parsing

generating deep structure based on the surface structure, and in spoken language prosodic cues are used in generation

sentence comprehension model possibilities

>syntactic analysis precedes semantic analysis >semantic analysis precedes syntactic analysis >syntactic and semantic analysis is simultaneous >syntax and semantics are closely associated and interact

garden-path model

(Frazier and Rayner) >only one syntactical structure initially considered >selection governed by minimal attachment (priority) and late closure >not influenced by meaning

minimal attachment in the garden-path model

longer fixations on more complex sentences shows more difficult processing for this

late closure in the garden-path model

longer fixations on object of first clause when incoming words not attached to current phrase

constraint-based Model

(MacDonald et al.) >connectionist model of comprehension >all information sources assumed to be available immediately >interpretation with the most supporting evidence is chosen

resolving ambiguity in the constraint-based model

>constrained by grammatical knowledge >different word types associated with any given word are not independent from each other >word less ambiguous in some ways than in others >some interpretations more frequent than other based on past experiences

verb bias in the constraint-based model

verbs appear in some syntactic structures more often than others

strengths of the constraint-based model

>sensible to use all relevant information >predicts flexibility in comprehension by accounting for individual differences

weaknesses of the constraint-based model

>predictions are general >syntactic constraint findings can also be interpreted using the garden-path model

unrestricted race model

(Van Gompel et al.) >combines garden-path and constraint models >semantic and syntactic information used to identify syntactic structure >violation of active syntactic structure leads to re-analyse and revision

globally ambiguous sentences

meaning isn't constrained by semantics as both verb-phrase and noun-phrase attachments are plausible interpretations

verb-phrase and noun-phrase sentences

syntactically ambiguous but disambiguated by semantics

sentence type predictions for garden-path model

all sentences are globally ambiguous syntactically so take the same amount of time to process

sentence type predictions for constraint-based model

>semantically constrained sentences should be easier to choose the preferred interpretation for >faster to read than unconstrained sentences

sentence type predictions for unrestricted race model

>globally ambiguous sentences read fastest >verb- and noun-phrase sentences take longer as the wrong initial analysis may be chosen

strengths of unrestricted race model

>combines best element of other models >explains experimental evidence so far

weaknesses of unrestricted race model

>not all predictions tested >assumes advantage from ambiguity but not always true

good-enough processing account

(Ferreira et al.) >heuristic and algorithmic routes

heuristic route

>quick solutions >minimal effort >error prone

algorithmic route

>careful analysis of word relations, properties, and syntactic rules >may be impacted by individual differences

strengths of good-enough processing account

>acknowledges error-prone language comprehension >allows flexibility for processing

weaknesses of good-enough processing account

>nature of the rule of thumb are not well defined >heuristic processing may involve limited processing of semantic information >no empirical support that re-analysis improves comprehension

pragmatics

>figurative language processing (metaphors, idioms, irony, sarcasm) >common ground and cooperative principle

standard pragmatic model

(Grice) >literal meaning accessed, then decision made whether literal meaning makes sense in context >if not, reader searches for non-literal meaning

metaphor interference effect of standard pragmatic model

>longer time to decide whether literal meaning of metaphors are false than in control sentences (factually false) >figurative meaning of a metaphor can influence interpretation of literal meaning

prediction model

(Kintsch) >2 components- latent semantic analysis and construction-integration

latent semantic analysis component of the prediction model

word meanings based on relations with other words

construction-integration component of the prediction model

direction of statement important and used to construct interpretation

strengths of prediction model

>empirical support

weaknesses of prediction model

>doesn't consider the many types of metaphors >doesn't differentiate between novel and familiar metaphors

common ground

shared knowledge and beliefs possessed by a speaker and listener

cooperative principle

participants in conversation agree to cooperate by saying things appropriate to the conversation and not ending it unilaterally, but violations can be informative too

common ground vs egocentric heuristic

(Keysar et al.) >listeners use rapid, non-effortful egocentric heuristic rather than common ground

strengths of egocentric heuristic

listeners make simultaneous use of both egocentric heuristics and common ground

weaknesses of egocentric heuristic

>experimental situations highly artificial and not ecologically valid >common ground more common with familiar people than strangers >processing limitations may put a cap on common ground use

discourse processing

>inferences during comprehension >brief overview on larger text comprehension

local inferences

inferences that follow necessarily from the meaning of words

bridging/backwards inferences

conclusions drawn to increase coherence between current and preceding parts of discourse

elaborative inferences

based on our world knowledge that adds detail to discourse

predictive inferences

expectations of what will happen next in discourse

constructionist approach

(Bransford et al.) >relatively complete mental model of text is constructed >many elaborative references

minimalist hypothesis

(McKoon and Ratcliff) >inferences are automatic or strategic (goal directed)

automatic inferences in the minimalist hypothesis

>some establish local coherence >others rely on informative readily available as it is explicitly stated in text

strategic inferences in the minimalist hypothesis

formed in pursuit of goals, most often being made at recall

schema theory for larger texts

(Bartlett) >schemas strongly influential in remembering texts >lead to error in recall

constructionist-integration model for larger texts

>sentences turned to propositions alongside inferences >at representation level it's hard to distinguish what is stated vs inferred >surface, propositional, and situation representation levels

RI-Val model for larger texts

(O'Brien and Cook) >improves on construction-integration model >3 stages- activation, integration, validation

activation stage of RI-Val model

discourse relevant LTM activated

integration stage of RI-Val model

activated concepts integrated with WM content

validation stage of RI-Val model

connection made during previous stage are validation against relevant real-world information in LTM