Sentence Processing

Question 1

Syntactic parsing

Answer 1

Process of identifying and using cues to discover how words in a sentence relate to one another

Question 2

Phrase structure trees

Answer 2

Sentence: noun phrase (node 1) + verb phrase (node 2)

Question 3

Temporarily ambiguous sentences

Answer 3

• Take longer to read and process, lowers comprehension accuracy
• Increased processing load occurs at the disambiguating point
• Evidence for immediacy of lexical info, incremental sentence interpretation, and syntactic structure errors
• Can cause garden-pathing: the need to turn back and read the sentence again

Question 4

Garden-Path Model (2-stage model)

Answer 4

• (1) identify syntactic word categories and build structure
• (2) assess outcome against context and plausibility –> meaning is understood
• Processing heuristics: simple structure, minimal attachment, emphasis on main assertion

Question 5

Constraint-Based Model (1-stage model)

Answer 5

• Sentences represented as patterns of activation in neural network
• Parsers build on or activate all acceptable structures simultaneously (parallel processing), which compete for activation and then are ranked
• Verbs sometimes require an additional word – some verbs “do” things, some do not, which affects structural predictions

Question 6

Argument Structure Hypothesis

Answer 6

• Long-term memory only stores argument-related structural information
• Between 5-10 structures stored for each verb; all other syntactic structures are generated in real time
• Evidence: bias towards expecting the direct object after verb; ambiguous constituents are interpreted as arguments; comprehenders supply missing arguments
  Limitations: parsers may not always favor more likely structures over simpler structures

Question 7

Construal Account Model

Answer 7

• Parsing occurs in discrete stages (like garden-path model) but context influences which structure the parser prefers
• Applies to relative clause attachment ambiguities
• Competition predicts more difficulty when structures are not possible

Question 8

Race-Based Parsing Model (2-stage model)

Answer 8

• Parser builds multiple structures in parallel, structures do not inhibit one another; they race
• First structure to rise above activation threshold is selected and interpreted

Question 9

Good-Enough Parsing Model

Answer 9

• Parsing is not always necessary, especially when redundant with lexical information
• Predicts parsing errors when lexical information contradicts structural information

Question 10

Long-Distance Dependencies

Answer 10

• When words have close syntactic relationships, but are not adjacent in sentence (Gaps and Traces Hypothesis, Chomsky)
• Processed by associating fillers with gaps

Question 11

Dependency Locality Theory

Answer 11

• Processing cost increases with distance between a gap and its partner
• Parser predicts how many categories needed to complete a phrase –> predictions load working memory –> sentences with long distance dependencies involve more predictions than local dependencies