Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Examining the neuroanatomical correlates of reading in developmental dyslexia
(USC Thesis Other)
Examining the neuroanatomical correlates of reading in developmental dyslexia
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
EXAMINING THE NEUROANATOMICAL CORRELATES OF READING IN
DEVELOPMENTAL DYSLEXIA
by
Rita Barakat
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(NEUROSCIENCE)
August 2022
Copyright 2022 Rita Barakat
Acknowledgements and Funding
The author would like to thank and acknowledge the mentorship, guidance and support of the
following individuals, without whom this dissertation would not have been possible to complete.
Dissertation Committee Members:
Megan Herting (PhD, Committee Chair)
Jason Zevin (PhD, Primary Advisor)
Kristi Clark (PhD, Outside Member and Mentor)
Frank Manis (PhD, Outside Member and Mentor)
Other Mentors:
Mara Mather (PhD, former Committee Chair and Rotation Mentor)
Rand Wilcox (PhD, Statistics Advisor)
Farshid Sepehrband (PhD, Neuroimaging Advisor)
Jeiran Choupan (PhD, Neuroimaging Advisor)
Ryan Cabeen (PhD, Neuroimaging Advisor/ Technical Support)
Carolina Makowski (PhD, OHBM Mentor)
Google (and specifically, Stack Overflow and GitHub
ii
Lab Colleagues and Research Groups:
Center for Aphasia and Related Disorders
(CARD) Group
Robert Knight (MD)
Nina Dronkers (PhD)
Juliana Baldo (PhD)
Krista Parker (PhD)
Connectivity and Network Development
(CANDL) Group
Kirsten Lynch (PhD)
Max Orozco
Hadley McGregor
Anisa Azad
Stephen Gonzalez
Maya Rajan
Funding Sources:
Kennedy Shriver National Institute of Child Health and Human Development
(R00HD065832)
University of Southern California Provost Graduate Fellowship
USC Neuroscience Graduate Program NIH T-32 Training Grant (5T32MH111360-02)
National Science Foundation (NSF) Graduate Research Fellowship (GRFP).
iii
TABLE OF CONTENTS
Acknowledgements and Funding ………………………………………….. ……………….ii
List of Tables ……………………………………………………………….. ……………….v
List of Figures ………………………………………………………………. ..…………….vii
Abstracts …………………………………………………………………… .....…………viii
Chapter 1: Background on Dyslexia, and the Behavioral and
Neuroscientific Literature …………………………………………………... ……………….1
Chapter 2: An Adaptive Behavioral Paradigm for assessing Orthographic
and Phonological Processing in Children with Dyslexia …………………… ……………...14
Chapter 3: Neuroanatomical Differences as a function of Development,
Reading Skill and Dyslexia ……………………………………………….... ……………...35
Chapter 4: Examining the Relationship between Brain Structure and
Reading Performance using an Adaptive Behavioral Task Paradigm ……… ……………...58
Chapter 5: Broader Implications, Limitations and Considerations when
Studying Developmental Dyslexia ………………………….…………….... ……………...80
Supplementary Materials ………………………………………………........ ……………...89
References ……………………………………………….............................. ...……………90
iv
List of Tables
Table 1: Neuropsychological and demographic summary data for subjects
included in the study of the behavioral task paradigm …………………………...
………...25
Table 2: Summary of behavioral data results for orthographic and phonological
tasks ………………………………………………………………………………
………...26
Table 3: Linear contrasts used to perform General Linear Model (GLM) analysis ………...44
Table 4: Statistical results from Analysis of Variance (ANOV A) of cortical
thickness, surface area and volume data (left hemisphere only) to examine
reading skill effect ………………………………………………………………..
………...45
Table 5: Statistical results from Analysis of Variance (ANOV A) of cortical
thickness, surface area and volume data (left hemisphere only) to examine
developmental effect ……………………………………………………………...
………...50
Table 6: Statistical results from Analysis of Variance (ANOV A) of cortical
thickness, surface area and volume data (left hemisphere only) to examine
dyslexia effect ……………………………………………………………………
………...53
Table 7: Significant correlations (which structural measures with which
behavioral measures for each region of interest) on the orthographic task ………
………...71
Table 8: Significant correlations (which structural measures with which
behavioral measures for each region of interest) on the phonological task ……....
………...72
v
List of Figures
Figure 1: Schematic illustration of the Triangle Model of Reading …………... …………...17
Figure 2: Two sample stimuli (image + target and foil), both at Difficulty
Level 1, for the orthographic (right) and phonological (left) behavioral tasks .. …………...22
Figure 3: Schematic illustration of the event-related task design implemented
for the orthographic and phonological task runs ……………………………… …………...23
Figure 4: Graphical representation of group by task performance for all three
subject groups on the orthographic and phonological tasks with average
difficulty level as the performance measure …………………………………... …………...28
Figure 5: Graphical representation of group by task performance for all three
subject groups on the orthographic and phonological tasks with average
reaction time (correct trials only) as the performance measure ……………….. …………...29
Figure 6: Graphical representation of group by task performance for all three
subject groups on the orthographic and phonological tasks with accuracy as
the performance measure ……………………………………………………… …………...30
vi
Figure 7: Graphical representation of group by task performance for all three
subject groups on the orthographic and phonological tasks with variability as
the performance measure ……………………………………………………… …………...31
Figure 8: Canonical left-hemisphere regions associated with reading and
language more broadly, viewed from the exterior cortical surface (sagittal
view) …………………………………………………………………………... …………...39
Figure 9: FreeSurfer inflated pial surface map depicting cortical thickness,
surface area and volume statistics (p-value) for reading skill effect ………….. …………...45
Figure 10: FreeSurfer inflated pial surface map depicting cortical thickness,
surface area and volume statistics (p-value) for developmental effect ………... …………...49
Figure 11: FreeSurfer inflated pial surface map depicting cortical thickness,
surface area and volume statistics (p-value) for dyslexia effect ………………. …………...53
vii
Abstracts
Chapter 2
It has been well established that children and adults with dyslexia show deficits in
integrating orthographic and phonological information at the level of higher sensory and lexical
processing networks during reading. Heterogeneity of dyslexia symptom presentation and
severity poses a significant obstacle in the implementation of practical remediation treatments,
and current behavioral tasks do not always address confounding factors related to accuracy of
task performance between different subject populations, making comparison between impaired
and non-impaired individuals difficult. Here we present an adaptive variation of the canonical
orthographic and phonological decision-making paradigm implemented in behavioral studies of
developmental dyslexia. This particular experimental design presents numerous advantages for
individual subject comparisons as well as for group-level analyses, and allows for more detailed
exploration of reading performance as it pertains to reading skill, development and dyslexia.
Analysis of task performance from 35 subjects (age-matched controls, reading level-matched
controls and subjects with dyslexia) showed that the tasks are effective at distinguishing
developmentally-driven factors that may impact reading ability from those that are specific to
individuals with dyslexia, while also revealing individual differences within subject groups on
task performance. Further validation of this design is made possible through the documentation
and release of the original stimulus repository, as well as the scripts to execute the orthographic
and phonological tasks for the purposes of behavioral and/ or neuroimaging research.
viii
Chapter 3
Developmental dyslexia is a disorder characterized by slow, effortful reading, and
numerous studies have postulated that this deficit is the result of poor integration of phonological
and orthographic information within the left-hemisphere reading network. Structural analyses
indicate that reductions in cortical thickness in key regions of the reading network provide
evidence for underlying structural differences in the reading networks of children with dyslexia
(relative to typical readers). The following study corroborates findings from the literature and
implements an HCP-style approach to assessing structural differences in the reading network of
children with dyslexia and age-matched/ reading level-matched control subjects. All subjects
were scanned in a Siemens Prisma 3T Scanner, and image data were pre-processed using the
Human Connectome Project’s (HCP) minimal preprocessing Structural Pipeline (PreFreeSurfer).
Cortical thickness, surface area and volume were calculated via the HCP Structural Pipelines 2
and 3 (FreeSurfer and PostFreeSurfer). A priori regions of interest from HCP-specified
FreeSurfer parcellation were identified and selected for linear regression analysis in R Studio and
Python to determine the relationship between dyslexia diagnosis/ qualification (subject group)
and cortical thickness, surface area and volume of these key brain regions implicated in reading.
Results show that dyslexia diagnosis/ qualification is a significant predictor (at the p < 0.05
level) of cortical thickness, surface area and volume in key left hemisphere reading network
regions (specifically, the superior temporal gyrus/ Heschl’s Gyrus and the inferior frontal gyrus,
among other regions).
ix
Chapter 4
Numerous studies incorporating different neuroimaging techniques and behavioral task
data have sought to elucidate the relationship between brain structure/ function and reading
performance, specifically in the context of reading disorders. However, several challenges arise
when attempting to draw meaningful conclusions about the relationship between a specific
region of interest (ROI) and reading ability more broadly. While this wealth of multimodal
imaging studies of reading has contributed to a vast knowledge of left-hemisphere brain regions
implicated in reading behavior, determining whether these same regions may be significantly
different in terms of their structure and/ or function in individuals with reading disorders requires
further exploration and replication. The following study seeks to contribute to the collective
understanding of structural brain differences between children with dyslexia and their
age-matched and reading-level matched counterparts, while also assessing the potential
correlations between these measures of brain structure, reading ability and other factors such as
the incorporation of remediation therapies. Cortical thickness, surface area and volume metrics
were extracted from T1- and T2-weighted magnetic resonance imaging (MRI) data in eleven a
priori left-hemisphere ROIs, then correlated with average difficulty level, average reaction time
(on correct trials only), accuracy and variability on the two behavioral tasks administered.
Correlation analyses were conducted such that the effects of development, inherent reading skill
and dyslexia could be effectively isolated. The results indicate that for all three effects, cortical
thickness was negatively correlated with reading ability, while surface area and volume were
positively correlated with reading ability. These findings generally corroborate the conclusions
x
from the existing literature illustrating that increased cortical thickness in select ROIs is often
associated with the early developmental stages of acquiring literacy.
xi
Chapter 1: Background on Dyslexia, and the Behavioral and Neuroscientific Literature
Barakat, Rita
University of Southern California, Neuroscience Graduate Program (NGP)
Leading Neuropsychological Theories
Dyslexia is a disorder characterized by difficulties in reading, specifically related to the
ability to accurately and efficiently recognize words (Peterson and Pennington, 2012). While the
ultimate purpose of reading is to comprehend what is being read, dyslexia is a disorder of
decoding individual words, leaving overall comprehension intact. Previous clinical definitions of
dyslexia have conflated the disorder with a more global “reading disability”, but this is in fact an
inaccurate representation of deficits unique to dyslexia, which are often distinctly different from
those of other reading disorders (Shaywitz and Shaywitz, 2005). In addition to confusion relating
to the comprehension versus decoding aspects of reading, there have also been many
misrepresentations of dyslexia as being a disorder associated with low IQ. Recent studies of
individuals with dyslexia have shown that this is not always necessarily the case, and in fact,
there are many instances in which dyslexia exists in conjunction with high IQ, leading
researchers to question whether in fact diagnostic measures of dyslexia should be related to IQ
measures (Peterson and Pennington, 2012).
Dyslexia has been studied extensively across many disciplines and cultural groups, and
the etiology remains consistent despite morphological and phonological differences between
global languages. More specifically, developmental dyslexia refers to dyslexia that is detected in
an individual or population in which literacy maturation has not yet been reached, yet
1
pronounced difficulties in word recognition are evident throughout the course of acquiring
literacy (Shaywitz and Shaywitz, 2005). Currently, there are no “one-size-fits-all” approaches for
diagnosing developmental dyslexia, which can ultimately lead to educational delays in children
who present the symptoms of dyslexia, but remain undiagnosed and are thus not treated early on
their literacy training. In addition, many studies have demonstrated several psychological and
neurobiological phenomena associated with dyslexia, however, few have yet to be isolated as a
particular causal candidate of the disorder.
Several notable researchers in the fields of neuropsychology and cognitive psychology
have presented compelling behavioral and cognitive evidence to support theories of the
underlying cause(s) of dyslexia, with the ultimate goal of characterizing behavioral and
biological observations as being either associated with/ co-morbid with dyslexia, or potentially
causal of dyslexia. The following paragraphs will give attention to each of these theories and
their supporting evidence, as well as the specific experimental designs that have been
implemented to examine these theories.
Phonological Deficit Theory
Currently the “reigning” theory explaining the underlying causal components of dyslexia,
the phonological deficit theory states that individuals with dyslexia have an impairment in the
representation, storage and retrieval of the phonological aspects of language, or put more simply,
the “speech sounds” (Ramus et al., 2003). Linguistic support for this theory stems from the
fundamental process of acquiring literacy, which involves learning the association between
graphemes and phonemes in the alphabetic system of a language: in other words, learning to
2
associate individual letters with their corresponding sounds. The phonological deficit theory
postulates that there is a deficit in the formation of these grapheme-phoneme associations, which
ultimately impacts the ability to process the alphabetic system more globally (Ramus et al.,
2003). Behavioral observations of individuals with dyslexia also provide strong support for this
theory, as it remains largely undisputed that dyslexia is a disorder of phonological processing,
though there are several competing theories explaining this particular deficit. In a way, the
phonological deficit theory provides the most direct association between the underlying
cognitive processes of reading and the behavioral manifestations observed in individuals with
dyslexia, namely those pertaining to deficits in phonological awareness (Share and Stanovich,
1995).
There have also been several structural and functional neuroimaging studies that provide
support for this theory, isolating left-hemisphere brain regions in the perisylvian cortex (temporal
lobe areas immediately surrounding the sylvian fissure) as being implicated in phonological
processing. However, the neuroimaging evidence supporting this theory is perhaps the most
contested, as other studies using structural and functional magnetic resonance imaging (MRI/
fMRI) have identified more dispersed white matter tracts, some of which cross through these
perisylvian regions, as well as more global networks pertaining to sensory and motor function, as
being implicated in the phonological deficits observed in dyslexia (Ramus et al., 2003). Overall,
the conflicting neurobiological evidence suggests that while the hallmark symptom of dyslexia
may be difficulties in phonological processing, the underlying regions/ networks responsible for
this deficit may not be isolated to those that are specifically implicated in phonological
processing.
3
A comprehensive study by Ramus et al. examining several competing theories of
dyslexia found particular support for the phonological deficit theory, when testing a population
of seventeen university students with diagnosed dyslexia. The students underwent a battery of
neuropsychological, phonological, auditory, visual and cerebellar tests to assess their cognitive
abilities in each of these particular domains, each of which is tied to a theory of the underlying
causality of dyslexia. The researchers found that a significant portion of their subjects performed
below the standardized average for their age/ IQ/ educational level on the phonological tests, and
the specific pattern of this low performance was homogeneous across the subject population (i.e.
scores on each of the individual phonological tests were highly similar). This led the group to
conclude that deficits in the phonological representation of words are “sufficient cause” of the
observed behavioral manifestations of dyslexia within their particular subject population. The
researchers do note that several of the subjects also presented deficits in auditory processing, as
evidenced from these subjects’ low performance on the auditory behavioral tasks, however, they
speculate that based on the heterogeneity of the auditory task scores, it is most likely not linked
to the observed phonological deficits in a meaningful way. Another aspect of this study that is
important to consider is the nature of the study design, which did not account for differences in
reading level or educational background between the subject population used for the study and a
more representative (random) pool of individuals with dyslexia, which the researchers noted was
a weakness of the study design.
4
Multiple Deficit Hypothesis
In the mid-21
st
century, many additional behavioral disorders were increasingly observed
as being comorbid with dyslexia (such as ADHD and Speech Sound Disorder, or SSD).
Researchers attempted to explain these comorbidities by hypothesizing that the underlying
causality of dyslexia may not be a single cognitive process, but rather, many cognitive processes
that, when individually disrupted, result in the downstream etiology of dyslexia. This hypothesis
is known as the Multiple Deficit Hypothesis, and was presented as a possible explanation for the
myriad behavioral deficits observed in individuals with dyslexia by Bruce Pennington in 2006.
Pennington argued that the framework for understanding the cognitive and behavioral
components of dyslexia was fundamentally in contradiction with itself, as dyslexia is often
characterized as a heterogeneous disorder in the context of the behavioral manifestations, but has
traditionally been thought of as homogeneous in the context of the underlying cognitive
processes affected (Pennington, 2006). In order to resolve this contradiction, Pennington and
others have presented a “multiple cognitive deficit model” that explains the comorbidity of
dyslexia and other behavioral/ attentional disorders, offering a contrast to the “single deficit”
(also known as the phonological deficit) theory formalized by Ramus et al. in 2003.
Pennington and others went on to examine the cognitive, behavioral, clinical and
statistical relationship between dyslexia and other disorders, primarily focusing on its
comorbidity with ADHD, and these studies provide extensive evidence for the multiple deficit
framework of dyslexia (Pennington, 2006). For example, in the original paper discussing this
multiple deficit hypothesis, Pennington reviewed the hereditary evidence linking dyslexia and
ADHD, to show that the overlap between these two disorders exists not in a diagnostic realm
5
(the diagnostic tests that are used to assess these disorders do not overlap at all), but in a genetic
realm. By studying monozygotic (MZ, or identical) and dizygotic (DZ, or fraternal) twins, one
can assess the “bivariate heritability” of two disorders, which exists when the relationship
between these disorders is stronger in MZ twins than DZ twins (Pennington, 2006). A
meta-analysis of studies using MZ and DZ twins showed there is high bivariate heritability
between dyslexia and ADHD, indicating that the underlying cognitive processes affected in both
of these disorders may overlap more than previously expected.
In a more recent paper revisiting the multiple deficit hypothesis and other theories of
dyslexia, Pennington’s group performed extensive statistical analyses to test the relative clinical
effectiveness of these models in diagnosing developmental dyslexia in pre-literate subject
populations (Pennington et al. 2012). The results of these analyses were surprising in that the
multiple deficit model did not in fact provide the greatest explanatory power for the deficits
observed in children with dyslexia, leading Pennington to conclude that this and other models for
understanding dyslexia should be considered as “probabilistic” predictors rather than
“deterministic” frameworks for examining individuals with dyslexia. Pennington discusses the
relative merits of the more contemporary “hybrid model” of dyslexia, and points out that various
factors related to a child’s family history (hereditary factors), age (important for considering
reading level) and potential comorbidity with ADHD should be considered when diagnosing
dyslexia.
6
Double Deficit Hypothesis
The double deficit hypothesis of dyslexia postulates that the behavioral outcomes
observed in individuals with dyslexia are related to deficits in two individual, yet inextricably
linked, word processing streams: the orthographic and the phonological streams. Maryanne Wolf
and Patricia Greig Bowers were among the first to note that the difficulties in phonological
processing in developmental dyslexia were also associated with slow symbol naming speed,
leading them to argue that orthographic processing, critical in the visual assessment of words,
should be considered in conjunction with phonological processing in the context of reading
ability more broadly (Wolf and Bowers, 1999). Wolf and Bowers point out that naming speed has
typically been considered a component of phonological processing, but it contributes to the
significant variance in reading ability across individuals, independent of phonological
processing, implying that the orthographic codes associated with words are necessary for fluent
reading, in conjunction with phonological codes.
This theory returns to one of the central tenants presented in the phonological (single)
deficit theory of dyslexia: that the grapheme-phoneme association between letters in the alphabet
must be established early on in a child’s literacy education, in order for the child to acquire
reading fluency (Ramus et al., 2003). In re-examining this aspect of the phonological deficit
theory, it is evident that in order for the correspondence between letters and sounds to be truly
effective in establishing literacy, the downstream orthography of an entire word must be linked to
the phonology of that word, as well as the individual morphemes that comprise the word.
Many behavioral studies provide support for the double deficit theory by demonstrating
how problems with orthographic processing, when examined separately from phonological
7
processing, result in lower reading performance than predicted for a given sample population.
Manis et al. (1990) found that a significant sub-population of individuals with diagnosed reading
deficits showed poorer orthographic processing skills than their reading recognition scores would
have originally predicted. In addition, these individuals appeared to continuously rely on a
phonological coding strategy that was not always effective in the absence of an orthographic
memory for words. By examining these findings in the context of this double deficit theory, one
can conclude that an attempt to compensate for a deficit in one of these word-coding strategies
by relying on the other will still result in behavioral symptoms characteristic of dyslexia, as
overall reading fluency requires both of these word coding strategies.
Visual Attention Hypothesis
In somewhat of a departure from the previous three theories of dyslexia, the visual
attention (V A) hypothesis proposes that a limitation in the number of visual elements that can be
processed in parallel within the visual field (also known as the visual attention span) result in the
behavioral manifestations of dyslexia, independent of deficits in phonological awareness. Bosse
et al. (2007) studied two independent populations of adults with dyslexia, one French and one
British, and observed that the visual attention (V A) span accounted for reading performance in
both groups, even after controlling for other factors such as IQ, verbal fluency, vocabulary and
single letter identification skills. This led the group to conclude that deficits beyond phonological
awareness not only contribute to the behavioral manifestations of dyslexia, could in fact be the
core elements of the disorder.
8
Other neuropsychological evidence supporting the V A hypothesis comes from studies of
the attentional blink (AB), or the deficit in the ability to identify a second target following the
first target, when both appear randomly within a sequence of other distractor items (foils). In a
particular study conducted by Judy Buchholz and Anne Aimola Davies, the AB of five adults
with dyslexia were compared to those of normal adult readers (Buchholz and Davies, 2007).
They found that the AB depends on the nature of the particular task administered, and this was
true for both, the dyslexia and control subject groups. They also concluded that the attentional
system more broadly is compromised in individuals with dyslexia, however, the authors did not
go so far as to claim that this attentional deficit may be a core factor of dyslexia, but rather, one
that may be associated with, or perhaps co-morbid with, dyslexia, and is thus important to
consider in the context of reading ability.
Amplitude Onset Deficit Hypothesis
The amplitude onset deficit hypothesis is one that is primarily concerned with the early
developmental (pre-literacy) processing of acoustic elements of speech and other auditory
stimuli, which are universal across all languages (Goswami et al., 2011). In order for infants to
begin to parse sounds as being specific to speech and prosody, auditory stimuli must be
segmented into what these infants will later perceive to be words and sentences. Deficits in this
early stage of processing auditory stimuli can lead to deficits in phonological processing once
children are at the stage of acquiring literacy, by which point effective intervention may be too
late. Usha Goswami and colleagues have argued in the last five years that a neurobiological
marker for this deficit in auditory sensory processing is necessary in order to form an effective
9
detection method for early developmental dyslexia, and that these fundamental sensory
processing deficits are the root cause of the difficulties in phonological processing characteristic
of dyslexia.
In a 2011 study, Goswami’s group examined preliterate children from three language
backgrounds: English, Spanish and Chinese. They used the rate of onset of the amplitude
envelope (rise time) as the measure for effective parsing of sensory information into
speech-related stimuli, as it is known that the rise time is critical in rhythmic processing of
speech (Goswami et al., 2011). The results show that rise time sensitivity was a significant
predictor of phonological processing ability, and was in fact the only significant predictor of
reading acquisition, regardless of language. These findings lead the authors to postulate that
early remediation therapies targeting the rhythm of speech sounds, as opposed to the
phonological aspects of speech, may be more effective at treating the root cause of dyslexia
preemptively.
Family Risk Factors and Heritability
A large body of literature suggests that there is a strong relationship between family risk
factors related to dyslexia and onset of the disorder during childhood. Logistically-speaking, the
ability to assess various aspects of reading ability in relation to family history of dyslexia
throughout the course of a child’s reading acquisition is a difficult experiment to conduct. A very
recent study from Viersen et al., using longitudinal data from the Dutch Dyslexia Program,
carried out this exact experiment, examining whether oral language skills were related to two
distinct pathways leading to reading fluency, those pathways being the early pre-literacy pathway
10
of developing phonological awareness, and the late literacy pathway of developing linguistic
comprehension (critical for downstream reading comprehension). In addition, they studied the
relationship between oral language skills and a history of family risk for dyslexia, in order to
determine whether one was a covariate of the other in one or both of these literacy pathways
(Viersen et al., 2018).
The group found that family risk did in fact have an influence on the acquisition of
literacy, in that there were effects observed between family risk and pre-literacy skills, word
decoding and reading comprehension. However, despite the family-risk population of children
having lower reading performance scores on the neuropsychological tests administered, there
was no significant relationship between family risk for the early or late literacy pathways
(Viersen et al., 2018). The associations between family risk and these various aspects of literacy
were assessed at multiple time points (grade levels) throughout the course of the study for each
individual subject, thus accounting for the temporal dimension of when family risk is considered
a contributing factor to the onset of dyslexia.
Neuroimaging Studies examining Reading Level and Dyslexia
In addition to careful consideration of family risk when assessing early precursors of
dyslexia, critically-examining the reading level and developmental trajectory of children prior to
literacy training is essential in understanding the neurobiology of dyslexia. Structural and
functional deficits observed in brain regions/ networks related to reading may not necessarily be
specific to dyslexia, but could in fact be due to variability in the process of acquiring literacy
(Clark et al., 2014). In order to parse dyslexia-specific effects from more general reading
11
development effects, an experimental design which separates control subjects into age-matched
and reading level-matched groups is necessary.
In a 2014 longitudinal study conducted by Kristi Clark, in conjunction with researchers
from Norway, this experimental design was implemented in order to assess whether
neuroanatomical changes in the reading network were specific to children diagnosed with
dyslexia, or were also observed in reading level-matched control subjects, indicating that these
neuroanatomical changes may be the result of a developmental effect on reading. 27 Norwegian
children were tested prior to the start of formal literacy training, and after the diagnosis of
dyslexia, conveying the full spectrum of literacy acquisition up until a clinical diagnosis was
made. The results illustrated, perhaps unsurprisingly given the plethora of theories related to
dyslexia, that the neuroanatomical abnormalities observed prior to acquiring literacy were not in
the reading network (perisylvian region of the left hemisphere), but were in fact dispersed
throughout the brain in other, indirectly connected networks related to more primary sensory
functions, as well as those related to executive functioning (Clark et al., 2014). Abnormalities in
the reading network itself were not observed until the children had acquired literacy and received
a diagnosis of dyslexia. These results suggest that abnormalities in sensory regions/ networks
that are necessary for the optimal functioning of the reading network may be the core cause of
dyslexia, rather than disorders of the reading network itself.
12
Summary
From this review, it is clear that several theories exist to explain the underlying
behavioral, psychological and neurobiological causes of developmental dyslexia. However, the
evidence for each of these theories is broadly dispersed across many disciplinary domains,
making it increasingly difficult to compare each method and assess whether the relative merits of
one may trump the other. More likely than not, considering multiple theories at once, and in a
sense following a “hybrid theory” of dyslexia, may provide the most accurate and comprehensive
understanding of the underlying, root causes of the disorder. In addition, when attempting to
generate clinically-meaningful measures for diagnosing dyslexia prior to literacy acquisition,
clinicians and researchers alike should consider multiple other factors related to dyslexia, such as
its comorbidity with disorders like ADHD, family risk factors and hereditary information, as
well as the reading trajectory of children with dyslexia, and whether or not reading
developmental may influence the behavioral and neurological manifestations of dyslexia.
13
Chapter 2: An Adaptive Behavioral Paradigm for assessing Orthographic and Phonological
Processing in Children with Dyslexia
Barakat, Rita
1
; Krafnick, Anthony J.
4
; Lynch, Kirsten
3
; McGregor, Hadley A.
5
; Orozco, Max; Manis, Frank
2
; Zevin,
Jason
1,2
; Clark, Kristi
1. University of Southern California, Neuroscience Graduate Program (NGP)
2. University of Southern California, Department of Psychology
3. University of Southern California, Department of Neurology
4. Dominican University, Department of Psychology
5. Loma Linda University, School of Behavioral Health
Introduction
Over the last several decades, researchers in the interdisciplinary fields of
neurolinguistics and neuropsychology have sought to characterize the underlying deficits in
higher-level language processing in individuals (both children and adults) with dyslexia (Boada
and Pennington, 2006; Norton, Beach & Gabrieli, 2015; Peterson & Pennington, 2012; Ramus,
2003; Shaywitz & Shaywitz, 2005; Wolf & Bowers, 1999, and many more). A number of
theories and computational models have emerged attempting to explain the causal disruptions
that result in dyslexia and other reading disabilities, as well as the fundamental neural
mechanisms that facilitate reading (Bosse, Tainturier & Valdois, 2007; Erbeli et al., 2017;
Pennington et al., 2012; Zevin, 2019; Zoubrinetzky et al., 2019). One of the earliest and most
widely referenced models of reading is the Dual Route Hypothesis/ Model, proposed by
Seidenberg and McClelland (1989). The “Seidenberg-McClelland Model”, as it is often referred
to, articulates and demonstrates a comprehensive algorithm that utilizes sets of orthographic and
phonological units, with an additional layer of “hidden units”, to systematically simulate reading
while taking into account key points of variation that influence an individual’s unique reading
14
ability: these include the level of word recognition skill one possesses, the relative processing
difficulty for individual words, and pronunciation of novel items, to name a few (Seidenberg &
McClelland, 1989). This model and others helped to solidify the core relationship between
deficits in phonological processing and dyslexia, ultimately affirming that phonological
awareness effectively predicts reading ability, mechanistically-speaking.
While no single model of reading ability (or disability) has been universally accepted by
researchers and clinicians alike, the Triangle Model of Reading (see Figure 1 for schematic
illustration) has proven to be highly effective for the purposes of understanding the many
linguistic elements necessary for modeling fluent reading (Clark et al., 2014; Manis et al., 1996;
Wolf & Bowers 1999; Zevin, 2019). The model posits that fluent reading requires the seamless
integration of phonological awareness, or the mapping of speech sounds to individual
morphemes and graphemes; orthographic awareness, which results from visual perception of
morphemes and graphemes during reading; and semantic awareness, the process of deriving
meaning from the former two elements. In the context of dyslexia, phonological awareness is
arguably the most crucial of these linguistic elements to understand and incorporate into a model
of reading. It is well established that phonology is essential for generalization to pseudo-word
and/ or novel word reading, and the ability to perform this kind of mapping is closely related to
the manifestation of dyslexia (Zevin, 2019). It is also important to note the complex and not
entirely understood interplay between these three elements, in terms of how they inform one
another and influence the degree to which reading comprehension can be achieved. However, as
the Triangle Model has continued to evolve, it has been shown to be particularly efficient at
distinguishing between “surface dyslexia” (or behavioral manifestations that appear to mirror
15
dyslexia, in terms of difficulty with reading) and developmental dyslexia (Manis et al., 1996;
Zevin, 2019).
In addition to the Triangle Model of Reading, other theories that focus on these same
linguistic components have come to the fore, with the Phonological Deficit Theory of dyslexia
being widely recognized as one of the leading theories for understanding how reading is
disrupted in individuals with dyslexia. (Ramus et al., 2003; Share and Stanovich, 1995;
Ozernov-Palchik et al., 2016). The theory states that individuals with dyslexia have an
impairment in the representation, storage and retrieval of the phonological aspects of language,
and this impairment ultimately results in a deficit in the formation of grapheme-phoneme
associations. Difficulty in forming associations between speech sounds and visual letter/ word
forms has the potential to impact the ability to process the alphabetic system more globally,
which can lead to the more characteristic behavioral trademarks of dyslexia such as slow and
inaccurate word reading, and pseudoword decoding (Ramus et al., 2003). Behavioral
observations of individuals with dyslexia also provide strong support for this theory, and placing
the “starting point” of dyslexia etiology at the level of phonology allows for a variety of clinical
practices to be implemented that specifically target improving phonological awareness. In this
way, the Phonological Deficit Theory provides the most direct association between the
underlying cognitive processes of reading and the behavioral manifestations observed in
individuals with dyslexia, namely those pertaining to deficits in phonological awareness (Share
and Stanovich, 1995).
16
Figure 1. Schematic illustration of the Triangle Model of Reading, one of a number of reigning and
continuously-evolving hypotheses explaining how meaning is derived from successful orthographic and
phonological integration.
Yet despite the consensus in the research and clinical communities regarding the role of
phonological deficits in dyslexia, there remain significant challenges to effectively studying and
treating individuals with dyslexia. The aforementioned theories and models do not always or
necessarily account for individuals with significant comorbidities (such as Attention-Deficit and
Hyperactivity Disorder (ADHD), Autism Spectrum Disorder (ASD), and others), individuals
with discrepancies between IQ and reading ability, and/ or the school environment context, but
these factors are clearly important in determining meaningful reading outcomes. These models
do have the potential to translate into effective reading interventions, provided there are ways in
which environmental factors can be effectively quantified and taken into consideration.
Perhaps the most significant challenge lies in the immense heterogeneity of symptoms,
even among individuals who receive a clinical diagnosis. Because individuals with dyslexia may
show vastly different degrees of poor phonological awareness, treating phonological deficits as
the singular cause of dyslexia, rather than a key part of the complex integration with other
linguistic factors, can leave individuals with inadequate support and ultimately fail to address
individual-specific symptoms that hinder reading performance. It has been well documented in
17
the literature that isolating a “pure dyslexia” sample is incredibly difficult, if not impossible, and
therefore, it is incredibly important that behavioral tasks targeting dyslexia be sensitive to
individual differences. As a result of this diversity in dyslexia behaviors, many researchers have
worked to generate performance-matching tasks that measure reading ability, with one such
example being the adaptive single-word reading task utilized in Brad Schlagger’s 2002 study
comparing the relative reading performance and corresponding functional brain connectivity of
adults and school-aged children (Schlagger et al., 2002).
The following paper presents an adaptive behavioral task that serves as a variation on the
canonical phonological identification task (“PID”) implemented by Stephen Frost et al. (2009).
We highlight the process of development and preliminary results from the implementation of a
behavioral paradigm that aims to elucidate the reading deficits that result from impaired
orthographic and phonological processing, as well as the critical points of integration for these
two processing streams. We chose to focus primarily on the novel variations of this paradigm,
and its strong potential to yield significant results when implemented conjunctively with
neuroimaging methods. Furthermore, the data from this orthographic/ phonological task
paradigm provide substantially more opportunities for examining developmental effects
(Age-matched Controls + Dyslexia > Reading-level matched Controls), reading skill effects
(Age-matched Controls > Reading-level matched Controls + Dyslexia) and dyslexia-specific
effects (Age-matched Controls + Reading-level matched Controls > Dyslexia), by nature of
randomizing the stimuli presented in real-time to ensure each subject sees a unique set of stimuli
of varying difficulty levels.
18
We have also included a reference to the GitHub webpage containing the scripts to
generate the stimulus sets and execute the task (for behavioral and neuroimaging studies), as well
as the full stimulus repository used for this paradigm, as a means of promoting experimental
transparency and enhancing replicability. We outline the composition of each of the behavioral
tasks (one orthographically-based and one phonologically-based) and the generation of stimulus
repositories for each task, discuss some of the notable findings to date from the implementation
of this task design at the behavioral level, and conclude with future implications and potential for
the use of this paradigm in studying dyslexia and developing therapies for the specific reading
impairments associated with the disorder.
Methods
Participants
In testing this behavioral paradigm, 18 children with dyslexia (ages 10-13 years; 10
females, 8 males), 18 age-matched controls (ages 8-13 years; 9 females, 9 males) and 16
reading-level-matched controls (ages 7-10 years; 6 females, 10 males) were recruited from the
local community in Southeast Los Angeles, California to participate in four hours of
neuropsychological testing and one hour of MRI scanning. Dyslexia diagnosis was qualified
using standard criteria of at least one standard deviation of discrepancy between the child’s
measures of IQ and reading ability, or with a prior clinical diagnosis of dyslexia. Control subjects
were placed in the reading level-matched control group based upon their scores on the
Letter-Word Identification subtest of the Woodcock-Johnson Test of Academic Achievement 3
rd
edition (WJ-III): subjects who scored in the bottom 15
th
percentile when correcting for the age
19
discrepancy between these younger subjects and the dyslexia group were assigned to the reading
level-matched control group. Standard exclusion criteria for all three subject groups were as
follows: mental retardation and/ or diagnoses of other impairments, including ADHD,
neurological impairment, and any visual or hearing impairments (including color blindness). To
mitigate any potential confounds of a second language, only children who were monolingual,
native English speakers were included in the study.
Parent participation involved completion of questionnaires regarding their child’s
executive functions (the Behavior Rating Inventory of Executive Function, BRIEF) and
childhood problem behaviors (the Child Behavior Checklist, CBCL), as well as a personal data
questionnaire regarding medical, social and academic information about their child (Gioia et al.,
2010; Achenbach & Rescorla, 2001). Data from the CBCL was used as a screening measure for
children with severe behavioral difficulties or psychiatric diagnoses.
Neuropsychological testing for each child was completed in two sessions of two hours
each, to reduce fatigue. The Wechsler Intelligence Scale for Children 4
th
Edition (WISC-1V) was
used to assess intelligence (IQ), the Woodcock-Johnson Test of Academic Achievement 3
rd
edition (WJ-III) was used to assess reading skill (only on the Word Attack and Word
Identification sub-tests), the Comprehensive Test of Phonological Processes was used to assess
phonological awareness (characteristically diminished in children with dyslexia), and the
Working Memory Test Battery for Children (WMTB-C) was used to assess components of
Baddeley and Hitch’s model of working memory. In addition, all subjects completed the Test of
Word Reading Efficiency (TOWRE) A and B to assess individual subjects’ speed-accuracy in
20
reading real and pseudowords (Yeates & Donders, 2005; Braden & Alfonso, 2003; Pickering &
Gathercole, 2001; Torgesen et al., 1999).
Stimuli
Real word stimuli were selected using the Educator's Word Frequency Guide (Zeno,
1995). Word length was limited to 3-12 letters, and frequency was set to greater than or equal to
three (frequency represents the total number of occurrences in the corpus of the Educator's Word
Frequency Guide). The word list was further trimmed by removing all of the following:
compound words, words that have natural homophones, words that might be inappropriate for
children in the age range of the study, and words that would not have easily identifiable images
for the task (e.g. “mystery”). Pseudowords were generated using the program Wuggy (Keuleers
& Brysbaert, 2010). The real word list was run through Wuggy which produced pseudowords
matched on the following features : the length of sub-syllabic segments, length in letters,
transition frequencies, and two out of three sub-syllabic segments (default option). Wuggy
generates multiple pseudowords for each real word, and one was randomly selected for inclusion
in each trial. Pseudohomophones were created from the real words similar to the protocol
described in Bruno et al. (2008). Three researchers from the study (Lynch, McGregor and
Orozco) checked pseudohomophones to ensure they were pronounced the same as the real word.
21
Task Design
Figure 2. Two sample stimuli (image + target and foil), both at Difficulty Level 1, for the orthographic (right) and
phonological (left) behavioral tasks. Subjects were instructed to give a key press indicating which word/
pseudoword most closely matched the image displayed, with judgment criteria varying by task.
Each participant was first presented with the orthographic matching task, containing 60
trials, followed by the phonological matching task containing the same number of trials. The
time of presentation for each token was limited to six seconds, and the subjects' responses were
recorded via a left or right button press (indicating the selection of either the target or foil
accompanying the stimulus) within this six second event window. Upon the subjects’ button
press, a fixation cross was presented for the duration of the trial period, before automatically
advancing to the next six-second trial. Figure 2 illustrates a sample stimulus/ target-foil pair
from each of the two tasks: for both tasks, the participants were presented with a pictorial
representation of the word, as well as the two options to match the representation, with the
criteria for selecting the correct choice based on the specific task (orthographic-matching versus
phonological-matching). Figure 3 shows a schematic representation of the experimental design
and timing for each trial on the orthographic task. Stimulus timing did not vary between the
orthographic and phonological tasks, only the nature of the word/ pseudoword stimuli
themselves.
22
Figure 3. Schematic illustration of the event-related task design implemented for the orthographic and phonological
task runs. The event-related design is optimized to be used in the context of neuroimaging research, and is also
effective at distinguishing between orthographic and phonological judgements in a stand-alone computer task
environment.
The adaptability of the task to the individual participant’s performance provides a critical
control for accuracy confounds when interpreting the data post-task administration. The
‘difficulty’ of the stimuli was determined through balancing various orthographic and
phonological features of each word relative to the sum of all potential words in the stimulus
repositories of each task, then pre-assigning each stimulus to a difficulty level based on the
reading (grade) level at which the corresponding word appears frequently in academic texts.
Both tasks were initiated with the first trial set to a median difficulty level of three, and each
subsequent trial could range anywhere from a difficulty level of one to five (with level one being
the least difficult, and level five being the most difficult). Changes in difficulty level occurred in
a stepwise fashion (without skipping levels between trials) based upon individual subject
performance during the task.
23
The orthographic and phonological tasks were designed to set a benchmark accuracy
level of approximately eighty percent for all subjects, though as evidenced from the graphs
plotting accuracy across both tasks, there were individuals who exceeded this accuracy
benchmark, illustrating a ceiling effect in both tasks. The determination of this benchmark was
based on previous work showing that an approximately eighty percent accuracy level would
encourage participation from subjects, particularly those in a pediatric or adolescent population,
while still challenging and motivating them, thus eliminating accuracy, effort and motivation as
significant confounds in the interpretation of behavioral results between subject groups (Kujala,
Richardson & Lyytinen, 2010).
In this way, quantification of each participant’s performance was completed through
measurement of the number of trials provided at each difficulty level, then averaged over the
total number of trials in each task, to determine the average difficulty level for each individual
subject on the two tasks. In addition, reaction time measures for each trial provided further
information regarding overall participant performance. Finally, a novel measurement to
characterize the number of instances in which subjects shifted difficulty level within a task,
variability, gives insight into which individual subjects or subject groups show a more varied
performance for a specific task.
Statistics
Individual subject data from the orthographic and phonological tasks were compiled into
the appropriate data structures in R for the purpose of performing linear regression analysis using
the built-in linear model (lm) function. Age and full-scale IQ were included as non-interest
24
(controlled) variables in the model. Full model: Performance Measure ~ Group + Age +
Full-scale IQ + Group*Task + (1|Task) + (1|Subject).
Results
Table 1. Neuropsychological and demographic summary data for subjects included in the study of the behavioral
task paradigm. There were no significant differences in the gender distribution across all three subject groups, and
the summary statistics indicate overlaps in age across all three subject groups. There were also no significant
differences in the WISC-V Full-scale Composite IQ distributions across all three groups, which is of particular
importance given historical definitions of dyslexia centering around discrepancies in IQ and reading ability. Select,
reading-related neuropsychological test results have been reported, including the TOWRE A and B mean scores and
standard deviations, which were significantly different across all three groups, as well as the Woodcock Johnson
Letter-Word Identification subtest mean scores and standard deviations, which were critical for calculating reading
level and also different across the subject groups.
Age-matched
Control Subjects
Reading Level-matched
Control Subjects
Dyslexia Subjects Statistics
Sample Size 18 16 18 N/A
Gender
Distribution
(M:F)
9:9 10:6 8:10
( (1) = 2.83, χ 2
p > 0.05)
Age Mean
(S.D.), Range
10.87 (1.39), 8-12.92 9.91 (1.85), 7.5-9.5 10.85 (1.32), 10-12.92
(F(2,49) = 2.153,
p > 0.1)
WISC-V
Full-scale
Composite IQ
Mean (S.D.)
102.83 (13.03) 98.38 (8.85) 101.78 (12.1)
(F(2,49) = 0.675,
p > 0.5)
TOWRE A
Score Mean
(S.D.)
109.39 (12.59) 100.88 (11.03) 87 (22.2)
(F(2,49) = 8.722,
p < 0.001)
TOWRE B
Score Mean
(S.D.)
109.89 (14.21) 100.07 (11.97) 86 (22.8)
(F(2,49) = 8.813,
p < 0.001)
Woodcock
Johnson III
Letter-Word
Identification
Score Mean
(S.D.)
109.5 (8.98) 100.5 (6.75) 94.39 (10.26)
(F(2,49) = 13.244
p < 0.001)
25
Table 2A. Summary results (including Means, Standard Error (“SE”), Adjusted R
2
(“R
2
”),
F-statistic (“F”) and p-value (“p”) from performing linear regression analysis in R to quantify
group differences on the orthographic and phonological tasks. Results from the linear regression
analysis comparing task performance across all three subject groups for the orthographic task.
Orthographic
Task
Age-matched
Control Subjects
Reading Level-matched
Control Subjects
Dyslexia
Subjects
Statistics
Average
Difficulty Level
4.56 3.13 3.28
SE: 0.1027
R
2
: 0.1932
F(2,52): 13.45
p = 0.00059
Average Reaction
Time (seconds,
correct trials only)
1.97 2.15 2.45
SE: 0.2945
R
2
: 0.01625
F(2,52): 1.859
p = 0.1787
Accuracy (%) 92.1 79.36 79.16
SE: 0.1027
R
2
: 0.1932
F(2,52): 13.45
p = 0.00069
Variability 10.27 19.25 19.25
SE: 0.01651
R
2
: 0.1442
F(2,52): 13.45
p = 0.0029
26
Table 2B. Summary results (including Means, Standard Error (“SE”), Adjusted R
2
(“R
2
”),
F-statistic (“F”) and p-value (“p”) from performing linear regression analysis in R to quantify
group differences on the orthographic and phonological tasks. Results from the linear regression
analysis comparing task performance across all three subject groups for the phonological task.
Phonological
Task
Age-matched
Control Subjects
Reading Level-matched
Control Subjects
Dyslexia
Subjects
Statistics
Average
Difficulty Level
4.67 3.03 3.45
SE: 0.1023
R
2
: 0.0694
F(2,52): 4.878
p = 0.03172
Average Reaction
Time (seconds,
correct trials only)
2.27 2.65 2.69
SE: 0.2399
R
2
: 0.09371
F(2,52): 6.377
p = 0.01471
Accuracy (%) 95.5 75.41 81.96
SE: 0.7873
R
2
: 0.1775
F(2,52): 10.71
p = 0.0021
Variability 7.18 14 17.63
SE: 0.01794
R
2
: 0.1843
F(2,52): 12.75
p = 0.00079
27
Average Difficulty Level
Figure 4. Graphical representation of group by task performance for all three subject groups on the orthographic and
phonological tasks with average difficulty level as the performance measure. Significance codes: * = p < 0.05, ** =
p < 0.01, *** = p < 0.005 and **** = p < 0.001.
In examining average difficulty level as a measure of performance on both tasks,
reading-level matched control subjects showed a qualitatively similar performance to dyslexia
subjects on both, the orthographic and phonological tasks, which confirms our initial hypothesis
that these groups would function at similar difficulty levels on one or both of the tasks.
Age-matched control subjects performed at substantially higher difficulty levels on both tasks,
and after accounting for individual variability between groups, showed a significantly greater
performance on the orthographic task (F(2,52) = 13.45, p < 0.001) and phonological task
(F(2,52) = , p < 0.05). There were no group-by-task interactions for the performance measure of
average difficulty level.
28
Average Reaction Time (correct trials only)
Figure 5. Graphical representation of group by task performance for all three subject groups on the orthographic and
phonological tasks with average reaction time (on correct trials only) as the performance measure. Significance
codes: * = p < 0.05, ** = p < 0.01, *** = p < 0.005 and **** = p < 0.001.
When looking at average reaction time (on correct trials only) as a measure of
performance on both tasks, a broader distribution across the three subject groups was observed
on the Orthographic task in comparison to the Phonological task. Again, reading level-matched
control subjects and dyslexia subjects showed more similar reaction times in comparison to the
age-matched control subjects, who showed faster reaction times on both the Orthographic and
Phonological tasks. After accounting for individual variability between groups, however, these
differences did not meet the threshold for statistical significance on the Orthographic task
(F(2,52 ) = 1.859, p > 0.1), but did meet threshold on the Phonological task (F(2,52) = 6.377, p <
0.05). There were no group-by-task interactions for the performance measure of average reaction
time.
29
Accuracy
Figure 6. Graphical representation of group by task performance for all three subject groups on the orthographic and
phonological tasks with accuracy as the performance measure. Significance codes: * = p < 0.05, ** = p < 0.01, ***
= p < 0.005 and **** = p < 0.001.
In order to validate that the original task design did in fact control for accuracy, the
accuracy (%) for both tasks was tabulated for each subject group. While the reading
level-matched controls and dyslexia subjects performed within the originally-intended 75-80%
accuracy window, age-matched controls showed a ceiling effect on both tasks, significantly
exceeding the benchmark accuracy threshold. Significant differences in accuracy were observed
between the age-matched control group and the reading-level matched control and dyslexia
groups on the Orthographic task (F(2,52) = 13.33, p < 0.001) and the Phonological task (F(2,52)
= 10.71, p < 0.005). There were no group-by-task interactions for the performance measure of
accuracy.
30
Variability
Figure 7. Graphical representation of group by task performance for all three subject groups on the orthographic and
phonological tasks with accuracy as the performance measure. Significance codes: * = p < 0.05, ** = p < 0.01, ***
= p < 0.005 and **** = p < 0.001.
For the purpose of this study, variability refers to the number of times an individual
subject shifted difficulty level within a single task run. These individual variability values were
then averaged within subject groups to assess differences across groups on both tasks. As with
the previous three performance measures (average difficulty level, average reaction time and
accuracy), the reading level-matched control subjects performed similarly to dyslexia subjects on
both tasks, but most notably on the Orthographic task (almost identical values), while the
age-matched control subjects switched difficulty level significantly fewer times in comparison,
and thus showed lower variability on the Orthographic task (F(2,52) = 9.765, p < 0.005) and the
31
Phonological task (F(2,52) = 12.75, p < 0.001). There were no group-by-task interactions for the
performance measure of variability.
Discussion
The patterns observed from the data collected using this adaptive paradigm are consistent
with the hypotheses and findings from the behavioral literature, in that they show subjects with
dyslexia demonstrate similar performance on single-word judgment tasks to their reading-level
matched control counterparts who are still in the early developmental stages of acquiring literacy.
The fact that no interactions between subject group and task performance (for all four
performance measures) on the phonological task were observed is not entirely surprising given
the adaptive nature of the paradigm. Previous studies showing evidence of a dyslexia-specific
phonological deficit on similar behavioral task paradigms did not set a benchmark for accuracy, a
threshold that in the case of the present study, likely masked the variability that would otherwise
be noted between the reading level-matched controls and dyslexia subjects (Bolger et al., 2008;
Bruno et al., 2008; Cao et al., 2006; Perrachione et al., 2016; Skeide, Brauer, & Friederici, 2015;
Temple et al., 2001). Thus, the data reveal the effects of development and acquired reading skill
without confirming any dyslexia-specific effects on reading ability, by nature of the
approximately eighty-percent accuracy threshold applied to the task design.
The ability to parse more nuanced performance measures (such as difficulty level and
variability), which are not traditionally assessed in similar linguistic designs, also optimizes this
behavioral paradigm for more robust individual subject analyses. Such an approach may be of
particular interest in explorative datasets in which the criteria for dyslexia are not clearly defined,
32
or hypothesis-driven research in which the interest is to observe more general (not limited to
dyslexia) reading impairments in pediatric populations.
In its current form, the task design presents numerous opportunities for modification to
suit the needs of a particular study population and methodology. While the existing task is
event-related and set to present each trial for a period of six seconds, this can be adjusted to
conform to a jittered event-related design in order to avoid issues of collinearity presented
through a general linear model analysis (as is standard in task-fMRI research). In addition, a
hybrid version of the independent task designs can be created using the existing task
infrastructure (code) such that alternating trials of Orthographic and Phonological stimuli/
target-foil pairs are interspersed throughout a single task run. Rest event periods can also be
incorporated into the existing design, in order to more efficiently model task-specific activation
for the purpose of an fMRI experimental design. Finally, the specific real word stimuli can be
modified for the study population (e.g. more advanced reading level words for older subject
populations) and reprocessed through Wuggy, then reincorporated into the existing task script to
make the design suitable for different subject groups, while still allowing for robust analysis of
alternative performance measures.
Conclusion
The results from these pilot data show that the task is 1) sensitive to behavioral
differences in dyslexia, 2) able to detect individual differences in preferred cognitive strategies
(specifically, differences in various performance measures on word decoding tasks) and 3) is
robust to confounds in motivation and difficulty due to the adaptive nature of the task.
33
Due to the vast heterogeneity in symptom severity and presentation among individuals
with dyslexia, it is critical that behavioral paradigms seeking to elucidate higher-order integration
of Orthographic and Phonological processing streams allow for numerous measures of reading
ability to be evaluated. By extracting information about difficulty level and its derivative
(variability), in addition to the more typically-examined reaction time and accuracy, one can
make individual and group comparisons on task performance and reading ability more generally.
Binning individuals into particular reading levels is a standard practice in western linguistic
culture, and thus, a task design that provides a link to this existing structure for assessing a
child’s reading ability provides significant potential for informing future remediation therapies.
34
Chapter 3: Neuroanatomical Differences as a function of Development, Reading Skill and
Dyslexia
Barakat, Rita
1
; Zevin, Jason
1,2
; Clark, Kristi
1,3
1. University of Southern California, Neuroscience Graduate Program (NGP)
2. University of Southern California, Department of Psychology
3. University of Southern California, Department of Neurology
Introduction
Recent advances in neuroimaging technology have given rise to a plethora of uni- and
multi-modal neuroimaging studies examining the structural and functional correlates of higher
cognitive processes, including reading and language processing. In addition to the technological
advances that have significantly improved image resolution and acquisition, processing of
magnetic resonance imaging (MRI) data has also progressed significantly such that more
functionally-relevant localization and parcellation of neuroanatomical regions is possible. All
these advances are highly relevant in the context of the present study, and set the stage for future
research that seeks to incorporate behavioral data that informs subsequent image analyses.
Examining the neuroanatomical correlates of reading in any population requires a
familiarity with the somewhat amorphous, typically left-hemisphere localized cortical areas that
are attributed to reading and language ability more broadly. Even prior to newer acquisition and
processing methods, several functional and structural MRI studies have corroborated the
pre-neuroimaging findings of historical figures in psycholinguistics and neurology (including,
but certainly not limited to: Paul Broca, Carl Wernicke, Ludwig Lichtheim and Norman
Geschwind) that implicated these left-hemisphere regions in reading and language function.
Turning to these more recent studies, a cluster of frontal-temporal and temporal-occipital cortical
35
areas continue to illustrate important contributions to reading. For the purpose of the present
study, we will focus on the roles of the following regions: the fusiform, supramarginal and
angular gyri, the inferior, middle and superior temporal lobes, the lateral occipital lobe, the
insular cortex, the pars triangularis and the superior frontal lobe.
Fusiform, Supramarginal and Angular Gyri
Functional MRI (fMRI) studies examining the automaticity of word recognition, distinct
from comprehension, have demonstrated that occipito-temporal regions, including the fusiform
gyrus, are active during masked priming tasks (Dehaene et al., 2004; Devlin et al., 2006). This is
significant in that it distinguishes the role of the (primarily posterior-middle) fusiform gyrus from
other object recognition-based behaviors, and additional studies showed that this posterior region
of the fusiform gyrus is also active when subjects are tasked with reading passages aloud,
suggesting that this region (along with other left-hemisphere regions associated with the broader
“reading network”) is critical for orthographic-phonological mapping as described in the “Dual
Route” Model of reading (Graves et al., 2004; Proverbio, Vecchi and Zani, 2004; Seidenberg and
McClelland, 1989).
These and other findings have given rise to a more recent moniker for this particular
portion of the fusiform gyrus, referring to it as the “Visual Word-Form Area” (VWF A), stemming
from evidence in fMRI and event-related potential (ERP) studies implicating this specific area in
distinguishing words from pseudowords during reading (Pammer et al., 2004; Xue et al., 2006).
In addition to the fusiform gyrus, the supramarginal and angular gyri, anatomically connected to
other left-hemisphere language centers via the arcuate fasciculus, have also been shown to play a
36
role in word recognition on single-word identification tasks (Graves et al., 2004; Stoeckel et al.,
2009).
Lateral Occipital Lobe
Posterior to the fusiform gyrus, the left lateral occipital lobe is also associated with
letter-word identification, and distinguishing words from pseudowords/ non-words: an early
lesion study from Sakurai, Ichikawa and Mannen showed that damage to the left posterior lateral
occipital lobe resulted in pure alexia (2001). More recent fMRI studies have confirmed that
indeed, this subregion of the occipital lobe, which overlaps with the generally-defined VWFA, is
critical for reading as well as broader categorization and identification of images (Borowsky et
al., 2007; Neudorf et al., 2022).
Inferior, Middle and Superior Temporal Lobes
There is a large body of literature supporting the functions of the inferior, middle and
superior temporal lobes in reading and language comprehension, though each of these subregions
are thought to play different roles in these complex processes. The left inferior temporal lobe in
particular has been shown to be active during lexical-based reading tasks, and is also associated
with the so-called “ventral visual attention stream”, thought to be responsible for object
recognition (Borowsky et al., 2007; Brown, 2009; Glezer et. al., 2016; Qu et al., 2022). The
posterior portion of the inferior temporal lobe also forms a nexus with the posterior fusiform
gyrus, forming the VWFA.
37
In contrast, the middle temporal lobe has long been associated with multimodal semantic
processing, as evidenced by functional imaging studies as well as lesion-based studies (Graves et
al. 2010; Joseph et al., 2006; Kronbichler et al., 2007; Pammer et al., 2004). Finally, the superior
temporal lobe has been shown to play a role in word versus pseudoword auditory judgment tasks
and is thus believed to play a role in phonological processing and speech-to-sound mapping,
unsurprising given its proximity to and overlap with Heschl’s gyrus (Baeck et al., 2015; Bolger
et al., 2008; Graves et al., 2010; Simos et al., 2000). In many respects, the left temporal lobe as a
whole provides an anatomical representation of the Triangle Model of reading, in that it serves as
a structural and functional intersection for orthographic and phonological information streams,
both of which are essential for comprehension.
Insular Cortex, Pars Triangularis and Superior Frontal Lobe
Sometimes referred to as the “fifth lobe”, the insular cortex has only recently been
implicated in reading and language function, stemming from early lesion work illustrating that
damage to the left insula (specifically the anterior portion) is significantly associated with
aphasia and other language impairments (Glezer et al., 2016; Ibañez,, Gleichgerrcht, and Manes,
2010). More recent fMRI studies suggest that the insula may be involved in word versus
pseudoword naming tasks, though this may be more indicative of the overall function of the
insula in speech production rather than phonological awareness specifically (Joseph et al., 2006;
Oh, Duerden and Pang, 2014; Randazzo et al., 2019; Woolnough et al., 2019). Similarly, the pars
triangularis, a region only recently identified as part of the reading network, is activated during
word recognition tasks, suggesting a potential role in orthographic processing (Elmer, 2016;
38
Randazzo et al., 2019). While not traditionally considered part of the reading network, the
superior frontal lobe’s role in language processing more broadly (and in decision-making, as it
relates to distinguishing between words and pseudo-/non-words) cannot be overlooked. In
addition to its role in controlling saccades produced during normal reading, the left superior
frontal lobe has also been shown to be active during word recognition tasks, indicating a
potential role in orthographic processing (Christodoulou et al., 2014; Siok et al., 2008; Sood and
Sereno, 2016).
Figure 8. Canonical left-hemisphere regions associated with reading and language more broadly, viewed from the
exterior cortical surface (sagittal view).
39
These cortical regions were selected for region-of-interest (ROI) analysis for the present
study based on their aforementioned roles in reading. Metrics of cortical thickness, surface area
and volume were extracted for each region and used to perform linear regression analyses that
sought to elucidate the effects of development, inherent reading skill and dyslexia specifically on
reading ability, as measured through neuropsychological tests but principally through an adaptive
behavioral task paradigm assessing individual subjects’ orthographic and phonological judgment.
Methods
Participants
25 children with dyslexia (ages 8-13 years; 13 females, 12 males), 22 age-matched
controls (ages 8-13 years; 9 females, 13 males) and 18 reading-level-matched controls (ages 7-10
years; 8 females, 10 males) were recruited from the local community in Southeast and
South-Central Los Angeles, California to participate in four hours of neuropsychological testing
and one hour of MRI scanning. Dyslexia diagnosis was qualified using standard criteria of at
least one standard deviation of discrepancy between the child’s measures of IQ and reading
ability, or with a prior clinical diagnosis of dyslexia. Control subjects were placed in the reading
level-matched control group based upon their scores on the Letter-Word Identification subtest of
the Woodcock-Johnson Test of Academic Achievement 3
rd
edition (WJ-III): subjects who scored
in the bottom 15
th
percentile when correcting for the age discrepancy between these younger
subjects and the dyslexia group were assigned to the reading level-matched control group.
Standard exclusion criteria for all three subject groups were as follows: mental retardation and/
or diagnoses of other impairments, including ADHD, neurological impairment, and any visual
40
or hearing impairments (including color blindness). To mitigate any potential confounds of a
second language, only children who were monolingual, native English speakers were included in
the study.
Parent participation involved completion of questionnaires regarding their child’s
executive functions (the Behavior Rating Inventory of Executive Function, BRIEF) and
childhood problem behaviors (the Child Behavior Checklist, CBCL), as well as a personal data
questionnaire regarding medical, social and academic information about their child (Gioia et al.,
2010; Achenbach & Rescorla, 2001). Data from the CBCL was used as a screening measure for
children with severe behavioral difficulties or psychiatric diagnoses.
Neuropsychological testing for each child was completed in two sessions of two hours
each, to reduce fatigue. The Wechsler Intelligence Scale for Children 4
th
Edition (WISC-1V) was
used to assess intelligence (IQ), the Woodcock-Johnson Test of Academic Achievement 3
rd
edition (WJ-III) was used to assess reading skill (only on the Word Attack and Word
Identification sub-tests), the Comprehensive Test of Phonological Processes was used to assess
phonological awareness (characteristically diminished in children with dyslexia), and the
Working Memory Test Battery for Children (WMTB-C) was used to assess components of
Baddeley and Hitch’s model of working memory. In addition, all subjects completed the Test of
Word Reading Efficiency (TOWRE) A and B to assess individual subjects’ speed-accuracy in
reading real and pseudowords (Yeates & Donders, 2005; Braden & Alfonso, 2003; Pickering &
Gathercole, 2001; Torgesen et al., 1999).
41
Structural MRI Acquisition Parameters
All subjects were scanned at the University of Southern California Dornsife
Neuroimaging Institute (DNI) on a Siemens 3.0 Tesla Prisma Scanner. Total of 3.83 minutes
acquisition time for T1- (MPRAGE) and T2-weighted scanning sequences. An MPRAGE
sequence was implemented, with TR = 2.4 seconds, TE = 0 milliseconds, flip angle = 8
o
, voxel
resolution = 0.7 mm
3
. A T2 sequence was also implemented (for more robust segmentation, as
per the Human Connectome Project’ s minimal preprocessing pipelines for structural MRI data,
as well as for incidental findings per DNI scanning policy), with TR = 10 seconds, TE = 88
milliseconds, flip angle = 120
o
, voxel resolution = 0.8 mm x 0.8 mm x 3.5 mm.
FreeSurfer/ HCP Pipeline Preprocessing
Subjects’ raw, high-resolution MPRAGE images were parcellated into white matter, gray
matter and cerebrospinal fluid (CSF) using the Human Connectome Project’ s modified
FreeSurfer protocol, implemented through the PreFreeSurfer and FreeSurfer pipeline scripts
(Glasser et al., 2013). This approach incorporates white matter intensity data from T2-weighted
images to perform a more robust segmentation and parcellation of cortex, using the
Desikan-Killiany anatomical atlas as reference. Regions-of-interest (ROIs) from the resulting
parcellated images were selected a priori, based on evidence in the literature illustrating the role
of these regions in reading and related cognitive functions.
42
Statistical Analysis and Qit Visualization
Table 3. Linear contrasts used to perform General Linear Model (GLM) analysis in Python (with parallel analyses in
R), parsing the differential effects of reading skill, development and dyslexia on brain morphology (specifically,
cortical thickness, surface area and volume) in left-hemisphere reading-associated brain regions.
Contrast Effect
Age-matched Controls - (Reading Level-matched Controls + Dyslexia)
AGE - (READ + DYS)
Reading Skill
Reading Level-matched Controls - (Age-matched Controls + Dyslexia)
READ - (AGE + DYS)
Development
(Age-matched Controls + Reading Level-matched Controls) - Dyslexia
(AGE + READ) - DYS
Dyslexia
After quality-checking the results from the HCP/ FreeSurfer segmentation, mean values
for volume (mm
3
), surface area (mm
2
) and cortical thickness (mm) of a priori ROIs were
extracted and incorporated into a one-way Analysis of Variance (ANOV A) performed in R
Studio, and subsequently verified through linear regression analysis in Python, to determine the
relationship between dyslexia diagnosis/ qualification (subject group) and cortical thickness,
surface area and volume of these key brain regions implicated in reading (RStudio Team, 2020).
For this analysis, as well as subsequent analyses in Chapter 4, three contrasts, or effects were
created in order to isolate structural differences that are indicative of a reading skill effect, a
developmental effect and a dyslexia effect (see Table 3 for summary of three effect contrasts).
Statistical significance was determined at the p < 0.05 level, and Bonferroni correction was
performed (at p < 0.005) to address multiple comparisons. The results from the linear regression
analysis were then mapped onto an inflated pial surface using the Quantitative Imaging Toolkit
(Qit), developed by Ryan Cabeen (2018).
43
Results
Reading Skill Effect
The following are the results showing significant differences (at p < 0.05) in cortical
thickness (Table 4A and Figure 9A), surface area (Table 4B and Figure 9B) and volume (Table
4C and Figure 9C) when focusing on the effect of reading skill (contrast of Reading
Level-matched Controls and Dyslexia subjects vs. Age-matched Controls). A few regions
showed a statistically-significant association with this effect (only the superior frontal and
supramarginal gyri, for this particular contrast). In examining the surface area data for this same
contrast, several more regions were shown to be significantly different across the two groups.
Finally, the volume results for this contrast showed substantial overlap with the surface area data,
in terms of which regions were implicated in the effect of reading skill.
Table 4A. Statistical results from Analysis of Variance (ANOV A) of cortical thickness data (left hemisphere only) to
examine reading skill effect (contrast: AGE - [READ + DYS]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus -0.02829 F(2,62) = 0.1195 0.8875
Inferior Temporal Lobe 0.01376 F(2,62) = 1.446 0.2432
Lateral Occipital Lobe 0.0245 F(2,62) = 1.804 0.1732
Lateral Orbitofrontal
Cortex
-0.0128 F(2,62) = 0.5957 0.5543
Middle Temporal Lobe -6.452e-05 F(2,62) = 0.9979 0.3745
Pars Triangularis -0.02831 F(2,62) = 0.1191 0.888
Superior Frontal Lobe 0.08363 F(2,62) = 3.921 0.02493
Superior Temporal Lobe -0.01403 F(2,62) = 0.5572 0.5756
Supramarginal Gyrus 0.08256 F(2,62) = 3.88 0.02585
Insula -0.02981 F(2,62) = 0.0738 0.9289
44
Figure 9A. FreeSurfer inflated pial surface map depicting cortical thickness statistics (p-value) for reading skill
effect (contrast: AGE - [READ + DYS]). The color bar indicates the statistical significance (p-value) of the
individual ANOV A F-tests performed.
Table 4B. Statistical results from Analysis of Variance (ANOV A) of surface area data (left hemisphere only) to
examine reading skill effect (contrast: AGE - [READ + DYS]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus 0.6012 F(2,62) = 49.25 1.568e-13
Inferior Temporal Lobe 0.5968 F(2,62) = 48.37 2.203e-13
Lateral Occipital Lobe 0.4 F(2,62) = 22.34 4.953e-08
Lateral Orbitofrontal
Cortex
0.58 F(2,62) = 45.19 7.82e-13
Middle Temporal Lobe 0.4983 F(2,62) = 32.78 1.937e-10
Pars Triangularis 0.2751 F(2,62) = 13.14 1.743e-05
Superior Frontal Lobe 0.6487 F(2,62) = 60.08 3.089e-15
Superior Temporal Lobe 0.623 F(2,62) = 53.87 2.758e-14
Supramarginal Gyrus 0.5043 F(2,62) = 33.56 1.33e-10
Insula 0.5401 F(2,62) = 38.58 1.302e-11
45
Figure 9B. FreeSurfer inflated pial surface map depicting surface area statistics (p-value) for reading skill effect
(contrast: AGE - [READ + DYS]). The color bar indicates the statistical significance (p-value) of the individual
ANOV A F-tests performed.
Table 4C. Statistical results from Analysis of Variance (ANOV A) of volume data (left hemisphere only) to examine
reading skill effect (contrast: AGE - [READ + DYS]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus 0.4787 F(2,62) = 30.39 6.338e-10
Inferior Temporal Lobe 0.5956 F(2,62) = 48.12 2.428e-13
Lateral Occipital Lobe 0.3787 F(2,62) = 20.51 1.459e-07
Lateral Orbitofrontal
Cortex
0.5488 F(2,62) = 39.92 7.232e-12
Middle Temporal Lobe 0.4954 F(2,62) = 32.41 2.318e-10
Pars Triangularis 0.2074 F(2,62) = 9.373 0.0002775
Superior Frontal Lobe 0.6701 F(2,62) = 66 4.387e-16
Superior Temporal Lobe 0.5805 F(2,62) = 45.28 7.551e-13
Supramarginal Gyrus 0.3961 F(2,62) = 21.99 6.07e-08
Insula 0.6514 F(2,62) = 60.78 2.435e-15
46
Figure 9C. FreeSurfer inflated pial surface map depicting volume statistics (p-value) for reading skill effect
(contrast: AGE - [READ + DYS]). The color bar indicates the statistical significance (p-value) of the individual
ANOV A F-tests performed.
47
Developmental Effect
The results presented below indicate regions in which significant differences (at p < 0.05)
in cortical thickness (Table 5A and Figure 10A), surface area (Table 5B and Figure 10B) and
volume (Table 5C and Figure 10C) were observed when focusing on the effect of development
(contrast of Age-matched Controls and Dyslexia subjects vs. Reading Level-matched Controls).
It is noteworthy that only the region which showed significant differences in cortical thickness
across the two groups is the superior frontal gyrus. In examining the surface area data for this
same contrast, several more regions (including the middle and superior temporal lobes, which
have been shown in numerous studies to be closely related to reading and visual word-form
processing) were significantly different across the two groups. Finally, the volume results for this
contrast showed substantial overlap with the surface area data, in terms of which regions were
implicated in the effect of development on reading.
48
Table 5A. Statistical results from Analysis of Variance (ANOV A) of cortical thickness data (left hemisphere only) to
examine developmental effect (contrast: READ - [AGE + DYS]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus -0.02167 F(2,62) = 0.3213 0.7264
Inferior Temporal Lobe 0.01313 F(2,62) = 1.426 0.2481
Lateral Occipital Lobe 0.01685 F(2,62) = 1.549 0.2207
Lateral Orbitofrontal
Cortex
-0.028 F(2,62) = 0.1284 0.8797
Middle Temporal Lobe -0.008746 F(2,62) = 0.7225 0.4896
Pars Triangularis -0.0183 F(2,62) = 0.425 0.6557
Superior Frontal Lobe 0.1051 F(2,62) = 4.759 0.01194
Superior Temporal Lobe -0.02669 F(2,62) = 0.168 0.8457
Supramarginal Gyrus 0.04522 F(2,62) = 2.516 0.08904
Insula -0.02816 F(2,62) = 0.1235 0.884
Figure 10A. FreeSurfer inflated pial surface map depicting cortical thickness statistics (p-value) for developmental
effect (contrast: READ - [AGE + DYS]). The color bar indicates the statistical significance (p-value) of the
individual ANOV A F-tests performed.
49
Table 5B. Statistical results from Analysis of Variance (ANOV A) of surface area data (left hemisphere only) to
examine developmental effect (contrast: READ - [AGE + DYS]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus 0.5983 F(2,62) = 48.67 1.959e-13
Inferior Temporal Lobe 0.5995 F(2,62) = 48.91 1.788e-13
Lateral Occipital Lobe 0.4168 F(2,62) = 23.87 2.055e-08
Lateral Orbitofrontal
Cortex
0.5846 F(2,62) = 46.04 5.544e-13
Middle Temporal Lobe 0.4972 F(2,62) = 32.65 2.066e-10
Pars Triangularis 0.2751 F(2,62) = 13.15 1.741e-05
Superior Frontal Lobe 0.6498 F(2,62) = 60.37 2.805e-15
Superior Temporal Lobe 0.6259 F(2,62) = 54.55 2.156e-14
Supramarginal Gyrus 0.5044 F(2,62) = 33.56 1.327e-10
Insula 0.5743 F(2,62) = 44.18 1.184e-12
Figure 10B. FreeSurfer inflated pial surface map depicting surface area statistics (p-value) for developmental effect
(contrast: READ - [AGE + DYS]). The color bar indicates the statistical significance (p-value) of the individual
ANOV A F-tests performed.
50
Table 5C. Statistical results from Analysis of Variance (ANOV A) of volume data (left hemisphere only) to examine
developmental effect (contrast: READ - [AGE + DYS]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus 0.4684 F(2,62) = 29.19 1.165e-09
Inferior Temporal Lobe 0.5949 F(2,62) = 48 2.544e-13
Lateral Occipital Lobe 0.4018 F(2,62) = 22.5 4.512e-08
Lateral Orbitofrontal
Cortex
0.5538 F(2,62) = 40.72 5.102e-12
Middle Temporal Lobe 0.4976 F(2,62) = 32.7 2.013e-10
Pars Triangularis 0.2072 F(2,62) = 9.361 0.0002802
Superior Frontal Lobe 0.6647 F(2,62) = 64.43 7.286e-16
Superior Temporal Lobe 0.5772 F(2,62) = 44.68 9.62e-13
Supramarginal Gyrus 0.3965 F(2,62) = 22.03 5.933e-08
Insula 0.6799 F(2,62) = 68.97 < 2.2e-16
Figure 10C. FreeSurfer inflated pial surface map depicting volume statistics (p-value) for developmental effect
(contrast: READ - [AGE + DYS]). The color bar indicates the statistical significance (p-value) of the individual
ANOV A F-tests performed.
51
Dyslexia Effect
Finally, the following results show the left hemisphere regions with significant
differences (at p < 0.05) in cortical thickness (Table 6A and Figure 11A), surface area (Table
6B and Figure 11B) and volume (Table 6C and Figure 11C) when focusing on the effect of
dyslexia (contrast of all control subjects vs. Dyslexia subjects). Only one of the a priori regions
showed a statistically-significant association with this effect (the superior frontal gyrus). In
examining the surface area data for this same contrast, several more regions were shown to be
significantly different across the two groups. Finally, the volume results for this contrast showed
substantial overlap with the surface area data, in terms of which regions were implicated in the
effect of dyslexia.
52
Table 6A. Statistical results from Analysis of Variance (ANOV A) of cortical thickness data (left hemisphere only) to
examine dyslexia effect (contrast: DYS - [AGE + READ]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus -0.02791 F(2,62) = 0.1313 0.8772
Inferior Temporal Lobe 0.01304 F(2,62) = 1.423 0.2488
Lateral Occipital Lobe 0.02047 F(2,62) = 1.669 0.1969
Lateral Orbitofrontal
Cortex
-0.02633 F(2,62) = 0.1791 0.8365
Middle Temporal Lobe -0.01337 F(2,62) = 0.5777 0.5642
Pars Triangularis -0.01021 F(2,62) = 0.6767 0.512
Superior Frontal Lobe 0.08888 F(2,62) = 4.121 0.02087
Superior Temporal Lobe -0.01207 F(2,62) = 0.6184 0.5421
Supramarginal Gyrus -0.004335 F(2,62) = 0.8619 0.4274
Insula -0.0254 F(2,62) = 0.2074 0.8132
Figure 11A. FreeSurfer inflated pial surface map depicting cortical thickness statistics (p-value) for dyslexia effect
(contrast: DYS - [AGE + READ]). The color bar indicates the statistical significance (p-value) of the individual
ANOV A F-tests performed.
53
Table 6B. Statistical results from Analysis of Variance (ANOV A) of surface area data (left hemisphere only) to
examine dyslexia effect (contrast: DYS - [AGE + READ]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus 0.6089 F(2,62) = 50.82 8.597e-14
Inferior Temporal Lobe 0.595 F(2,62) = 48.01 2.538e-13
Lateral Occipital Lobe 0.4322 F(2,62) = 25.36 8.961e-09
Lateral Orbitofrontal
Cortex
0.5696 F(2,62) = 43.34 1.673e-12
Middle Temporal Lobe 0.4995 F(2,62) = 32.94 1.794e-10
Pars Triangularis 0.2767 F(2,62) = 13.24 1.624e-05
Superior Frontal Lobe 0.6452 F(2,62) = 59.2 4.18e-15
Superior Temporal Lobe 0.6192 F(2,62) = 53.03 3.757e-14
Supramarginal Gyrus 0.5043 F(2,62) = 33.56 1.328e-10
Insula 0.5636 F(2,62) = 42.33 2.558e-12
Figure 11B. FreeSurfer inflated pial surface map depicting surface area statistics (p-value) for dyslexia effect
(contrast: DYS - [AGE + READ]). The color bar indicates the statistical significance (p-value) of the individual
ANOV A F-tests performed.
54
Table 6C. Statistical results from Analysis of Variance (ANOV A) of volume data (left hemisphere only) to examine
dyslexia effect (contrast: DYS - [AGE + READ]).
Brain Region
(Left Hemisphere)
Adjusted R
2
F-Statistic p-value
Fusiform Gyrus 0.4776 F(2,62) = 30.25 6.784e-10
Inferior Temporal Lobe 0.589 F(2,62) = 46.86 4.002e-13
Lateral Occipital Lobe 0.4105 F(2,62) = 23.28 2.87e-08
Lateral Orbitofrontal
Cortex
0.5371 F(2,62) = 38.13 1.597e-11
Middle Temporal Lobe 0.4983 F(2,62) = 32.78 1.936e-10
Pars Triangularis 0.2063 F(2,62) = 9.315 0.0002902
Superior Frontal Lobe 0.6711 F(2,62) = 66.28 4.01e-16
Superior Temporal Lobe 0.5678 F(2,62) = 43.04 1.904e-12
Supramarginal Gyrus 0.3844 F(2,62) = 20.98 1.097e-07
Insula 0.6607 F(2,62) = 63.3 1.053e-15
Figure 11C. FreeSurfer inflated pial surface map depicting volume statistics (p-value) for dyslexia effect (contrast:
DYS - [AGE + READ]). The color bar indicates the statistical significance (p-value) of the individual ANOV A
F-tests performed.
55
Discussion
A qualitative comparison of the results for the three contrasts examined shows overlap in
the left-hemisphere regions that were significantly different in terms of the structural measures
tested, suggesting that these regions do not change substantially over the course of development,
and/ or when one begins to acquire more advanced reading skills, nor do they serve as effective
markers for differentiating dyslexia from other reading-related disorders.
However, these results do indicate that there may be meaningful associations between
brain morphology as it relates to surface area and volume measures of cortical regions and
reading performance irrespective of the effects examined in the present study. These potential
associations are further probed in subsequent work (Chapter 4: Examining the Relationship
between Brain Structure and Reading Performance using an Adaptive Behavioral Task
Paradigm). As previously stated, the overlap in left-hemisphere regions that were significantly
different in terms of these three structural measures across all three contrasts examined indicates
that there is likely no distinct pattern of brain morphology that could be used to differentiate
between dyslexia and the impact of development and reading ability on the left-hemisphere
reading network.
Alternatively, the approach of a priori selection of left-hemisphere cortical regions may
have also contributed to the lack of differentiation between the contrasts examined, and other
methods for performing a region-of-interest (ROI) analysis may be more appropriate for this
specific dataset and/ or for comparing the pediatric participant groups included. A more robust,
multimodal approach could include selection of ROIs from functional neuroimaging (fMRI)
results from the same participant sample, or even further, adopting a machine learning (ML)
56
approach to selection of ROIs from fMRI data in an equivalent training set, then applying these
same ROIs in subsequent structural analyses in the experimental dataset.
Conclusion
The present study sought to elucidate neuroanatomical group differences (in cortical
thickness, surface area and volume) in key left-hemisphere regions associated with reading, and
whether these differences may serve as structural markers to differentiate the effects of
development and/ or inherent reading ability from dyslexia specifically. While the results were
largely inconclusive as it relates to the linear contrasts (effects) examined, the findings from this
work do suggest that cortical thickness in particular may not serve as the most accurate measure
of structural integrity of isolated reading network regions, whereas surface area and volume
showed far more significant results and thus may serve as more appropriate indicators of
neuroanatomical development. Future work will seek to determine how these structural measures
may or may not correlate to performance on a reading-oriented behavioral task (see Chapter 4),
as well as explore the possible relationship between left hemisphere brain morphology and the
analogous right hemispheric regions.
57
Chapter 4: Examining the Relationship between Brain Structure and Reading Performance
using an Adaptive Behavioral Task Paradigm
Barakat, Rita
1
; Zevin, Jason
1,2
; Clark, Kristi
1,3
1. University of Southern California, Neuroscience Graduate Program (NGP)
2. University of Southern California, Department of Psychology
3. University of Southern California, Department of Neurology
Introduction
Overview of Past and Current Reading Models
The last several decades have seen a rise in prominence of theoretical models of reading
behavior, particularly in the context of research examining the underlying mechanisms of reading
deficits such as dyslexia. Perhaps one of the oldest and most famous of these models is the Dual
Route Hypothesis/ Model of reading, proposed by Seidenberg and McClelland (1989), which
describes a fundamental algorithm of reading behavior that utilizes orthographic and
phonological “units”, as well as a layer of other “hidden” linguistic units, to model the typical
reading process. While this model has proven to be effective in developing our collective,
conceptual understanding of reading behavior, it fails to capture other non-linguistic aspects of
reading that ultimately determine one’s reading ability later in life, including inherent word
recognition skill and phonological awareness and mapping, among others.
Other theoretical models that have attempted to illustrate the integrative process of
reading include the Triangle Model of Reading, as well as the Phonological Deficit Theory, the
latter being more specific to describing the underlying deficits seen in developmental dyslexia.
As with the Seidenberg-McClelland Model, these similar yet distinct conceptual frameworks
provide a strong foundation for understanding the relationship between different linguistic
58
elements essential for reading, but do not account for the environmental, genetic and other
non-linguistic factors that influence reading ability. This is of particular concern in the context of
developmental dyslexia, as numerous genetic studies have shown a heritable component to
dyslexia and potentially other reading disabilities. In addition, a significant body of research has
confirmed the broader societal understanding that greater access to high-quality literacy
education and resources in childhood often correlates with enhanced reading ability later in life,
a factor which is not explicitly represented in any of the aforementioned reading models. Finally,
these models do not account for the high prevalence of specific comorbidities observed in
individuals with reading disorders, such as Attention Deficit Hyperactivity Disorder (ADHD) that
often affect reading ability, particularly in young children still acquiring literacy.
In short, these theoretical frameworks for understanding the cognitive and linguistic
aspects of reading are valuable tools that help to simplify the complex integration of
orthographic, phonological and other essential units in reading, but should be incorporated into a
more holistic interpretation and definition of reading that recognizes the extra-linguistic factors
that influence reading ability. For a more detailed discussion of the relative strengths and
limitations of these models, see Barakat et al. (2022).
59
Findings from the Behavioral Literature
Behavioral scientists, neuropsychologists and other professionals in the field have had
great success and interest in applying the aforementioned models of reading to studies of
developmental disorders of reading, including dyslexia. Some of the most notable figures in this
field include Franck Ramus, whose work applying and modifying theoretical frameworks of
reading demonstrates that deficits in phonological processing and integration of phonological
information with other linguistic units leads to the hallmark behaviors seen in individuals with
dyslexia. Bennett and Sally Shaywitz have also published several studies building on the
phonological deficit model of dyslexia, demonstrating that children with dyslexia are slower on
timed reading tasks than their typical reading counterparts, and show a significant degree of
difficulty with reading that is not explained by a broader intellectual deficit as measured through
validated IQ tests. Other research groups have since confirmed the findings of Ramus and
Shaywitz and Shaywitz, showing that these unique deficits observed in individuals with dyslexia
can be measured across cultures and in a variety of populations around the world.
The most common, generalized task paradigm researchers have used to study the specific
behavioral deficits observed in dyslexia is the Phonological Identification (“PID”) Task, or a
variation of this paradigm, originally developed by Stephen Frost in 2009. The task can easily be
administered via computer or modified to be used in a neuroimaging (functional MRI) study, and
requires subjects to distinguish between correct and incorrect phonological representations of
common words, with the “common-ness” of the word typically being determined through
broader analysis of word frequency in the literature consumed by a specific age-group
population. Variations of this task paradigm may utilize visual stimuli or feedback, and can also
60
be accompanied by a “control” task that measures an individual’s ability to make similar
judgements based on orthographic (or spelling) differences in word and pseudoword stimuli.
The present study utilizes a novel version of this original “PID” task paradigm, modifying the
task so that it is adaptive to individual subject performance and includes orthographic and
phonological judgment trials. A more detailed description of the adaptive behavioral task
paradigm and stimuli can be found in Barakat et al. (2022).
Left-hemisphere Brain Regions Implicated in Reading
Parallel to the generation of theoretical models and behavioral studies of reading,
numerous multimodal neuroimaging studies have revealed the relationship between brain
structure and activity in typical as well as atypical reading behavior. Research from the Haskins
Laboratory, lead by Kenneth Pugh, has shown that children with dyslexia show a reduced Blood
Oxygen Level Dependent (BOLD) signal in left-hemisphere reading network brain regions, in
comparison to age-matched, typical reading control subjects, when participating in a similar
reading task to the previously-described “PID” paradigm. What’s more, these dyslexia subjects
also show reduced functional connectivity between a priori left-hemisphere reading network
brain regions, particularly the inferior and middle temporal gyri, as well as the
temporal-occipital junction and supramarginal gyrus, as compared to control subjects.
Similarly, Clark et al. (2014) examined measures of brain structure (specifically, cortical
thickness) in children with dyslexia in left-hemisphere reading network regions overlapping with
Pugh et al. (2013). The investigators found that dyslexia subjects showed reduced cortical
thickness in specific left-hemisphere brain regions when compared to age-matched and reading
61
level-matched controls, a novel between-groups analytic approach that allows for more nuanced
distinction between structural differences that may result from normal development and reading
skill, and those that are unique to dyslexia. These and other studies have also shown that older,
more “fluent” readers often exhibit a posteriorization of reading in the brain, or a progressive
shift of reading function to more posterior regions (specifically, regions localized to the
temporal-occipital lobes) as literacy increases, providing evidence for a general cognitive shift
from more effortful, focused reading that requires substantial input from prefrontal and
orbitofrontal cortices to relatively faster, less effortful processing in temporal-occipital regions.
Thus, some variation in brain activity and structure as a function of normal development is
expected, often resulting in significant heterogeneity across individual subjects in a study of
brain structure/ function and reading that can make drawing meaningful conclusions about
dyslexia-specific differences difficult if not nearly impossible.
Other studies have followed a more “data-driven” approach to examining brain structure
and function in the context of reading. Research by Nicholson, Fawcett and Dean suggests that
subcortical and brainstem-adjacent structures, including the cerebellum, may play a more critical
role in literacy acquisition and reading than previously thought, though more research in this area
is needed (2001). In addition, more recent research has examined the within-subject differences
across hemispheres in order to determine whether structural and functional differences observed
in individuals with dyslexia (and presumably other reading deficits) may extend to the
contralateral hemisphere.
The present study seeks to expand upon this foundation of behavioral and structural
neuroimaging studies of dyslexia by examining the putative correlations between brain structure
62
(specifically, cortical thickness, surface area and volume in a priori left-hemisphere reading
network regions) and reading ability as measured by an adaptive behavioral task paradigm
(Barakat et al., 2022). By utilizing this adaptive variation of the traditional “PID” task, a wider
array of reading performance metrics can be incorporated into subsequent statistical analyses,
also accounting for individual differences in task performance that are not related to reading
ability. In addition, separating subjects into three groups (dyslexia subjects, age-matched controls
and reading level-matched controls) allows for analysis of between- and within-group
differences that correspond to the effects of normal development, inherent reading skill and
dyslexia on reading ability.
Methods
Participants
18 children with dyslexia (ages 10-13 years; 10 females, 8 males), 18 age-matched
controls (ages 8-13 years; 9 females, 9 males) and 16 reading-level-matched controls (ages 7-10
years; 6 females, 10 males) were recruited from the local community in Southeast and
South-Central Los Angeles, California to participate in four hours of neuropsychological testing
and one hour of MRI scanning. Dyslexia diagnosis was qualified using standard criteria of at
least one standard deviation of discrepancy between the child’s measures of IQ and reading
ability, or with a prior clinical diagnosis of dyslexia. Control subjects were placed in the reading
level-matched control group based upon their scores on the Letter-Word Identification subtest of
the Woodcock-Johnson Test of Academic Achievement 3
rd
edition (WJ-III): subjects who scored
in the bottom 15
th
percentile when correcting for the age discrepancy between these younger
63
subjects and the dyslexia group were assigned to the reading level-matched control group.
Standard exclusion criteria for all three subject groups were as follows: mental retardation and/
or diagnoses of other impairments, including ADHD, neurological impairment, and any visual
or hearing impairments (including color blindness). To mitigate any potential confounds of a
second language, only children who were monolingual, native English speakers were included in
the study.
Parent participation involved completion of questionnaires regarding their child’s
executive functions (the Behavior Rating Inventory of Executive Function, BRIEF) and
childhood problem behaviors (the Child Behavior Checklist, CBCL), as well as a personal data
questionnaire regarding medical, social and academic information about their child (Gioia et al.,
2010; Achenbach & Rescorla, 2001). Data from the CBCL was used as a screening measure for
children with severe behavioral difficulties or psychiatric diagnoses.
Neuropsychological testing for each child was completed in two sessions of two hours
each, to reduce fatigue. The Wechsler Intelligence Scale for Children 4
th
Edition (WISC-1V) was
used to assess intelligence (IQ), the Woodcock-Johnson Test of Academic Achievement 3
rd
edition (WJ-III) was used to assess reading skill (only on the Word Attack and Word
Identification sub-tests), the Comprehensive Test of Phonological Processes was used to assess
phonological awareness (characteristically diminished in children with dyslexia), and the
Working Memory Test Battery for Children (WMTB-C) was used to assess components of
Baddeley and Hitch’s model of working memory. In addition, all subjects completed the Test of
Word Reading Efficiency (TOWRE) A and B to assess individual subjects’ speed-accuracy in
64
reading real and pseudowords (Yeates & Donders, 2005; Braden & Alfonso, 2003; Pickering &
Gathercole, 2001; Torgesen et al., 1999).
Task Design
Figure 2. Two sample stimuli (image + target and foil), both at Difficulty Level 1, for the orthographic (right) and
phonological (left) behavioral tasks. Subjects were instructed to give a key press indicating which word/
pseudoword most closely matched the image displayed, with judgment criteria varying by task.
Each participant was first presented with the orthographic matching task, containing 60
trials, followed by the phonological matching task containing the same number of trials. The
time of presentation for each token was limited to six seconds, and the subjects' responses were
recorded via a left or right button press (indicating the selection of either the target or foil
accompanying the stimulus) within this six second event window. Upon the subjects’ button
press, a fixation cross was presented for the duration of the trial period, before automatically
advancing to the next six-second trial. Figure 2 illustrates a sample stimulus/ target-foil pair
from each of the two tasks: for both tasks, the participants were presented with a pictorial
representation of the word, as well as the two options to match the representation, with the
criteria for selecting the correct choice based on the specific task (orthographic-matching versus
phonological-matching). Figure 3 shows a schematic representation of the experimental design
65
and timing for each trial on the orthographic task. Stimulus timing did not vary between the
orthographic and phonological tasks, only the nature of the word/ pseudoword stimuli
themselves.
Figure 3. Schematic illustration of the event-related task design implemented for the orthographic and phonological
task runs. The event-related design is optimized to be used in the context of neuroimaging research, and is also
effective at distinguishing between orthographic and phonological judgements in a stand-alone computer task
environment.
The adaptability of the task to the individual participant’s performance provides a critical
control for accuracy confounds when interpreting the data post-task administration. The
‘difficulty’ of the stimuli was determined through balancing various orthographic and
phonological features of each word relative to the sum of all potential words in the stimulus
repositories of each task, then pre-assigning each stimulus to a difficulty level based on the
reading (grade) level at which the corresponding word appears frequently in academic texts.
Both tasks were initiated with the first trial set to a median difficulty level of three, and each
subsequent trial could range anywhere from a difficulty level of one to five (with level one being
66
the least difficult, and level five being the most difficult). Changes in difficulty level occurred in
a stepwise fashion (without skipping levels between trials) based upon individual subject
performance during the task.
The orthographic and phonological tasks were designed to set a benchmark accuracy
level of approximately eighty percent for all subjects, though as evidenced from the graphs
plotting accuracy across both tasks, there were individuals who exceeded this accuracy
benchmark, illustrating a ceiling effect in both tasks. The determination of this benchmark was
based on previous work showing that an approximately eighty percent accuracy level would
encourage participation from subjects, particularly those in a pediatric or adolescent population,
while still challenging and motivating them, thus eliminating accuracy, effort and motivation as
significant confounds in the interpretation of behavioral results between subject groups (Kujala,
Richardson & Lyytinen, 2010).
In this way, quantification of each participant’s performance was completed through
measurement of the number of trials provided at each difficulty level, then averaged over the
total number of trials in each task, to determine the average difficulty level for each individual
subject on the two tasks. In addition, reaction time measures for each trial provided further
information regarding overall participant performance. Finally, a novel measurement to
characterize the number of instances in which subjects shifted difficulty level within a task,
variability, gives insight into which individual subjects or subject groups show a more varied
performance for a specific task.
67
Structural MRI Acquisition Parameters
All subjects were scanned at the University of Southern California Dornsife
Neuroimaging Institute (DNI) on a Siemens 3.0 Tesla Prisma Scanner. Total of 3.83 minutes
acquisition time for T1- (MPRAGE) and T2-weighted scanning sequences. An MPRAGE
sequence was implemented, with TR = 2.4 seconds, TE = 0 milliseconds, flip angle = 8
o
, voxel
resolution = 0.7 mm
3
. A T2 sequence was also implemented (for more robust segmentation, as
per the Human Connectome Project’ s minimal preprocessing pipelines for structural MRI data,
as well as for incidental findings per DNI scanning policy), with TR = 10 seconds, TE = 88
milliseconds, flip angle = 120
o
, voxel resolution = 0.8 mm x 0.8 mm x 3.5 mm.
FreeSurfer/ HCP Pipeline Preprocessing
Subjects’ raw, high-resolution MPRAGE images were parcellated into white matter, gray
matter and cerebrospinal fluid (CSF) using the Human Connectome Project’ s modified
FreeSurfer protocol, implemented through the PreFreeSurfer and FreeSurfer pipeline scripts
(Glasser et al., 2012). This approach incorporates white matter intensity data from T2-weighted
images to perform a more robust segmentation and parcellation of cortex, using the
Desikan-Killiany anatomical atlas as reference. Regions-of-interest (ROIs) from the resulting
parcellated images were selected a priori, based on evidence in the literature illustrating the role
of these regions in reading and related cognitive functions.
68
Statistical Analysis
After quality-checking the results from the HCP/ FreeSurfer segmentation, mean values
for volume (mm
3
), surface area (mm
2
) and cortical thickness (mm) of a priori ROIs were
extracted and correlated with the four behavioral outcome measures from the orthographic and
phonological tasks in R Studio (R Studio Team, 2020). To address the non-normal,
heteroscedastic nature of the data, Spearman’ s correlation coefficients were calculated, with
results being statistically significant at the p < 0.05 level. All correlations were calculated within
groups (i.e. correlations between task performance and structural measures were determined for
each of the three effects of interest examined). To determine if there were significant differences
between the correlation matrices that resulted from this within-group analysis, Wilcoxon Rank
Tests were performed along with Bonferroni correction for multiple comparisons.
Further analyses were performed to determine whether additional covariates of interest
were predictive of group differences in behavioral task performance and structural MRI metrics.
Mixed-effects linear regression models were generated and tested, with each model applying
additional covariates in a stepwise manner. Figure 4 summarizes the stepwise regression
approach with each linear model. Statistical significance for interactions between these
covariates and reading performance/ brain structure was determined at p < 0.05.
69
Results
Orthographic Task
In examining the within-group results on the orthographic task, several significant
positive and negative correlations were observed, particularly among control subjects
(age-matched and reading level-matched controls), as summarized in Table 7A. There were far
fewer significant correlations between the three structural metrics and four behavioral metrics for
the older subject group (dyslexia subjects and age-matched controls, summarized in Table 7B),
as well as the poor readers group (dyslexia subjects and reading-level matched controls,
summarized in Table 7C).
Interestingly, all but one of the ten left-hemisphere a priori ROIs analyzed showed
significant correlations between brain structure and behavior for all three subject groups. The
lateral occipital lobe, a broad area that has been shown in previous studies to play a role in
decoding pertinent visual (morphological) information during reading, was not significantly
correlated with performance on this task. A pattern can be discerned from these data showing
that cortical thickness is negatively correlated with performance on the orthographic task, while
surface area and volume are positively correlated with performance on the orthographic task.
70
Table 7A. Significant correlations (which structural measures with which behavioral measures for each region of
interest) for all control subjects (age-matched and reading level-matched controls).
ROI Positive Correlations Negative Correlations
Fusiform Gyrus
● Area and difficulty level
● Area and accuracy
● V olume and difficulty level
● V olume and accuracy
● Area and reaction time
● V olume and reaction time
● V olume and variability
Inferior Temporal Lobe ● Area and reaction time
Insula
● Area and difficulty level
● V olume and difficulty level
● V olume and accuracy
● Area and reaction time
● V olume and reaction time
Lateral Orbitofrontal Cortex
● Area and variability
● V olume and variability
Middle Temporal Lobe
● Area and difficulty level
● Area and accuracy
● Area and reaction time
Pars Triangularis
● Thickness and reaction
time
● Area and reaction time
Superior Frontal Lobe
● Area and difficulty level
● Area and accuracy
● V olume and difficulty level
● Thickness and reaction
time
● Area and reaction time
● V olume and reaction time
● Thickness and difficulty
level
Supramarginal Gyrus
● Area and difficulty level
● Area and accuracy
● Thickness and reaction
time
● Area and reaction time
● Thickness and difficulty
level
Superior Temporal Lobe
● Area and difficulty level
● Area and accuracy
● V olume and accuracy
● Area and reaction time
Table 7B. Significant correlations (which structural measures with which behavioral measures for each region of
interest) for older subjects (age-matched controls and dyslexia subjects).
ROI Positive Correlations
Insula Thickness and accuracy
Supramarginal Gyrus Thickness and reaction time
71
Table 7C. Significant correlations (which structural measures with which behavioral measures for each region of
interest) for poor readers (reading level-matched controls and dyslexia subjects).
ROI Positive Correlations Negative Correlations
Inferior Temporal Lobe Area and variability
Lateral Orbitofrontal Cortex V olume and variability
Pars Triangularis Thickness and reaction time
Phonological Task
In examining the within-group results on the phonological task, fewer positive and
negative correlations were observed across all three subject groups (results summarized in
Tables 8A, 8B and 8C below). However, a different subset of the ten left-hemisphere a priori
ROIs showed significant correlations between brain structure and task performance, most
notably the lateral occipital lobe. In contrast to the orthographic task results, there were no
significant correlations between any of the three structural measures and reaction time for any of
the three subject groups.
These results are consistent with the pattern observed in the orthographic task
correlations (namely, that cortical thickness is negatively correlated with task performance, while
surface area and volume are positively correlated with task performance).
Table 8A. Significant correlations (which structural measures with which behavioral measures for each region of
interest) for all control subjects (age-matched and reading level-matched controls).
ROI Positive Correlations Negative Correlations
Lateral Occipital Lobe
● Area and difficulty level
● Area and accuracy
● Thickness and difficulty
level
Superior Frontal Lobe
● Area and difficulty level
● Area and accuracy
● Thickness and difficulty
level
72
Table 8B. Significant correlations (which structural measures with which behavioral measures for each region of
interest) for older subjects (age-matched controls and dyslexia subjects).
ROI Positive Correlations Negative Correlations
Inferior Temporal Lobe ● Thickness and accuracy
Lateral Occipital Lobe
● Area and accuracy
● Thickness and variability
Middle Temporal Lobe
● Area and variability
● V olume and variability
Table 8C. Significant correlations (which structural measures with which behavioral measures for each region of
interest) for poor readers (reading level-matched controls and dyslexia subjects).
ROI Positive Correlations Negative Correlations
Inferior Temporal Lobe ● Thickness and accuracy
Lateral Occipital Lobe ● Thickness and variability
● Thickness and difficulty
level
Middle Temporal Lobe ● Thickness and accuracy
Pars Triangularis ● Thickness and accuracy
Superior Frontal Lobe
● Area and accuracy
● V olume and difficulty level
Supramarginal Gyrus ● Thickness and accuracy
Superior Temporal Lobe ● Area and accuracy
Between-group Comparisons
After performing Wilcoxon Rank Tests on the group correlation matrices and correcting
for multiple comparisons, there were no significant differences between typical readers and poor
readers (comparing age-matched control subjects to the reading level-matched control and
dyslexia subjects), suggesting that none of the ten left-hemisphere a priori ROIs demonstrate a
reading skill effect as it relates to brain structure and task performance. However, when
comparing older subjects to younger subjects (age-matched control and dyslexia subjects versus
73
reading level-matched control subjects), there were significant differences observed between the
correlations of brain structure and task performance in the left pars triangularis on the
orthographic task. This same difference was also observed when comparing dyslexia subjects to
all control subjects.
Additional Covariates and Interactions
Finally, Mixed Effects Linear Regression Models were applied to a subset of the original
subject pool for which data on participation in remediation therapies (specifically, reading and
speech-language therapies) existed. Regression analyses were performed in an analogous
manner to the aforementioned correlation analyses, such that group comparisons allowed for the
identification of developmental, reading skill and dyslexia-specific effects. Results from these
stepwise linear regression analyses show that participation in remediation therapies was not
significantly predictive of task performance and brain structural differences observed between
groups.
Discussion
Within-group Correlations
Taken together, the results from the within-group correlation analyses across all three
subject groups, and for both behavioral task paradigms, confirm findings from the literature
demonstrating that cortical thickness is negatively correlated with task performance, whereas
surface area and volume are positively correlated with task performance. Despite these overall
trends remaining consistent for all three subject groups, or “effects”, it is worth noting that
74
different left hemisphere brain regions implicated in reading ability showed significant
correlations with different task performance metrics across all three effects. This suggests that
some left hemisphere reading areas may differ in structure as a function of normal development,
reading ability and perhaps even provide a marker for dyslexia, though further study on a larger
population is necessary to further discriminate between these three effects. In addition, the partial
overlap of brain regions showing significant correlations across the two behavioral tasks
(orthographic and phonological) indicates that some regions are more closely associated with
phonological processing. In the present study, the lateral occipital lobe structural measures were
significantly correlated with task performance measures across all three subject groups for the
phonological task only, suggesting a unique role in phonological processing regardless of age,
reading skill or dyslexia diagnosis.
While the adaptive nature of the behavioral tasks, as well as the subtle but meaningful
linguistic differences between the two tasks, allows for some level of distinction between brain
areas closely associated with the two different processing “streams” (orthographic and
phonological), creation of fMRI-based anatomical maps that are unique to the subjects studied
would likely provide a more accurate representation of the left-hemisphere brain regions
recruited during task performance. In other words: the a priori regions studied, while supported
by extensive background research illustrating their respective roles in reading and language
processing, may not be the most pertinent to examine in the context of this specific behavioral
task paradigm and for this particular subject population. Thus, replicating the present analyses
with the addition of functional neuroimaging data can only serve to refine and confirm the
75
validity of these correlational findings by informing the selection of brain regions from which
structural measures would be extracted.
Between-group Differences
The lack of significant differences between the correlation matrices for poor readers
(reading level-matched controls and dyslexia subjects) and typical readers (age-matched control
subjects) on both behavioral tasks suggests that none of the ten a priori left-hemisphere brain
regions studied show a reading skill effect, meaning that the relationship between brain structure
and task performance does not differ significantly as it relates to reading ability. However,
comparing the correlation matrices for the developmental (older versus younger subject groups)
and dyslexia (dyslexia versus all control subjects) effects showed that there were differences in
the correlations between task performance and brain structure for the left pars triangularis on the
orthographic task only. This finding is complicated on two levels, the first of which being that
the result indicates that the pars triangularis shows developmental and dyslexia effects
simultaneously, making it difficult if not impossible to use structural measures of this region as
an indicator of dyslexia in the absence of any reading performance data.
The second complication arises from the limited evidence in the literature regarding the
specific role of the pars triangularis in orthographic processing: while functional neuroimaging
studies have shown that the left pars triangularis is recruited during reading-based tasks in typical
reading controls, these tasks are not similar enough to those implemented in the present study to
confirm that the pars triangularis is in fact critical for orthographic processing during reading. As
with the within-group correlation analyses, a primary limitation of the analytical approach for
76
conducting between-group comparisons stems from the a priori selection of left-hemisphere
ROIs to be included in statistical analyses. Incorporation of functional neuroimaging data during
the initial structural image processing stages would serve to identify specific left-hemisphere
regions recruited during task performance, and these active regions could then be selected as
ROIs for further structural processing and analysis.
Additional Factors affecting Reading Performance and Brain Structure
Separate but parallel analyses were performed to assess the predictive potential of
additional demographic and other factors on reading performance and brain structure. Mixed
effects linear models which included variables of interest for reading problems, participation in
reading therapy and participation in speech-language therapy were applied to a subset of the
original subject pool for which sufficient data was available to perform these analyses. An
additional factor of non-interest was included to control for any potential differences that may be
attributed to sex, however, previous examination of the original subject population indicated that
there were no significant differences in the distribution of male and female subjects across all
three original subject groups. Interestingly, the results from applying these linear models in a
stepwise fashion showed that there were no statistically significant interactions between existing
reading problems and/ or participation in remediation therapies and reading performance and
brain structure. One possible explanation for this lack of significant interactions for both the
orthographic and phonological tasks is the relatively small sample size of the population subset
examined. More specifically, the individual subject groups (age-matched controls, reading
level-matched controls and dyslexia subjects) were each considerably smaller in this parallel
77
analysis as compared to the sample sizes used in the correlation analyses, likely resulting in a
substantial drop in statistical power.
Another limitation extends from the way in which these additional factors were coded in
order to be compatible with a mixed-effects regression approach. While sex is traditionally
treated as a binary categorical variable in this type of analysis, the presence of reading problems
as well as participation in remediation therapies were also treated as binary categorical variables,
which arguably does not effectively capture the impact these factors have on reading
performance. A replication of this analysis that utilizes a larger sample size and/ or relies on a
different statistical approach capable of determining the variance explained by these additional
factors would likely provide more clarity on the role these more complex demographic factors
play in reading performance.
Conclusion
The present study sought to examine the statistical relationship between reading
performance and brain structure, and how this relationship may differ as a function of
development, reading skill and dyslexia. Within-group correlation analyses showed that indeed,
structural measures in several left-hemisphere brain regions implicated in reading are
significantly correlated with performance on the two behavioral tasks administered, each
targeting different yet integrated “streams” of processing. However, few brain regions showed
definitive correlations that could be interpreted as signifying structural differences unique to one
of the three effects examined, and between-group comparison of the individual group correlation
matrices illustrate that overall, correlations between brain structure and reading performance are
78
similar across groups. What’s more, sex, existence of reading problems and/ or participation in
remediation therapies targeting reading/ speech and language do not appear to play a critical role
in predicting task performance from brain structure, as shown through application of
mixed-effects linear models incorporating these factors.
The absence of significant group differences at the level of individual brain structure-task
performance correlations does not mean that there are in fact no differences that can be attributed
to the developmental stage, reading ability and/ or diagnosis of dyslexia. Rather, these results
speak to the complexity of the neurological processing that underlies reading, and the incredible
heterogeneity observed at the level of brain structure and behavior as it relates to reading ability.
Increasingly large and diverse subject samples, as well as replication in other populations, will
help to clarify the existence or lack of these brain-behavior relationships. Finally, multimodal
neuroimaging studies which utilize functional brain data as a means for generating more subject/
population-specific ROI atlases will allow for a more accurate approach to studying brain
structure in the context of reading.
79
Chapter 5: Broader Implications, Limitations and Considerations when Studying
Developmental Dyslexia
Barakat, Rita
University of Southern California, Neuroscience Graduate Program (NGP)
Significance of Behavioral and Neuroimaging Findings
Taken together, the results presented in Chapters 2 - 4 of this dissertation reaffirm the
findings from the existing neuropsychological and neuroimaging literature, as well as the
reigning theories and models of developmental dyslexia. In first examining the within- and
between-group differences observed in performance data from the two adaptive behavioral tasks
(the orthographic and phonological judgment tasks presented in Chapter 2), several trends align
with the expected performance outcomes for children with dyslexia and their reading
level-matched control counterparts. Specifically, these individuals showed a markedly poorer
performance on both tasks as compared to the age-matched control subjects, and while there
were no significant group-by-task interactions, the main effects for each task were statistically
significant (at the p < 0.05 level), indicating that these differences in performance were not likely
due to chance.
However, in somewhat of a departure from the leading Phonological Deficit Theory of
developmental dyslexia, the data showed that dyslexia subjects and their reading level-matched
control counterparts did not perform significantly better on the orthographic task relative to the
phonological task, suggesting that deficits in task performance are not exclusively related to a
deficit in phonological processing (Ramus et al., 2003; Share and Stanovich, 1995). While it
would certainly be speculative to draw this conclusion from these data alone, perhaps it is not
80
entirely surprising that dyslexia subjects (and the younger, reading level-matched subjects) did
not show a significantly poorer performance on the phonological task given that other, more
recent models of reading (namely the Triangle Model) emphasize the importance of integration
of processing streams for facilitating fluent reading (Clark et al., 2014; Manis et al., 1996; Wolf
and Bowers 1999; Zevin, 2019). Further commentary on the neurobiological correlates of these
integrated processing streams is presented in the final section of this chapter.
The structural neuroimaging analyses are also closely aligned with and confirm findings
from previous studies of children with dyslexia, particularly as it relates to the utility of cortical
thickness as a marker for structural integrity and development of brain regions implicated in
reading (Beelen et al., 2019; Clark et al., 2014; Cone et al., 2008; Elnakib et al., 2014; Ma et al.,
2015). It has been well established in the developmental neuroimaging literature that overall
cortical thickness increases moderately during the early stages of development (roughly birth to
age four/ five), then begins to show slow and gradual declines with normal aging (Burgaleta et
al., 2014; Fjell et al., 2015; Vijayakumar et al., 2016; Wierenga et al., 2014). This trend is
observed cross-culturally and is not limited to specific brain areas/ structures, which calls into
question the appropriateness of using cortical thickness as a proxy for structural- and ultimately
functional- integrity of a brain region as it relates to a specific cognitive function (Menary et al.,
2013; Perdue et al., 2020).
Nevertheless, the findings from comparing cortical thickness in left hemisphere brain
regions involved in reading and language (Chapters 3 and 4) are consistent with the general
findings from the developmental neuroimaging literature. Cortical thickness in many of these
regions was not significantly correlated with reading performance on either the orthographic or
81
phonological tasks (Chapter 4), and the few regions for which there were significant correlations
between cortical thickness and task performance showed a clear negative association between
these two measures, not surprising given that by around age eight individuals would already be
showing some slight declines in overall cortical thickness. Additional commentary on the
biological implications of cortical thinning are presented in the final section of this chapter.
Limitations and Future Directions
Chapters 2 - 4 of this dissertation contain brief remarks addressing the practical
limitations of the present study, and some of the ways in which these limitations can be
addressed in future research. The following section summarizes these limitations and provides
some guidance for future directions of this work.
Collinearity in Adaptive Behavioral Task
One critical limitation of the experimental design utilized for the behavioral portion of
this study (Chapter 2) centered around the interstimulus interval (ISI) established between
individual trials in each of the two adaptive tasks. While a constant ISI for an exclusively
behavioral study does not produce collinearity effects, this becomes a significant problem when
attempting to analyze functional MRI (fMRI) data from these same tasks. It is difficult to
distinguish between task versus rest brain activity as modeled through the canonical
Hemodynamic Response Function (HRF), though there are means of modeling brain activity
from fMRI data that do not rely on the more standard (and admittedly less biologically-accurate)
HRF.
82
To address this issue of collinearity in the subsequent fMRI data, a jittered ISI (in this
case ranging from 4 to 6 seconds, randomly alternating between each trial on the two tasks) can
be introduced, creating more temporal dynamics in the resulting data. A modified version of the
original adaptive task paradigm code is available on GitHub (see Supplementary Materials
section) and includes a jittered ISI as well as a hybrid version of the two tasks in which
orthographic and phonological trials are interspersed in a single task run. Including rest events in
this hybrid, jittered event-related design also allows for more neurologically-relevant modeling
of task versus rest activation.
Functional Applications of Anatomical Atlases
Another significant limitation of the completed analyses for this research stems from the
region-of-interest (ROI)-based analytical approach taken. The Human Connectome Project
(HCP)’ s Minimal Preprocessing Pipelines for structural MRI (sMRI) data rely primarily on
built-in FreeSurfer toolboxes, including the recon-all reconstruction protocol, among others. As
part of the segmentation preprocessing step, the pipelines utilize T1- and T2-weighted images to
generate gray- and white-matter maps which then inform subsequent anatomical parcellation.
The current FreeSurfer standard parcellation uses the Desikan-Killiany Anatomical Atlas to
define and label cortical and subcortical regions/ structures, and while this atlas is widely used in
neuroimaging research more broadly, it presents its own challenges in the context of the present
study.
First, the atlas is not specifically designed for anatomical parcellation of pediatric
neuroimaging data: while not necessarily a significant limitation, it still warrants consideration
83
for future replication and validation research. More significant is the degree to which this atlas
accurately captures (identifies) cortical regions important for reading, particularly those that are
less well-studied and not typically distinguished from broader areas on standard anatomical
atlases (such as the Visual Word-Form Area, or VWF A). A more robust and functionally-relevant
approach for ROI-analyses in the context of developmental dyslexia would begin with an
inductive (data-driven) analysis of the fMRI data for the present study cohort. From these results,
custom ROI masks would be generated, indicating the cortical (and any subcortical) regions
recruited during the two adaptive behavioral tasks. From these masks, an averaged atlas would
be created and incorporated into subsequent structural (FreeSurfer-based) analyses, thus drawing
a direct connection between the functional and structural data for the specific subject sample
analyzed. Alternatively, multi-voxel pattern analysis (MVP A) or other machine-learning methods
could be used to generate a more functionally-applicable anatomical atlas, though these
approaches would almost certainly require larger sample sizes for each of the three subject
groups in the present study.
Streamlining and Generalizing Multimodal Processing Pipelines
In addition to analyzing and incorporating the fMRI data from this subject population
into replications of the structural analyses, further study of the diffusion MRI (dMRI) data would
undoubtedly shed light on the role of major white matter connections between “hub” left
hemisphere cortical regions in reading ability. These dMRI data were not incorporated into the
present study largely due to the inflexible nature of the pipelines used to process the structural
data. The HCP pipelines were originally designed to process functional, structural and diffusion
84
data collected as part of the large-scale, global HCP studies. Thus, the pipelines perform
multimodal processing and analyses, but require specific types and qualities of data that are not
always collected in non-HCP acquisition protocols. In addition, the pipeline scripts are not meant
to be adapted to other data types, with any modification beyond standard adjustments of data
directories resulting in complete failure of the parent pipeline script.
For this reason, future multimodal analyses of this particular dataset would almost
certainly require the creation of custom processing pipeline scripts that utilize toolboxes from
different, freely-available neuroimaging software (including, but not necessarily limited to: FSL
for functional/ structural processing and analysis, FreeSurfer for structural segmentation and
parcellation, and FSL/ Qit for diffusion image processing and tractography).
Biological, Genetic and Sociocultural Factors
As alluded to in the previous sections of this chapter, there are a number of biological,
genetic and sociocultural factors that should be taken into consideration in order to accurately
and thoroughly examine the effects of development, inherent reading skill and dyslexia on
reading ability. Beginning with the neurobiological underpinnings of reading, it is important to
draw a clear distinction between direct studies of brain structure and function at the micro- and
meso-scales and the present study, which among many other studies of developmental dyslexia,
is limited to indirect analysis of these micro- and meso-scale correlates of reading due to
methodological and resolution limitations with neuroimaging technology.
Many studies have sought to specify the degree to which macroscale measures of brain
structure -including cortical thickness, surface area and volume- correspond to microstructural
85
changes of interest, such as proliferation of neurons and glial cells in localized regions of the
brain and strengthening of the structural and functional connectivity of white matter pathways
between localized regions (Afzali et al., 2021; Frye et al., 2010; Gennatas et al., 2017;
Koelkebeck et al., 2014; Lemaitre et al., 2012; Lyall et al., 2014). However, a consensus has yet
to be reached regarding the biological accuracy of attributing macrostructural changes to these
specific micro- and meso-scale correlates. As previously stated, cortical thickness has been
shown to decline with normal aging, and can decline more precipitously as a result of
neurodegeneration, toxicity and other gross tissue damage (Gennatas et al., 2017; Lemaitre et al.,
2012). It is believed that reduction in cortical thickness may serve as an indicator of long-term
loss of neurons and/ or glial cells, and among the three macrostructural measures examined in
this dissertation research, cortical thickness is most closely associated with gray matter (i.e.
neurons and glial cells). Surface area and volume changes, however, are not as limited to
changes in gray matter and may also correspond to development and/ or decline in white matter
connectivity between cortical regions.
The findings in Chapters 3 - 4 of this dissertation also suggest that surface area and
volume are more biologically-relevant markers of structural integrity in left hemisphere regions
implicated in reading, as these measures were significantly positively correlated with reading
performance.
More broadly, microstructural changes to white matter integrity are more
biologically-relevant to cognitive function than analogous changes to gray matter, as many
cognitive processes (particularly reading) involve significant integration between cortical
regions, with this integration largely reliant on signaling via major white matter tracts (Afzali et
86
al., 2021). In the left hemisphere specifically, the arcuate, uncinate, superior longitudinal and
inferior longitudinal fasciculi have been shown to be essential in connecting several cortical
regions implicated in reading and language processing (Cheema and Cummine, 2018; Sihvonen
et al., 2021). While it is valuable to carefully study and understand the way in which structural
changes associated with changes in gray matter relate to reading ability, it is perhaps even more
essential to examine the role of white matter changes, indirectly measured through dMRI, on
reading ability, as these changes are more closely associated with the underlying neurobiology of
reading and language.
Furthermore, developmental dyslexia is known to be a significantly heritable disorder, as
observed in mono- and di-zygotic twin studies of children with family histories of dyslexia
(Bishop, 2015; Erbeli, Rice and Paracchini, 2021). One complication in studying the specific
genetic underpinnings of dyslexia is that several of the most common comorbidities associated
with dyslexia (including ADHD) also show a high degree of heritability, and developmental
psychologists continue to debate whether or not dyslexia itself may be inextricably linked to
these comorbidities (see the Multiple Deficit Hypothesis of dyslexia summarized in Chapter 1).
Only a limited number of genomic studies have sought to isolate individual loci and/ or genetic
polymorphisms that could lead to a predisposition for dyslexia, and of these, the majority of this
research has been conducted in animal models (Erbeli, Rice and Paracchini, 2021; Lampis et al.,
2021; Nöthen et al., 1999).
Finally, and perhaps most significantly, a variety of sociocultural variables must be
accounted for in order to accurately capture the environmental influence on reading ability,
particularly for a pediatric population such as the one included in the present study. As with
87
many other complex and uniquely-human behaviors, reading is acquired relatively late in
development and while certain aspects of reading behavior are consistent across individuals
(such as the repetitive saccades produced during reading), the overall process is quite
heterogeneous as evidenced by individual differences in reading behavior. This heterogeneity
likely manifests in part from the environmental factors that differ between individuals, including
(but not limited to): one’s exposure to literature and language early in life, one’s access to
literature and education, and the demographic factors that ultimately lead to exposure and access
to literacy (most notably socioeconomic status). In circumstances in which an individual has a
biological and/ or genetic predisposition for developing dyslexia, early interventions such as
remediation therapy and/ or access to intensive literacy education can significantly affect the
individual’s likelihood of ultimately being diagnosed with dyslexia later in life. Thus, early
detection of reading difficulties and signs of dyslexia, in conjunction with the availability of
trained educational staff and remediation resources, can effectively treat dyslexia and prevent
subsequent reading disability, ultimately leading to better educational outcomes overall.
88
Supplementary Materials
Chapter Resource (Linked)
Chapter 2
CNS/ OHBM 2019 Poster
GitHub Repository
Chapter 3
OHBM 2021 Virtual Poster
OHBM 2021 Recorded “Mini” Talk
Full Data Analysis and Results
Chapter 4 Full Data Analysis and Results
Resource (Linked)
PSYC-502 (Behavioral Statistics) Final Project
PSYC-555 (Functional Neuroimaging) Final Project
Springer Nature (SN) Social Sciences STEM Education Research Publication
Full Dissertation, Coursework and Teaching Document Folder
89
References
1. Achenbach, T.M. and Rescorla, L.A. (2001). Manual for the ASEBA School-Age Forms & Profiles.
Burlington, VT: University of Vermont, Research Center for Children, Youth, & Families.
https://store.aseba.org/MANUAL-FOR-THE-ASEBA-SCHOOL-AGE-FORMS-PROFILES/productinfo/50
5/
2. Afzali, M., Pieciak, T., Newman, S., Garyfallidis, E., Özarslan, E., Cheng, H., & Jones, D. K. (2021). The
sensitivity of diffusion MRI to microstructural properties and experimental factors. Journal of
Neuroscience Methods, 347, 108951. https://doi.org/10.1016/j.jneumeth.2020.108951
3. Altarelli, I., Monzalvo, K., Iannuzzi, S., Fluss, J., Billard, C., Ramus, F., & Dehaene-Lambertz, G. (2013).
A Functionally Guided Approach to the Morphometry of Occipitotemporal Regions in Developmental
Dyslexia: Evidence for Differential Effects in Boys and Girls. Journal of Neuroscience, 33(27),
11296–11301. https://doi.org/10.1523/jneurosci.5854-12.2013
4. Amunts, K., Lenzen, M., Friederici, A. D., Schleicher, A., Morosan, P., Palomero-Gallagher, N., & Zilles,
K. (2010). Broca’s Region: Novel Organizational Principles and Multiple Receptor Mapping. PLoS
Biology, 8(9), e1000489. https://doi.org/10.1371/journal.pbio.1000489
5. Baeck, A., Kravitz, D., Baker, C., & Op de Beeck, H. P. (2015). Influence of lexical status and orthographic
similarity on the multi-voxel response of the visual word form area. NeuroImage, 111, 321–328.
https://doi.org/10.1016/j.neuroimage.2015.01.060
6. Beelen, C. et al. (2019). Atypical Gray Matter in Children with Dyslexia before the Onset of Reading
Instruction. Cortex, 121, 399–413. https://doi.org/10.1016/j.cortex.2019.09.010
7. Bishop, D. V . (2002). Cerebellar Abnormalities in Developmental Dyslexia: Cause, Correlate or
Consequence? Cortex, 38(4), 491–498. https://doi.org/10.1016/s0010-9452(08)70018-2
8. Bishop, D. V ., McDonald, D., Bird, S., & Hayiou-Thomas, M. E. (2009). Children Who Read Words
Accurately Despite Language Impairment: Who Are They and How Do They Do It? Child Development,
80(2), 593-605. https://doi.org/10.1111/j.1467-8624.2009.01281.x
90
9. Bishop, D. V . M. (2015). The interface between genetics and psychology: lessons from developmental
dyslexia. Proceedings of the Royal Society B: Biological Sciences, 282(1806), 20143139.
https://doi.org/10.1098/rspb.2014.3139
10. Boada, R., and Pennington, B. F. (2006). Deficient implicit phonological representations in children with
dyslexia. Journal of Experimental Child Psychology, 95(3), 153-193.
https://doi.org/10.1016/j.jecp.2006.04.003
11. Bolger, D.J. et al. (2008). Differential Effects of Orthographic and Phonological Consistency in Cortex for
Children with and without Reading Impairment. Neuropsychologia, 46(14), 3210–3224.
https://dx.doi.org/10.1016%2Fj.neuropsychologia.2008.07.024
12. Booth, J. R. et al. (2004). Development of Brain Mechanisms for Processing Orthographic and
Phonological Representations. Journal of Cognitive Neuroscience, 16(7), 1234-1249.
https://doi.org/10.1162/0898929041920496
13. Borowsky, R., Esopenko, C., Cummine, J., & Sarty, G. E. (2007). Neural representations of visual words
and objects: a functional MRI study on the modularity of reading and object processing. Brain topography,
20(2), 89–96. https://doi.org/10.1007/s10548-007-0034-1
14. Bosse, M., Tainturier, M. J., & Valdois, S. (2007). Developmental dyslexia: The visual attention span
deficit hypothesis. Cognition, 104(2), 198-230. https://doi.org/10.1016/j.cognition.2006.05.009
15. Braden, J.P., and Alfonso, V .C. (2003). The Woodcock-Johnson III Tests of Cognitive Abilities in
Cognitive Assessment Courses. WJ III Clinical Use and Interpretation, pp. 377–401.
https://doi.org/10.1016/B978-012628982-4/50013-1
16. Brown J. M. (2009). Visual streams and shifting attention. Progress in brain research, 176, 47–63.
https://doi.org/10.1016/S0079-6123(09)17604-5
17. Bruno, J.L., et al. (2008). Sensitivity to Orthographic Familiarity in the Occipito-Temporal Region.
NeuroImage, 39(4), 1988–2001. https://doi.org/10.1016/j.neuroimage.2007.10.044
91
18. Burbridge, T., Wang, Y ., V olz, A., Peschansky, V ., Lisann, L., Galaburda, A., lo Turco, J., & Rosen, G.
(2008). Postnatal analysis of the effect of embryonic knockdown and overexpression of candidate dyslexia
susceptibility gene homolog Dcdc2 in the rat. Neuroscience, 152(3), 723–733.
https://doi.org/10.1016/j.neuroscience.2008.01.020
19. Buchholz, J. & Davies, A. A. (2007). Attentional blink deficits observed in dyslexia depend on task
demands. Vision Research, 47(10), 1292-1302. https://doi.org/10.1016/j.visres.2006.11.028
20. Burgaleta, M., Johnson, W., Waber, D.P., Colom, R., Karama, S., (2014). Cognitive ability changes and
dynamics of cortical thickness development in healthy children and adolescents. NeuroImage, 84 810-819.
https://doi.org/10.1016/j.neuroimage.2013.09.038
21. Cabeen, R. P., Laidlaw, D. H., and Toga, A. W. (2018). Quantitative Imaging Toolkit: Software for
Interactive 3D Visualization, Data Exploration, and Computational Analysis of Neuroimaging Datasets.
Proceedings of the International Society for Magnetic Resonance in Medicine (ISMRM), 2854.
22. Cao, F., et al. (2006). Deficient orthographic and phonological representations in children with dyslexia
revealed by brain activation patterns. Journal of Child Psychology and Psychiatry, 47(10), 1041-1050.
https://doi.org/10.1111/j.1469-7610.2006.01684.x
23. Caravolas, M., Lervåg, A., Mousikou, P., Efrim, C., Litavský, M., Onochie-Quintanilla, E., Salas, N.,
Schöffelová, M., Defior, S., Mikulajová, M., Seidlová-Málková, G., & Hulme, C. (2012). Common Patterns
of Prediction of Literacy Development in Different Alphabetic Orthographies. Psychological Science,
23(6), 678–686. https://doi.org/10.1177/0956797611434536
24. Catani, M., Jones, D. K., & Ffytche, D. H. (2004). Perisylvian language networks of the human brain.
Annals of Neurology, 57(1), 8–16. https://doi.org/10.1002/ana.20319
25. Caverzasi, E. et al. (2018). Abnormal Age-Related Cortical Folding and Neurite Morphology in Children
with Developmental Dyslexia. NeuroImage: Clinical, 18, 814–821.
https://doi.org/10.1016/j.nicl.2018.03.012
26. Cheema, K. and Cummine, J. (2018). The Relationship between White Matter and Reading Acquisition,
Refinement and Maintenance. Developmental neuroscience, 40(3), 209–222.
https://doi.org/10.1159/000489491
92
27. Christodoulou, J. A., Del Tufo, S. N., Lymberis, J., Saxler, P. K., Ghosh, S. S., Triantafyllou, C.,
Whitfield-Gabrieli, S., & Gabrieli, J. D. (2014). Brain bases of reading fluency in typical reading and
impaired fluency in dyslexia. PloS one, 9(7), e100552. https://doi.org/10.1371/journal.pone.0100552
28. Clark, K. A., Helland, T., Specht, K., Narr, K. L., Manis, F. R., Toga, A. W., & Hugdahl, K. (2014).
Neuroanatomical precursors of dyslexia identified from pre-reading through to age 11. Brain, 137(12),
3136–3141. https://doi.org/10.1093/brain/awu229
29. Cone, N.E. et al. (2008). Developmental Changes in Brain Regions involved in Phonological and
Orthographic Processing During Spoken Language Processing. NeuroImage(41), 623–635.
https://doi.org/10.1016/j.neuroimage.2008.02.055
30. Dalby, M. A., Elbro, C., & Stødkilde-Jørgensen, H. (1998). Temporal Lobe Asymmetry and Dyslexia: An
in Vivo Study Using MRI. Brain and Language, 62(1), 51–69. https://doi.org/10.1006/brln.1997.1887
31. Dehaene, S et al. (2010). How Learning to Read Changes the Cortical Networks for Vision and Language.
Science(330),1359–64. https://doi.org/10.1126/science.1194140
32. Dehaene, S., Cohen, L., Morais, J., & Kolinsky, R. (2015). Illiterate to literate: behavioral and cerebral
changes induced by reading acquisition. Nature Reviews Neuroscience, 16(4), 234–244.
https://doi.org/10.1038/nrn3924
33. Deutsch, G. K., Dougherty, R. F., Bammer, R., Siok, W. T., Gabrieli, J. D., & Wandell, B. (2005).
Children’s Reading Performance is Correlated with White Matter Structure Measured by Diffusion Tensor
Imaging. Cortex, 41(3), 354–363. https://doi.org/10.1016/s0010-9452(08)70272-7
34. Devlin, J. T., Jamison, H. L., Gonnerman, L. M., & Matthews, P. M. (2006). The role of the posterior
fusiform gyrus in reading. Journal of cognitive neuroscience, 18(6), 911–922.
https://doi.org/10.1162/jocn.2006.18.6.911
35. D’Mello, A. M., & Gabrieli, J. D. E. (2018). Cognitive Neuroscience of Dyslexia. Language, Speech, and
Hearing Services in Schools, 49(4), 798–809. https://doi.org/10.1044/2018_lshss-dyslc-18-0020
36. Elmer, S. (2016). Broca Pars Triangularis Constitutes a “Hub” of the Language Control Network during
Simultaneous Language Translation. Front. Human Neuroscience .
https://doi.org/10.3389/fnhum.2016.00491
93
37. Elnakib, A., Soliman, A., Nitzken, M., Casanova, M. F., Gimel’farb, G., & El-Baz, A. (2014). Magnetic
Resonance Imaging Findings for Dyslexia: A Review. Journal of Biomedical Nanotechnology, 10(10),
2778–2805. https://doi.org/10.1166/jbn.2014.1895
38. Erbeli, F., Hart, S. A., Wagner, R. K., and Taylor, J. (2018). Examining the Etiology of Reading Disability
as Conceptualized by the Hybrid Model. Scientific Studies of Reading, 22(2), 167-180.
https://dx.doi.org/10.1080%2F10888438.2017.1407321
39. Erbeli, F., Rice, M., & Paracchini, S. (2021). Insights into Dyslexia Genetics Research from the Last Two
Decades. Brain Sciences, 12(1), 27. https://doi.org/10.3390/brainsci12010027
40. Fiez, J. A., & Petersen, S. E. (1998). Neuroimaging studies of word reading. Proceedings of the National
Academy of Sciences, 95(3), 914–921. https://doi.org/10.1073/pnas.95.3.914
41. Fischl B. (2012). FreeSurfer. NeuroImage, 62, 774-781. https://doi.org/10.1016/j.neuroimage.2012.01.021
42. Fjell, A. M., Grydeland, H., Krogsrud, S. K., Amlien, I., Rohani, D. A., Ferschmann, L., Storsve, A. B.,
Tamnes, C. K., Sala-Llonch, R., Due-Tønnessen, P., Bjørnerud, A., Sølsnes, A. E., Håberg, A. K., Skranes,
J., Bartsch, H., Chen, C. H., Thompson, W. K., Panizzon, M. S., Kremen, W. S., Dale, A. M., … Walhovd,
K. B. (2015). Development and aging of cortical thickness correspond to genetic organization patterns.
Proceedings of the National Academy of Sciences of the United States of America, 112(50), 15462–15467.
https://doi.org/10.1073/pnas.1508831112
43. Franceschini, S., Gori, S., Ruffino, M., Pedrolli, K., & Facoetti, A. (2012). A Causal Link between Visual
Spatial Attention and Reading Acquisition. Current Biology, 22(9), 814–819.
https://doi.org/10.1016/j.cub.2012.03.013
44. Frost, S.J., et al. (2009). Phonological awareness predicts activation patterns for print and speech. Ann. of
Dyslexia(59), 78–97. https://dx.doi.org/10.1007%2Fs11881-009-0024-y
45. Frye, R. E., Liederman, J., Malmberg, B., McLean, J., Strickland, D., & Beauchamp, M. S. (2010). Surface
Area Accounts for the Relation of Gray Matter V olume to Reading-Related Skills and History of Dyslexia.
Cerebral Cortex, 20(11), 2625–2635. https://doi.org/10.1093/cercor/bhq010
46. Gabel, L. A., Gibson, C. J., Gruen, J. R., & LoTurco, J. J. (2010). Progress towards a cellular neurobiology
of reading disability. Neurobiology of Disease, 38(2), 173–180. https://doi.org/10.1016/j.nbd.2009.06.019
94
47. Galaburda, A. M., & Kemper, T. L. (1979). Cytoarchitectonic abnormalities in developmental dyslexia: A
case study. Annals of Neurology, 6(2), 94–100. https://doi.org/10.1002/ana.410060203
48. Galaburda, A. M., LoTurco, J., Ramus, F., Fitch, R. H., & Rosen, G. D. (2006). From genes to behavior in
developmental dyslexia. Nature Neuroscience, 9(10), 1213–1217. https://doi.org/10.1038/nn1772
49. Gennatas, E. D., Avants, B. B., Wolf, D. H., Satterthwaite, T. D., Ruparel, K., Ciric, R., Hakonarson, H.,
Gur, R. E., & Gur, R. C. (2017). Age-Related Effects and Sex Differences in Gray Matter Density, V olume,
Mass, and Cortical Thickness from Childhood to Young Adulthood. The Journal of Neuroscience, 37(20),
5065–5073. https://doi.org/10.1523/jneurosci.3550-16.2017
50. Gioia, GA, et al. (2000). Test Review Behavior Rating Inventory of Executive Function. Child
Neuropsychology(6), 235-238. https://doi.org/10.1076/chin.6.3.235.3152
51. Giraud, A. L., & Ramus, F. (2013). Neurogenetics and auditory processing in developmental dyslexia.
Current Opinion in Neurobiology, 23(1), 37–42. https://doi.org/10.1016/j.conb.2012.09.003
52. Glasser, M.F. et al. (2013). The minimal preprocessing pipelines for the Human Connectome Project.
Neuroimage, 80, 105-124. https://doi.org/10.1016/j.neuroimage.2013.04.127
53. Glezer, L. S., Eden, G., Jiang, X., Luetje, M., Napoliello, E., Kim, J., & Riesenhuber, M. (2016).
Uncovering phonological and orthographic selectivity across the reading network using fMRI-RA.
NeuroImage, 138, 248–256. https://doi.org/10.1016/j.neuroimage.2016.05.072
54. Goswami, U., Wang, H. S., Cruz, A., Fosker, T., Mead, N., & Huss, M. (2011). Language-universal
Sensory Deficits in Developmental Dyslexia: English, Spanish, and Chinese. Journal of Cognitive
Neuroscience, 23(2), 325-337. https://doi.org/10.1162/jocn.2010.21453
55. Graves, W. W., Desai, R., Humphries, C., Seidenberg, M. S., & Binder, J. R. (2010). Neural systems for
reading aloud: a multiparametric approach. Cerebral cortex (New York, N.Y. : 1991), 20(8), 1799–1815.
https://doi.org/10.1093/cercor/bhp245
56. Hoeft, F., McCandliss, B. D., Black, J. M., Gantman, A., Zakerani, N., Hulme, C., Lyytinen, H.,
Whitfield-Gabrieli, S., Glover, G. H., Reiss, A. L., & Gabrieli, J. D. E. (2010). Neural systems predicting
long-term outcomes in dyslexia. Proceedings of the National Academy of Sciences, 108(1), 361–366.
https://doi.org/10.1073/pnas.1008950108
95
57. Hoeft, F., Meyler, A., Hernandez, A., Juel, C., Taylor-Hill, H., Martindale, J. L., McMillon, G.,
Kolchugina, G., Black, J. M., Faizi, A., Deutsch, G. K., Siok, W. T., Reiss, A. L., Whitfield-Gabrieli, S., &
Gabrieli, J. D. E. (2007). Functional and morphometric brain dissociation between dyslexia and reading
ability. Proceedings of the National Academy of Sciences, 104(10), 4234–4239.
https://doi.org/10.1073/pnas.0609399104
58. Ibañez, A., Gleichgerrcht, E. & Manes, F. (2010). Clinical effects of insular damage in humans. Brain
Struct Funct, 214, 397–410. https://doi.org/10.1007/s00429-010-0256-y
59. Jobard, G., Crivello, F., & Tzourio-Mazoyer, N. (2003). Evaluation of the dual route theory of reading: a
meta analysis of 35 neuroimaging studies. NeuroImage, 20(2), 693–712.
https://doi.org/10.1016/s1053-8119(03)00343-4
60. Keuleers, E., & Brysbaert, M. (2010). Wuggy: A multilingual pseudoword generator. Behavior Research
Methods 42(3), 627-633. https://doi.org/10.3758/BRM.42.3.627
61. Kibby, M. Y., Fancher, J. B., Markanen, R., & Hynd, G. W. (2008). A Quantitative Magnetic Resonance
Imaging Analysis of the Cerebellar Deficit Hypothesis of Dyslexia. Journal of Child Neurology, 23(4),
368–380. https://doi.org/10.1177/0883073807309235
62. Koelkebeck, K., Miyata, J., Kubota, M., Kohl, W., Son, S., Fukuyama, H., Sawamoto, N., Takahashi, H.,
& Murai, T. (2014). The contribution of cortical thickness and surface area to gray matter asymmetries in
the healthy human brain. Human Brain Mapping, 35(12), 6011–6022. https://doi.org/10.1002/hbm.22601
63. Kraft, I., Cafiero, R., Schaadt, G., Brauer, J., Neef, N. E., Müller, B., Kirsten, H., Wilcke, A., Boltze, J.,
Friederici, A. D., & Skeide, M. A. (2015). Cortical differences in preliterate children at familial risk of
dyslexia are similar to those observed in dyslexic readers. Brain, 138(9), e378.
https://doi.org/10.1093/brain/awv036
64. Kraft, I., Schreiber, J., Cafiero, R., Metere, R., Schaadt, G., Brauer, J., Neef, N. E., Müller, B., Kirsten, H.,
Wilcke, A., Boltze, J., Friederici, A. D., & Skeide, M. A. (2016). Predicting early signs of dyslexia at a
preliterate age by combining behavioral assessment with structural MRI. NeuroImage, 143, 378–386.
https://doi.org/10.1016/j.neuroimage.2016.09.004
96
65. Kronbichler, M., Bergmann, J., Hutzler, F., Staffen, W., Mair, A., Ladurner, G., & Wimmer, H. (2007).
Taxi vs. taksi: on orthographic word recognition in the left ventral occipitotemporal cortex. Journal of
cognitive neuroscience, 19(10), 1584–1594. https://doi.org/10.1162/jocn.2007.19.10.1584
66. Lampis, V ., Ventura, R., Di Segni, M., Marino, C., D’Amato, F. R., & Mascheretti, S. (2021). Animal
models of developmental dyslexia: Where we are and what we are missing. Neuroscience & Biobehavioral
Reviews, 131, 1180–1197. https://doi.org/10.1016/j.neubiorev.2021.10.022
67. Lemaitre, H., Goldman, A. L., Sambataro, F., Verchinski, B. A., Meyer-Lindenberg, A., Weinberger, D. R.,
& Mattay, V . S. (2012). Normal age-related brain morphometric changes: nonuniformity across cortical
thickness, surface area and gray matter volume? Neurobiology of Aging, 33(3), 617.e1–617.e9.
https://doi.org/10.1016/j.neurobiolaging.2010.07.013
68. Leonard, C. M. (2001). Imaging Brain Structure in Children: Differentiating Language Disability and
Reading Disability. Learning Disability Quarterly, 24(3), 158–176. https://doi.org/10.2307/1511241
69. Lyall, A. E., Shi, F., Geng, X., Woolson, S., Li, G., Wang, L., Hamer, R. M., Shen, D., & Gilmore, J. H.
(2014). Dynamic Development of Regional Cortical Thickness and Surface Area in Early Childhood.
Cerebral Cortex, 25(8), 2204–2212. https://doi.org/10.1093/cercor/bhu027
70. Ma, Y. et al. (2015). Cortical Thickness Abnormalities Associated with Dyslexia, Independent of
Remediation Status. NeuroImage: Clinical, 7, 177–186. https://doi.org/10.1016/j.nicl.2014.11.005
71. Malins, J. G. et al. (2018). Individual Differences in Reading Skill Are Related to Trial-by-Trial Neural
Activation Variability in the Reading Network. The Journal of Neuroscience, 38(12), 2981-2989.
https://doi.org/10.1523/jneurosci.0907-17.2018
72. Manis FR, et al. (1996). On the bases of two subtypes of developmental [corrected] dyslexia. Cognition(2),
157-95. https://doi.org/10.1016/0010-0277(95)00679-6
73. Mascheretti, S., de Luca, A., Trezzi, V ., Peruzzo, D., Nordio, A., Marino, C., & Arrigoni, F. (2017).
Neurogenetics of developmental dyslexia: from genes to behavior through brain neuroimaging and
cognitive and sensorial mechanisms. Translational Psychiatry, 7(1), e987.
https://doi.org/10.1038/tp.2016.240
97
74. Menary, K., Collins, P. F., Porter, J. N., Muetzel, R., Olson, E. A., Kumar, V ., Steinbach, M., Lim, K. O., &
Luciana, M. (2013). Associations between cortical thickness and general intelligence in children,
adolescents and young adults. Intelligence, 41(5), 597–606. https://doi.org/10.1016/j.intell.2013.07.010
75. Merz, E. C., Maskus, E. A., Melvin, S. A., He, X., & Noble, K. G. (2019). Socioeconomic Disparities in
Language Input Are Associated With Children’s Language ‐Related Brain Structure and Reading Skills.
Child Development, 91(3), 846–860. https://doi.org/10.1111/cdev.13239
76. Morris, R. D. et al. (1998). Subtypes of reading disability: Variability around a phonological core. Journal
of Educational Psychology, 90(3), 347-373. https://doi.org/10.1037//0022-0663.90.3.347
77. Neudorf, J., et al. (2022) Unique, Shared, and Dominant Brain Activation in Visual Word Form Area and
Lateral Occipital Complex during Reading and Picture Naming. Neuroscience, 481, 178–196.
https://doi.org/10.1016/j.neuroscience.2021.11.022
78. Nicolson, R.I, Fawcett, A.J. and Dean, P. (2001). Developmental Dyslexia: The Cerebellar Deficit
Hypothesis. Trends in Neurosciences, 24(9), 508–511. https://doi.org/10.1016/s0166-2236(00)01896-8
79. Norton, E. S., Beach, S. D., & Gabrieli, J. D. (2015). Neurobiology of dyslexia. Current Opinion in
Neurobiology, 30, 73–78. https://doi.org/10.1016/j.conb.2014.09.007
80. Norton, E. S., Black, J. M., Stanley, L. M., Tanaka, H., Gabrieli, J. D., Sawyer, C., & Hoeft, F. (2014).
Functional neuroanatomical evidence for the double-deficit hypothesis of developmental dyslexia.
Neuropsychologia, 61, 235–246. https://doi.org/10.1016/j.neuropsychologia.2014.06.015
81. Nöthen, M.M., Schulte-Körne, G., Grimm, T. et al. Genetic linkage analysis with dyslexia: Evidence for
linkage of spelling disability to chromosome 15. European Child & Adolescent Psychiatry 8, S56 (1999).
https://doi.org/10.1007/PL00010696
82. Ozernov-Palchik, O. & Gaab, N. (2016). Tackling the ‘dyslexia paradox’: reading brain and behavior for
early markers of developmental dyslexia. Wiley Interdisciplinary Reviews: Cognitive Science, 7(2),
156–176. https://doi.org/10.1002/wcs.1383
83. Ozernov-Palchik, O., et al. (2016). Longitudinal stability of pre-reading skill profiles of kindergarten
children: Implications for early screening and theories of reading. Developmental Science, 20(5).
https://doi.org/10.1111/desc.12471
98
84. Pennington, B. F. & Lefly, D. L. (2001). Early Reading Development in Children at Family Risk for
Dyslexia. Child Development, 72(3), 816–833. https://doi.org/10.1111/1467-8624.00317
85. Pennington, B.F. (2006). From single to multiple deficit models of developmental disorders. Cognition,
101(2), 385-413. https://doi.org/10.1016/j.cognition.2006.04.008
86. Pennington, B. F., et al. (2012). Individual prediction of dyslexia by single versus multiple deficit models.
Journal of Abnormal Psychology, 121(1), 212-224. https://doi.org/10.1037/a0025823
87. Perdue, M. V ., Mednick, J., Pugh, K. R., & Landi, N. (2020). Gray Matter Structure Is Associated with
Reading Skill in Typically Developing Young Readers. Cerebral cortex (New York, N.Y. : 1991), 30(10),
5449–5459. https://doi.org/10.1093/cercor/bhaa126
88. Perrachione, T.K. et al. (2016). Dysfunction of Rapid Neural Adaptation in Dyslexia. Neuron 92(6),
1383–1397. https://doi.org/10.1016/j.neuron.2016.11.020
89. Peterson, R. L. & Pennington, B. F. (2015). Developmental Dyslexia. Annual Review of Clinical
Psychology, 11(1), 283–307. https://doi.org/10.1146/annurev-clinpsy-032814-112842
90. Peterson, R. L., and Pennington, B. F. (2012). Developmental Dyslexia. The Lancet, 379(9830),
1997-2007. https://doi.org/10.1016/s0140-6736(12)60198-6
91. Phinney, E., Pennington, B., Olson, R., Filley C., & Filipek, P. (2007). Brain Structure Correlates of
Component Reading Processes: Implications for Reading Disability. Cortex, 43(6), 777–791.
https://doi.org/10.1016/s0010-9452(08)70506-9
92. Pickering, S., and Gathercole, S. (2001). Working memory test battery for children (WMTB-C): Manual.
London: Pearson. verbio, A. M., Vecchi, L., & Zani, A. (2004). From orthography to phonetics: ERP
measures of grapheme-to-phoneme conversion mechanisms in reading. Journal of cognitive neuroscience,
16(2), 301–317. https://doi.org/10.1162/089892904322984580
93. Proverbio, A. M., Vecchi, L., & Zani, A. (2004). From orthography to phonetics: ERP measures of
grapheme-to-phoneme conversion mechanisms in reading. Journal of cognitive neuroscience, 16(2),
301–317. https://doi.org/10.1162/089892904322984580
99
94. Pugh, K.R., et al. (2000). The Angular Gyrus in Developmental Dyslexia: Task-Specific Differences in
Functional Connectivity within the Posterior Cortex. Psychological Science,11(1), 51–56.
https://doi.org/10.1111/1467-9280.00214
95. Pugh, K. R. et al. (2013). The relationship between phonological and auditory processing and brain
organization in beginning readers. Brain and Language, 125(2), 173-183.
https://doi.org/10.1016/j.bandl.2012.04.004
96. Qu, J., Pang, Y ., Liu, X., Cao, Y ., Huang, C., & Mei, L. (2022). Task modulates the orthographic and
phonological representations in the bilateral ventral Occipitotemporal cortex. Brain imaging and behavior,
10.1007/s11682-022-00641-w. Advance online publication. https://doi.org/10.1007/s11682-022-00641-w
97. Ramus, F. (2003). Theories of developmental dyslexia: Insights from a multiple case study of dyslexic
adults. Brain, 126(4), 841-865. https://doi.org/10.1093/brain/awg076
98. Randazzo, M., Greenspon, E. B., Booth, J. R., & McNorgan, C. (2019). Children With Reading Difficulty
Rely on Unimodal Neural Processing for Phonemic Awareness. Frontiers in human neuroscience, 13, 390.
https://doi.org/10.3389/fnhum.2019.00390
99. Raschle, N. M., Becker, B. L. C., Smith, S., Fehlbaum, L. V ., Wang, Y ., & Gaab, N. (2015). Investigating
the Influences of Language Delay and/or Familial Risk for Dyslexia on Brain Structure in 5-Year-Olds.
Cerebral Cortex, bhv267. https://doi.org/10.1093/cercor/bhv267
100.Richlan, F., Kronbichler, M., & Wimmer, H. (2012). Structural abnormalities in the dyslexic brain: A
meta-analysis of voxel-based morphometry studies. Human Brain Mapping, 34(11), 3055–3065.
https://doi.org/10.1002/hbm.22127
101.Rimrodt, S. L., Peterson, D. J., Denckla, M. B., Kaufmann, W. E., & Cutting, L. E. (2010). White matter
microstructural differences linked to left perisylvian language network in children with dyslexia. Cortex,
46(6), 739–749. https://doi.org/10.1016/j.cortex.2009.07.008
102.Roux, F. E., Durand, J. B., Jucla, M., Réhault, E., Reddy, M., & Démonet, J. F. (2012). Segregation of
Lexical and Sublexical Reading Processes in the Left Perisylvian Cortex. PLoS ONE, 7(11), e50665.
https://doi.org/10.1371/journal.pone.0050665
100
103.RStudio Team (2020). RStudio: Integrated Development for R. RStudio, PBC, Boston, MA URL
http://www.rstudio.com/
104.Sakurai, Y., Ichikawa, Y . and Mannen, T. (2001). Pure Alexia from a Posterior Occipital Lesion.
Neurology, 56(6), 778–781. https://doi.org/10.1212/wnl.56.6.778
105.Schlaggar, B.L. et al., (2002). Functional Neuroanatomical Differences between Adults and School-Age
Children in the Processing of Single Words. Science, 296(5572), 1476–9.
https://doi.org/10.1126/science.1069464
106.Seidenberg, M. S., and McClelland, J. L. (1989). A distributed, developmental model of word
recognition and naming. Psychological Review, 96(4), 523–568.
https://doi.org/10.1037/0033-295X.96.4.523
107.Shaywitz, B. A., Lyon, G. R., & Shaywitz, S. E. (2006). The Role of Functional Magnetic Resonance
Imaging in Understanding Reading and Dyslexia. Developmental Neuropsychology, 30(1), 613–632.
https://doi.org/10.1207/s15326942dn3001_5
108.Shaywitz, S.E., et al. (1998). Functional Disruption in the Organization of the Brain for Reading in
Dyslexia. Proceedings of the National Academy of Sciences of the United States of America, 95(5),
2636–41. https://doi.org/10.1073/pnas.95.5.2636
109.Shaywitz, S. E., and Shaywitz, B. A. (2005). Dyslexia (Specific Reading Disability). Biological
Psychiatry, 57(11), 1301-1309. https://doi.org/10.1016/j.biopsych.2005.01.043
110.Sihvonen, A. J., Virtala, P., Thiede, A., Laasonen, M., & Kujala, T. (2021). Structural white matter
connectometry of reading and dyslexia. NeuroImage, 241, 118411.
https://doi.org/10.1016/j.neuroimage.2021.118411
111. Simos, P. G., Breier, J. I., Wheless, J. W., Maggio, W. W., Fletcher, J. M., Castillo, E. M., & Papanicolaou,
A. C. (2000). Brain mechanisms for reading: the role of the superior temporal gyrus in word and
pseudoword naming. Neuroreport, 11(11), 2443–2447. https://doi.org/10.1097/00001756-200008030-00021
112.Siok, W.T. et al. (2008). A Structural–Functional Basis for Dyslexia in the Cortex of Chinese Readers.
Proceedings of the National Academy of Sciences, 105(14), 5561–5566.
https://doi.org/10.1073/pnas.0801750105
101
113.Skeide, M., J. Brauer, and D. Friederici. (2015). Brain Functional and Structural Predictors of Language
Performance. Cerebral Cortex, 1–13. https://doi.org/10.1093/cercor/bhv042
114.Sood, M. R., & Sereno, M. I. (2016). Areas activated during naturalistic reading comprehension overlap
topological visual, auditory, and somatomotor maps. Human brain mapping, 37(8), 2784–2810.
https://doi.org/10.1002/hbm.23208
115.Sun, Y. F., Lee, J. S., & Kirby, R. (2010). Brain Imaging Findings in Dyslexia. Pediatrics & Neonatology,
51(2), 89–96. https://doi.org/10.1016/s1875-9572(10)60017-4
116.Stoeckel, C., Gough, P. M., Watkins, K. E., & Devlin, J. T. (2009). Supramarginal gyrus involvement in
visual word recognition. Cortex; a journal devoted to the study of the nervous system and behavior, 45(9),
1091–1096. https://doi.org/10.1016/j.cortex.2008.12.004
117.Temple, E., et al. (2001). Disrupted neural responses to phonological and orthographic processing in
dyslexic children: An fMRI study. Neuroreport, 12(2), 299-307.
https://doi.org/10.1097/00001756-200102120-00024
118.Thomas, T., Perdue, M. V ., Khalaf, S., Landi, N., Hoeft, F., Pugh, K., & Grigorenko, E. L. (2021).
Neuroimaging genetic associations between SEMA6D, brain structure, and reading skills. Journal of
Clinical and Experimental Neuropsychology, 43(3), 276–289.
https://doi.org/10.1080/13803395.2021.1912300
119.Torre, G., Matejko, A., & Eden, G. (2020). The relationship between brain structure and proficiency in
reading and mathematics in children, adolescents, and emerging adults. Developmental Cognitive
Neuroscience, 45, 100856. https://doi.org/10.1016/j.dcn.2020.100856
120.Torgesen, J.K., et al. (1999). TOWRE, Test of Word Reading Efficiency: Examiner's Manual. PRO-ED.
121.Vandermosten, M., Hoeft, F., & Norton, E. S. (2016). Integrating MRI brain imaging studies of
pre-reading children with current theories of developmental dyslexia: a review and quantitative
meta-analysis. Current Opinion in Behavioral Sciences, 10, 155–161.
https://doi.org/10.1016/j.cobeha.2016.06.007
102
122.Viersen, S. V ., Bree, E. H., Zee, M., Maassen, B., Leij, A. V ., & Jong, P. F. (2018). Pathways Into Literacy:
The Role of Early Oral Language Abilities and Family Risk for Dyslexia. Psychological Science, 29(3),
418-428. https://doi.org/10.1177/0956797617736886
123.Vijayakumar, N., Allen, N.B., Youssef, G., Dennison, M., Yücel, M., Simmons, J.G. and Whittle, S.
(2016), Brain development during adolescence: A mixed-longitudinal investigation of cortical thickness,
surface area, and volume. Hum. Brain Mapp., 37: 2027-2038. https://doi.org/10.1002/hbm.23154
124.Vogel, A. C., Petersen, S. E., and Schlaggar, B. L. (2014). The VWFA: It's not just for words anymore.
Frontiers in Human Neuroscience(8). https://doi.org/10.3389/fnhum.2014.00088
125.Werth R. (2021). Is Developmental Dyslexia Due to a Visual and Not a Phonological Impairment?. Brain
sciences, 11(10), 1313. https://doi.org/10.3390/brainsci11101313
126.Wierenga, L. M., et al. (2014). Unique Developmental Trajectories of Cortical Thickness and Surface
Area. NeuroImage, 87, 20–126. https://doi.org/10.1016/j.neuroimage.2013.11.010
127.Williams, V . J. et al. (2017). Cortical Thickness and Local Gyrification in Children with Developmental
Dyslexia. Cerebral Cortex, 28(3), 963–973. https://doi.org/10.1093/cercor/bhx001
128.Wolf, M., and Bowers, P. G. (1999). The double-deficit hypothesis for developmental dyslexias. Journal of
Educational Psychology, 91(3), 415-438. https://psycnet.apa.org/doi/10.1037/0022-0663.91.3.415
129.Woolnough, O., Forseth, K.J., Rollo, P.S., Tandon, N. (2019). Uncovering the functional anatomy of the
human insula during speech. eLife. https://doi.org/10.7554/eLife.53086
130.Xue, G., Chen, C., Jin, Z., & Dong, Q. (2006). Language experience shapes fusiform activation when
processing a logographic artificial language: an fMRI training study. NeuroImage, 31(3), 1315–1326.
https://doi.org/10.1016/j.neuroimage.2005.11.055
131.Yeates, K.O., and Donders, J. (2005). The WISC-IV and Neuropsychological Assessment. WISC-IV
Clinical Use and Interpretation, 415–434., https://psycnet.apa.org/doi/10.1016/B978-012564931-5/50014-1
132.Zeno, S. (1995). The Educator's Word Frequency Guide. Touchstone Applied Science Associates.
133.Zevin, J. (2019). “Modeling Developmental Dyslexia across Languages and Writing Systems.”
Developmental Dyslexia across Languages and Writing Systems, by Ludo Th. Verhoeven et al. Cambridge
University Press, 372–387.
103
134.Zoubrinetzky, R., et al. (2019). Remediation of Allophonic Perception and Visual Attention Span in
Developmental Dyslexia: A Joint Assay. Frontiers in Psychology(1).
https://doi.org/10.3389/fpsyg.2019.01502
104
Abstract (if available)
Abstract
It has been well established that children and adults with dyslexia show deficits in integrating orthographic and phonological information at the level of higher sensory and lexical processing networks during reading. Heterogeneity of dyslexia symptom presentation and severity poses a significant obstacle in the implementation of practical remediation treatments, and current behavioral tasks do not always address confounding factors related to accuracy of task performance between different subject populations, making comparison between impaired and non-impaired individuals difficult. In the first portion of this doctoral research, I present an adaptive variation of the canonical orthographic and phonological decision-making paradigm implemented in behavioral studies of developmental dyslexia. This particular experimental design presents numerous advantages for individual subject comparisons as well as for group-level analyses, and allows for more detailed exploration of reading performance as it pertains to reading skill, development and dyslexia. Analysis of task performance from 35 subjects (age-matched controls, reading level-matched controls and subjects with dyslexia) showed that the tasks are effective at distinguishing developmentally-driven factors that may impact reading ability from those that are specific to individuals with dyslexia, while also revealing individual differences within subject groups on task performance. Further validation of this design is made possible through the documentation and release of the original stimulus repository, as well as the scripts to execute the orthographic and phonological tasks for the purposes of behavioral and/ or neuroimaging research. In the second portion, I corroborate findings from the literature and implements an HCP-style approach to assessing structural differences in the reading network of children with dyslexia and age-matched/ reading level-matched control subjects. All subjects were scanned in a Siemens Prisma 3T Scanner, and image data were pre-processed using the Human Connectome Project’s (HCP) minimal preprocessing Structural Pipeline (PreFreeSurfer). Cortical thickness, surface area and volume were calculated via the HCP Structural Pipelines 2 and 3 (FreeSurfer and PostFreeSurfer). A priori regions of interest from HCP-specified FreeSurfer parcellation were identified and selected for linear regression analysis in R Studio and Python to determine the relationship between dyslexia diagnosis/ qualification (subject group) and cortical thickness, surface area and volume of these key brain regions implicated in reading. Results show that dyslexia diagnosis/ qualification is a significant predictor (at the p < 0.05 level) of cortical thickness, surface area and volume in key left hemisphere reading network regions (specifically, the superior temporal gyrus/ Heschl’s Gyrus and the inferior frontal gyrus, among other regions). In the final portion of this research, I contribute to the collective understanding of structural brain differences between children with dyslexia and their age-matched and reading-level matched counterparts, while also assessing the potential correlations between these measures of brain structure, reading ability and other factors such as the incorporation of remediation therapies. Cortical thickness, surface area and volume metrics were extracted from T1- and T2-weighted magnetic resonance imaging (MRI) data in eleven a priori left-hemisphere ROIs, then correlated with average difficulty level, average reaction time (on correct trials only), accuracy and variability on the two behavioral tasks administered. Correlation analyses were conducted such that the effects of development, inherent reading skill and dyslexia could be effectively isolated. The results indicate that for all three effects, cortical thickness was negatively correlated with reading ability, while surface area and volume were positively correlated with reading ability. These findings generally corroborate the conclusions from the existing literature illustrating that increased cortical thickness in select ROIs is often associated with the early developmental stages of acquiring literacy.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The neural correlates of skilled reading: an MRI investigation of phonological processing
PDF
Neurobiological correlates of skilled and disabled reading
PDF
Relationships among cortical thickness, reading skill, and print exposure in adult readers
PDF
P-center perception in children with developmental dyslexia: do low level auditory deficits underlie reading, spelling, and language impairments?
PDF
Neuroimaging in complex polygenic disorders
PDF
Functional neuroimaging of the processes involved in reading for non-impaired and dyslexic children
PDF
Biological and behavioral correlates of emotional flexibility and associations with exposure to family aggression
PDF
Cognitive-linguistic factors and brain morphology predict individual differences in form-sound association learning: two samples from English-speaking and Chinese-speaking university students
PDF
Neurobiological correlates of fathers’ transition to parenthood
PDF
Motor cortical representations of sensorimotor information during skill learning
PDF
Considering the effects of disfluent speech on children’s sentence processing capabilities and language development
PDF
Independent and interactive effects of depression genetic risk and household socioeconomic status on emotional behavior and brain development
PDF
Morphological and microstructural models of typical development in the hippocampus and white matter
PDF
Beatboxing phonology
PDF
Sequential decisions on time preference: evidence for non-independence
PDF
Testosterone’s effect on reward, cognition, and myelination: rat brain and behavior
PDF
Brain and behavior correlates of intrinsic motivation and skill learning
PDF
Individual differences in phonetic variability and phonological representation
PDF
Homeostatic imbalance and monetary delay discounting: effects of feeding on RT, choice, and brain response
PDF
Relationships between lifetime chronic stress exposure, vascular risk, cognition, and brain structure in HIV
Asset Metadata
Creator
Barakat, Rita
(author)
Core Title
Examining the neuroanatomical correlates of reading in developmental dyslexia
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Neuroscience
Degree Conferral Date
2022-08
Publication Date
07/22/2022
Defense Date
06/06/2022
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
behavior,cognitive neuroscience,Development,dyslexia,Language,MRI,neuroimaging,OAI-PMH Harvest,Reading
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Herting, Megan (
committee chair
), Clark, Kristi (
committee member
), Manis, Frank (
committee member
), Zevin, Jason David (
committee member
)
Creator Email
rbarakat@usc.edu,rbarakat1800@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC111373707
Unique identifier
UC111373707
Legacy Identifier
etd-BarakatRit-10909
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Barakat, Rita
Type
texts
Source
20220722-usctheses-batch-960
(batch),
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
behavior
cognitive neuroscience
dyslexia
MRI
neuroimaging