he spread the warm bread with socks
TRANSCRIPT
Spelling out the N400 integration hypothesis(reminder)
Semantic anomalies elicit increased N400 amplitudes.
He spread the warm bread with socks.
If the amplitudes of the N400 reflect the difficulty ofintegrating the meaning of a word to the existing semanticrepresentation, then semantic anomaly would have to be oneof the factors that affects the process of integration. How would that work?
The ease of integration would be affected by whether or not thecomposed meaning describes a plausible state of affairs in theworld. In order to evaluate that, you’d of course first first need toperform the integration.
John saw the happy rockthe property ofbeing a rock
the property ofbeing happy
the unique(contextuallydetermined)individual suchthat it has theproperty X
an event suchthat it is aseeing eventand it has x asits Agent and yas its Theme
John
the property of beinghappy and a rock
For our ‘happy rock’ example this would mean that formingthe intersection of ‘happiness’ and ‘rockness’ would beharder if no or few things are characterized by both ofthese properties (I.e., fall into the intersection).
Spelling out the N400 integration hypothesis(reminder)
Stimuli
Under the integration hypothesis, what would itmean to get different effects for ‘white’ and ‘sour’?
(Somewhat idealizing:) In both cases there should be no cases in thereal world of Dutch trains carrying the target property (sour or white).
So if N400 amplitude is modulated by how well the sentence meaningmaps onto how things are in the world, both should elicit similareffects.
However, if the effort of integration is further modulated by ourknowledge of what could be true under (some set of) standardassumptions about how the world works, then maybe the sour casesmight pattern differently.
Main result: identical N400 effects forthe two violations
Under the integration hypothesis, what would itmean to get different effects for ‘white’ and ‘sour’?
(Somewhat idealizing:) In both cases there should be no cases in thereal world of Dutch trains carrying the target property (sour or white).
So if N400 amplitude is modulated by how well the sentence meaningmaps onto how things are in the world, both should elicit similareffects.
However, if the effort of integration is further modulated by ourknowledge of what could be true under (some set of) standardassumptions about how the world works, then maybe the sour casesmight pattern differently.
Conclusion
“While reading a sentence, the brain retrievesand integrates word meanings and worldknowledge at the same time …
… it does not take longer to discover that asentence is untrue than to detect that it issemantically anomalous.”
Under the lexical access hypothesis ofthe N400…
… this result is explainable entirely in terms of clozeprobability: both white and sour probably have zeroprobability of occurring in the context below. Thereforeneither are primed by the preceding context. ‘Yellow’, onthe other hand would be heavily primed.
A prediction of the N400 integrationhypothesis
Any factor that affects the ease of semanticintegration should affect N400 amplitudes.
Complement coercionComplement coercion
The author began reading the bookThe author began reading the book
Complement coercionComplement coercion
reading /reading /writingwriting
The author began the bookThe author began the book
CoercionCoercion
• Coercion is behaviorally costly in a number of paradigms
(McElree et al., 2001; Traxler, Pickering & McElree
(2002).
• McElree et al. (2001):
Coercion is costlyCoercion is costly
Self-paced reading and eye-tracking measures on:a. The author was starting the bookstarting the book in his house... (Coerced)(Coerced)
b. The author was writing the bookwriting the book in his house... (Simple, Preferred)(Simple, Preferred)
c. The author was reading the bookreading the book in his house... (Simple, Dispreferred)(Simple, Dispreferred)
Not a general difficulty with aspectual verbsNot a general difficulty with aspectual verbs+ NP+ NP
• Traxler, Pickering & McElree (2002):
The boy started the fight after school today. (event verb & event NP)The boy saw the fight after school today. (neutral verb & event NP)The boy started the puzzle after school today. The boy started the puzzle after school today. (event verb & entity NP)(event verb & entity NP)The boy saw the puzzle after school today. (neutral verb & entity NP)
costly costly →→
Not just ambiguityNot just ambiguity
Context SentenceContext SentenceThe student was readingreading all morning. (Relevant Context)(Relevant Context)The student was relaxingrelaxing all morning. (Neutral Context)(Neutral Context)
Target SentenceTarget SentenceAfter a while, he started a bookstarted a book about health care… (Coercing Target)(Coercing Target)After a while, he read a bookread a book about health care… (Control Target)(Control Target)
• Traxler, McElree, Seely & Pickering (submitted):
• Coercion cost is not eliminated or even attenuated by
lack of ambiguity.
• Can we identify a neural correlate of the
coercion cost?
• Does coercion elicit neural activity that is distinct
from the N400 generator?
The pizza was too hot to The pizza was too hot to drink drink ←← N400N400
The pizza was too hot to The pizza was too hot to eateat
QuestionQuestion
What to expect of the N400 in MEG?What to expect of the N400 in MEG?
• Helenius et al 1998:• Classic N400 paradigm in MEG.
• Source of the N400 localizes where the M350localizes: in the vicinity of the left auditory cortex.
(Helenius, P, Salmelin, E, et al. 1998. Distinct time courses of word andcontext comprehension in the left temporal cortex. Brain, 121, 1133-1142.)
MaterialsMaterials
Coerced: Coerced: the professor began the bookbegan the book before his evening tea
Implausible:Implausible:the professor disgusted the bookdisgusted the book before his evening tea
Control:Control:the professor read the bookread the book before his evening tea
• Control verbs determined by determined by a “fill-in-the-blank” test (The author began ___ the book).
• Sentences matched for length and frequency of the verb,and the coerced and control stimuli were matched forrated plausibility.
MaterialsMaterials
Coerced: Coerced: the professor began the bookbegan the book before his evening tea
Implausible:Implausible:the professor disgusted the bookdisgusted the book before his evening tea
Control:Control:the professor read the bookread the book before his evening tea
• 70 stimuli per condition
• Word by word presentation
• Sensicality judgment task
• 16 subjects
Sensicality Sensicality judgment datajudgment data
93%540AnomalousAnomalous
86%763ControlControl
86%757CoercionCoercion
AccuracyAccuracyRTRT
• Sensicality judgments easiest for anomalousstimuli.
• No difference between coerced and controlsentences.
All sourcesAll sources
Multi-source modelling
N400 effectN400 effect
Coercion effect, Anterior Midline Field (AMF)Coercion effect, Anterior Midline Field (AMF)
Where does the AMF come from?Where does the AMF come from?
-15
-10
-5
0
5
10
15
20
1 51 101 151 201 251 301 351 401 451
nA
m
COERCION NO COERCION
Mean AMF dipole inside a standard MRIMean AMF dipole inside a standard MRI
S1 S2 S10
Coercion
Control
Anomalous
AMF
Estimation of current source density at the time of the AMFpeak.
L
R
Where does the AMF come from?Where does the AMF come from?
• Ventromedial prefrontal cortex likely source.
• What do we know about this area?
– Contains secondary taste cortex & secondary and tertiary
olfactory cortices.
– Involved in controlling and correcting reward-related and
punishment related behavior.
– Decision making.
– Theory of mind.
– Not a traditional language area, but has been found to be activated
by linguistic stimuli in several previous studies as well (e.g.
Halgren et al., 2002; Marinkovic et al., 2003).
– Potential interface between language and other higher cognitive
skills.
Connections to Connections to orbitofrontal orbitofrontal cortexcortex
• The area receives direct input from:
– Inferior temporal cortex
– Superior temporal sulcus
• The AMF follows temporal activity by 20-50ms.
Goal
To study the localization of coercion operationsby studying them in Broca’s and Wernicke’saphasics. Backgroud assumption: Wernicke’s aphasics have a
“semantic” deficit -- therefore would be impaired oncoercion.
Two tasks
Complement coercion Coerced: The boy began the book.
Transparent: The boy began reading the book.
Aspectual coercion Transparent: The horse jumped over the fence yesterday.
Coerced: The horse jumped for an hour yesterday.
Results: complement coercion
Results: aspectual coercion
Alleged localization of coercion:Wernicke’s area
But our results indicate a frontal lobe source forcoercion. How to reconcile these results?
Possible answer: Wernicke’s area (roughly)provides the input for coercion. If the input isblocked, coercion is blocked.