Lexical access

! The process by which the basic sound-meaning

connections of language, i.e., lexical entries, are
activated.
! How to find lexical access in the brain (of a healthy
human)?
A. Manipulate the presence of lexical access.
Condition 1: + Lexical Access
Condition 2: - Lexical Access
But setting up a situation where only the presence of
lexical access is manipulated is tricky because most
stimuli that even remotely resemble words activate actual
lexical entries.
B. Manipulate the difficulty of lexical access.
Condition 1: Hard Lexical Access
Condition 2: Easy Lexical Access
This manipulation is more straightforward to set up since
behavioral research has provided us with a detailed
understanding of what factors affect lexical access.
How to find lexical access in the brain?
! The ease of accessing a lexical entry always depends on
the level of activation in that entry prior to access.
! The pre-access level of activation primarily depends on two
factors:
The resting level of activation in the entry.
! Each lexical entry has certain resting level of activation.
! This resting level is primarily a function of how frequently you
access the word. Frequent access leads to a higher resting level.
Context (i.e., what other entries youve just accessed)
What affects lexical access?
Cognitive processes involved in
Cognitive processes involved in
lexical access
lexical access
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
Stimulus: TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition Competition
Spread of activation on the basis of meaning
! Frequent words are accessed faster than rare
words.
Comprehension: Faster lexical decision times.
Production: Faster naming times.
! Higher resting level allows them to reach a
threshold for recognition faster.
Effect of frequency
1 2 3 4 5 6
Frequency Category (Frequent -- Infrequent)
Behavioral Data: Reaction Time
Categories (n/Million):
1: 700
2: 140
3: 30
4: 6
5: 1
6: .2
1: number
2: ask
3: wheel
4: candle
5: clam
6: snarl
Effect of frequency on lexical decision
(Embick et al. 2000)
Context effects
! Repetition priming.
! Semantic priming
! Phonological priming.
Semantic priming
PRIME TARGET
Semantic priming
doctor
Repetition
! doctor (obviously) activates the lexical representation DOCTOR very
strongly.
! Because of this robust activation, a subsequent presentation of
doctor will show repetition priming for quite some time, even if
there are intervening words in the list.
Phonological priming
! Complicated. Effect depends on the stimulus
onset asynchrony (SOA)
SOA: interval between prime and target
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition
Competition
TURF
PRIME TARGET
TURF is presented before its activation starts to decrease due to
inhibition from TURN " Positive priming compared to unrelated
control (e.g., CLOCK - TURF)
TURF is presented after its activation has been suppressed by
TURN " No priming or even slower processing times than in an
unrelated control (e.g., CLOCK - TURF).
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition
Competition
PRIME
TURF
TARGET
A. Manipulate the presence of lexical access.
Condition 1: + Lexical Access
Condition 2: - Lexical Access
But setting up a situation where only the presence of
lexical access is manipulated is tricky because most
stimuli that even remotely resemble words activate
active actual lexical entries.
B. Manipulate the difficulty of lexical access.
Condition 1: Hard Lexical Access
Condition 2: Easy Lexical Access
This manipulation is more straightforward to set up
since behavioral research has provided us with a
detailed understanding of what factors affect lexical
access.
How to find lexical access in the brain?
Millisecond by millisecond MEG measurement of the brain
activity elicited by visual words in the lexical decision task:
CAT
0 200 400 600 800 1000
Time [msec]
Response
! What part of this activity is sensitive to
Frequency?
Repetition?
Semantic relatedness?
Phonological relatedness?
MEG components elicited by visual words
What is the time course of lexical access?
M100 M170 M250 M350
100-150ms 150-200ms 200-300ms 300-400ms
Averaged response to
visual words
100170 250 350
1st MEG
component
showing a
reliable effect
of frequency
1 2 3 4 5 6
Frequency Category (Frequent -- Infrequent)
Behavioral Data: Reaction Time
Categories (n/Million):
1: 700
2: 140
3: 30
4: 6
5: 1
6: .2
1: number
2: ask
3: wheel
4: candle
5: clam
6: snarl
Effect of frequency on lexical decision
(Embick et al. 2000)
M350 data collected during the same experiment:
1 2 3 4 5
6
Frequency Category (Frequent -- Infrequent)
Latency of m350 Component
Categories (n/Million):
1: 700
2: 140
3: 30
4: 6
5: 1
6: .2
1: number
2: ask
3: wheel
4: candle
5: clam
6: snarl
CAT
0 200 400 600 800 1000
Time [msec]
Response
! Finding: Left temporal activity around ~350ms is sensitive to
lexical frequency in the lexical decision task.
! What follows from this?
Either: the M350 reflects lexical access.
Or: the M350 reflects some process the follows lexical
access.
! If lexical access occurs faster due to high frequency, then all
subsequent processing is presumably speeded up, too.
! M350 could be related to the experimental task, I.e., it could reflect
the lexical decisions.
Is the M350 task-related?
! Pylkknen, Stringfellow, & Marantz (2002):
Set up a situation where activation is speeded up while
lexical decisions are simultaneously slowed down. Does
the M350 show a speed-up or a slow-down?
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
Stimulus: TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition Competition
Activation is Activation is
facilitated facilitated
Selection is Selection is
slowed down slowed down
Which way would Which way would
the M350 move? the M350 move?
Method: Manipulate the degree of competition by varying
the phonotactic probability of the stimuli
! Phonotactic probability: How common the sounds
and the sound sequences of the stimulus are.
! Words and nonwords with a high phonotactic
probability:
mile, pick, fan, line
sipe, tane, rean, cade
! Words and nonwords with a low phonotactic
probability:
house, lock, peep, town
theeg, yush, nirg, veige
Effect of phonotactic probability
! On the on the hand, stimuli that have a high phonotactic
probability activate the lexicon faster.
Just like frequent words are faster to process, frequent sounds are
also faster to process.
! On the other, stimuli that have a high phonotactic probability
activate lots of lexical entries.
This slows down selection (or in the case of a nonword, the decision
that the stimulus is not a word) since there are many alternatives to
consider.
! High phonotactic probability correlates tightly with the density
of the words similarity neighborhood.
Neighborhood density.
Effect of
Effect of
phonotactic
phonotactic
probability:
probability:
early facilitation
early facilitation
RT
Same/different task (low level)
Are these two stimuli the same or different?
RTs to words and nonwords with a high phonotactic
probability are speeded up.
High probability:
MIDE
RT
YUSH
Low probability:
Sublexical
frequency
effect
(Vitevich and Luce 1998, 1999)
Effect of
Effect of
phonotactic
phonotactic
probability:
probability:
later inhibition
later inhibition
RT
High probability:
MIDE
YUSH
RT
Low probability:
mile mild might
migrate mike mime
mine mire mind
mite migraine micro
neighborhood activated
yuppie yucca
yuck yum
neighborhood activated
Competition
effect
(Vitevich and Luce 1998, 1999)
Lexical decision (high level)
Requires searching through the lexicon.
RTs to nonwords with a high phonotactic probability
are slowed down.
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
Stimulus: TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition Competition
Facilitates Facilitates
activation activation
slows down slows down
selection selection
induces intense induces intense
competition competition
High High phonotactic phonotactic probability probability
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
Stimulus: TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition Competition
Then high probability/ Then high probability/
density should delay density should delay
M350 latencies M350 latencies
If M350 = Selection If M350 = Selection (or later processing) (or later processing)
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
Stimulus: TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition Competition
If M350 = Activation If M350 = Activation
Then high probability/ Then high probability/
density should speed up density should speed up
M350 latencies M350 latencies
JIZE, YUSH MIDE, PAKE Nonword
PAGE, DISH BELL, LINE Word
Low probability High probability
Four categories of 70 stimuli:
Lexical decision.
(Pylkknen, Stringfellow, Marantz, Brain and Language, 2002)
Materials (visual)
Materials (visual)
(Pylkknen, Stringfellow, Marantz, Brain and Language, 2002)
Effect of probability/density Effect of probability/density
100
200
300
400
500
600
700
800
M170 M250 M350 RT
High probability word Low probability word
n.s.
n.s.
*
*
Words
100
200
300
400
500
600
700
800
M170 M250 M350 RT
High probability nonword Low probability nonword
n.s.
n.s.
*
*
Nonwords
M350: (i) 1
st
component sensitive to
lexical frequency
(ii) not affected by competition
time
l
e
v
e
l

o
f

a
c
t
i
v
a
t
i
o
n
resting level
Stimulus: TURN
TURN
T
U
R
N
IP
T
U
R
F
T
U
R
T
L
E
Activation Activation
Selection Selection Competition Competition
M350: (i) 1
st
component sensitive to
lexical factors
(ii) not affected by competition
The M350 also is also sensitive to repetition and both
phonological and semantic relatedness, as one would expect
if it indexed the access to sound-meaning connections
(Pylkknen et al. 2006).
Further, the M350 is elicited in the auditory modality, as one
would expect if it indexed the access to modality independent
representations.
M350 localizations from Pylkknen
et al. (2006)
Localizations of M350 current generator
M350 localizations from Pylkknen
et al. (2006)
Location of electrode sites where
TSA was induced.
Intact: Repetition, syllable
discrimination, speech, naming and
word reading.
M350 vs. induced TSA
Boatman et al. (2000)
M350 vs. induced TSA
The localization of TSA and M350 localizations
suggest a similar degree of between-subjects
variance.