Grand Rounds 5/15/2012
Department of Neurology
P Dr. John Shelley-Tremblay, USA Psychology
P I have no financial disclosures
P I discuss no medications nore off-label uses
of medications
An Introduction to the use
of Event-related Potentials
for the Study of Semantic
Memory
John Shelley-Tremblay
University of South Alabma Dept.
Of Psychology
Types of memory
P Within Long-term memory
< Episodic
< Semantic
P Why study semantic memory?
< Tells us about how the majority of our knowledge
is stored
< May help us to understand and remediate deficits,
as in Epilepsy, Alzheimer’s, Aphasia, and other
disorders with language deficits
< Tells us about an important part of the reading
process: Our goal is to read for understanding
Schematic Representation of the Reader
General
knowledge
Comprehension processes
Text representation
Integration
with prior context
Linguistic
system
Text representation
Phonology,
Syntax,
Morphology
Meaning and
form selection
Word representation
Word
identification
Ortographic
Units
Phonological
Units
Lexicon
Lexical
Meaning,
Morphology,
Syntax,
-argument
structure
-thematic
roles
Othographic
System
Mapping to
phonolgy
Visual input
From Perfetti (2000), Adapted from Levelt (1989)
Schematic Representation of the Reader
General
knowledge
Comprehension processes
Text representation
Integration
with prior context
Post - Lexical
Text representation
Linguistic
system
Phonology,
Syntax,
Morphology
Meaning and
form selection
Word representation
Word
identification
Ortographic
Units
Phonological
Units
Lexicon
Meaning,
Morphology,
Syntax,
-argument
structure
-thematic
roles
Othographic
System
Mapping to
phonolgy
Visual input
From Perfetti (2000), Adapted from Levelt (1989)
Semantic Priming
P Provides clues about the structure and
proceses underlying semantic memory
P Behavioral priming is the reaction time
advantage when responding to a target
when it is related to the prime
P Common Tasks
< Task 1: Word/non-word judgement
(lexical decision)
< Task 2: “Are these words related?”
(Semantic relatedness)
< Task 3: “Is prime in same category as target?”
(Category membership)
SOA = Stimulus onset asynchrony
Example:
SOA =
1000ms
0ms
“DOCTOR”
PRIME
1000ms
“NURSE”
TARGET
SOA =
1000ms
0ms
“TRUCK”
Response
Priming
1000ms
“NURSE”
Response
P Behavioral priming is the reaction time
advantage when responding to a target
when it is related to the prime
P Priming effects differ at short (<250 ms),
versus long (>750 ms) SOA’s
< Short SOA: AUTOMATIC PROCESSES
– Not under conscious control; use no “resources”
– Spreading activation (Meyers & Schvanneveldt,1971)
– Semantic memory as a network, any node must reach threshold
in order to be retreived for further processing
“DOCTOR”
“NURSE”
< Short SOA: AUTOMATIC PROCESSES
– Not under conscious control; use no “resources”
– Spreading activation (Meyers & Schvanneveldt,1971)
– Semantic memory as a network, any node must reach threshold
in order to be retreived for further processing
“TRUCK”
“DOCTOR”
“NURSE”
< Long SOA: CONTROLLED PROCESSES
– Under conscious control; conflict with other controlled tasks
– Expectancy generation; although controlled, is thought
to be pre-lexical (occurs before target’s meaning is
accessed);
– The generation of a possible related set of words after prime
presentation.
“SCALPEL”
“SURGEON”
“DOCTOR”
“HOSPITAL”
“NURSE”
Seems straightforward...but:
P Dagenbach, Carr, & Wilhelmson (1990)
“XXXXX” “DOCTOR” “XXXXX”
“NURSE”
Response
P Masking makes prime identification difficult
P When primes are not identifiable, pattern of
results changes
Negative Priming
P In addition to (+) priming, there is negative
priming as well.
< RT is slowed for a related prime, in comparison
to a neutral (unrelated) baseline
“XXXXX”
“XXXXX”“DOCTOR”
“DOCTOR” “XXXXX”
“XXXXX” “NURSE”
“NURSE”
“XXXXX” “TRUCK” “XXXXX”
Slowing Response
Response
“NURSE” Response
P How do we account for negative priming?
< Dagenbach et al. (1990) posit that prime
becomes only weakly activated in lexicon.
< Spreading activation would predict NO priming
for weak primes, no inhibitory mechanisms
“DOCTOR”
< Dagenbach et al. (1990)
suggest an alternative mechanism
“NURSE”
Attentional Center-Surround
Mechanism (CSM)
Dagenbach & Carr (1994)
< Lexical
Level Effect
ROCK
(+)
DOCTOR
(-)
NURSE
Retrospective Prime Clarification
Theory (RPC)
Kahan (1999)
P when a masked word is presented that is
difficult to identify
P then the target will be used to aid in the
identification of the prime
P A version of semantic matching;
occurs after meaning of prime and target are
processed
P Task: Is nurse a word? (*prime unretreivable)
Unrelated
“
“NURSE”
”
DOCTOR? NO
“
– Generate possible matches (few)
– Response = NO
”
HOSPITAL?NO
“
”
End search
– Relatively fast
Related
“
”
+
“NURSE”
HOSPITAL? NO SCALPEL? NO
BED PAN? NO
– Generate possible matches (many)
– Relatively slow
NO
P How does RPC account for negative
priming?
“
”
+
“NURSE”
P When prime is unclear, S uses target to aid
in its retrieval
P Subject matches task-relevant codes of the
target to those of the prime
P Greater number of partial matches resulting
in a slow down in decision time
P Competing Hypotheses:
< RPC: negative priming due delay in “matching”
process: POST LEXICAL
< CSM: negative priming due to inhibition of
target’s level of activation: LEXICAL
P Competing Hypotheses:
< RPC: negative priming due delay in “matching”
process: POST LEXICAL
< CSM: negative priming due to inhibition of
target’s level of activation: LEXICAL
P RT cannot easily differentiate these
processes
P Event-related potentials provide a window
into the sub-stages of mental processes
Event-related Potential (ERP)
P Necessary Components
< Stimulus delivery device
< Electrode Array
X1,Y1 X2,Y2 X3,Y3 X4,Y4
– Receives brain waves
< Amplifier
– Magnifies tiny (~5 - 100 υV) signals
< EEG recording computer
– W/ A to D sampling card
< Post-recording software
– Turns EEG into ERP’s
SUBJECT!
ERP Methods
P Collection of continuous EEG
1
2
3
4
1
2
P From multiple points on the scalp
NOSE
ERP Methods
TREE
DOG
CAT
DOCTOR
NURSE
(Preceded by “SHRUB”)
Functionally equivalent stimuli are combined.
TREE
CAT
TREE
NURSE
CAT
NURSE
Stimuli of the same type are combined
(related with related) (unrelated with unrelated)
P Individual segments are sorted by stimulus type and
averaged
P Event-related activity becomes magnified (because its
time-locked), while noise is reduced.
The N400 Component
P The N400 is a late negative
component, elicited by words and
legal non-words, which is smaller for
words which are semantically primed
than for those which are not (Bentin,
McCarthy, & Wood 1985; Kutas &
Hillyard, 1989).
P First discovered in complete
sentences, such as:
“She spread the warm bread with butter.”
N400 Response
+3.5µV
Congruent
Incongruent
-3.5µV
butter
1000ms
“She spread the warm bread with socks.”
N400 Response
+3.5µV
Congruent
Incongruent
-3.5µV
butter
socks
1000ms
Priming
N400 Response
+3.5µV
Congruent
Incongruent
-3.5µV
1000ms
P N400 also elicited by single words,
and is sensitive to priming by a single
prime word
P DOCTOR - NURSE (Primed)
P TRUCK - NURSE (Unprimed)
What does the presence of the
N400 component tell us?
(I.E., what is its functional significance?)
P N400 reflects lexical/semantic processing.
< N400 priming occurs at short SOA (Boddy,1986)
< N400 priming when word is masked to below
recognition threshold (Deacon, Hewitt, Yang, &
Nagata,2000)
< N400 priming when prime and target are being
actively ignored (Shelley-Tremblay & Deacon, in
preparation)
ERP Evidence for an Attentional
Center-surround Mechanism
John Shelley-Tremblay & Diana Deacon, PhD
City University of New York
P How can we produce weakly activated
primes without the trouble of masking?
– Barnhardt, Glisky, Polster, & Elam (1996)
P Experimental Design
< Training Phase: Learn new words (very rare)
< ERP Phase: new (rare) words serve as primes
< Final Recognition Test
P Training Phase
< Subjects presented with 104 novel
words
< One at a time for 7 seconds
< Subjects train until 50% recal criterion
Ready?
P Factors involved
Priming
Unprimed
Claver-Rock
Primed
Associate
Claver-Gossip
CHAT
CLAVER
Synomym
Claver-Chat
+
ROCK
CLAVER
P HYPOTHESES
< LH will show priming for synonyms, but reverse
priming for associates
– If N400 negative priming occurs (lexical level), CSM
is supported
RH
LH
< Lexical
Level Effect
CLAVER
(+)
CLAVER
CHAT
CHAT
(-)
GOSSIP
ROCK
GOSSIP
ROCK
< RH will show priming for associates, and may
show priming for synonyms (NO INHIBITION)
LH Synonyms
N400
F3
FZ
F4
T3
C3
CZ
C4
T4
T5
P3
PZ
P4
T6
O1
OZ
O2
Primed
+3.5µV
Unprimed
450 ms
LH Associates
N400
F3
FZ
F4
T3
C3
CZ
C4
T4
T5
P3
PZ
P4
T6
O1
OZ
O2
Primed
+3.5µV
Unprimed
450 ms
HEOG
RH Associates
VEOG
N400
F3
FZ
F4
T3
C3
CZ
C4
T4
T5
P3
PZ
P4
T6
O1
OZ
O2
Primed
+3.5µV
Unprimed
450 ms
RH Synonyms
N400
F3
FZ
F4
T3
C3
CZ
C4
T4
T5
P3
PZ
P4
T6
O1
OZ
O2
Primed
+3.5µV
Unprimed
450 ms
Conclusions
P N400 shows LH priming for synonyms (center)
P N400 shows LH reverse priming for associates (surround)
P In conjunction, this dissociation supports existence of
CSM in LH
P RH associative priming suggests a rich language
network with qualitative differences from LH
P *Excellent learners didn’t show reverse priming....
– Does it activate and dissipate very rapidly, or do
excellent learners lack this mechanism?
– Would adoption of good learners’ strategies lead to a
dissapearance of reverse priming?
Why ERP’s
How do ERP’s complement RT as a cognitive measure?
P Provides alternative source of evidence.
P Gives clues about the neural substrates
of cognitive processes.
N400 Subtraction wave for LH Associates
– What areas of scalp reveal most electrical activity?
P Allows for fine-grained analysis of time
course of proposed cognitive operations.
P Can be combined with neuroimaging to
determine sources of brain activity.
Priming Effect: Voltage Map
Unprimed
Primed
Clinical Application
Visual Evoked Potentian (VEP)
Origins of Dyslexia
Neurobiological Origin
Impairment of Visual System
Deficit in Magnocellular
Pathway (M-Pathway)
Dyslexia
M-Pathway in Visual System
P
P
P
P
One of three pathways in visual system
motion detection pathway
originates from the rod receptors located on the retina
very sensitive to low contrast and low spatial frequencies
Area MT
PATH Training
P Uses a series of grayscale
stripes that differ in width with
each trial.
P A target area in the shape of a
fish is located in the middle of
the background stripes.
P The stripes in the fish move left
or right; the child must
discriminate which direction the
stripes are moving in each trial
VEP RESULTS
E le c tr o d e : O1
1 2 .5
1 0 .0
10 . 0
7. 5
7 .5
5. 0
5 .0
2 .5
2. 5
FV
- 10 0 .0
0. 0
0.0
- 2. 5
FV
1 00 .0
2 00 .0
3 0 0.0
4 0 0.0
50 0 .0
- 10 0 .0
T im e 1 V EP .a v g
T im e 2 V EP .a v g
ms
1 0 .0
7. 5
7. 5
5. 0
5. 0
2. 5
-1 2 .5
50 0 .0
E l e ctr o d e : P O2
1 2 .5
1 0 .0
-1 0 .0
4 00 . 0
ms
E l e ctr od e : O2
1 2 .5
- 7 .5
3 0 0 .0
- 12 . 5
-1 2 .5
- 5 .0
20 0 .0
- 10 . 0
-1 0 .0
- 10 0 .0
1 00 .0
- 7 .5
- 7. 5
0. 0
0.0
- 2 .5
0 .0
0. 0
- 2 .5
- 5 .0
- 5. 0
FV
E l e ctro d e : O Z
12 . 5
2. 5
1 0 0. 0
20 0 .0
30 0 .0
40 0 .0
FV
5 0 0.0
- 10 0 .0
0. 0
0. 0
- 2. 5
1 0 0. 0
2 0 0. 0
- 5. 0
- 7. 5
-1 0 .0
-1 2 .5
m s
3 0 0. 0
40 0 .0
50 0 .0
Questions?