Lecture 3 Notes: Syntax (Grammar Acquisition)

Syntax Acquisition Notes

Introduction

This lecture focuses on syntax acquisition, exploring how infants learn the rules for ordering words to create meaning. Key questions include whether humans possess an innate understanding of syntax or if it is learned through environmental interaction.

Core Readings
  • Sedivy (2014, 1st ed.), Chapter 6, Section 6.2 (pp. 200-206).
  • Sedivy (2020, 2nd ed.), Chapter 6, Section 6.2 (pp. 219-226).
Lecture Outline
  • W8: Perceiving the sounds of language
  • W9: Acquiring words and their meaning
  • W10: Syntax
  • W11: Developmental disorders of communication

Main Aims

  • Define syntax.
  • Review accounts of how infants learn syntax.
  • Discuss the nativist (innate understanding) vs. empiricist (learned from environment) perspectives.

Defining Syntax

  • Syntax involves conventions or rules for ordering words to change the meaning of an utterance.
    • Example:
      • "Mary loves John" vs. "John loves Mary"
  • A finite number of words and rules can generate an infinite number of sentences.
  • Syntax is generative, meaning it generates semantics (meaning).

Syntax and Apes

  • Kanzi (bonobo): Demonstrated word learning (symbolism, association) and some understanding of word order but with limitations compared to humans.
  • Kanzi's output was relatively poor despite extensive explicit training, suggesting syntax acquisition is not the same across species.

Syntactic Structures and Recursivity

Syntax involves a hierarchy of structures:

Examples of Sentences with Varying Complexity:

  • a: a chair and a bicycle
  • b: The chair is next to the bicycle.
  • c: Santa gave the productive professor a shiny new bicycle.
  • d: That is the bicycle that I like best.
  • e: The mechanic said that he repaired the bicycle.
  • f: John believed Mary to have stolen the bicycle.
  • g: John persuaded Mary to steal the bicycle.
  • h: The cyclist with the lovely bicycles was happy.
  • i: The cyclist who owned the lovely bicycles was happy.
  • j: The cyclist near the lovely bicycles was happy.
  • k: The cyclists in the Tour were fast, weren't they?

Examples of Syntactic structures:

  • Conjoined noun phrase.

  • Locative clause.

  • Double-object dative clause.

  • Cleft sentence with relative clause.

  • Sentence with sentence complement.

  • Object-raising construction.

  • Object-control construction.

  • Subject-verb agreement over a prepositional phrase.

  • Subject-verb agreement over a relative clause.

  • Subject-verb agreement over a less semantically integrated prepositional phrase.

  • Pronoun-antecedent agreement

  • By age 5, children typically achieve mastery of syntax equivalent to that of adults.

  • Recursivity: A key feature of syntax, allowing infinite meaning generation.

Key Features of Syntax

  1. Word order
  2. Syntactic categories: verbs, nouns, adjectives, prepositions, etc.
  3. Abstract rules: conjugation

Early Comprehension of Word Order

  • Hirsh-Pasek and Golinkoff (1993): 19-month-olds understand sentences like "Big Bird is tickling Cookie Monster".
    • Preferential looking paradigm: Infants looked longer at the video that correctly matched the sentence.
  • The ability to understand how word order creates meaning develops around 19 months.

Inference of Syntactic Categories

  • Mintz (2006): 12-month-olds can infer syntactic categories after a few examples.
    • Head-turn preference procedure used.
    • Infants distinguish between grammatical and ungrammatical sentences, with a stronger effect for verb frames.
  • Understanding that words fall into different categories is present around 12 months.

Abstract Rule Learning

  • Gomez and Gerken (1999): 12-month-olds demonstrate abstract rule learning.
    • Exposure to grammatical "sentences" (e.g., VOT PEL JIC RUD TAM RUD).
    • Test with new grammatical (e.g., PEL TAM PEL JIC RUD TAM) and non-grammatical sentences (e.g., PEL RUD JIC RUD VOT TAM).
    • Infants showed a familiarity effect, listening longer to grammatical sentences (6.716.71 sec vs. 4.614.61 sec).
  • Sensitivity to syllable sequence patterns emerges around 12 months.

Language Learning Approaches

  1. Vygotsky: Language mastery emerges through practical activities in the "zone of proximal development" within a social environment.
  2. Skinner: Children imitate and associate words, with positive and negative reinforcement shaping language use (Hebbian learning).
  3. Piaget: Language development is connected to cognitive development, with cognitive prerequisites influencing language acquisition.
  • These approaches are not mutually exclusive and likely contribute to language development to varying extents.
  • They focus on different driving forces and generally claim that language is learned.
  • Other approaches claim that language is "acquired".

Syntactic Development: Nativist Perspective

  • Chomsky: Argues for innate language abilities due to:

    • Poverty of the input: Children are rarely exposed to ungrammatical sequences or corrected for syntactic errors.
    • Children have a hard-wired ability to learn and process abstract rules.

  • ChatGPT vs. 3-year-old:

    • ChatGPT is trained on ~400400 billion words (~2020 billion sentences) and knows ~500,000500,000 words.
    • A 3-year-old has heard ~3030 million words (~1010 million sentences) and knows ~500500 words.
    • Chomsky argues that this disparity supports the idea of an innate language acquisition device.

Recursivity as a Key Issue

  • Recursivity is a hallmark of syntax, allowing infinite meaning generation.
    • Example: "The boy who saw the dog that chased the cat that scratched the girl left."

Critical Period Hypothesis

  • Chomsky acknowledges the importance of the linguistic environment but believes it primarily triggers an innate syntax-acquisition device during a critical period (early years of life).
  • The critical period hypothesis suggests a gradual decline in the ability to acquire syntax with age.
Evidence from Isolated Children

  • Genie: A case study of a child with severe language deprivation.

    • Genie's sentences lacked key syntactic elements, even after years of training.
    • Examples:
      • "Genie bad cold live father house."
      • "Father hit Genie cry long time ago."
      • "Genie have Mama have baby grow up."

  • Genie’s language age:

    • When found: 1-2y
    • After years of training: 2-3y

  • Genie's case suggests that language exposure after the critical period results in language learning (effortful, general-learning mechanisms) rather than automatic language acquisition.

Language Development Stages

  • 24 Months: Short MLU, few function words, single-clause sentences, present tense.
    • Example utterances: "Hot. This hot. This hot this time. Mommy, help me. This hot."
  • 36 Months: Longer MLU, more function words, multiple-clause sentences, various tenses.
    • Example utterances: "Mom fixed this for me and I don’t like it. Abe's gonna eat rest of it."

Compatibility of Genie's Case with Different Scenarios

Genie's case is compatible with the following scenarios:

  • Gradual decline in ability to acquire syntax
  • Critical period hypothesis
  • Early and sharp drop down to absolute inability to acquire syntax

Summary of Perspectives

  • Vygotsky, Skinner, and Piaget view syntax as an emergent property of social, linguistic, and cognitive input.
  • Chomsky emphasizes the need to explain syntax acquisition given impoverished input, viewing language as special.

Dominant View: Universal Grammar (UG)

  • Chomsky (1965): Children possess a biological endowment for language processing (Universal Grammar).
  • UG principles:
    1. Continuity of linguistic representations: Innate linguistic knowledge does not change qualitatively; expression evolves quantitatively with more input.
    2. Parameter setting: UG has parameters that vary across languages (e.g., word order).
      • Percentages of different word orders in languages:
        • Subject-Object-Verb: 44%
        • Subject-Verb-Object: 35%
        • Verb-Subject-Object: 19%
        • Verb-Object-Subject: 2%
        • Object-Verb-Subject: 0%
        • Object-Subject-Verb: 0%
    3. Modularity: Fodor (1981) proposes language is a stand-alone, human-specific faculty, encapsulated from the rest of cognition.
      • Evidence: Developmental Language Disorder (DLD) primarily affects language functions while leaving cognition relatively intact.

Alternative Views

  • Pinker (1984, 1987): Believes in an innate capacity for language but is open to additional mechanisms and that language is not that modular!
  • A theory of language acquisition must take into account a “language instinct” (Chomsky & Fodor), the linguistic input (Skinner & Chomsky), and the cognitive ability of the child (Piaget).
  • Semantic bootstrapping (Pinker, 1984): Children have preconceptions about links between roles (semantic) and grammatical position (syntactic).
Semantic Bootstrapping:
  • There are regularities in the observable world.
  • Examples: "The boy kicks the ball" and "The girls pull the rope".
  • Syntactic knowledge emerges from the ability to compute co-occurrences between language and the environment.
  • Agent-action-patient is equivalent to subject-verb-object.

Radically Different View

  • Seidenberg et al. (2002): Syntactic knowledge is not innate but represented in the input, which is richer than believed.
    • Examples:
      • Adjectives usually precede nouns.
      • Certain verbs predict certain nouns.
      • -ing is associated with the end of a verb.
  • Semantic bootstrapping (Pinker) can help but is not always necessary.

Statistical Learning and Algebraic Rules

  • Marcus et al. (1999): Innate mechanisms detect structure in the input, not to compute syntax but rather open-ended abstract relationships.
    • S = NP + VP + NP
    • This can be tested with simple rules of how syllables are organised in the input: XYX, XXY.

Is Syntax Simply Algebra?

  • Study: Exposure to XYX utterances (ga-ti-ga, li-na-li, etc.) or XYY (la-ti-ti, li-na-na, etc.) to 16 7-month-olds.
  • Test on new XYX utterances (bu-da-bu) and new XYY utterances (bu-da-da) with head-turn preference procedure.
  • Novelty effect indicated sensitivity to algebraic rules that could be the biological endowment Chomsky alluded to.

Summary

  • Language is based on rules difficult to learn explicitly from impoverished input.
  • Infants may have innate knowledge of linguistic computation (Chomsky, Marcus, Fodor).
  • The input is richer than originally thought (Pinker, Seidenberg).
  • Structure in language maps onto structure in the world (Pinker).
  • Acquiring syntax may require a device that can spot patterns and map them across domains, without being specific to language.