Lecture 12: Neurocognitive models of language processing Flashcards
What is another name for serial models, and why does it have this name? How does the processing work?
Processing goes in one direction and each one of these processes is autonomous. Another name is “syntax-first models”, because syntax comes before semantics.
Phonological -> Lexical -> Syntax -> Semantics, context
What is meant by “late interaction” regarding serial language processing models?
Semantic and syntactic aspects are only compared in the end of the serial processing.
Garden path sentences (sentences that are incorrectly parsed at first, and then only correctly read upon re-evaluation) are a very good example to show this.
Example: The horse raced past the barn fell.
Your brain has found the most likely syntax choosing the tree with the least nodes using minimal attachment. Your brain uses that syntax to determine the semantics When the semantics are incorrect, the sentence is evaluated using a new, less probable syntax.
What is “Late closure”?
Words attach to the most recent words.
Example: “You said you would stop in the car.”
-> “You said [you would stop in the car].”
-> “You said in the car [you would stop].”
What is another name of “parallel models”? Why do they have this name?
In this model, processing goes in all directions, and modules have access to all information. This is also called an “early interaction” model.
What are “constraint-satisfaction models”?
Constraint-satisfaction models are serial or parallel models with probability and frequency being taken into account.
Explain ‘Phase 1: Local structure building’ of Friederici’s model.
Phase 1: Local structure building (ELAN -> 150-200 ms)
○ What kind of word are there?
Beginning of the experiment:
Participants were shown sentences, that were either grammatically correct, or had an extra preposition.
When participants read the incorrect sentence, Friedrici measured ELAN (Early LAN).
This was proof of syntax coming before semantics: the identification of a word category (syntax). If the syntax is violated at this point, the semantics of the sentence will not be re-evaluated.
If there are no problems during word category detection, we move onto phase 2.
Explain ‘Phase 2: Semantic and syntactic/thematic relations’ of Friederici’s model
Phase 2: Semantic and syntactic/thematic relations(LAN -> 300-500 ms / N400)
○ Does this sentence make syntactic and semantic sense?
Beginning of the experiment:
Participants were shown sentences, that were either grammatically correct, or had an extra preposition.
This phase detects morpho-syntactic violations (LAN). It is also the phase in which the working memory can interact with language.
If the sentence doesn’t contain syntactic violation but does not make sense semantically, you would detect N400.
WATCH OUT: If both would be the case (semantic AND syntactic violation), N400 would not occur because the syntactic evaluation comes before the semantic one and gets stuck there.
Explain ‘Phase 3: Syntactic integration’ of Friederici’s model
Phase 3: Syntactic integration
○ Re-evaluating morpho-syntactic violations (P600)
○ Interaction between semantics and syntax, making this part an interactive (parallel) model (whereas the rest of the model is serial).
○ Second parsing phase, first one is automatic in phase 2
What are the interactions between prosody and syntax/semantics in friederici’s model?
- Interacts with syntax as an early interaction (phase 2)
- Interacts with semantics (phase 2)
- Interacts with syntax again as a late interaction (phase 3)
What does “Prosody incongruent” mean?
When a sentence had a period but it was pronounced as if it continued, or the other way around.
Give a proof for early interaction between syntax-prosody.
- EEG of participants was recorded when shown correct sentences and sentences with only incorrect prosody.
- Found a negative peak at 300-500 ms in either left or right hemisphere
- Then, presented participants with sentences having syntactic violations as well, so both at the same time
- Elicited a bilateral (in both hemispheres at the same time) negativity
Give a proof for late interaction between syntax-prosody.
- Participants were shown sentences with statistically likely and unlikely syntax (Note: both syntax forms are correct, but one is more likely than the other. This is why it’s late syntax interaction)
- These sentences either had regular or irregular prosody.
- Because this is a case of late interaction, P600 was measured. For BOTH irregular sentences, P600 was bigger than the regular ones
Give a proof for interaction between semantics-prosody.
- Presented participants with sentences that were stressed and were either correct or had a semantic violation
- Both regular and irregular stress elicit N400, but the irregular pattern has a greater amplitude
- In other words: regular prosody makes the strange semantics easier to integrate.
In which brain hemispheres happen which components of prosody?
Prosody is processed mostly in the right side of the brain. The right side mostly processed the melodic aspect of prosody, while the left side processes anything to do with timing (so the rhythmic aspect).
What is the problem about the ELAN in Friederici’s model? Does it make Friederici’s model useless?
ELAN was found to be an artifact. It peaked after a syntactic violation due to an unnecessary preposition, which means that it is a reaction to that preposition, rather than to the syntactic violation, because the presence of “am” prepares participants to react to syntactically incorrect sentences.
-> This is a problem of experimental design, as two completely different things are being compared. Participants may develop an experimental strategy, as they see a pattern.
Does this make the model completely useless?
No:
ELAN is used to explain Phase 1, so Phase 1 is not valid, but, we can still use Phases 0, 2 and 3.
