Title: Neuroimaging Studies of Adaptive Learning in typically and atypically developing children
1Neuroimaging studies of language development
reading and reading disabilities
- Kenneth R. Pugh, PhD
- President and Director of Research,
- Haskins Laboratories, and
- Associate Professor, Yale University School of
Medicine
2- Perhaps the most practical, near-term synergy
between education and cognitive neuroscience is
in the service of predicting reading difficulty
and then offering intervention to avoid reading
failure. - John Gabrielli, Science 17 July, 2009
3Language Reading and Brain
- Spoken language is a biological specialization
but written language is largely a cultural
invention. Moreover, spoken language is mastered
naturally in almost all people, without direct
instruction - but reading is difficult and reading failure
occurs in large numbers of children across all
written languages. - No brain specialization for reading.
- Implication Literacy acquisition is a major
challenge to brain plasticity.
4Language Reading and Brain
- The development of fluent reading skill is
essential for success in the modern world. - Significant numbers of children in all countries
fail to acquire adequate literacy skills. - For many this is due largely to lack of good
learning opportunities but for some will reflect
difficulties that are brain-based (Specific
Reading Disability).
5Pre-literacy Risk for Reading Disability
- Language milestones early on are critical for
adequate reading development later on - 1) Early receptive and expressive language
development predicts reading acquisition. - Phonological Awareness is critical in
preschoolers. - Key Pre-school exposure to phonological (spoken
language) and simple orthographic experience is
very important in potentially preventing later
reading difficulties.
6Reading Disability Behavioral phenotype
- Phonological deficits are universal, but what is
the underlying cause? - Proposals include sensory deficits, compromised
neural systems for language, metalinguistic
deficits, attentional deficits. - It is possible that there are multiple
sub-types, with different pathways but a common
end-state (phonological processing deficit). - Neurobiological (brain) research may help address
this heterogeneity.
7Behavioral phenotype in reading disability
- Word identification is slow, labored, and error
prone in RD (bottleneck for comprehension). - Early deficits in developing fine-grained
phonemic awareness predict word reading
difficulties later on. - These deficits in phonological awareness impede
the development of efficient phonological
assembly routines (grapheme to phoneme mapping)
which, in turn, places severe limits on word (and
pseudoword) reading fluency.
8How do skilled readers recognize words
- 1) Skilled readers can read words fast (approx.
200 msec.) - 2) Pseudoword reading is nearly as fast!
- QUESTION Do skilled readers continue to rely on
sub-lexical phonological assembly?
9The primacy of sublexical phonology
- Lukatela Turvey 1992Pseudo-associative masked
priming in naming. - Prime Mask TARGET
- TOAD FROG
- TOLD FROG
10The primacy of sublexical phonology
- Lukatela Turvey 1992Pseudo-associative masked
priming - Prime Mask TARGET
- TOAD FROG
- TOLD FROG
- TOWED FROG
- TOWEL FROG
- TODE FROG
- TOGE FROG
11The primacy of sublexical phonology
- Identity Priming (Lukatela Frost and Turvey,
2003) - While the best prime for a word is the word
itself (identity priming) under fast masked
priming this effect is diminished for
orthographic to phonological inconsistent (BOWL)
words relative to consistent (BEND) words. - Implication of these studies lexical access in
skilled readers is sensitive to the phonological
code
12Implications for reading development
- Struggling readers have difficulty establishing
phonological coherence. Fluent lexical access
requires adequate binding of orthography,
phonology, morphology, and semantics. Phonology
may be the glue that holds this all together (Van
Orden et al., 1999)
13The Neurobiology of Reading Disability
- Genetics
- A) heritability (co-twin) studies establish a
genetic linkage in RD - B) A number of molecular-genetic investigations
(e.g., Fisher DeFries, 2002) have reported
linkages between reading-related processes and
regions on chromosomes 2, 3, 6, 15, and 18. These
studies are at an early stage thus far. - Gene/environment interaction shapes cognitive
development!
14The Neurobiology of Reading Disability
- Brain organization in RD
- 1) Post-mortum studies implicate cell
abnormalities (ectopias) in LH language zones.
Relevant genes have been targeted (Gruen et
al.,). - 2) Morphometry studies are complex but some
suggest differences in laterality ratios in
auditory cortex (Hynd and others). - 3) New directions examine connectivity with DTI,
suggest fiber tract anomalies (Klingberg,
McCandliss).
15(No Transcript)
16(No Transcript)
17The Language Brain
- Damage/lesion studies
- Brocas area Thought to be specialized for
speech production, grammar - Wernickes area Thought to be specialized for
speech comprehension, semantics
18Limitations of lesion studies
- Extreme localization view is simplistic.
- Language processes involve many regions operating
in dynamic fashion. - WE NEED TO EXAMINE SYSTEMS LEVEL PROCESSING. NEW
IMAGING TECHNOLOGIES CAN BE HELPFUL.
19The Neurobiology of Reading Disability
- Functional brain imaging
- two major classes of techniques
- electrophysiological (EEG MEG)
- hemodynamic (fMRI, PET).
- The former give information on timing of brain
activity while the latter provide information on
localization.
20Potential roles for neuroimaging
- Neurobiological measures provide mediating levels
of analysis between gene and behavioral
phenotype. - 2) Sensitivity A potentially deeper account of
individual differences in either typical or
atypical development and individual differences
in optimal intervention strategies for at-risk
children. What works for whom. - 3) Early detection of biomarkers predictive of
risk for atypical development.
21Reading Language and Brain Brain mapping
- Q) What are the underlying neurobiological
mechanisms associated with the development of
reading skill? - Q) What are the neurobiological underpinnings of
reading disability (RD), and how does training
and remediation modify brain organization for
printed language in children and adolescents with
RD?
22Slice Locations
23Auditory vs Visual Sentence Task
common
print(red) speech(blue)
RH on left side
Constable, Pugh et al. (2004)
24The Reading Circuit (Pugh et al., 2005)
- Hypothesized Role of component circuits
- Phonological
- IFG
- SMG/STG
- Semantic
- MTG/ITG
- AG
- Putative Visual word form Area
- Skill Zone is phonologically and
morphologically tuned
25Reading Disability
- Frequent finding A large number of studies
indicate that RD readers tend to under-activate
both LH temporoparietal and LH ventral
(occipitotemporal) regions during reading- and
language tasks this has been seen in several
languages to date (Paulesu et al., 2001). - RH and frontal compensatory shift in RD often
reported
26TD RD Reading Children (Temple et al., 2003)
Normal Readers
Frontal Temporo-parietal
Frontal but NO Temporo-parietal
Dyslexic Readers
27Pugh, et al., (2000), Psychological Science
- A fuller account requires focus not only on
within-region group differences, but also on
group differences in patterns of inter-regional
correlations or functional connectivity (e.g.,
Horwitz, Rumsey Donahue, 1998). - 32 adult NI readers vs. 29 adult RD readers.
- Task Hierarchy Two alternative forced choice
same/different Line (//\\, //\\) Letter case
(BTbT, BTbT) Single Letter Rhyme (B, T)
Non-word Rhyme (LETE,JEAT) Word Category
(CORN,RICE).
28(No Transcript)
29Key Finding Given that NWR shares with C
multi-letter processing and with SLR Phonological
analysis the breakdown in FC for NWR appears
relational. Implication LH posterior systems are
weak but not fundamentally disrupted In RD. This
encourages optimism with respect to intervention
effects.
30Plasticity and learning Developmental
trajectories in NI and RD children (Shaywitz,
Shaywitz, Pugh et al., 2002) N144 (ages 7-17
74 NI controls, 70 RD)
31Correlating activation with age and skill.
- 1) Correlating brain activation with
chronological age in NI and RD to examine
development. - 2) Correlating activation with reading skill
(after co-varying out age effects) allows us to
isolate those neural systems that support fluent
and accurate reading.
32(No Transcript)
33Plasticity in Reading Development
Temporoparietal
Anterior
Occipitotemporal
34Reading Development Skill Zone in Putative
VWFA
35Initial reading acquisition
- We need to examine development longitudinally
with integrated brain/behavior designs. - We have two ongoing NIH funded longitudinal
studies asking What are the behavioral and
neurobiological preconditions for successful
literacy acquisition? - Study 1) from 7-9 years examines at risk children
with multiple levels of analysis (genetics,
neuroanatomy, neurochemistry, neurocircuitry,
behavior). - Study 2) From 5-11 years examines brain/behavior
trajectories in three languages that vary in
orthographic depth (English, Finnish, Mandarin
Chinese).
36(No Transcript)
37MR Spectroscopy scanning
- Neurochemicals examined in a region of interest
centered on medial occipital cortex - GABA, N-acetyl aspartate (NAA), Creatine (Cr),
and Glutamate - Initial analyses of Time 1 data indictate
significant correlations between MRS, FMRI, AND
BEHAVIORAL MEASURES. WE ARE NOW TRACKING
DEVELOPMENT
38Schematic of fMRI Task
Event-related fMRI match/mismatch judgment via
button press of speech and print targets to a
picture
Mismatch/Match ratio (80-20) Only mismatch
trials included in analyses
39Key Contrasts
- Modality print vs speech
- Lexicality words vs. pseudowords
- Pronounceability printed pseudowords vs
consonant strings -
40Effects of Early Language Development Preston et
al., in prep.
- Several studies of late talkers have suggested
residual deficits in reading in young school-age
children (Rescorla 2002, 2005, 2009 Scarborough
Dobrich, 1990) - Parents complete questionnaire on child/family
background and childs development - Asked to report on when child began to Speak 2-3
word sentences. - Parent rated child as Early, On Time or Late
41Talker Group Comparisons on Reading
MeasuresWoodcock-Johnson Tests of Achievement
42Talker Group Comparisons on Reading
MeasuresGray Oral Reading Test (GORT)
43Late versus early talkers Brain activation in
reading and speech at at 7.5
p lt .025, FDR corrected
44(No Transcript)
45(No Transcript)
46Phonemic awareness and speech/print integration
(Frost et al,. 2009)
- Phonemic awareness scores reflect reading
readiness. - How do children with higher reading readiness
differ in initial brain organization?
47Modality Effects
gtPrint gtSpeech Overlap
p lt .001
48PA x Modality
p lt .01
49PA x Modality
r 0.44
p lt .01
50Correlation of PA with BOLD Modality Effect
51PA x Pronounceability
p lt .01
52PA x Pronounceability
r 0.57
p lt .01
53Key finding
- This finding suggests that children who are
developing normally in reading modify brain
systems for spoken language processing to become
available for processing visual graphemes. - It is crucial to promote learning experiences in
beginning readers that help to connect spoken
language and printed language to one another.
54Comprehension is the goal of most acts of reading.
- But comprehension is not easy to pin down.
- Comprehension reflects
- the reader,
- the text,
- the purpose of reading.
55Comprehension isnt just about reading
- Depends on general language abilities and memory.
- Implicates all structural elements of language
that reading shares with speech - phonologic, morphologic/lexical, syntactic,
semantic. - Comprehension is the outcome of multiple
processes that bridge between knowledge of
language and knowledge of the world.
56If a reader does poorly on a reading
comprehension test,
- The reader may lack
- a) Background knowledge and
vocabulary? - b) Requisite language abilities?
- c) Sufficient word decoding skills?
- d) Some combination of a, b, and c.
57- Sentence processing study Shankweiler et al.,
2008. - Reading is possible because the language brain is
supramodal - Questions
- Where do speech and print streams merge during
- sentence processing?
- 2. Does convergence co-vary with reading skill?
58Speech-Print Intersection and Differences Non-Ano
malous Baseline Sentences. Skilled readers.
R L
59Illustrating high and low values of the
Convergence Index
60Correlation of convergence and skill at dorsal IFG
r .7
61Plasticity and Remediation in Reading Development
Temporoparietal
Anterior
Occipitotemporal
62- RD readers do not tend to show this
neurodevelopmental trend. - Trajectory is rightward and frontward.
- Question Does remediation normalize this
trajectory?
Temporoparietal
Anterior
Occipitotemporal
63Remediation in RD
- Are these under-engaged LH systems fundamentally
disrupted, or does observed de-activation reflect
an unstable but potentially trainable state? - Can remediation focused on training up phonemic
awareness (PA) skills modulate the neurocognitive
risk profile in beginning reading.
64Testing effects of intensive phonological
remediation in RD in emergent readers
- Overview In collaboration with Dr. Benita
Blachman (Syracuse University) we examined
neurobiological changes associated with a nine
month intervention emphasizing phonemic
awareness, alphabet principle, and vocabulary
development in young children (Shaywitz et al.,
2004). - 3 Groups NI (N 28) RD control (N 12), RD
Treatment (N 32). Each group scanned at
baseline (average age 6.5), one year later
(post-treatment), and for the RD Treatment Group
at one year follow up. - (see Simos et al.,2002 Temple et al.,2003 for
similar findings with different phonological
training protocols)
65Training
- 50 min tutoring, 5 days per week, 9 months (105
hours total) - 5 step plan (unscripted) individualized
- Letter-sound associations
- Phoneme manipulation
- Reading words
- Reading text
- Assessment
66- Key behavioral result Reliable improvement on a
battery of reading-related tests for the
treatment relative to the control RD group
(Blachman et al., 2005) after nine months of
intensive evidence based training. - Effects stable at one year follow up.
67Treatment Year 2 Vs RD Control Year 2
Treatment Year 2 Vs Treatment Year 1
RD Control Year 2 Vs RD Control Year 1
68Treatment Group Year 3 (follow-up) minus Year 1
(Pre-Treatment)
69Temple et al. (2003) fMRI Data
L. Inferior frontal and L. temporo-parietal
activation
Some L.inferior frontal but no L.
temporo-parietal activation
Increases in L.inferior frontal and L.
temporo-parietal activation and right hemisphere
comologues
70- We thus have evidence that appropriate training
has a normalizing effect on the neurobiological
trajectory in emergent at risk readers. - Plasticity is strong even in struggling beginning
readers. LH posterior system appears to be
unstable but trainable in young at risk readers.
71- What about older RD readers?
- Children whose RD persists may be more severely
congenitally effected. - Also, years of reading experience may limit
plasticity in older RD kids. - First question Can we demonstrate potential
reading responsiveness in LH systems in
adolescent RD readers?
72Adolescent RD vs. NI study Repetition (learning)
Effects (Pugh et al., 2009, JOCN)
- To examine potential LH posterior functionality
in RD readers we examined repetition effects on
words (seen 6 times over the course of a scanning
run). - We know that RD readers will benefit from
repetition behaviorally.
73- Effects of repetition on behavior will be similar
for both groups, but - For NI with generally more stable LH system,
repeated words will be associated with reduced
activation (increased neural efficiency). - For RD, with generally unstable LH system,
repeated words will show increased activation in
unstable LH posterior regions.
74characterization
- Repetition Study
- present 56 words in each 5-minute imaging run
judge animacy - 6 of these are presented 6 times pseudorandomly
throughout - 20 intermixed novel words serve as controls
- repeat up to 10 runs with different words
- subjects
- NI readers RD readers
- N 18 14
- age 14.7 12.8
- WA 71 24
- FSIQ 112 104
75(No Transcript)
76Group X Rep Interaction
77LH Occipitotemporal
78LH STG
79LH IFG
80Theoretical characterization
Learning Curve Hypothesis Activation levels
(BOLD) follow an inverted-U shaped function with
respect to familiarity and/or skill in processing
words.
good readers
poor readers, high-freq words
U
good readers, high-freq words
poor readers
gt activation gt
good readers, repeated words
gt familiarity gt
81Adaptive learning beyond repetition
- These findings suggest that even in older readers
the LH language areas are potentially trainable
and capable of learning and plasticity. - Repetition demonstrates latent functionality in
RD but hardly constitutes an ideal remediation
strategy. - We are currently investigating how different
kinds of training contexts modulate brain
responses in these LH regions. Focus is on what
works for whom (e.g., Sandak et al., 2004).
82Looking ahead
How does speech-print integration come about?
How is the brain structured by different
learning experiences and instructional
environments? Which structures within the
inferior frontal region and elsewhere support the
skill-induced convergence of speech and print?
How are plastic connections within the
language network modulated by skill
differences?
83Conclusions
- Early language development is key to later
reading success. - LH circuit develops with learning.
- Early and intense remediation can promote good
brain development even in severe reading
disabled children.
84Collaborators
- Haskins Laboratories Einar Mencl, Rebecca
Sandak, Stephen Frost, Dina Moore, Nicole Landi,
Leonard Katz, Jay Rueckl, Jim Magnuson, Donald
Shankweiler, Jun Ren Lee, Carol Fowler, Alvin
Liberman, Mark Seidenberg - Yale Reading Center Ken Pugh (Director), Gina
Della Porta, Eleanor Tejada, Kelley Delaney,
Ashley Zennis, Anish Kurian, Heatherly Carlson,
Priya Pugh, -
- Yale Center for the Study of Learning and
Attention Bennett Shaywitz, Sally Shaywitz,
Karen Marchione, John, Holahan, Jack Fletcher - Yale University/Diagnostic Radiology John Gore,
Todd Constable, Robert Fulbright, Doug Rothman,
Graeme Mason, Pawel Skudlarski, Cheryl Lacadie - Yale University/Psychiatry Leslie Jacobsen
- Yale Child Study Center Elena Grigorenko