Chapter 6: Sentence Structure and Processing
Chapter 6 Lecture Notes
6.1 Sentence Structure
Discuss how thematic roles are mapped onto syntactic positions such as subject and object.
Discuss processing issues involved in reversible and irreversible sentences.
Sentence Processing
Sentence Processing involves three levels:
Conceptual Level
Intended message
Syntactic Level
Structure of the sentence
Phonological Level
Spoken sentence form
Components of sentence comprehension and production:
Intended Message
Syntactic Structure
Spoken Sentence
Conceptual Level
Intended message is characterized as:
Event + Participants
Example: {RIDE, CLOWN, UNICYCLE}
Thematic Roles corresponding to various participants in an event include:
Agent:
Definition: The agent causes the event.
Example: CLOWN in the event of riding.
Patient:
Definition: The patient is acted upon in the event.
Example: UNICYCLE in the event of riding.
Syntactic Level – Assigning Roles
Syntax:
Definition: A set of rules for ordering words in a sentence.
Canonical Word Order: Typical sequence of sentence elements.
Example for English: Subject-Verb-Object (SVO).
Thematic Role Assignment: Mapping of thematic roles onto syntactic positions.
Example:
Agent → Subject → CLOWN
Patient → Object → UNICYCLE
Sentence structure formed:
CLOWN – RIDE – UNICYCLE
(where CLOWN is the Subject, RIDE is the Verb, and UNICYCLE is the Object).
Syntactic Level – Satisfying the Grammar
Breakdown of how to satisfy grammatical structures:
Subject:
Singular or plural? → Just one, so CLOWN.
Already mentioned? → Yes, so add "the" → CLOWN → The clown.
Verb:
Present or past? → Present, so RIDE.
Subject agreement? → Yes, so add -s → RIDE → ride-s.
Object:
Singular or plural? → Just one, so UNICYCLE.
Already mentioned? → No, so add "a" → UNICYCLE → The unicycle.
Final structure: The clown rides a unicycle.
Phonological Level
Syntactic sequence is converted into pronunciation:
Syllables and word boundaries come into play.
Example:
"rides a" → sounds like "ride-za."
Stress patterns:
CLOWN = emphasized
ride-s = pronounced as RIDE-z
UNICYCLE = U-ni-cy-cle
Prosodic Contour:
Starts mid, peaks at U, drops sharply.
Table 6.2 Sentence Tree Abbreviations
S: Sentence
NP: Noun Phrase
VP: Verb Phrase
D: Determiner
N: Noun
Pro: Pronoun
A: Adjective
V: Verb
PP: Prepositional Phrase
Triangle: Indicates that detail has been omitted.
Figure 6.2 Sentence Tree for "The clown rides a unicycle"
Tree breakdown example:
S → NP VP → D N → V D N
Components:
The
clown
rides
a
unicycle
Active and Passive Voice
Subject: What the sentence is about; focus of attention.
Active Voice: Agent maps onto subject.
Example: The clown rides the unicycle.
Passive Voice: Patient maps onto subject.
Example: The unicycle is ridden by the clown.
Reversible and Irreversible Sentences
Irreversible Sentence: Does not maintain sense if agent and patient swap subject and object positions.
Example: The clown rides the unicycle→ The unicycle rides the clown (?) (nonsensical).
Reversible Sentence: Maintains sense but alters meaning if agent and patient swap positions.
Example: The clown chased the lion → The lion chased the clown.
Reversible Passive: Reversible sentence in passive voice.
Example: The lion was chased by the clown.
Notably difficult to process, especially for young children or patients with aphasia.
6.2 Comprehending Sentences
Differentiate between constraint-based and garden-path models of sentence comprehension.
Explain the minimal attachment strategy, the predictions it makes, and the evidence testing it.
Discuss syntactic priming and the empirical evidence supporting it.
Parsing as a Means of Comprehension
Parse: Process of assigning words to a phrase structure.
Example: While Sarah bathed her, baby played on the floor.
First interpretation: Something wrong.
Reinterpretation needed.
Models of Sentence Comprehension
Sentence comprehension involves two processes:
Syntactic analysis of sentence structure
Semantic interpretation of sentence meaning
Two discussed models:
Garden path model
Constraint-based model
Garden Path Model of Sentence Comprehension
Garden Path Model:
A two-stage model comprising:
Syntactic analysis
Semantic interpretation
Utilization of heuristics (mental shortcuts) for quick syntax analysis:
Heuristic 1: Late Closure
Heuristic 2: Minimal Attachment
Garden Path Sentence: Deviates from expected structure, causing processing difficulty.
Example: While Sarah bathed, her baby played on the floor.
Questions arose: Did Sarah bathe her baby? Who played on the floor?
Heuristic 1 – Late Closure
Late Closure: A syntactic parsing strategy where one continues adding new words to the current structure until sufficient evidence suggests that a new structure should begin.
Examples of confusion that can occur:
"Because he runs a mile is nothing."
"We painted the walls with cracks."
"He knew her as a young boy."
"I told the girl the cat scratched Bill would help her."
"The cat returned home was hungry."
Work on rewriting these sentences to ease understanding.
Heuristic 2 – Minimal Attachment
Minimal Attachment: A strategy focusing on the simplest possible sentence structure.
Example: I shot an elephant in my pajamas.
Potential Confusion: Who was wearing the pajamas—the speaker or the elephant?
High Attachment: Attach prepositional phrase to the verb, leading to a simpler structure (the speaker wearing pajamas).
Low Attachment: Attach prepositional phrase to the object (the elephant wearing pajamas).
Notably, high attachment yields simpler structure.
Sentence Trees for High and Low Attachment
Figure 6.7 illustrates:
High Attachment
Simplified Interpretation.
Low Attachment
More complex interpretation with additional ambiguity.
Breakdown of sentence trees for analytic clarity: (a) High attachment:
NP
NP
S
VP
Pro
V
NP
PP
Testing Minimal Attachment
The Garden Path model indicates that high attachment is the default option.
Comic Effect Example:
One morning I shot an elephant in my pajamas. How she got into my pajamas I'll never know.
Context can influence listeners to opt for either high or low attachment based on syntactic preference:
High Attachment: The thief opened the safe with the stick of dynamite.
Low Attachment: The thief opened the safe with the rusty lock.
Constraint-Based Model
Constraint-based Model: Proposes that syntactic analysis and semantic interpretation occur simultaneously.
Syntax and semantics constrain one another based on intended meaning interpretation during comprehension.
Elements of Constraint-Based Approach include:
Syntax
Word Meaning
Context
Understanding Sentences: Context
Tanenhaus & Trueswell (1995): Utilize a visual world paradigm to assess contextual influence on interpretation.
Observation through eye movements indicating structural reinterpretation necessity due to new information sources. Understand how extra linguistic information contributes to understanding
ambiguous and non-ambiguous sentences.
Syntactic Priming
Syntactic Priming: Refers to a tendency to replicate previously encountered sentence structures.
Provides empirical support for the garden path model.
Observes a lexical boost, indicating syntactic priming increase when the verb is repeated in both prime and target sentences.
Anticipation & Cloze Probability
Anticipation: The likelihood that a person will finish a sentence with a specific word.
Example: I take my coffee with cream and…
Cloze probability of sugar is nearly 100%.
Anticipation and Visual World Paradigm
Processing delays noted (time intervals) as a function of predictive disambiguation.
ERP and Anticipation
N400: ERP component elicited after the presentation of semantically inconsistent stimuli.
Example: She spreads her toast with socks.
P600: ERP component elicited post-syntactic inconsistency.
Example: The broker persuaded to sell the stock.
6.3 Producing Sentences
Distinguish between serial, parallel, and incremental models of sentence production.
Discuss empirical evidence relating to planning scope during sentence production.
Models of Sentence Production
Serial Model: All processing at one step must complete before proceeding to the next step.
Parallel Model: Processing at one stage occurs simultaneously with other stages.
Incremental Model: Processing of one stage is concurrent with proceeding to the next stage.
Most evidence currently favors the incremental model indicating initiation of the sentence occurs prior to entire planning.
Visualization of Sentence Production Models
Figure 6.10 illustrates how sentence processing occurs in parallel, incrementally, and serially with emphasis on structure supremacy at conceptual, lexical/syntactic, and phonological levels.
Planning Scope
Inquiry into how far ahead individuals plan their sentences shows data discrepancies:
Content word, phrase, clause considerations.
Explanatory factors for inconsistencies include:
Experimental biases toward specific planning scopes.
Variability of planning scope contingent on processing demands.
Multifaceted structure within advanced planning.
Visual Attention & Sentence Production
Visual Attention observed sequentially via saccades and fixations, with subject focality observed preferentially.
Referential Priming: Prior exposure of an item prior to full display presentation leads to selection bias in picture description tasks.
Scrambling: Syntactic manipulation placing the object before the subject observed; used predominately in some languages (e.g., Russian, Chinese).
Sentence Production and the Brain
Dual Stream Model:
Ventral “What” Stream: Lexical selection facilitated through the temporal lobe.
Dorsal “How” Stream: Syntactic processing via parietal and frontal lobes.
Syntactic Priming: Previous exposure biases structures within sentence production tasks.
Repetition Suppression: Demonstration of reduced brain activity during the processing of syntactically primed sentences.
6.4 Learning Syntactic Structure
Processes to describe include:
Prosodic, syntactic, and lexical bootstrapping.
Differentiate between generativist and usage-based approaches in language acquisition.
Bootstrapping
Intonational Phrase Boundary: Prosodic cues indicating the termination of a syntactic phrase.
Characterized by pitch drop, lengthening of final syllbles; facilitates children’s detection of phrase boundaries.
Prosodic Bootstrapping: Children use prosodic patterns to discern syntactic structures.
Infants utilize pauses for this purpose; 3-year-olds are sensitive to cues even innately occurring drops in pitch.
Syntactic Bootstrapping: Leveraging syntactic information for word meaning inference.
Lexical Bootstrapping: Word meanings aiding in inferring syntactic structure.
Models of Syntax Acquisition
Generativist Approach:
Core premise: Syntax acquisition is driven by innate mechanisms.
Poverty-of-the-Stimulus Argument: Insufficient linguistic input for children to inherently learn language.
Language Acquisition Device: A hypothesized brain module containing universal grammar rules guiding language development.
Usage-Based Framework:
Presumes children utilize cognitive mechanisms such as pattern detection for building language comprehension incrementally.
U-Shaped Learning Curve
Illustrates differences in children’s mastery of plural and past tense inflections, transitioning from correct use to overgeneralization, and resolution of form distinctions.
Incremental Structure Building
Children notably say: “There he goes” but may resort to “He go now”; these discrepancies challenge straightforward rule-based understandings.
Collocation: Word sequences that frequently appear together (e.g., Brush your teeth).
Young children may utilize collocations but erratically form words in less frequent contexts (e.g., tooths, mouses).
Poverty of the Stimulus Revisited
Usage-Based Theorists posit adult language errors may enhance linguistic learning rather than detract (e.g., simplified English yes-no questions).
Primacy of Meaning
Children adopt syntactic structures relevant to their personal experience, with action verbs driving developmental syntax progression (SVO structure and embedding).
Passive voices often misinterpreted actively, underscoring the challenges in understanding syntax relevance based on contextual knowledge.