Upload
erika-carr
View
217
Download
0
Tags:
Embed Size (px)
Citation preview
Neuromagnetic Evidence for the Timing of Lexical Activation: An MEG Component Sensitive to
Phonotactic Probability but Not to Neighborhood Density
Sarah LaszloThe Cognition and Brain Laboratory
For Psych 593SG
9.22.2005
Liina Pylkkänen, Andrew Stringfellow, and Alec Marantz
30 Second MEG
MEG is like EEG, but with magnetic fields instead of electrical current, much better spatial resolution and much more expensive equipment
+ + MEG
10 Minute MEG
Wherever there is electrical current, there is a magnetic field:
There is bioelectric currentin post-synaptic neurons,so there is a magnetic field associated with their firing.
10 Minute MEG
Wherever there is electrical current, there is a magnetic field:
There is bioelectric currentin post-synaptic neurons,so there is a magnetic field associated with their firing.
Just as we measure that bioelectric current with EEG,we measure the associatedmagnetic field with MEG
10 Minute MEG
MEG shares some of the pitfalls of EEG:- These magnetic fields are small and tend to cancel each other out, so we can only measure them when they are produced by a large population of neurons firing nearly simultaneously and in a serendipitous configuration
10 Minute MEG
And has additional pitfalls all of its own:- The magnetic field of the earth is many orders of magnitude greater than any signal made in the brain, so MEG facilities have to be magnetically shielded (adding to the $$ factor)
10 Minute MEG
The Payoff:-In addition to EEG quality temporal information, MEG can provide millimeter quality spatial information
10 Minute MEG
The Payoff:-In addition to EEG quality temporal information, MEG can provide millimeter quality spatial information
(Because the skull and skin do not distort the magnetic signal the way they do the electrical signal)
The Matter at Hand
The M350 MEG component- Left Temporal generator
- 300-450 msec peak latency
Pylkkänen et al, 2002
The Matter at Hand
The M350 MEG component- Left Temporal generator
- 300-450 msec peak latency
- Sensitive to many of the same factors as the N400
- NOT entirely homologous to the N400
- peaks earlier, narrower waveform
Pylkkänen et al, 2002
N400
Federmeier and Laszlo, in prep
Not Homologous, but Informative
Elbow-nudging, winking, sneaky, critical assumption of this paper:- Even though the M350 is not directly homologous to the N400, it is at least an early component of the N400, so knowing more about the M350 can inform theories of the N400.
The goal of the study
“ … to determine whether the M350 MEG response component reflects automatic
lexical activation or subsequent processing.”
The tumultuous past
There is no past in MEG research
*wink, wink*
There are however several ERP studies which have asked this same question about the N400
The tumultuous past
There is no past in MEG research
*wink, wink*
There are however several ERP studies which have asked this same question about the N400
AND there are behavioral tasks which can identify and discriminate between automatic lexical and controlled post-lexical processes
The tumultuous past
The Masked Priming Paradigm- Behavioral evidence suggests that semantic priming does not require awareness of the prime:
The tumultuous pastThe Masked Priming Paradigm
- Behavioral evidence suggests that semantic priming does not require awareness of the prime:
&&&
DOG
&&&
+
CAT
150 msec
10 msec
150 msec
500 msec
LEXICAL DECISION
The tumultuous pastThe Masked Priming Paradigm
- Behavioral evidence suggests that semantic priming does not require awareness of the prime:
Conclusion: Behavioral semantic priming must be an automatic process
The tumultuous pastThe Masked Priming Paradigm
- Behavioral evidence suggests that semantic priming does not require awareness of the prime:
Conclusion: Behavioral semantic priming must be an automatic process
Can semantic priming effects on the N400 also be elicited without awareness of the priming stimulus?
+
unprimed
primed
The tumultuous past
Can semantic priming effects on the N400 also be elicited without awareness of the priming stimulus?
If yes, the N400 must be automatic (lexical)
If no, the N400 must be controlled
(post-lexical) ?
The tumultuous past
The N400 is a controlled process
The N400 is an automatic process
The N400 DOESN’T show a semantic priming effect when primes are masked below perceptual threshold (Brown and Hagoort, 1993)
The tumultuous past
The N400 is a controlled process
The N400 is an automatic process
The N400 DOESN’T show a semantic priming effect when primes are masked below perceptual threshold (Brown and Hagoort, 1993)
The N400 DOES show a semantic priming effect when primes are masked below perceptual threshold (Deacon et al, 2000)
The tumultuous past
The N400 is a controlled process
The N400 is an automatic process
The N400 DOESN’T show a semantic priming effect when primes are masked below perceptual
threshold (Brown and Hagoort, 1993)
The N400 DOES show a semantic priming effect when primes are masked below perceptual
threshold (Deacon et al, 2000)
Deacon et al’s explanation:
-We threw out subjects who showed no priming effect for unmasked primes
The tumultuous past
The N400 is a controlled process
The N400 is an automatic process
The N400 DOESN’T show a semantic priming effect when primes are masked below perceptual
threshold (Brown and Hagoort, 1993)
The N400 DOES show a semantic priming effect when primes are masked below perceptual
threshold (Deacon et al, 2000)
Deacon et al’s explanation:- We threw out subjects who showed no priming effect for unmasked primes
- Brown and Hagoort used a between subjects design, preventing a similar exclusion system, and obfuscating the semantic priming effect they surely would have seen to masked primes with a similar system in place
The enlightened present (2002)
Let’s do away with this barbaric and uninterpretable masked priming procedure
The enlightened present (2002)
Let’s do away with this barbaric and uninterpretable masked priming procedure
New Paradigm:
Lexical decision with a twist
Lexical decision with a twist
Lexical decision:Participants see words and nonwords and push one button for ‘word’ and one button for ‘nonword’
CAT
+
CIT
Lexical decision with a twist
All stimuli vary simultaneously inphonotactic probability and phonological neighborhood density
Lexical decision with a twist
All stimuli vary simultaneously inphonotactic probability and phonological neighborhood density
High probability / density
Low probability / density
Word [bell line] [page dish]
Nonword [mide pake] [jize yush]
Lexical decision with a twist
High probability / density
Low probability / density
Word [bell line] [page dish]
Nonword [mide pake] [jize yush]
High phonotactic probability: /be/ is more likely than /dZi/High neighborhood density: more words sound like ‘line’ than dish
Lexical decision with a twist
All stimuli vary simultaneously inphonotactic probability and phonological neighborhood density
High probability / density
Low probability / density
Word [bell line] [page dish]
Nonword [mide pake] [jize yush]
*There were no high probability / low density or low probability /high density items
Lexical decision with a twist
All stimuli vary simultaneously inphonotactic probability and phonological neighborhood density
High probability / density
Low probability / density
Word [bell line] [page dish]
Nonword [mide pake] [jize yush]
*There were no high probability / low density or low probability /high density items
Lexical decision with a twist
WHY make up this complicated task?
Because high probability items facilitate lexical activation, but high density items inhibit lexical decision (Vitevich and Luce, 1998, 1999)
Lexical decision with a twist
WHY make up this complicated task?
Because high probability items facilitate lexical activation, but high density items inhibit lexical decision
Remember the goal of the study:
“ … to determine whether the M350 MEG response component reflects automatic lexical activation or subsequent processing.”
Lexical decision with a twist
If the M350 is a lexical component, it should onset earlier for high than low probability items.If the M350 is a post-lexical component, it should onset later for those same items because they are also high density.
The predictions:
Lexical decision with a twist
If the M350 is a lexical component, it should onset earlier for high than low probability items.If the M350 is a post-lexical component, it should onset later for those same items because they are also high density.
WHY make up this complicated task?
What a sophisticated and informative design!
Informative!
The behavioral results:- HIGH items were responded to more slowly in the lexical decision task
This is expected, because high density items inhibit lexical decision
Informative!
The behavioral results:- HIGH items were responded to more slowly in the lexical decision task
This is expected, because high density items inhibit lexical decision
The big question: Does M350 latency track RT for HIGH items?
Informative!
Peak latency is reduced for high probability words
Remember: high probability items facilitate lexical processing, but inhibit post-lexical
processing
Words Data
Informative!
Wow! Even though RT to HIGH items was LONGER, M350 latency was SHORTER!
Conclusion: The M350 must be an automatic lexical component
Informative AND thought-provoking
- Nope, sorry, I still think this design is overly complicated and / or entirely uninformative?
Informative AND thought-provoking
- Nope, sorry, I still think this design is overly complicated and / or entirely uninformative?
- You left out my favorite result! Why didn’t you talk about it??
Informative AND thought-provoking
- Nope, sorry, I still think this design is overly complicated and / or entirely uninformative?
- You left out my favorite result! Why didn’t you talk about it??
- I’m still unclear about aspect X of the MEG recording and optimistically hope you might be able to elaborate?