Neurobiology of Language

1
Improving Language and Literacy is a Matter of
Time and Neuroplasticity
Ian Creese, Ph.D.
Rutgers University, NewarkCenter for Molecular
Behavioral Neuroscience
2
The neural basis of language

• Understanding how the brain learns is important
  for understanding how students learn in the
  classroom.
• And how the brain learns language and reading
  depends crucially on time and timing.

3
Time matters for learning speech, language and
reading
say
Amplitude
stay
100 ms
Time (milliseconds)
These waveforms are identical except for an
inserted 100ms silent gap, yet we hear two
different words. In order to be able to read and
spell we need to hear these small acoustic
differences in words.
4
Neurobiology of Language

• Language is uniquely human. As such, it has been
  difficult to apply the majority of systems
  neuroscience techniques to the study of human
  language.
• Traditionally, what we know about the neural
  basis of language has been derived primarily from
  observing the effects of lesions to specific
  areas of the brain.
• More recently, the advent of functional
  neuroimaging technologies has revolutionized our
  ability to study brain activation patterns
  generated by language tasks.

40
39
46
41/42
45
44
22
47
Sylvian Fissure
Brocas Area Wernickes Area
Auditory Cortex Brodmanns numbers
5
Neurobiology of Language
LanguageSpoken and written
Circuit mechanisms
Despite these advances, it has remained difficult
to connect the study of cyto-architectural
regions involved in language, with the study of
neural circuitry and ultimatel language.
6
Immediate Objectives

• to describe, in general terms, how cortical
  processing machinery is initially organized by
  input exposure.
• to further elaborate studies of the development
  of abnormal cortical (system) processing as
  models for developmental language/reading
  impairment.
• to further define risk factors that interplay
  with developmental processes (such as inherited
  faults) to generate abnormal outcomes.
• to determine whether, and how we might prevent or
  correct these developmental disabilities in
  infancy, or in later life.

7
Neurobiology of Language

• Our research has focused on understanding the
  neurobiological basis of language development and
  disorders.
• We began our research program with the
  observation that many children with specific
  developmental language impairments (SLI) and
  reading deficits have particular difficulty at
  the phonological (speech) level of language.

8
Neurobiology of Language
LanguageSpoken and written
SpeechPerception and articulation
Circuit mechanisms
9
Neurobiology of Language
LanguageSpoken and written
Speech Perception and articulation
Sensorimotor systems
Circuit mechanisms
In order to study phonological deficits in more
detail, we began to focus on the speech signal
itself, from a sensory/motor systems perspective
10
Neurobiology of Language
Our earliest studies led to the discovery that
language impaired children have particular
difficulty in both perceiving and producing
brief, rapidly successive signals, specifically
in the tens of millisecond time range.
11
Children with language impairmentcant sequence
2 tones at rapid presentation rates
Tone Duration 75 msecTone 1 100 Hz, Tone 2
300 Hz
FAST
SLOW
Tallal Piercy (1973) Nature.
12
Neurobiology of Language
LanguageSpoken and written
Speech Perception and articulation
Sensorimotor systems
Temporal dynamicsSpecifically in the 10s of
milliseconds
Circuit mechanisms
13
Neurobiology of Language
Our data support an experience-dependent,
developmental model in which temporal dynamics
(specifically in the tens of millisecond time
range) serves as the conduit between some of the
most basic circuit level mechanisms and many of
the fundamental components of language.
14
10s of milliseconds can determine which syllable
we hear
Many speech sounds (phonemes) differ only by
brief spectral and/or temporal changes,
specifically within 10s of milliseconds
15
Not all speech sounds contain rapid
spectrotemporal changes
Steady-state vowel portions of syllables do not
incorporate brief spectral or temporal changes
16
Language impaired children have selective
deficits in discriminating those speech sounds
that differ by rapidly changing acoustic cues
Tallal Piercy (1974) Neuropsychologia, 12.
17
Speech can be computer modified to slow down
rapid spectrotemporal changes
18
Children with language leaning impairments show
significant improvement in syllable
discrimination when brief spectrotemporal changes
are extended in time
Tallal Piercy (1975) Neuropsychologia, 13.
19
Children with language learning impairments have
more difficulty distinguishing between words that
differ by phonemes incorporating rapid acoustic
changes
e.g. tie/pie, girl/curl
e.g. lawn/yawn, me/knee
Fellbaum et al. (1995).
20
Children with language learning impairmentshave
more difficulty producing rapid speech sounds
Tallal, Stark Curtiss (1976) Brain Language,
3.
21
Over two-thirds of children with language
learning impairments also meet the diagnostic
criteria for dyslexia
Language Impairment Only
Language Impairment Reading Impairment
Flax et al. (2003) JSLHR, 46.
22
There is a highly significant correlation between
nonsense word reading and rapid auditory
processing in dyslexics
Tallal (1980) Brain Lang. 9.
23
Evoked magnetic (MEG) waveform response to 2-tone
stimulus pairs with 200ms ISI
Individuals with dyslexia show physiological
deficits in rapid processing
reading impaired
MEG recording targeting A-1.
normal readers
Nagarajan, Merzenich et al (1999) PNAS
24
The left hemisphere exhibits specialization for
temporal dynamic processing
Temporal dynamic processing (specifically in the
TENS of milliseconds) has been shown to activate
left hemisphere-specific brain regions
traditionally associated with language. This was
not seen in adult dyslexic adults or children
with MRI. Training ameliorated the responses in
children (not tried in adult dyslexics yet).
25
Non-speech analog spectrotemporal stimuli with
rapid vs. slow onsets and offsets
Rapid Formant Transitions
Slow Formant Transitions
Frequency
Frequency
40
500
0
200
500
Time (ms)
hi
low
hi
low
Temple et al. (2000) PNAS, 97.
26
Left hemisphere-specific response to rapid
non-speech analog spectrotemporal acoustic stimuli
fMRI Results Fast gt Slow
20
16
12
Right
Left

• Left Middle Frontal Gyrus
• BA 46

Temple et al. (2000) PNAS, 97.
27
Dyslexics fail to show left hemisphere-specific
response to rapid non-speech spectrotemporal
acoustic stimuli
fMRI Results Fast gt Slow
Control Group
Dyslexic Group
16
16
Right
Left
Frontal Regions MFG (BA 46)
No Activity
Temple et al. (2000) PNAS, 97.
28
Rapid auditory processing (RAP) can be studied in
infants born into families with or without a
history of language learning impairments
An operantly conditioned head-turn procedure is
used to reward an infant for discriminating a
change in a 2-tone sequence
29
This baby has learned that something interesting
happens to her left when the two tones are
different
30
Rapid auditory processing threshold at 7.5
monthspredicts language comprehension at 36
months
100
Control
LI Family History
90
r2 .57 P lt .0001
80
70
PLS Comprehension Percentile
60
50
40
30
20
25
50
75
100
125
150
175
200
225
250
275
300
Auditory Temporal Processing Threshold in ms
Adapted from Benasich Tallal (2002) Behav
Brain Res. 136.
31
Rapid auditory processing (RAP) at 6 months
predicts developmental language delay at 3 years

• RAP thresholds at 6 months and male gender
  together accurately classified 91.4 of 3 year
  olds who scored in the impaired range on the
  Verbal Reasoning Scale of the Stanford-Binet
• There were no significant correlations between
  RAP and non-verbal outcomes

Benasich Tallal (2002) Behav Brain Res. 136.
32
Dr. April Benasich records electrophysiological
brain activity (event-related potentials - ERPs)
from infants
33
Electrophysiological differences (mismatched
response - MMR) to rapid tone sequences are
observed in infants with a family history of
language learning impairment
Control
Family History

• No significant group difference in mismatch
  response at 300ms ISI
• Infants with LI family history show significantly
  reduced MMR at 70ms ISI
• Significant group differences at 70ms ISI occur
  primarily in left hemisphere

Benasich et al. (in press) Neuropsychologia.
34
Goals for intervention
Improve reading
Strengthen language development
Enhance speech perception and production
Refine sensorimotor systems
Improve rapid auditory processing
- but how????
35
Grammatical morphology
Segmentation of speech into words
Phonological representations
Social interaction
LanguageSpoken and written
Speech Perception and articulation
Sensorimotor systems
Temporal dynamicsSpecifically in the 10s of
milliseconds
Circuit mechanisms
Synchronous activity
Gamma oscillations
Synaptic integration
Spike timing dependent plasticity
36
Cortical Plasticity
The ability of neurons to change their
responsiveness/connections via intensive periods
of experience and interaction with the
environment.

• The cortical representation of sensory stimuli,
  such as touch and sound, is well organized
  topographically.
• Research at UCSF first identified that this
  organization could be altered in ADULT primates
  through carefully designed behavioral
  experiences.

37
Brain Plasticity Research
Plasticity refers to the ability of the brain to
change through experience and learning.

• Synchronous Neural Activity
• Competition for Neural Space
• Rewarded Neural Activity
• Discriminating Neural Activity

Jenkins et al (1990).
38
Neuroscience Principles

• Synchronous Neural Activity
• Rewarded Neural Activity
• Repetition

39
Cortical Plasticity in Auditory Development in Rat

• P12-P30 the adult tonotopic frequency map
  develops by passive refinement of receptive
  fields and reduction in cortical area (CRITICAL
  PERIOD).
• Passive exposure to pulsed noise or complex tone
  patterns disrupted normal development which
  persisted in adults.
• Exposure to constant noise delayed normal
  development but was passively modifiable in
  adult.
• In normal adults active exposure to tones is
  required to modify the maps.

40
Normal Primary Auditory Cortex (A1) critical
period development
Zhang, Bao Merzenich, Nature Neurosci 2001
Zhang et al, Nature Neuroscience, 2001
41
A1 processing is specialized as the infant
rat is exposed to specific sound stimuli
A learning context is NOT required --- as
it is after the end of the critical
period.
Zhang et al, Nature Neuroscience, 2001
42
Changes endure into adulthood.
Zhang et al, Nature Neuroscience, 2001
43
A1 does NOT mature in rats raised in
moderate-level CONTINUOUS noise.
Chang et al. (2003) Science
44
In continuous-noise reared rats, the critical
period remains open, indefinitely.
Chang et al. (2003) Science
45
Designing computer exercises to drive
neuroplastic changes

• Scientific learning principles derived from
  neuroplasticity-based training studies with
  animals
• Frequent, intense input
• Individually adaptive (easy to hard) trials
• Sustained attention
• Timely rewards

46
Necessary Elements

• Motivation the person must attend closely to the
  stimuli.
• Repetition thousands of stimuli must be
  presented.
• Adaptive Difficulty Level the difficulty of the
  tasks must adapt so that the tasks are always
  near the persons threshold.
• Feedback on whether response was correct
  (reward).

47

• The goal of this exercise is to detect whether
  the two tones are both rising, both falling, or
  rising and falling
• As training progresses the rate of presentation
  increases

48
(No Transcript)
49
Non-linguistic auditory training significantly
improves syllable discrimination
See Lakshmi-Narayanan Tallal (2005) SFN Poster
408.7 Sunday
50
Goals for intervention
Improve reading
Strengthen language development
Enhance speech perception and production
Sharpen sensorimotor systems
Improve rapid auditory processing
51
Goals for intervention
Strengthen the underlying cognitive building
blocks for learning, which include memory,
attention, processing and sequencing.
52
Goals for intervention
Improve reading
Strengthen language development
Enhance speech perception and production
Sharpen sensorimotor systems
Improve rapid auditory processing
53
Acoustically enhanced speech

• Speech was computer modified by an algorithm in
  which the rate of the speech signal was prolonged
  and the fast transition elements (3-30 Hz
  components) were differentially amplitude
  enhanced
• This algorithm was applied to a wide variety of
  linguistic tasks
• The acoustic modification is reduced as
  linguistic performance improves, until
  age-appropriate levels of processing are reached
  with normal speech

Tallal et al. (1996) Science, 271.
54
(No Transcript)
55
(No Transcript)
56
Goals for intervention
Improve reading
Strengthen language development
Enhance speech perception and production
Sharpen sensorimotor systems
Improve rapid auditory processing
57
Intervention studies BEFORE training
Will be trained with

• Language impaired children (N22) were assigned
  to two groups matched on age, IQ, and severity of
  language impairment
• One group received intervention with acoustically
  enhanced speech and temporal modification
• Second group received same intervention but with
  natural speech and no temporal modification

Tallal et al. (1996) Science, 271.
58
Intervention studies outcomes after training
Changes in standardized test scores after training

• After intervention, the enhanced speech group
  made significantly greater gains compared to the
  natural speech control group
• P lt .015

Tallal et al. (1996) Science, 271.
59
Improvement in temporal threshold is highly
correlated with improvement in language
comprehension
Merzenich et al. (1996) Science, 271.
60
Clinical field trial of Fast ForWord with
language impaired children
Tallal (2000) in Bishop Leonard, Speech
Language Impairments in Children, Psychology
Press.
61
Fast ForWord significantly improved reading
scores in struggling readers in the School
District of Philadelphia, PA

• 300 low performing 4th 5th grade students from
  16 schools participated
• On average Fast ForWord Language and Reading
  products were used for 32 school days
• Fast ForWord software more than doubled the
  benefits of classroom reading instruction

MAPS for Learning Educator Reports (2004)
8(21)1-6.
62
Randomized control trial in Cherry Hill, NJ
schools Fast ForWord compared to classroom
intervention
Performance on standardized language assessment
battery

• 73 students identified as struggling with
  phonemic awareness were randomly assigned to one
  of two groups matched on degree of language
  impairment
• The participant group used Fast ForWord Language
  for an average of 34 school days, while the
  control group received regular school
  intervention
• Pre- and post-assessments show that, on average,
  Fast ForWord participants made significantly
  greater gains in language ability than controls

MAPS for Learning Educator Reports (2004) 8(4)1-4.
63
Intervention field trial of at-risk children in
inner-city public school shows normalization of
score distribution
Score Distribution on a Standardized Language
Assessment (n 288)
60
50
40
Number of Cases
30
20
10
0
0
4
-2
2
1
3
-1
-3
-4
0
4
-2
2
1
3
-1
-3
-4
Pre-Fast ForWord z-score
Post-Fast ForWord z-score
Tallal (2000) Speech Language Impairments in
Children, Psychology Press.
64
Goals for intervention
Improve reading
Strengthen language development
Enhance speech perception and production
Sharpen sensorimotor systems
Improve rapid auditory processing
65
Images of the Reading Brain
66
Goal To Determine if Language Instruction
Changes Reading Performance and Brain
ActivityDesign Pre-Post test comparison of an
Experimental and Control Group. Students
received the program at their school district.
Stanford University conducted the pre- and
post-testing.
School-based University StudyfMRI Intervention
StudyPrincipal InvestigatorsElise Temple, John
Gabrieli
67
Neural Effects of Remediation

• Determine if behavioral remediation can alter the
  disrupted response to phonological processing
  seen in children with reading problems
• 1. Can the disrupted response in temporo-parietal
  cortex be ameliorated?
• 2. Are changes in other brain areas apparent
  after training?

68
Neural Effects of Remediation

• Reading Impaired Children
• N20 8-12 years old
• MRI and behavioral testing
• before and after 8 weeks of training
• Normal Reading Children
• N12 8-12 years old
• MRI and behavioral testing
• 2 times 6-12 weeks apart

69
MRI
70
fMRI activation while viewing two letters and
determining whether their names rhyme
Control
Frontal AND Temporo-parietal
Dyslexic
Frontal but NOT Temporo-parietal
Example B D Rhyme B K
Do Not Rhyme
Temple et al. (2003) PNAS, 100.
71
Effects of Intervention

• Training improved reading behavior
• Training resulted in increased brain activity
• During phonological processing
• regions normally involved in processing
• left frontal
• left temporo-parietal
• regions not normally involved
• right frontal and temporal regions
• compensating brain response

72
Reading improvements after intervention
Real Word Reading
Non-Word Decoding
Passage Comprehension
120
115
110
105
100
Average
p lt .0001
Standard Score (Mean 100, SD 15)
95
p lt .005
90
p lt .0005
-1 SDBelow Average
85
80
75
70
Control
Dyslexic
Control
Dyslexic
Control
Dyslexic
Group
Before Intervention
After Intervention
Temple et al. (2003) PNAS 100.
73
Neural effects of intervention in dyslexic
children
Pre-Intervention
Frontal but NOT Temporo-parietal
Post-Intervention
Increased activity in Frontal AND Temporo-parietal
After training, metabolic brain activity in
dyslexics more closely resembles that of normal
readers.
Temple et al. (2003) PNAS, 100.
74
Improving Language and Literacy is a Matter of
Time
To date, over 1,000,000 children in over 4,000
schools nationwide have received Fast ForWord
Language and/or Reading intervention programs
75
Collaborators Paula Tallal
ElectrophysiologicalStudies April
Benasich Naseem Choudhury Paavo Leppanen Jennifer
Thomas-Friedman
Intervention Studies Mike Merzenich Steve
Miller Bill Jenkins Barbara Calhoun Gail
Bedi Thanassi Protopappas Joe Hardy Henry Mahnke
Animal Studies Holly Fitch Mike
Merzenich Christine Brown Matthew Clark Albert
Galaburda Bill Jenkins Itzel Orduña Glen Rosen
Imaging studies John Gabrieli Elise
Temple Mirella Dapretto Julie Fiez Nadine
Gaab Gary Glover Steve Petersen Russ
Poldrack Mark Raichle Terry Jernigan Monty
Buchsbaum Frank Wood Sri Nagarajan Urs
Ribrary Rudolpho Llinas
fMRI MRI PET MEG
Behavioral and Linguistic Studies April
Benasich Susan Curtiss Judy Flax Heeso Kim Kala
Lakshmi-Narayan Rachael Stark
Gene Linkage Studies Linda Brzustowicz Christopher
Bartlett Judy Flax Teresa Realpe
76
Memory problems in the elderly are exacerbated by
their slower processing speed
Young (18-21 yrs)
Elderly (64-83 yes)
p lt .005
Fast presentation rate (0-150ms ISI)
Slow presentation rate (500ms ISI)
Serial memory span
Lakshmi-Narayanan Tallal (2005) CNS Annual
Meeting
77
A brain plasticity based listening training
program significantly enhanced memory in elderly
adults
Overall neuropsychological function pre- and
post-training Age-normed scores
n.s.
p lt .0005

• Results from a pilot randomized controlled trial
  with 95 participants aged 64-95
• Intensive 8-week listening training program
• Memory improvement in training group is
  equivalent to 10 years

See Hardy et al. (2005) SFN Poster 408.7
78
Listening, Language and the Reading Brain

• At birth, we have an equal potential to learn any
  language.
• By 6 months, we begin to build the phonemes
  specific to our native language based on
  experience.

79
Hart and Risley (1995) longitudinal study of
language experience on language development
children

• Compared different economic levels
• Professional families
• Working-class families
• Families on welfare

80
Research
Words Heard per hour Affirmatives per hour Prohibitions per hour
Professional family child 2153 32 5
Working class child 1251 12 7
Welfare child 616 5 11
(Hart and Risley, 1995)
81
Experience and Vocabulary Size
Families Language and Use Differ Across Income Groups Hart Risley, 1995 Families Language and Use Differ Across Income Groups Hart Risley, 1995 Families Language and Use Differ Across Income Groups Hart Risley, 1995 Families Language and Use Differ Across Income Groups Hart Risley, 1995 Families Language and Use Differ Across Income Groups Hart Risley, 1995 Families Language and Use Differ Across Income Groups Hart Risley, 1995 Families Language and Use Differ Across Income Groups Hart Risley, 1995
Professional Professional Working-class Working-class Welfare Welfare
Measures Scores Parent Child Parent Child Parent Child
Recorded vocabulary size 2,176 1,116 1,498 749 974 525
Average different words per hour 382 297 251 216 167 149
82
30 Million Word Difference Language Experiences
by Group
(Hart and Risley, 1995)
83
Neuroscience Principles
Synchronous Neural Activity The systematic use
of acoustically modified speech, minimal pairs,
word families and systematic levels help to
selectively activate different brain regions
that struggling readers have difficulty in
activating.
84
Neuroscience Principles
Competition for Neural Space -Examples
85
Neuroscience Principles
Competition for Neural Space Cortical neuronal
changes in learning are competitive by nature and
thus dominant inputs, either by frequency or
numbers dominate cortical representations.
86
Reading Instruction Essential Elements from the
National Reading Panel Report

• 1. Phonemic Awareness Recognizing the smallest
  units of speech that affect meaning (phonemes).
• 2. Systematic Explicit Phonics Skills that
  help students to understand how phonemes, or
  speech sounds, are connected to print. Cracking
  the Code
• 3. Background knowledge and vocabulary -
  Sufficient background information and vocabulary
  to foster reading comprehension.
• 4. Fluency - The ability to derive meaning from
  print with minimal effort spent on decoding the
  individual words.
• 5. Comprehension - The development of
  appropriate cognitive strategies to construct
  meaning from print.

87
Dyx1c1 gene knockdown in rats results in rapid
auditory processing deficits
Rats with knockdown of the Dyx1c1 gene that has
been linked with dyslexia show similar acoustic
discrimination deficits to tone sequences using
the acoustic startle reflex reduction paradigm.
See Threlkeld et al. (2005) SFN Poster 829.3
88
Neuroplastic physiological changes are induced by
spectrotemporal training in mature rats

• The cortical area responding to complex sound was
  much smaller in naïve rats (A) than in trained
  rats (B)
• Cortical responses evoked by periodic trains of
  FM sweeps were less in number and persistent in
  naïve rats (C) than in trained rats (D)

A
16
kHz
2
1 sec.
B
C
Mean of spikes
D
Post-stimulus Time (s) Histograms
Mercado et al. (2001) Neuroreport.
89
(No Transcript)
90
Rapid 2-tone sequence discrimination in rats with
cortical migrational anomalies
Acoustic startle reflex reduction paradigm

P lt .05
Long (gt 89 msec)
Short (lt 64 msec)

• Attenuated response is given in with lower
  values representing better detection
• Lesioned rats were significantly different from
  shams in detecting short, but not longer,
  duration 2-tone sequences.

Clark, Rosen, Tallal Fitch (2000) Journal of
Cognitive Neuroscience, 12.
91
The left hemisphere exhibits specialization for
temporal dynamic processing
Temporal dynamic processing (specifically in the
TENS of milliseconds) has been shown to activate
left hemisphere-specific brain regions
traditionally associated with language. This was
not seen in adult dyslexic adults or children
with MRI. Training ameliorated the responses in
children (not tried in adult dyslexics yet).