Linköping
University
medical
dissertations,
No.
1358
Mind your Language, All Right? Performance‐dependent
neural
patterns
of
language
Helene van Ettinger‐Veenstra
Center
for
Medical
Image
Science
and
Visualization
Division
of
Radiological
Sciences
Department
of
Medical
and
Health
Sciences
Linköping
University,
Sweden
Linköping
2013
©
Helene
van
Ettinger‐Veenstra,
2013
[email protected]
Published
papers
have
been
reprinted
with
permission
of
the
copyright
holders
Cover
design:
Tjeerd
Veenstra
www.tjeerdveenstra.nl
Printed
in
Sweden
by
LiU
Tryck,
Linköping,
Sweden,
2013
ISSN
0345‐0082
ISBN
978‐91‐7519‐668‐8
voor mijn lieve Lucas Levi
They say the left side of the brain Dominates the right And the right side has to labor through The long and speechless night … Maybe I think too much
‘Think
Too
Much
(b)’
‐
Paul
Simon
ABSTRACT
The
main
aim
of
this
dissertation
was
to
investigate
the
difference
in
neural
language
patterns
related
to
language
ability
in
healthy
adults.
The
focus
lies
on
unraveling
the
contributions
of
the
right‐hemispheric
homologues
to
Broca’s
area
in
the
inferior
frontal
gyrus
(IFG)
and
Wernicke’s
area
in
the
posterior
temporal
and
inferior
parietal
lobes.
The
functions
of
these
regions
are
far
from
fully
understood
at
present.
Two
study
populations
consisting
of
healthy
adults
and
a
small
group
of
people
with
generalized
epilepsy
were
investigated.
Individual
performance
scores
in
tests
of
language
ability
were
correlated
with
brain
activation
obtained
with
functional
magnetic
resonance
imaging
during
semantic
and
word
fluency
tasks.
Performance‐dependent
differences
were
expected
in
the
left‐hemispheric
Broca’s
and
Wernicke’s
area
and
in
their
right‐hemispheric
counterparts.
PAPER
I
revealed
a
shift
in
laterality
towards
right‐hemispheric
IFG
and
posterior
temporal
lobe
activation,
related
to
high
semantic
performance.
The
whole‐brain
analysis
results
of
PAPER
II
revealed
numerous
candidate
regions
for
language
ability
modulation.
PAPER
II
also
confirmed
the
finding
of
PAPER
I,
by
showing
several
performance‐dependent
regions
in
the
right‐hemispheric
IFG
and
the
posterior
temporal
lobe.
In
PAPER
III,
a
new
study
population
of
healthy
adults
was
tested.
Again,
the
right
posterior
temporal
lobe
was
related
to
high
semantic
performance.
A
decrease
in
left‐ hemispheric
IFG
activation
could
be
linked
to
high
word
fluency
ability.
In
addition,
task
difficulty
was
modulated.
Increased
task
complexity
showed
to
correlate
positively
with
bilateral
IFG
activation.
Lastly,
PAPER
IV
investigated
anti‐correlated
regions.
These
regions
are
commonly
known
as
the
default
mode
network
(DMN)
and
are
normally
suppressed
during
cognitive
tasks.
It
was
found
that
people
with
generalized
epilepsy
had
an
inadequate
suppression
of
regions
in
the
DMN,
and
showed
poorer
performance
in
a
complex
language
test.
The
results
point
to
neural
adaptability
in
the
IFG
and
temporal
lobe.
Decreased
left‐lateralization
of
the
IFG
and
increased
right‐ lateralization
of
the
posterior
temporal
lobe
are
proposed
as
characteristics
of
individuals
with
high
language
ability.
I
II
SAMMANFATTNING
Som
vuxna
människor
är
vi,
även
då
vi
är
friska,
väldigt
olika,
med
olika
förmågor.
Så
är
det
också
med
språklig
förmåga.
Det
varierar
betydligt
mellan
olika
personer
hur
bra
läsförståelse
man
har,
eller
hur
lätt
man
har
att
hitta
på
ord.
Denna
avhandling
bygger
på
att
dessa
mätbara
språkliga
skillnader
också
kan
synliggöras
i
hjärnan
med
hjälp
av
hjärnscanning,
så
kallad
funktionell
magnetresonanstomografi.
Hjärnaktivering
vid
språkfunktion
är
ofta
koncentrerad
i
den
vänstra
hjärnhalvan;
i
nedersta
delen
av
pannloben
samt
i
bakre
delen
av
tinningloben,
men
även
den
högra
hjärnhalvan
kan
aktiveras
av
flera
olika
språkfunktioner.
Speciellt
finns
de
funktioner
som
får
en
person
att
förstå
komplicerade
språkkomponenter,
till
exempel
bildspråk
eller
andra
typer
av
underliggande
betydelser
i
språket,
i
den
högra
hjärnhalvan.
I
studierna
som
ligger
till
grund
för
denna
avhandling
förväntades
att
hjärnaktiveringen
i
vanliga
språkområden
i
den
vänstra
hjärnhalvan
skulle
variera
med
språklig
förmåga.
Om
personer
som
är
bättre
på
språk
har
en
hjärna
som
fungerar
mer
effektivt,
så
skulle
det
visa
sig
som
mindre
aktivering
i
vänstersidiga
språkområden.
Å
andra
sidan,
om
personer
som
presterar
bra
har
bättre
kognitiv
förmåga
än
sämre
presterande,
skulle
det
kunna
synas
som
mer
aktivering
i
de
understödjande
språkområdena
i
höger
hjärnhalva.
Resultaten
som
framgår
i
denna
avhandling
är
framför
allt
att
aktivering
i
höger
tinninglob
är
involverad
i
bättre
språklig
förmåga.
Det
finns
också
antydningar
att
nedre
delen
av
den
högra
pannloben
är
mer
aktiverad
när
man
är
bra
på
språk.
Resultaten
visade
sig
dock
att
variera
med
språkuppgift;
det
finns
bevis
för
mer
aktivering
i
höger
pannlob
i
samband
med
bättre
språkförståelse
och
för
mindre
aktivering
i
vänster
pannlob
i
samband
med
bättre
förmåga
att
generera
ord.
Dessutom
är
den
nedre
delen
av
pannloben
mer
aktiv
vid
svårare
språkförståelseuppgifter.
Slutsatsen
av
dessa
studier
är
att
aktivering
i
den
nedre
pannloben
är
beroende
av
kognitiv
kapacitet,
men
att
aktivering
i
den
högersidiga
bakre
tinningloben
är
specifik
för
språkförståelse.
De
studier
som
är
inkluderade
i
avhandlingen
visar
att
desto
bättre
man
är
på
språk,
desto
mindre
använder
man
enbart
den
vänstra
hjärnhalvan
när
man
läser
eller
genererar
ord.
III
IV
LIST OF PUBLICATIONS
This dissertation is based on the following original papers, which are referred to throughout the text by their Roman numerals: PAPER I
Van
Ettinger‐Veenstra
HM,
Ragnehed
M,
Hällgren
M,
Karlsson
T,
Landtblom
A‐M,
Lundberg
P,
and
Engström
M
(2010).
Right‐hemispheric
brain
activation
correlates
to
language
performance.
NeuroImage
49(4):
3481–3488.
PAPER II
Van
Ettinger‐Veenstra
HM,
Ragnehed
M,
McAllister
A,
Lundberg
P,
and
Engström
M
(2012).
Right‐hemispheric
cortical
contributions
to
language
ability
in
healthy
adults.
Brain
and
Language
120(3):
395–400.
PAPER III
Gauffin
H*,
Van
Ettinger‐Veenstra
HM*,
Landtblom
A‐M,
Ulrici
D,
McAllister
A,
Karlsson
T,
and
Engström
M.
Impaired
language
function
in
generalized
epilepsy:
Inadequate
suppression
of
the
default
mode
network.
Accepted
in
Epilepsy
&
Behavior,
2013.
PAPER IV
Van
Ettinger‐Veenstra
HM,
Karlsson
T,
McAllister
A,
Lundberg
P,
and
Engström
M.
Laterality
shifts
in
neural
activation
coupled
to
language
ability.
Submitted
to
PLoS
ONE,
2013.
* The first two authors contributed equally to this paper
Related Peer‐Reviewed Conference Abstracts
Veenstra
HM,
Ragnehed
M,
Hällgren
M,
Lundberg
P,
and
Engström
M.
Brain
lateralization
assessed
by
fMRI
and
dichotic
listening.
Paper
presented
at
the
15th
Annual
Meeting
of
the
Organization
for
Human
Brain
Mapping,
California,
USA,
2009.
Veenstra
HM,
Pettersson
J,
Nelli
C,
Ragnehed
M,
McAllister
A,
Lundberg
P,
and
Engström
M.
Influence
of
performance‐related
language
ability
on
cortical
activation.
Paper
presented
at
the
15th
Annual
Meeting
of
the
Organization
for
Human
Brain
Mapping,
California,
USA,
2009.
Van
Ettinger‐Veenstra
H,
Karlsson
T,
Ulrici
D,
Gauffin
H,
Landtblom
AM,
and
Engström
M.
Language
ability
in
healthy
and
epilepsy
participants:
an
fMRI
investigation.
Paper
presented
at
the
43rd
European
Brain
and
Behaviour
Society
Meeting,
Seville,
Spain,
2011.
Van
Ettinger‐Veenstra
H,
Gauffin
H,
McAllister
A,
Lundberg
P,
Ulrici
D,
Landtblom
A‐M,
and
Engström
M.
Language
deficits
in
Epilepsy,
an
fMRI
study.
Paper
presented
at
the
18th
Annual
Meeting
of
the
Organization
for
Human
Brain
Mapping,
Beijing,
China,
2012.
V
AT A GLANCE
PAPER (study)
METHODS
I
14
healthy
adults.
fMRI:
Lateralization
Index
(A)
from
sentence
reading
(SENCO)
task
was
correlated
with
Read,
BeSS,
FAS
&
BNT
performance
scores.
Also,
Dichotic
Listening
laterality
measurements
were
investigated.
II
(A)
III
18
healthy
adults.
Whole‐brain
analyses
from
sentence
reading
(SENCO)
and
word
fluency
(WORGE);
activation
was
correlated
with
Read,
BeSS,
FAS
&
BNT
performance
scores.
27
healthy
adults.
Lateralization
Index
from
(B)
ROI
analyses
of
sentence
reading
(SEN)
and
word
fluency
(WORD),
correlated
with
performance
scores
on
BeSS
and
FAS.
Also,
task
difficulty
related
brain
activation
was
investigated
with
multiple
regression.
27
healthy
&
11
Generalized
Epilepsy
IV
(B)
participants.
Investigated
for
deactivation
in
the
default
mode
network
during
sentence
reading
(SEN).
Also,
language
performance
measurements
of
the
epilepsy
group.
VI
CONCLUSIONS
RESULTS
Both
dichotic
listening
and
fMRI
results
point
to
a
right‐hemispheric
activation
as
a
characteristic
for
high
language
ability.
Activation
in
the
right‐hemispheric
ROIs
was
more
pronounced
for
high
performance.
This
correlated
with
the
dichotic
listening
results.
Especially
high
BeSS
and
Read
scores
correlated
with
increased
right‐lateralization.
Regions
in
inferior
frontal
gyrus
(BA
47)
and
middle
temporal
gyrus
(BA
21)
are
related
to
high
semantic
language
ability.
Several
clusters
in
right
IFG
and
temporal
lobe
showed
to
correlate
with
BeSS
and
Read
on
the
sentence
reading
fMRI
task.
No
such
results
for
word
fluency.
Activation
in
the
inferior
frontal
gyrus
is
modulated
by
semantic
difficulty,
while
right
temporal
lobe
activation
is
specific
for
semantic
ability.
Activation
in
the
temporal
lobe
was
more
right‐lateralized
for
high
BeSS
performance.
Activation
in
left
IFG
was
less
left‐lateralized
for
high
FAS
performance.
The
difficult
incongruent
sentence
reading
condition
was
characterized
by
bilateral
IFG
activation
People
with
Generalized
Epilepsy
experience
language
difficulties.
This
could
be
explained
by
aberrant
suppression
of
activation
in
the
default
mode
network.
A
failure
to
suppress
default
mode
network
activation
is
disturbing
for
cognitive
functioning.
People
with
Generalized
Epilepsy
showed
worse
performance
in
BeSS
than
healthy
controls.
They
also
showed
diminished
DMN
deactivation,
notable
was
the
decreased
left
temporal
lobe
deactivation
and
increased
hippocampal
activation.
VII
VIII
ABBREVIATIONS
BA
BeSS
BNT
Brodmann
Area
“Bedömning
av
Subtila
Språkstörningar”
–
Assessment
of
Subtle
Language
Deficits
Boston
Naming
Test
BOLD
Blood
Oxygen
Level
Dependent
DMN
Default
Mode
Network
fMRI
functional
Magnetic
Resonance
Imaging
FWE
Family‐Wise
Error
GE
Generalized
Epilepsy
GLM
General
Linear
Model
IFG
Inferior
Frontal
Gyrus
LI
Laterality
Index
MNI
Montreal
Neurological
Institute
MRI
Magnetic
Resonance
Imaging
P‐FIT
Parieto‐Frontal
Integration
Theory
ROI
Region
of
Interest
SEN
sentence
reading
fMRI
task
used
in
PAPER
III
&
PAPER
IV
SENCO
sentence
completion
fMRI
task
used
in
PAPER
I
&
PAPER
II
WORD
word
generation
fMRI
task
used
in
PAPER
III
WORGE
word
generation
fMRI
task
used
in
PAPER
II
IX
CONTENTS
ABSTRACT
I
SAMMANFATTNING
III
LIST OF PUBLICATIONS
V
AT A GLANCE
VI
ABBREVIATIONS
IX
1
INTRODUCTION 1.1
LANGUAGE ABILITY 1.1.1
Language
Abilities
1.1.2
Language
Dysfunctions
1.2
NEURAL CORRELATES TO LANGUAGE 1.2.1
Language
Models
1.2.2
Semantics
1.2.3
Word
Fluency
1.2.4
Right‐Hemispheric
Influences
1.2.5
Laterality
1.2.6
Anti‐correlated
Brain
Activation
1.3
INTELLIGENCE MODELS FOR LANGUAGE ABILITY 1.3.1
Relation
Language
Ability
and
Intelligence
1.3.2
Intelligence
Models
1.4
AIMS 2
METHODS 2.1
NEUROLINGUISTIC MEASURES 2.1.1
Tests
of
Language
Ability
2.1.2
Dichotic
Listening
2.1.3
fMRI
Language
Paradigms
2.1.4
Study
Population
2.1.5
Generalized
Epilepsy
2.2
FUNCTIONAL MRI 2.2.1
Properties
of
Functional
MRI
1
2
2
3
4
4
8
8
8
9
10
11
11
11
13
15
15
15
16
16
17
17
18
18
2.2.2
Region
of
Interest
Analysis
2.2.3
Laterality
Index
Analysis
3
RESULTS 3.1
3.2
3.3
3.4
19
20
23
MULTIPLE REGRESSION ANALYSES LATERALITY ANALYSES TASK DIFFICULTY MODULATION LANGUAGE DYSFUNCTIONS IN EPILEPSY
4
DISCUSSION
24
27
28
29
31
4.1
NEURAL CORRELATES TO PERFORMANCE 4.1.1
Multiple
Regression
Analyses
4.1.2
Laterality
Analyses
4.1.3
Task
Difficulty
Modulation
4.1.4
Language
Dysfunctions
in
Epilepsy
4.2
HEALTHY ADULTS 4.3
INTERPRETATION OF ACTIVATION PATTERNS 4.4
FUTURE DIRECTIONS
31
31
33
34
35
36
37
42
5
CONCLUSIONS
45
ACKNOWLEDGMENTS
46
REFERENCES
49
PAPER I
PAPER II PAPER III PAPER IV
Big black cloud On a yellow plain Sure enough it Looks like rain Packin' up all our Faith and trust Me and the wanderlust ‘Wanderlust’
‐
Mark
Knopfler
1 INTRODUCTION
Mapping
of
language
disability
patterns
requires
a
thorough
understanding
of
language
ability
patterns.
The
neural
pathways
for
perceiving
and
generating
language
are
slowly
being
unraveled,
but
the
exact
contributions
of
typical
left‐hemispheric
language
areas
(Broca’s
and
Wernicke’s
area)
are
not
yet
completely
clear.
Neither
is
the
role
of
language‐related
regions
in
the
–
usually
non‐ dominant
–
right
hemisphere.
The
opinion
about
how
right‐hemispheric
regions
influence
language
has
changed.
In
the
past,
activation
in
the
right
hemisphere
during
language
tasks
was
largely
overlooked;
but
over
time,
researchers
gained
an
understanding
of
the
emotional
content
processing
aspects.
At
present,
additional
roles
of
the
right
hemisphere
in
language
are
being
explored,
including
language
comprehension
aspects.
Evidence
of
these
right‐hemispheric
comprehensive
aspects
is
presented
in
this
dissertation
within
a
framework
of
manifestations
of
language
ability
in
the
brain.
This
dissertation
presents
four
papers
that
investigated
language
ability,
which
was
defined
as
language
production
and
comprehension
abilities.
The
first
three
papers
describe
how
healthy
adults
were
tested
for
brain
activation
evoked
by
neurolinguistic
functional
magnetic
resonance
imaging
(fMRI)
tasks.
These
fMRI
tasks
measured
semantic
processing
and
word
fluency
activations.
The
results
were
related
to
individual
performance
measurements
in
various
tests
of
language
ability,
including
reading,
word
fluency,
picture
naming
and
use
of
complex
language.
The
fourth
paper
discusses
how
the
brains
of
people
with
generalized
epilepsy
can
express
altered
activation
patterns
in
relation
to
lower
language
ability.
1
1.
INTRODUCTION
1.1 Language Ability
1.1.1 Language Abilities The
ability
to
produce
language
enables
one
to
communicate
one’s
own
thoughts
and
express
oneself.
Comprehension
of
language
will
enable
one
to
perceive
information
that
might
be
new
or
interesting.
As
in
all
skills;
individual
differences
are
present.
The
origins
of
these
differences
might
be
attributed
to
the
amount
of
exposure
to
language,
or
to
one’s
own
interests
in
reading
or
verbal
expression.
Whenever
people
manifest
differences
in
behavior,
neuroimagers
will
look
for
the
neural
correlates
to
these
differences.
Indeed,
the
rationale
behind
the
performed
experiments
that
led
to
this
dissertation
was
to
visualize
language
ability
differences
in
healthy
subjects.
The
current
sub‐chapter
will
present
previous
research
on
language
ability
variation.
In
the
following
sub‐chapter,
‘Neural
Correlates
to
Language’,
a
more
detailed
framework
for
language
ability
will
be
introduced.
Language
discussions
often
refer
to
the
classical
language
areas
of
Broca’s
area
in
the
left
inferior
frontal
gyrus
(IFG)
and
Wernicke’s
area
in
the
left
posterior
temporal
lobe.
It
is
also
known
that
other
functional
regions
are
involved
in
language
processes;
these
will
be
explored
in
the
next
sub‐chapter.
It
seems
that
differences
in
language
performance
can
be
–
at
least
partly
–
explained
by
differentiations
in
activation
in
Broca’s
and
Wernicke’s
language
areas,
although
their
exact
contribution
is
not
yet
clear.
Studies
investigating
high
performance
in
word
fluency
have
shown
an
increase
of
left‐hemispheric
IFG
activation
for
high
performance
(Wood
et
al.,
2001),
but
also
no
difference
at
all
(Dräger
et
al.,
2004).
When
semantic
tasks
are
studied,
increased
activation
of
posterior
temporal
and
parietal
regions
is
shown
for
high
performance
(Booth
et
al.,
2003;
Meyler
et
al;
2007;
Weber
et
al.,
2006).
However,
an
opposing
view
emerges
from
an
increasing
number
of
works
revealing
a
relationship
between
reading
and
sentence
comprehension
and
decreased
activation
in
left
hemispheric
language
areas
(Reichle
et
al.,
2000;
Prat
et
al.,
2007;
2011,
Prat
&
Just,
2011).
The
mechanism
behind
this
activation
reduction
is
thought
to
be
a
more
efficient
neural
functioning.
Efficacy
in
recruiting
neural
regions
or
pathways
enables
a
person
to
re‐attribute
cognitive
resources
guided
by
task
demand.
Thus,
a
person
skilled
in
language
may
use
his
or
her
brain
in
a
more
optimal
way
for
the
presented
task.
Furthermore,
there
is
evidence
of
a
specific
role
of
the
right‐hemispheric
homologues
of
Broca’s
and
Wernicke’s
area
in
high
language
performance.
Many
of
the
results
presented
in
the
papers
that
are
included
in
this
dissertation
point
also
to
a
right‐hemispheric
contribution
to
high
language
ability.
If
people
with
a
high
language
ability
recruit
additional
language‐supporting
areas,
this
may
indicate
that
a
high
adaptability
of
neural
resources
is
an
explanatory
mechanism
for
language
ability
differences.
Research
supporting
the
theories
of
neural
adaptability
and
neural
efficiency
as
2
1.
INTRODUCTION
explicatory
for
high
language
ability
will
be
presented
in
the
sub‐chapter
‘Intelligence
models
for
Language
Ability’
1.1.2 Language Dysfunctions The
introduction
started
out
by
stating
that
knowledge
of
language
ability
will
lead
to
an
understanding
of
language
disability.
PAPER
IV
presents
a
group
of
people
with
epilepsy
showing
subtle
language
disabilities,
and
compares
them
with
healthy
subjects
performing
on
a
normal
level.
The
reverse
statement
to
the
one
above
is
also
true;
upon
investigating
language
disabilities,
a
model
for
language
abilities
can
be
created.
Much
of
our
knowledge
about
the
language
system
has
been
gained
from
lesion
studies
notably
those
on
left‐hemispheric
lesioned
patients
showing
word
production
problems,
as
presented
a
little
later
in
this
section.
Language
impairment
can
have
a
variety
of
underlying
causes;
impaired
language
functioning,
cognitive
ability,
or
sensory/motoric
abilities,
or
lack
of
training
or
exposure
to
language.
A
disruption
in
any
component
of
language
production
or
comprehension
in
the
language
model1
evidently
will
result
in
a
disruption
of
language
ability.
Since
the
studies
included
in
this
dissertation
measure
word
generation
and
sentence
reading,
this
section
discusses
reading
impairment
(dyslexia)
and
production
problems.
Developmental
dyslexia
is
characterized
by
various
neurological
differences
throughout
the
brain,
probably
caused
by
anomalies
during
the
development
of
language
systems
in
the
brain
(Catts
&
Kamhi,
2005;
Démonet
et
al.,
2005).
It
has
been
suggested
that
this
type
of
dyslexia
is
related
to
abnormal
dominance
patterns
or
abnormal
development
of
dominance
(Heim
et
al.,
2010),
but
the
causes
are
though
probably
multiple
and
more
complex
(Crystal
2010).
Acquired
dyslexia
can
occur
after
a
lesion
in
one
out
of
various
brain
regions
(Price
et
al.,
2003).
Functional
imaging
studies
on
the
neurological
differences
between
people
with
dyslexia
and
normal
performers
show
a
diminished
activation
in
temporal
and
parietal
regions
(Salmelin
et
al.,
1996;
Shaywitz
et
al.,
1998),
and
an
increase
in
inferior
frontal
activation
(Shaywitz
et
al.,
1998).
Both
the
presence
of
expected
activation
and
the
absence
of
unexpected
activation
in
the
right
hemisphere
have
been
observed
to
act
as
distinguishers
of
people
with
dyslexia
from
people
without
reading
impairment
(Paulesu
et
al.,
1996;
Simos
et
al.,
2000).
Word
production
problems
are
often
not
development‐related
but
result
from
lesions
in
the
language‐dominant
hemisphere.
Problems
with
word
fluency
are
seen
in
people
with
dementia
and
with
left
temporal
lobe
epilepsy
(Ruff
et
al.,
1997).
Named
after
the
location
of
brain
damage,
aphasia
1
e.g.
the
space
station
model
presented
in
the
following
sub‐chapter
‘Brain
Functioning’
3
1.
INTRODUCTION
can
be
classified
as
Broca’s
aphasia,
Wernicke’s
aphasia
or
global
aphasia
–
the
latter
being
a
combination
of
Broca’s
and
Wernicke’s
aphasia.
It
is
now
known
that
in
Broca’s
aphasia,
brain
regions
posterior
to
Broca’s
area
are
often
damaged;
and
that
in
Wernicke’s
aphasia
the
location
of
damage
can
vary
(Crystal
2010).
Broca’s
aphasia
results
in
deficits
in
expressive
abilities
and
is
characterized
by
non‐fluent
speech
which
is
grammatically
incorrect.
Wernicke’s
aphasia
occurs
when
receptive
systems
are
damaged
and
results
in
both
comprehension
problems
and
problems
producing
intelligible
speech,
even
though
it
appears
to
be
fluent.
Furthermore,
word
retrieval
problems
are
a
common
deficiency
(Crystal
2010).
Studies
on
language
disabilities
can
help
us
to
find
regions
of
interest
for
the
investigation
of
language
abilities.
Lesion
studies
that
have
led
to
an
understanding
of
language
disabilities
have
shown
that
disruption
of
language
functioning
in
the
language‐dominant
hemisphere
has
a
much
higher
impact
than
a
disruption
in
the
non‐dominant
hemisphere.
Thus,
the
language
functions
in
the
non‐dominant
hemisphere
may
not
be
compulsory
for
language
production,
but
may
support
complex
processing.
1.2 Neural Correlates to Language
1.2.1 Language Models There
are
many
possible
theoretical
models
to
describe
the
complex
structure
of
language.
Often,
these
models
use
similar
distinctions
between
word
forms,
word
structure,
word
meaning
and
understanding
of
text
or
speech.
In
other
words,
many
models
describe
language
as
a
process
defining
the
range
of
linguistic
information
from
small
building
blocks
to
complex
meaningful
communication.
To
understand
language
in
the
context
of
this
dissertation,
a
useful
model
is
the
space
station
model
as
presented
by
Crystal
(2010),
and
represented
in
Figure
1.
This
model
describes
an
interactive
framework
integrating
the
components
of
language
that
are
investigated
in
the
papers
included
in
this
dissertation.
The
different
components
are:
phonetics
(pronunciation
attributes)
and
phonology
(sounds
that
convey
different
meanings),
morphology
(word
structure)
and
syntax
(sentence
structure),
semantics
(meaningful
content)
and
pragmatics
(discourse
information).
The
connection
between
these
components
is
not
uni‐directional,
but
rather
interconnected
as
represented
in
the
space
station
model.
This
is
consistent
with
the
neural
organization
of
language,
where
both
top‐down
and
bottom‐up
processes
can
be
observed
during
language
processes
(Friederici
2012).
4
1.
INTRODUCTION
Figure
1. Representation of the Space Station Language Model. The linguistic levels presented in the circles are interconnected, indicating free exchange of linguistic information between levels; thus all information is available at once for an external researcher. Figure adapted from Crystal (2010).
Measures
of
language
ability
preferably
test
for
many
linguistic
components,
including
production
and
perception
of
language,
and
have
a
high
enough
difficulty
level
to
measure
variability
in
language
skills.
On
the
other
hand,
the
total
test
duration
should
be
kept
to
a
minimum
as
to
impose
only
minimally
on
the
participants,
especially
on
those
with
cognitive
disabilities.
The
tests
used
in
our
studies,
(see
also
Methods
section
for
their
description),
show
two
approaches
towards
this
goal.
First;
established
tests
such
as
the
Boston
Naming
Test
(Kaplan
et
al.,
1983)
or
word
fluency
tests
– testing
word
retrieval
and
word
production
skills
–
are
used
in
many
research
studies
that
describe
the
neural
mechanisms
that
lie
behind.
Moreover,
these
tests
are
easily
translated
to
the
magnetic
resonance
scanner
environment
without
much
adapting.
However,
both
tasks
are
very
focused;
they
do
not
test
for
the
full
spectrum
of
language
ability.
Other
tests,
such
as
comprehensive
reading,
investigate
language
perception
and
comprehension
and
could
be
translated
to
the
scanner
environment
with
some
modification.
A
second
approach
is
to
gather
multiple
language
ability
tests
in
a
battery,
such
as
the
Assessment
of
Subtle
Language
Deficits
or
BeSS
test
(Laakso
et
al.,
2000).
This
relatively
new
complex
language
ability
test
is
not
yet
established,
but
can
detect
subtle
language
dysfunctions
without
showing
a
ceiling
effect
(as
the
results
of
our
papers
will
show).
Moreover,
this
is
a
compact
test,
so
that
language
ability
can
be
assessed
quickly
without
too
much
imposing
on
the
5
1.
INTRODUCTION
concentration
skills
of
people
with
language
dysfunctions
(such
as
the
people
with
generalized
epilepsy
from
our
PAPER
IV).
However,
this
test
is
less
practical
in
a
scanner
environment.
Neurological
models
are
often
based
on
the
classical
Wernicke‐Geschwind
model
(Geschwind
1965),
which
describes
the
neurological
dissociation
between
language
production/speech
attributed
to
Broca’s
area,
and
language
semantic
comprehension
(semantics)
attributed
to
Wernicke’s
area.
Many
later
studies
have
shown
that
this
description
is
insufficient,
as
it
does
not
take
into
account
other
functional
areas,
nor
does
it
describe
accurately
the
precise
boundaries
of
linguistic
functional
areas
(Price
2000;
2012;
Démonet
et
al.,
2005;
Smits
et
al.,
2006).
An
overview
of
the
segregation
in
left‐hemispheric
language
areas
is
given
in
Figure
2.
For
instance,
Broca’s
area
contains
regions
involved
in
semantics
as
well
as
in
syntax
processing
(cf.
Price
2012).
Interestingly,
although
language
studies
often
focus
on
the
language‐dominant
left
hemisphere
(Vigneau
et
al.,
2006),
the
right
hemisphere
often
shows
a
similar
activation
pattern
(Démonet
et
al.,
2005).
Nevertheless,
aspects
of
neural
correlates
to
the
Wernicke‐Geschwind
model
are
supported
by
recent
lesion
studies
investigating
aphasia
(Yang
et
al.,
2008)
and
by
functional
imaging
studies
(Price
2000;
Bookheimer
2002).
Therefore,
Broca’s
and
Wernicke’s
area
are
used
as
regions
of
interest
in
several
of
our
analyses,
in
combination
with
other
regions
that
were
found
in
relation
to
semantic
and
word
fluency
tasks.
When
using
the
labels
of
Broca’s
and
Wernicke’s
areas,
it
is
important
to
define
their
extent;
the
definition
of
Wernicke’s
area
in
particular
can
vary
from
including
only
the
posterior
superior
temporal
gyrus
to
the
inclusion
of
large
parts
of
the
parietal
and
temporal
cortex.
Throughout
this
dissertation,
including
all
articles,
the
definition
used
is
as
follows:
Broca’s
area
comprises
the
left
IFG;
specifically
Brodmann
areas
(BA)
44
and
45.
Wernicke’s
area
comprises
the
left
posterior
superior
temporal
gyrus
(BA
22)
and
the
posterior
part
of
BA
21,
as
well
as
the
posterior
perisylvian2
region
which
consists
of
the
left
angular
gyrus
and
the
supramarginal
gyrus
(BA
39
&
inferior
BA
40).
The
right‐hemispheric
counterparts
of
these
areas
are
referred
to
as
Broca’s
and
Wernicke’s
area
homologues.
Language
production
and
perception
are
by
no
means
controlled
solely
by
these
regions3.
The
regions
important
for
language
will
be
discussed
in
the
following
sections
which
introduce
an
overview
of
activation
related
to
semantic
and
word
fluency
tasks.
Since
the
topic
of
this
dissertation
is
language
ability,
neural
processes
not
directly
related
to
language
are
not
introduced
here.
2
Perisylvian
indicates
the
region
around
the
Sylvian
fissure.
This
fissure
divides
the
frontal
and
parietal
lobules
from
the
temporal
lobe.
3
An
example
is
given
by
(Dronkers
et
al.,
2007),
who
found
that
the
patients
of
Paul
Broca
–
whose
brains
evidenced
the
theory
of
speech
production
located
in
left
IFG
–
had
lesions
that
were
spread
over
a
wider
region
than
just
Broca’s
area.
6
1.
INTRODUCTION
Figure
2. Finite overview (based on imaging studies by Cathy Price) of the segregation of functional languagerelated areas in the left hemisphere. The colored areas each refer to different tasks, either differing in modality (auditory/visual) or in linguistic component. Figure reprinted with permission. See Price (2012) for details.
7
1.
INTRODUCTION
1.2.2 Semantics Our
studies
have
used
semantic
sentence
reading
fMRI
tasks,
either
requiring
completion
of
sentences
or
reading
of
congruent/incongruent
sentences.
Semantic
tasks
such
as
reading
(Price
2000),
and
sentence
and
story
comprehension
(Sakai
et
al.,
2001;
Kaan
&
Swaab,
2002)
typically
activate
Broca’s
and
Wernicke’s
area
in
the
left
hemisphere
(Price
et
al.,
2003;
overview
in
Binder
et
al.,
2009).
In
the
left
IFG,
BA
47
plays
also
a
role
in
semantic
processing
(Dapretto
&
Bookheimer,
1999;
Bookheimer
2002).
Furthermore,
the
anterior
temporal
cortex
and
the
fusiform
gyrus
are
involved
in
semantic
processing
(Price
et
al.,
2003;
overview
in
Price
2012).
Activation
in
the
parietal
perisylvian
region
has
been
shown
to
correlate
with
linguistic
complexity
in
sentences
(Carpenter
et
al.,
1999)
and
semantic
associating
(Price
2000).
Semantic
processing
often
also
activates
right‐ hemispheric
IFG
and
temporal
lobe
(Bookheimer
2002),
which
will
be
discussed
in
the
section
‘Right‐ Hemispheric
Influences’.
1.2.3 Word Fluency Word
generation
(or:
word
fluency)
tasks
are
frequently
used
to
determine
language
lateralization
by
fMRI
(Cuenod
et
al.,
1995;
Hertz‐Pannier
et
al.,
1997).
The
generation
of
words
evokes
activation
in
the
left
middle
and
inferior
frontal
gyrus
(Fu
et
al.,
2002;
Costafreda
et
al.,
2006),
with
a
particularly
important
role
for
the
pars
opercularis
(Price
2000).
Furthermore
is
activation
observed
in
the
inferior
temporal
cortex
and
in
the
adjacent
fusiform
area
(Price
2000),
and
in
the
anterior
cingulate
cortex
(Fu
et
al.,
2002)
The
sub‐regions
in
the
IFG
have
specific
roles
and
the
activation
pattern
is
dependent
on
the
nature
of
the
fluency
task
(Heim
et
al.,
2009).
1.2.4 Right‐Hemispheric Influences Most
language
tasks
evoke
activation
in
bilateral
frontal,
temporal
or
parietal
areas;
the
specific
role
of
right‐hemispheric
language
areas
is
often
interpreted
as
abstract
linguistic
functioning.
Although
lesion
studies
often
indicate
that
the
right‐hemisphere
is
not
indispensable
for
language
production,
neuroimaging
studies
show
that
the
right
hemisphere
plays
an
important
and
often
distinct
role,
something
we
found
evidence
of
in
our
studies
as
well.
Vigneau
and
colleagues
(2011)
discuss
in
their
meta‐analysis
the
right
hemisphere
in
relation
to
language
processing.
They
conclude
that
the
right‐ hemispheric
IFG
seems
to
have
no
access
to
phonemic
representations,
unlike
the
left
IFG.
Activation
in
the
right
IFG
is
observed
during
processing
of
metaphors
(Schmidt
&
Seger,
2009)
and
the
perception
of
prosody
(Buchanan
et
al.,
2000).
Furthermore,
the
right
IFG
is
active
when
information
is
conflicting
during
complex
language
tasks;
this
is
related
to
figurative
language
and
increasing
8
1.
INTRODUCTION
ambiguity
(Bookheimer
2002;
Snijders
et
al.,
2009).
Bookheimer
suggests
that
the
role
of
the
right
IFG
might
be
to
help
making
decisions
based
on
linguistic
information.
The
right
hemisphere
is
also
important
for
understanding
and
integrating
spoken
and
written
information
(Bookheimer
2002).
In
particular,
the
understanding
of
context
processing
or
pragmatics
–
which
is
necessary
for
interpreting
for
example
ambiguous
or
emotionally
loaded
information
–
is
attributed
to
the
right
temporal
lobe
(Vigneau
et
al.,
2011).
Examples
of
right
temporal
lobe
activation
are
seen
in
studies
investigating
the
interpretation
of
prosody
(Vigneau
et
al.,
2011),
the
integration
of
semantic
information
(Caplan
&
Dapretto,
2001),
or
the
processing
of
metaphors
(Bottini
et
al.,
1994;
Mashal
et
al.,
2005;
Ahrens
et
al.,
2007).
The
neural
activation
resulting
from
the
processing
of
metaphors
is
possibly
related
to
the
metaphors
being
perceived
as
nonsensical
or
containing
novel
semantic
information
(Mashal
et
al.,
2009).
The
right
hemisphere
is
thus
involved
in
pragmatic
processing
on
a
meta‐syntactic
level
(Mitchell
&
Crow,
2005).
1.2.5 Laterality The
dominance
of
a
hemisphere
in
language
processing
can
be
quantified
as
the
degree
of
lateralization.
A
non‐typical
degree
of
lateralization
has
been
attributed
to
both
language
abilities
and
disabilities
(cf.
the
first
section
‘Language
Abilities’).
Knecht
and
colleagues
(2000)
tested
188
healthy
right‐handed
adults
for
language
lateralization
in
the
brain
with
a
word
generation
fMRI
task.
This
task
has
been
widely
reported
to
be
a
powerful
and
effective
paradigm
for
generating
language
production
(Neils‐Strunjas
1998).
Language
lateralization
study
results
have
indicated
that
there
is
no
difference
in
language
lateralization
ratios
between
males
and
females.
Furthermore,
a
left‐
to
right‐hemispheric
dominance
ratio
of
13
to
1
was
established
(Knecht
et
al.,
2000).
Besides
fMRI,
dichotic
listening
is
an
alternative
and
feasible
non‐invasive
method
to
test
for
language
lateralization
(Hugdahl
2011).
The
dichotic
listening
method
is
based
on
the
notion
that
bi‐aural
auditory
stimuli
travel
more
easily
to
the
contralateral
rather
than
ipsilateral
hemisphere,
due
to
more
extensive
contralateral
than
ipsilateral
pathways
from
the
ear
to
the
auditory
cortex.
Also,
there
is
a
blocking
of
ipsilateral
pathways
during
conflicting
input.
After
travelling
to
the
contralateral
cortex,
the
auditive
signals
are
processed
more
automatically
in
the
hemisphere
that
is
dominant
for
language.
Ergo,
the
language‐dominant
hemisphere
presumably
resides
contralateral
to
the
ear
that
processes
more
stimuli
during
bi‐aural
stimulation
(Kimura,
2011).
9
1.
INTRODUCTION
Differences
between
methods
to
test
for
laterality
are
discussed
by
Abou‐Khalil
(2007),
who
concluded
that
fMRI
was
one
of
the
most
realizable
techniques4.
The
clear
advantage
of
fMRI
over
dichotic
listening
is
that
fMRI
can
localize
activation.
Nonetheless,
dichotic
listening
is
superior
in
practicality,
both
in
terms
of
costs
and
of
convenience.
It
is
also
important
to
realize
that
the
laterality
measurements
obtained
by
fMRI
are
very
much
dependent
on
which
language
task
is
chosen.
Both
word
fluency
and
sentence
comprehension
seem
to
be
indicative
of
determining
language
lateralization
(Niskanen
et
al.,
2012).
Besides
ear
dominance,
hand
dominance
is
also
seen
to
have
a
direct
connection
to
the
contralateral
hemispheric.
Right‐handedness
is
highly
correlated
with
left‐hemispheric
language
dominance
(in
94
–
96
%
of
right‐handers).
In
left‐handers,
it
is
slightly
more
common
to
have
right‐hemispheric
dominance,
yet
78
%
of
the
left‐handed
population
is
also
left
dominant
for
language
(Szaflarski
et
al.,
2002).
Language
lateralization
is
thought
to
correlate
with
differences
in
gray
matter
between
hemispheres,
and
when
the
cortex
is
damaged,
language
lateralization
for
expressive
language
functions
can
change
(Lee
et
al.,
2008).
Josse
and
colleagues
(2009)
investigated
how
gray
matter
differences
could
predict
language
lateralization,
and
showed
that
when
gray
matter
is
analyzed
with
a
voxel‐by‐voxel
method,
structural
asymmetry
correlated
well
with
language
lateralization.
However,
these
correlations
were
lost
when
global
lateralization
was
compared
with
regional
gray
matter
asymmetries.
Nowadays,
local
lateralization
is
of
interest
and
many
researchers
prefer
to
investigate
the
lateralization
of
separate
regions
(Seghier
et
al.,
2011b).
A
strong
lateralization
of
cognition
has
been
linked
to
high
cognitive
performance
(Güntürkün
et
al.,
2000).
Recently,
an
opposing
view
has
emerged,
namely
that
the
optimal
degree
of
lateralization
for
high
cognitive
performance
was
small.
In
other
words;
a
higher
degree
of
bilaterality
might
be
more
favorable
for
performance
(Hirnstein
et
al.,
2010).
1.2.6 Anti‐correlated Brain Activation In
PAPER
IV
we
examine
activation
that
is
correlated
negatively
with
language
tasks;
this
can
be
labeled
as
deactivation.
Deactivation
is
the
decrease
of
signal
in
regions
that
are
activated
during
rest
but
not
during
task
condition,
thus
functions
in
these
regions
are
thought
to
be
suppressed.
Some
of
these
regions
form
a
network
that
is
consistently
activated
during
rest
and
deactivated
during
tasks;
this
is
called
the
Default
Mode
Network
(DMN).
DMN
activation
is
associated
with
‘free
thinking’
4
cf.
(Medina
et
al.,
2007),
who
presents
an
overview
of
the
reliability
of
fMRI‐obtained
laterality
measurement.
10
1.
INTRODUCTION
processes
–
often
referred
to
as
thinking
about
the
day,
shopping
lists,
and
what’s
for
dinner
–
therefore
the
suppression
of
DMN
activation
enables
a
person
to
allocate
more
cognitive
power
to
the
task.
Heterogeneity
of
the
anti‐correlation
during
a
semantic
task
in
the
different
regions
of
the
DMN
is
to
be
expected
(Seghier
&
Price,
2012).
A
difference
in
suppression
of
the
DMN
between
the
task
and
control
condition
can
also
be
expected,
depending
on
how
engaging
the
control
condition
is.
Deactivation
patterns
might
be
just
as
necessary
as
activation
patterns
to
explain
brain
functioning
(Binder
2012).
1.3 Intelligence models for Language Ability
1.3.1 Relation Language Ability and Intelligence There
is
an,
although
limited,
correlation
between
language
ability
and
intelligence
(e.g.
word
fluency:
Haier
et
al.,
1992;
Roca
et
al.,
2010;
semantics:
Prat
et
al.,
2007).
Some
intelligence
models
describe
processes
that
can
be
applied
to
language
ability
as
well,
and
help
to
understand
the
differences
in
language
performance
observed
in
previous
and
our
current
work.
Intelligence
is
attributed
to
a
parieto‐frontal
network
that
includes
several
regions
and
connections
that
are
shared
with
language
processing
functions.
This
network
is
described
in
the
Parieto‐Frontal
Integration
Theory
of
intelligence
(Jung
&
Haier,
2007).
A
second
intelligence
theory
is
the
neural
efficiency
hypothesis
of
intelligence
(Haier
et
al.,
1992).
This
theory
describes
how
well‐developed
skills
can
be
characterized
by
a
more
effective
manner
of
processing
in
the
brain.
Thus;
high‐skilled
individuals
will
show
a
decreased
brain
activation
compared
with
lower‐skilled
persons.
This
reasoning
can
be
applied
to
language
skills
as
well,
as
will
be
put
forward
in
the
next
section.
Lastly,
neural
adaptability
is
discussed;
this
is
a
trait
observed
in
high‐skilled
individuals.
These
theories
together
may
explain
the
functional
activation
patterns
observed
in
high
performers
(e.g.
Prat
2011;
Langer
et
al.,
2012).
1.3.2 Intelligence Models The
ParietoFrontal Integration Theory (PFIT) of intelligence
is
a
summation
of
regions
in
a
network
found
to
show
activation
dependent
on
intelligence
level
(Jung
&
Haier,
2007).
It
has
been
known
that
neural
correlates
to
high
intelligence
are
located
in
the
prefrontal
cortex
(Thompson
et
al.,
2001),
and
that
increased
gray
and
white
matter
is
observed
in
both
frontal
and
parietal
regions
in
correlation
11
1.
INTRODUCTION
with
high
intelligence
(Neubauer
&
Fink,
2009).
The
P‐FIT
of
intelligence
states
that
it
takes
a
network
of
interactive
regions
to
provide
high
abilities.
The
functions
are
divided
within
this
network
from
caudally
located
rule
generating
processes,
to
rostral
functions
such
like
selecting,
and
testing
of
answers.
The
network
includes
the
language
processing
areas
in
the
posterior
perisylvian
region.
The
Neural efficiency hypothesis of intelligence
states
that
networks
for
cognitive
functions
work
in
a
more
efficient
manner
in
intelligent
brains.
Therefore,
intelligent
brains
will
show
less
activation
in
task‐specific
networks
during
imaging
studies.
Haier
and
colleagues
(1992)
state
that
the
mechanism
behind
neural
efficiency
might
be
deactivation
of
irrelevant
brain
areas,
or
a
more
specific
use
of
task‐related
areas.
The
neural
efficiency
hypothesis
of
intelligence
appears
to
be
limited
to
frontal
regions,
and
conditional
on
task
as
well
as
task‐difficulty
(Neubauer
&
Fink,
2009).
Predominantly
frontal
activation
patterns
in
high
performers
show
efficient
behavior
during
easy
to
moderately
difficult
tasks.
Activation
in
the
frontal
region
has
previously
been
shown
to
decrease
upon
automation
of
processes
(Ramsey
et
al.,
2004).
When
demands
get
high,
this
is
no
longer
true;
high
performers
then
recruit
more
brain
regions
to
solve
the
task.
The
high
intelligent
individuals
might
have
more
adaptive
strategies
than
low
performers
and
can
–
depending
on
task
demand
–
either
use
their
brain
efficiently
or
call
in
the
help
of
supporting
brain
regions
(Doppelmayr
et
al.,
2005).
Neural
efficiency
patterns
have
been
observed
in
high
capacity
readers
during
sentence
comprehension
(Maxwell
et
al.,
1974;
Prat
et
al.,
2007;
Prat
&
Just,
2011).
The
additional
recruitment
of
supporting
neural
resources
whenever
a
task
is
difficult
may
be
described
as
Neural adaptability
(Prat
et
al.,
2007).
It
is
hypothesized
that
individuals
highly
proficient
in
language
show
more
neural
adaptability
compared
with
people
with
lower
proficiency.
This
can
be
observed
as
activation
in
language‐related
regions,
either
in
main
language
regions
or
in
additional
supportive
regions.
Evidently,
the
theories
above
outline
a
varied
pattern
of
the
relation
between
high
performance
and
neural
activation
or
deactivation.
This
pattern
is
dependent
on
task,
task
demands
and
functional
region.
In
the
Discussion
the
considerations
concerning
the
interpretation
of
brain
activation
will
be
further
explored.
12
1.
INTRODUCTION
1.4 Aims
Language
ability
in
healthy
adults
was
expected
to
be
visualized
as
a
modulation
of
activation
in
language‐related
regions,
with
respect
to
the
level
of
activation,
but
also
the
degree
of
lateralization
between
hemispheres.
PAPER
I
aimed
to
determine
regional
lateralization
of
semantic
language
functions
in
relation
to
performance
in
tests
of
language
ability.
It
was
expected
to
find
laterality
differences
related
to
performance
in
the
IFG
and
posterior
temporal
lobe,
for
both
fMRI‐obtained
laterality
and
for
dichotic
listening.
PAPER
II
aimed
to
find
the
neural
correlates
to
language
ability
throughout
the
whole
brain.
The
expectation
was
to
find
specific
regions
in
the
right
IFG
and
posterior
temporal
lobe
activated
during
from
a
semantic
task
that
were
related
to
high
performance
in
tests
of
language
ability.
Furthermore,
brain
activation
during
word
fluency
was
investigated
and
compared
with
semantic
results,
in
order
to
find
whether
there
were
similarities
in
activation
patterns
related
to
high
language
ability.
PAPER
III
aimed
to
reproduce
the
findings
of
PAPER
I
and
PAPER
II
in
a
new
study
population.
Thus,
activation
during
semantic
and
word
fluency
tasks
that
emerged
in
the
right‐hemispheric
homologues
of
Broca’s
and
Wernicke’s
area
were
investigated
for
their
correlation
with
high
performance
in
tests
of
language
ability.
In
addition,
activation
related
to
task
demand
was
investigated.
Brain
activation
patterns
related
to
high
performance
were
expected
to
show
neural
efficiency
for
low‐demand
tasks
in
the
IFG.
Furthermore,
high
language
ability
was
expected
to
be
characterized
by
neural
adaptability;
i.e.
increased
right‐hemispheric
contributions.
PAPER
IV
aimed
to
investigate
language
deficits
in
people
with
generalized
epilepsy.
This
group
was
also
expected
to
show
an
inadequate
suppression
of
the
default
mode
network
that
is
normally
highly
anti‐correlated
with
the
task.
13
Strength and courage overrides The privileged and weary eyes Of river poet search naiveté Pick up here and chase the ride The river empties to the tide All of this is coming your way ‘Find
the
River’
–
Bill
Berry,
Michael
Stipe,
Peter
Buck,
Michael
Mills
14
2 METHODS
2.1 Neurolinguistic Measures
2.1.1 Tests of Language Ability In
PAPER
I
and
PAPER
II,
four
tests
to
measure
language
ability
were
used:
FAS
and
BNT
measured
word
retrieval
abilities,
and
BeSS
and
Read
measured
language
comprehension
abilities.
In
PAPER
III
and
IV,
only
BeSS
and
FAS
were
used.
FAS
is
a
phonemic
word
generation
test
in
which
participants
are
cued
with
a
letter
(F,
A,
S),
and
have
to
generate
as
many
words
as
possible,
starting
with
the
cue
letter.
Total
score
is
the
number
of
generated
words
for
all
three
letters.
BNT
is
the
established
Boston
Naming
Test.
During
the
test,
the
participant
is
presented
with
60
pictures
that
have
to
be
named.
BeSS
(
“Bedömning
av
Subtila
Språkstörningar”
or
Assessment
of
Subtle
Language
Deficits)
tests
for
the
use
of
complex
language
by
means
of
seven
subtasks
(Laakso
et
al.,
2000).
Those
subtasks
are:
REP
repetition
of
long
sentences
(9‐16
words)
CON
sentence
construction
(from
three
words,
with
given
context,
under
time
pressure)
INF
inferential
reasoning
(based
on
a
read
text)
COM
comprehension
of
complex
embedded
sentences
GAR
comprehension
of
garden‐path
or
ambiguous
sentences
MET
comprehension
of
metaphors
VOC
vocabulary
–
word
definition
15
2.
METHODS
Maximum
score
was
210
points.
The
Read
test
is
selected
from
a
Swedish
exam
for
university
students.
Participants
had
to
read
three
texts
and
answer
four
questions
on
each
text.
The
total
score
was
the
number
of
correctly
answered
questions.
2.1.2 Dichotic Listening Dichotic
Listening
scores
were
acquired
in
PAPER
I
with
the
use
of
a
version
of
the
Bergen
Dichotic
Listening
Test
(Hugdahl
1995),
which
is
a
consonant‐vowel
test.
Auditive
stimuli
created
from
the
combination
of
a
stop
consonant
and
the
vowel
‘a’
(e.g.
ba
–
ga
–
pa)
were
presented
bi‐aurally
to
the
participants.
Depending
on
the
instructions,
the
participants
had
to
report
the
stimuli;
either
heard
in
the
left
or
the
right
ear;
in
both
ears;
or
the
most
salient
stimulus.
The
results
were
calculated
as
a
right
ear
advantage;
subtracting
correct
responses
perceived
by
the
left
ear
from
those
heard
in
the
right
ear,
then
dividing
this
figure
by
the
number
of
total
correct
responses.
A
high
right
ear
advantage
meant
that
the
subject
was
better
at
reproducing
stimuli
heard
in
the
right
ear,
compared
with
the
left
ear.
This
was
interpreted
as
a
lateralization
index
for
language;
a
high
right
ear
advantage
meant
strong
left‐hemispheric
lateralization.
2.1.3 fMRI Language Paradigms The
word
generation
task
WORGE
from
PAPER
II
was
as
described
in
(Engström
et
al.,
2010)
but
with
moderation
of
the
control
condition.
The
participants
were
cued
with
a
letter
taken
from
the
Swedish
alphabet,
excluding
C,
Q,
W,
X,
Y,
Z,
Å,
Ä,
and
Ö.
They
were
instructed
to
generate
words
with
the
cued
letter,
as
many
as
possible
within
the
given
time
of
5
s.
The
cue
letters
were
varied
and
presented
in
blocks
containing
three
to
five
letters,
pseudorandomly
ordered.
The
baseline
or
control
task
consisted
of
presentation
of
an
asterisk
alternated
with
a
row
of
asterisks.
The
word
generation
task
WORD
is
described
in
PAPER
III.
Similarly
to
WORGE,
a
cue
letter
was
presented,
but
this
time
the
cue
letters
were
divided
into
two
difficulty
categories;
‘easy’
(frequent
starting
letter
in
a
Swedish
word
list)
and
‘hard’
(infrequent
starting
letter).
The
letters
were
presented
per
category
in
a
block
of
seven
letters,
alternating
with
control
blocks.
The
control
block
differed
from
WORGE
in
the
sense
that
only
one
asterisk
was
presented
each
trial.
The
sentence
completion
task
SENCO
is
described
in
PAPER
I.
This
was
a
cloze
task;
the
participant
had
to
silently
generate
the
missing
last
word
of
a
sentence.
The
sentences
were
presented
in
blocks,
16
2.
METHODS
the
presentation
duration
of
a
sentence
was
3
s
followed
by
display
of
an
asterisk
for
2
s.
The
control
condition
consisted
of
asterisks
mimicking
a
short
sentence.
The
congruent/incongruent
sentence
reading
task
SEN
is
described
in
PAPER
III.
The
participants
were
presented
with
blocks
differing
in
difficulty
level;
either
congruent
(‘easy’
condition)
or
incongruent
(‘hard’
condition)
sentences,
or
control
blocks
containing
a
row
of
asterisks
and
arrows.
The
participants
had
to
judge
whether
the
situation
described
in
the
sentence
took
place
inside
or
outside.
During
the
control
condition,
the
participants
had
to
report
in
which
direction
the
arrow
was
pointing.
2.1.4 Study Population Study
A
investigated
a
healthy
adult
population
of
18
participants:
nine
females
and
nine
males
aged
21‐64
(mean
age:
40).
For
PAPER
1,
a
subset
of
14
participants
(seven
females,
seven
males)
were
investigated,
aged
21‐55
(mean
age:
36.9).
Study
B
investigated
two
groups.
First,
a
healthy
adult
population
of
27
participants:
14
females
and
13
males
aged
18‐35
(mean
age:
25.5)
was
investigated.
The
analyses
from
PAPER
III
were
performed
on
data
from
this
group.
For
PAPER
IV;
the
healthy
control
group
was
compared
with
a
group
of
11
people
with
generalized
epilepsy:
six
females
and
five
males,
with
an
age
range
of
20‐35
years
(mean
age:
26.5).
In
both
the
healthy
control
group
and
in
the
group
of
people
with
generalized
epilepsy
there
was
a
left‐handed
individual.
All
participants
had
Swedish
as
their
first
language
and
were
screened
by
means
of
a
questionnaire
on
the
absence
of
neurological,
cognitive
or
psychiatric
disorders
and
magnetic
resonance
contra‐ indications.
2.1.5 Generalized Epilepsy The
different
types
of
epilepsy
can
be
classified
according
to
etiology.
This
results
in
a
distinction
between
generalized
epilepsies
with
genetically
inherited
origin,
and
focal
epilepsies
(Berg
et
al.,
2010;
Poduri
&
Lowenstein,
2011).
People
with
generalized
epilepsy
(GE)
show
a
widespread
atypical
cortical
activity
(Marini
et
al.,
2003)
and
may
experience
language
problems
(Chaix
et
al.,
2006;
Caplan
et
al.,
2009).
GE
is
also
related
to
an
abnormal
connectivity
in
the
default
mode
network
(McGill
et
al.,
2012).
17
2.
METHODS
2.2 Functional MRI
2.2.1 Properties of Functional MRI Functional
MRI
can
detect
susceptibility
changes
in
the
blood
that
arise
depending
on
the
amount
oxygen
that
is
present.
Neurons
that
are
activated
exchange
neurotransmitters,
and
this
exchange
process
consumes
oxygen.
This
is
overcompensated
by
transport
of
an
abundance
of
oxygenated
blood5
to
the
activated
area,
the
oxygenated
blood
differs
from
the
surrounding
deoxygenated
blood
in
magnetic
properties.
This
process
is
called
the
blood
oxygen
level
dependent
(BOLD)
response
and
is
measured
using
susceptibility
sensitive
magnetic
resonance
sequences6.
Since
changes
in
blood
flow
are
slow,
the
fMRI
signal
has
a
low
temporal
aspect.
Furthermore,
the
magnetization
difference
is
very
subtle,
with
a
low
signal‐to‐noise
ratio.
Therefore,
a
common
approach
is
to
repeat
the
action
or
stimulus
that
evokes
the
pattern
of
interest
many
times,
and
calculate
the
average
of
the
response.
The
highest
power
is
obtained
when
stimuli
are
presented
in
blocks,
and
the
blocks
for
different
conditions
and
the
baseline
are
presented
in
an
alternating
sequence.
To
get
a
measure
of
neural
activation
per
condition
in
each
spatial
unit
(i.e. voxel),
a
standard
approach
is
to
model
the
expected
BOLD
response
with
the
general
linear
model
(GLM),
which
is
then
fitted
to
the
data.
This
model
is
time‐variant.
An
equation
for
the
GLM
is
given
as
Y
=
Xβ
+
ε
,
in
which
Y
is
the
data
represented
by
the
design
matrix
X
(the
design
matrix
models
aspects
of
the
experiment
such
as
conditions
or
performance
covariates)
times
the
parameter
estimates
β
(estimates
for
the
data
that
explain
as
much
as
possible).
The
ε
is
the
residual
error
term.
In
our
studies,
we
used
statistical
parametric
mapping
(SPM)7
to
model
the
GLM
on
our
data.
All
our
studies
were
collected
with
a
Philips
Achieva
1.5
tesla
scanner,
using
gradient‐echo
planar
imaging
sequences.
The
obtained
images
were
all
normalized
to
a
standard
brain
with
coordinates
in
Montreal
Neurological
Institute
(MNI)
space.
The
activation
pattern
for
each
condition
can
be
quantified
by
subtracting
the
number
of
activated
voxels
in
one
condition
from
another.
Most
often
are
task
conditions
compared
to
a
baseline
condition.
Subsequently,
the
significance
of
the
first‐level
analysis
results
(testing
individuals)
can
be
tested
by,
for
example,
t‐tests.
Thus,
testing
for
activation
related
to
a
certain
condition
can
be
done
by
subtracting
baseline
activation
from
activation
during
the
condition.
Testing
for
deactivation
can
5
To
be
precise;
it
is
the
hemoglobin
protein
that
transports
oxygen
in
the
blood.
Hence
the
term
‘hemodynamic
response
function’
that
is
used
to
describe
the
overcompensation
of
oxygen
transport
to
active
neurons.
6
Paramagnetic
deoxygenated
blood
disturbs
the
magnetic
resonance
signal,
by
hastening
the
dephasing
of
protons
that
emit
this
signal.
If
the
amount
of
oxygenated
blood
increases,
the
measured
signal
increases
as
well.
7
www.fil.ion.ucl.ac.uk/spm/software
18
2.
METHODS
be
done
by
subtracting
condition
activation
from
baseline
activation.
The
resulting
statistical
maps
can
be
entered
into
a
second‐level
analysis
to
test
on
group
level.
Two
groups
can
be
compared
with
a
two‐sample
t‐test,
or
if
data
fluctuation
depending
on
individual
performance
scores
is
investigated,
a
multiple
regression
approach
can
be
taken.
For
our
multiple
regression
analysis,
we
corrected
for
age
by
modeling
age
as
a
covariate
and
tested
for
individual
performance
differences
by
modeling
performance
score
as
a
covariate
of
interest.
2.2.2 Region of Interest Analysis If
the
location
of
expected
activation
is
reasonably
certain,
and
an
analysis
of
the
whole
brain
is
not
required,
the
analysis
can
be
restricted
to
regions
of
interest
(ROIs).
In
our
studies,
ROIs
were
obtained
in
different
ways,
a posteriori
and
a priori,
to
answer
different
questions.
In
PAPER
II,
the
whole‐brain
analysis
results
were
used
to
guide
placement
of
small
spherical
ROIs
at
significant
peaks
of
activation.
Parameter
estimates
were
calculated
from
an
ROI
analysis
and
then
tested
for
their
correlation
strength
with
performance.
To
report
the
strength
of
these
correlations
as
a
measure
of
significance
would
give
an
inflated
measurement,
since
this
is
a
second
correlation
of
fMRI
data
with
performance
scores.
Therefore,
our
posthoc
results
were
merely
used
to
filter
out
low‐significant
correlations
from
the
regions
that
were
significant
in
the
multiple
regression
analysis
with
a
p‐value
threshold
of
0.01,
corrected
for
multiple
measurements
by
means
of
the
false
discovery
rate
In
Study
B
that
led
to
PAPER
III
and
PAPER
IV,
we
had
an
expectation
of
which
regions
would
be
active.
Therefore,
we
were
able
to
restrict
our
statistical
tests
to
include
only
the
voxels
in
the
predicted
regions
and
thus
correct
the
significance
calculation
for
the
small
volumes
used.
For
the
unpublished
results
related
to
the
healthy
population
in
Study
B
that
are
discussed
in
this
dissertation,
we
used
the
following
ROIs:
the
IFG
pars
opercularis
(BA
44),
IFG
pars
triangularis
(BA
45),
IFG
pars
orbitalis
(BA
47),
the
middle
and
superior
temporal
gyri
–
described
as
the
‘posterior
temporal
lobe’,
and
the
angular
gyrus
(BA
39).
Here,
only
results
significant
at
p