the horse raced past the barn fell - when reach fell we realise our initial interpretation wrong, so have to go back and reanalyse.
1 of 22
what is garden path model?
autonomous 2 stage parser, initial attachments made on basis of syntactic info w/ no reference to what words mean
2 of 22
what are 2 stages of garden path model?
1. initial attachments made on basis of principles minimal attachment and late closure. 2. attachments confirmed/revised on basis of semantic/pragmatic info
3 of 22
Rayner and frazier 1987
compared eye movements where minimal attachment should or shouldn't result in gp. criminal confessed that his sins harmed many people. without that it would result in temporary gp with blip at harmed
4 of 22
Frazier and Rayner 82
compared reading times. since j always jogs a mile and a half this seems a short distance to him late closure). without 'this' late closure closed sooner so takes longer to read
5 of 22
Ferreira and Clifton 86
compared sentences where MA was or wasn't semantically anomalous. the defendant/evidence examined by the lawyer was unreliable. examined treated as main verb. subjects backed up in both cases.semantic info didn't prevent from being gp'd but solved qu
6 of 22
Mitchell 87
compared reading times where late closure was or wasn't anomalous. as child had visited/sneezed the doctor prescribed course of injections. doctor treated as object. both cases showed garden path
7 of 22
problems with garden path
no such thing as neutral semantic context, gp reflects need to make additional assumptions in order to interpret certain sentences, relies on arguments from null effects
8 of 22
more problems with garden path
influenced by punctuation and prosodic cues. suggested methods used not sensitive enough to detect interactions. doesn't take language differences in to account. don't always follow late closure
9 of 22
constraint based?
product of interaction between multiple constraints/sources of info (semantic, pragmatic). current interpretation is one most strongly supported by multiple constraints
10 of 22
when would gp occur according the constraint models?
if correct analysis has too little activation
11 of 22
how are constraint based theories different to gp?
constraint consider all info at same time, 1 stage model
12 of 22
altmann and steedman (88)
measured reading times. bulglar blew open the safe with the dynamite/new lock and made off with the loot. prior context mentioned 1 or 2 safes.
13 of 22
results?
if told 1 safe, sentence with dynamite easier. if 2 sentence with new lock easier. contextual info affects how build understanding
14 of 22
truesell tanenhaus and garnsey 94
sentence: the author/book that was read by the student was very hard to understand. compared reading times where animacy/inanimacy or initial np implied main verb. gpm predicts gp regardless of animacy
15 of 22
results?
eye movement data showed when semantic ocnstraints strong, reduced relatives were no harder to process than unreduced
16 of 22
trueswell tanenhaus and kello? 93
compared self paced reading times and eye movement data for gp's with verbs for which direct object reading is more or less likely. eg. general pretended/revealed the weapon was ready to be used. gpm would state that weapon treated as direct object
17 of 22
results?
subjects only gp'd by first kind of sentence (with revealed)
18 of 22
issues with constraint based model?
no evidence for parallel representations, gpm states second stage begins quicky so interactions observed are just 2nd stage. constraint fail to account for full range of data. doesn't explain why subjects sometimes appear to prefer certain interpreta
19 of 22
tanenhous et al 95
in reading experiments, context not immediately available and so must be retrieved from memory. looked at eye movements in response to gp's in context of visual array (put apple on towel in box)
20 of 22
results?
gp predicts on the gtowel is destination of apple until hear on the box. 1 referent on towel intepreted in line of GPM. 2 referents towel and apple hardly looked at empty towel. context can be used if immediately available
21 of 22
is constraint based model framework or theory?
interaction effects suggest gpm is wrong. constraint based poorly specified - could accoutn for any results. need to be developed so can make more precise predictions about when interaction effects will occur
22 of 22
Other cards in this set
Card 2
Front
what is garden path model?
Back
autonomous 2 stage parser, initial attachments made on basis of syntactic info w/ no reference to what words mean
Comments
No comments have yet been made