Article: Understanding sentences in context

• model: semantics first... then syntax, and then you understand. But its not so simple. 
• some words have more than one meaning, and the context of the sentences change the meaning of a sentence... and the background information you know about the author or the kind of writing.. the whole context effect the meaning!
• So, the major question is: when and how does all this get processed and put together?
• why is it advantageous to use ERP to study this?
• language processing happens quickly, so you need a technique that can keep up. ERP is very good for WHEN. its a speed issue.
• what ERP component reflects a surprise about meaning (semantics)?
• n400 - negative response at 400 ms after the word.
• so maybe 400 ms. is what it takes for us to recognize the meaning of a word.. or 150-200 ms?
• But syntax may also be processed by then.
• even just saying the first syllable of a word brings about the N400.
• but the N400 is not just an anomoly detection...
• ex. sack and promote. diagram. participants were not expecting to promote someone who did something wrong, but 'sack' would make better sense. 
• It turns out that the N400 was still involed in milder surprises... its the prior information that makes it a surprise.. its all about context.
• What determines the n400 strength?
• how things fit for what is being talked about right now.
• How soon does info about the speaker have an effect?
• about the same... 200-400 ms. Ex. a small child's voice saying "I like olives". We draw on stereotypes about children and likes and dislikes. Even this context can invoke the same response.
• we take those into account in less than  400 ms.. very early.
• " the linguistic brain seems more "messy" and opportunistic, taking any partial cue that seems to bear an interpretation into account as such an item.
• When does a listener show evidence/knowledge of ambiguity?
• ex. the lecturer..
• in the frontal area, the Nref effect: 300-400 ms. (in a different part of the brain than the N400). Here there is greater negativity when we have ambiguity vs. certainty.
• people can rapidly take into account previous context to reduce ambiguity.
• How can ambiguity about the identity of the person lead to syntactical surprises?
• ex. ambiguity about which lecturer... the lecturer that commited plagarism.. the lecturerer that got fired.
• This surprise brings out a p600 effect
• Garden path sentences in syntax first: wikibook
• ex.  because he always jogs a mile seems a short distance.
• a "rule" is ambiguous in a way, its deceiving
• you momentarily think there was a syntactic error, so it brings about a p600 effect.
• we think there are errors when there aren't any
• ex. david prsed linda because he => we think we are going to get more information about Linda, no david. This also brings about the P600 effect.
• So, syntactic stuff is also contextually based and is processed very rapidly.
• This suggests that we are dealing with multiple levels of linguistic structure simultaneously... not just syntax first or semantics first. Doing a lot really fast.
• So, how are we doing things so quickly?
• we anticipate and guess ahead of time. 
• sometimes that may lead us to errors that slow us down, but overall, our guesses are confirmed.
• ex. family safe behind a painting: expected vs. a bookcase...
• it occurs quicky, in these two instances we get different responses.
• "an oportunistic proactive brain at work"
• taking cues, not following rules, but making predictions.... this gives us speed.
• The author says we need tailed computational models to continue studying this.