Sentence Processing

Human parsing mechanism

understanding the structure of the sentence by breaking it down into its components

Garden path sentences

a grammatically correct sentence that leads readers to an incorrect interpretation (a "dead end") based on its initial phrasing, forcing a re-evaluation to understand the true meaning

Garden path model

takes place in 2 stages

• 1st stage: only syntax no semantics

  • processor draws on syntactic information only

  • intitally driven by principles of minimal attachment and late closure

• 2nd stage: if the initial parse is incompatible with semantic info then the parse is revised

Constraint-based model

• different constraints mutually interact (syntax, semantic context, prosody, verb bias, frequency information)

• these constraints operate at every step

• eventually parsing reflects the structure most strongly supported by different sources of information

Garden path principle Late Closure

whenever possible, we prefer to attach new items to the current consitituent

Garden Path principle Minimal Attachment

new words are attached to the structure using the fewest nodes possible

• relates to tree diagram

Lexical ambiguity

the word has more than one meaning (fell, idle, etc)

What happens when late closure and minimal attachment conflict

minimal attachment goes before late closure

Self paced reading study taraban and McClelland

• evidence for interactivity

• violates the principle of minimal attachment

• semantic bias was found, meaning semantic information must play a role in how sentences are parsed

pros and cons of constraint based model

• pros

  • accounts for the role of factors other than syntax

  • evidence we can consider more than one syntatic analysis at a time

  • my account for inidividual differences

• Cons

  • difficult to falsify bc it doesnt make precise predictions

bilingual syntactic parsing

parsing is shaped by experience with each language

they show different parsing than native speakers

the attachment to preferences is not universal (english prefers late closure and spanish prefers earlier attachment)

Bilingual language experiences

their L1 effects how they view their L2 (or your more dominant language)

Language co-activation

• semantic relationships can mediate cross-language activation in bilinguals even without shared speech phonology

arguments for both models benefit

• garden path: effects that supposedly support constraint-based models arise because 2nd stage of parsing begins quickly

• Constraint: techniques used by garden path researchers are not sensitive enoguh to detect the role of semantics