authorship, and/or publication of this article.
References
Bakker, D. J., Bouma, A., & Gardien, C. J. (1990). Hemisphere-
specific treatment of dyslexia subtypes: A field experiment.
Journal of Learning Disabilities, 23(7), 433–438.
Beyond the numbers: A guide to interpreting and using the results
of standardized-achievement tests. (2012). New York, NY:
CTB/McGraw-Hill. Retrieved from www.nxtbook.com/nxt-
books/eschoolmedia/ctb.cgraw-hill/#/2
Doehring, D. G. (1976). Acquisition of rapid reading responses.
Monograph of the Society for Research on Child Development,
165(2),1–54.
Fry, E. (n.d.-a). Fry instant phrases. Retrieved from http://www.
timrasinski.com/presentations/fry_600_instant_phrases.pdf
Fry, E. (n.d.-b). Fry word list. Retrieved from http://mcaswiki.
mcas.k12.in.us/@api/deki/files/225/=High_Frequency_
Words.pdf
Goswami, U. (2008). Reading, dyslexia and the brain. Educational
Research, 50(2), 135–148.
Haier, R. J., & Benbow, C. P. (1995). Sex differences and later-
alization in temporal lobe glucose metabolism during math-
ematical reasoning. Developmental Neuropyschology, 11,
405–414.
Haier, R. J., Siegel, B. V., Nuechterlein, K. H., Hazlett, E., Wu,
J., Pack, J., . . . Buchsbaum, M. S. (1988). Cortical glucose
metabolic rate correlates of abstract reasoning and attention
studied with positron emission topography. Intelligence, 12,
199–217.
Koen et al. 17
Hebb, D. O. (1949). The organization of behavior. New York,
NY: Wiley.
Kame’enui, E. J. (2007). A new paradigm. Teaching Exceptional
Children, 39(5), 6–7.
Kame’enui, E. J., Simmons, D. C., Good, R. H., & Harn, B. A.
(2000). The use of fluency-based measures in early identifica-
tion and evaluation of intervention efficacy in schools. In M.
Wolf (Ed.), Dyslexia, fluency, and the brain (pp. 307–332).
Parkton, MD: York Press.
Katzir, T., Kim, Y., Wolf, M., O’Brien, B., Kennedy, B., Lovett,
M., & Morris, R. (2006). Reading fluency: The whole is more
than the parts. Annals of Dyslexia, 56(1), 52–82.
The Kid’s Stuff People. (1980). Yellow pages for students and
teachers. Chicago, IL: World Book/Incentive.
LaBerge, D., & Samuels, S. J. (1974). Toward a theory of automatic
information processing in reading. Cognitive Psychology, 6,
293–323.
Licht, R., Bakker, D. J., Kok, A., & Bouma, A. (1988). The
development of lateral event-related potentials (ERPs)
related to word naming: A four-year longitudinal study.
Neuropsychologia, 26, 327–340.
Lorusso, M., Facoetti, A., Paganoni, P., Pezzani, M., & Molteni,
M. (2006). Effects of visual hemispheric-specific stimula-
tion versus reading-focused training in dyslexic children.
Neuropyschological Rehabilitation, 16(2), 194–212.
Masutto, C., & Fabbro, F. (1995). FlashWord: Training neu-
ropsicologico per la dislessia. Gorizia, Italy: Editrice
TecnoScuola.
Misra, M., Katzir, T., Wolf, M., & Poldrack, R. A. (2004). Neural
systems for rapid automatized naming in skilled readers:
Unraveling the RAN-reading relationship. Scientific Studies
of Reading, 8(3), 241–256.
Mrazik, M., & Dombrowski, S. C. (2010). The neurobiological
foundations of giftedness. Roeper Review, 32, 224–234.
Pugh, K. R., Mencl, W. E., Jenner, A. R., Lee, J. R., Latz, L.,
Frost, S. J., . . . Shaywitz, B. A. (2001). Neuroimaging stud-
ies of reading development and reading disability. Learning
Disabilities Research and Practice, 16(4), 240–249.
Richards, T. L., Aylward, E. H., Field, K. M., Grimme, A. C.,
Raskind, W., Richards, A. L., . . . Berninger, V. W. (2006).
Converging evidence for triple word form theory in chil-
dren with dyslexia. Developmental Neuropsychology, 30(1),
565–589.
Sandak, R., Mencl, W. E., Frost, S. J., & Pugh, K. R. (2004). The
neurobiological basis of skilled and impaired reading: Recent
findings and new directions. Scientific Studies of Reading,
8(3), 273–292.
Shaywitz, B., Shaywitz, S., Blachman, B., Pugh, K., Fulbright, R.,
& Skudlarski, P., . . . Lyon, G. R. (2004). Development of left
occipitotemporal systems for skilled reading in children after
a phonologically based intervention. Biological Psychiatry,
55(9), 926–934.
Shaywitz, S., & Shaywitz, B. (1999). Cognitive and neurobio-
logic influences in reading and in dyslexia. Developmental
Neuropsychology, 16(3), 383–385.
Shaywitz, S. E., Shaywitz, B. A., Fulbright, R. K., Skudlarski,
P., Mencl, W. E., Constable, R. T., . . . Gore, J. C. (2002).
Disruption of posterior brain systems for reading in children
with developmental dyslexia. Biological Psychiatry, 52,
101–110.
Simos, P. G., Breier, J. I., Fletcher, J. M., Bergman, E., &
Papanicolaou, A. C. (2005). Early development of neu-
rophysiological processes involved on normal reading
and reading disability: A magnetic source imaging study.
Neuropsychology, 19(6), 787–798.
Steffener, J., Tabert, M., Reuben, A., & Stern, J. (2009).
Investigating hemodynamic response variability at the group
level using basis functions. NeuroImage, 49(3), 2113–2122.
Tressoldi, P. E., Lorusso, M. L., Brenbati, F., & Donini, R. (2007).
Fluency remediation in dyslexic children: Does age make a
difference? Dyslexia, 14, 142–152.
Wolf, M., & Katzir-Cohen, T. (2001). Reading fluency and its
intervention. Scientific Studies of Reading, 5(3), 211–239.
World Health Organization. (1992). International statistical clas-
sification of diseases and related health problems (10th rev.).
Geneva, Switzerland: Author.