Introduction to Syntax

As a cognitive science, language is vital to how abstract notions are conceived and expressed.

The dominant model of syntax is Generative Grammar (by Chomsky et al.). There are 2 approaches:

• Principles and Parameters
  • Principles constitute language universals
  • Parameters constitute language variation
    • Subconsciously set parameters by listening to the language
    • We have not come up with many examples of params.
• Minimalism (more recent)

Language cannot exist independently of the human mind. Syntax is the study of our capacity of building/decoding phrases/sentences.

An example

• Who did Sally take the largest photo of?
  • Ben
  • #Sally took the largest photo of Ben
  • These two statements are NOT EQUIVALENT!!!

The fragment answer is valid, but the full answer is semantically not equivalent!! This is a quirk of English-speaking human brains.

Generative Grammar

Thesis: The subconcious operates a set of procedures yielding sentences. These procedures are called rules.

The method of studying syntax involves 3 steps in the cycle:

1. Gathering data
2. Generalizing information from data
3. Developing hypotheses

Hypotheses are proposed rules, which hold until proven otherwise by data. Hence hypotheses must be falsifiable.

No one hypothesis can predict correctness/wrongness of every sentence i.e. there are no universal hypotheses.

Whether rules actually exist or not is an ontological question; for our purposes it suffices to say that rules merely form a model of our psychology of language.

Example of scientific method

Anaphor: noun in the form of (pronoun)-self

1. Hyp 1: Anaphors must have an antecedent and agree in gender.
   1. Bill kissed himself.
   2. *Bill kissed herself.
   3. *Jane kissed himself.
   4. Jane kissed herself.
   5. *Kissed himself.
2. Hyp 2: Anaphors must agree in gender and number.
   1. The Joneses think themselves the best family on the block.
   2. *The Joneses think himself the wealthiest on the block.
   3. Gary and Kevin ran themselves into exhaustion.
   4. *Gary and Kevin ran himself into exhaustion.
3. Hyp 3: Anaphors must also agree in person.
   1. *The man believed myself the best at sports.
   2. *The man believed themselves the best at sports.
   3. The man believed himself the best at sports.

Corpora are insufficient as data sources for linguists. To make hypotheses useful, ungrammatical/non-well formed sentences must be used (falsifying purposes).

Knowledge of native language is subconscious. It allows us to perform the grammaticality judgement task based on our intuition.

Q: Can intuition be considered scientific data?

• It is a real psychological effect
• It is replicable under test conditions
• It is also known as native speaker judgement

Types of ill-formedness

• Syntatic ill-formedness: Form itself unacceptable
• Semantic ill-formedness: Form is ok, but not with the intended meaning!

Competence vs Performance

• Garden path sentences: grammatically correct statements that are hard to parse.
• Centre embedding:
  • Cheese mice love stinks
  • #Cheese mice cats catch love stinks
  • Stacking too many reduced relative clauses (not relative marker) makes it hard to understand
  • Though grammatical, indicates constraints on either/both:
    • short term memory
    • mental ability to break apart sentences

Competence: What we know about the language.

Performance: The actual kinds of language being produced.

Psycholinguistics is about the real-time processing of sentences, studies performance.

Syntax studies the forms of language itself, studies correctness.

Where do rules come from

Learning vs Acquisition

Learning: Conscious knowledge

Acquiring: Subconscious knowledge

Classes in formal grammar of a foreign language fail abysmally to train people but immersion in an environment allows for the subconscious to acquire.

Innate: Chomsky claims that many facts about Language itself is instinctual or innate.

i.e. you are born with some capacity to learn language.

• Universal Grammar (UG)
• Innateness “proof”:
  • Syntax is a productive, recursive and infinite system.
  • (?) Productive and infinite systems are unlearnable.
  • It follows that some parts of syntax must be unlearnable and thus innate.
• Example of why infinite systems are unlearnable:
  • understanding of language is based on mapping of representation to situation.
  • Since there are infinite stimuli/situations and infinite representations
  • The representation-situation mapping cannot be exhaustively defined.
  • Hence humans must already “know” some of this mapping from birth
  • This is the logical argument of Language Acquisition
  • The underdetermination of data

Grammar universal examples:

1. Distribution of word orders not equal;
2. Every language has subjects and predicates.
3. No language has “good; better; goodest”:
   1. AAA: tall; taller; tallest
   2. ABB: good; better; best
   3. ABC: bonus; melior; optimus
   4. *ABA ???

Explaining language variation

Grammars of languages differ because of innate parameters that select between variants.

example: Word order (V-S-O and variants)

Adequacy levels:

• Observationally adequate: accounts for corpora
• Descriptively adequate: accounts for corpora and native speaker judgements
• Explanatorily adequate: accounts for corpora, native speaker judgements and how children acquire language

Parts of Speech

aka Word Classes. Traditionally includes:

1. Lexical Categories: Open Class [can take on new coinages]
   1. Nouns (N)
   2. Verbs (V)
   3. Adjectives (Adj)
   4. Adverbs (Adv)
2. Functional Categories: Closed class [fixed set]
   1. Determiners (D)
   2. Prepositions (P)
   3. Conjunctions (CONJ)
   4. Tense (T)
   5. Negation (Neg)
   6. Complementizers (C)

The POS (part of speech) of a word is determined by its place in the sentence and its morphology. (Proof: you can identify which POS of a sentence with nonsense words replacing all lexical words).

Distributional criteria

1. (The/A) (teacher/prof) (teaches/helps) that class

Hyp 1: Grammar treats words from the same POS similarly.

Morphological distribution:

• derivational morphemes: affixes making words of other words often resulting in a different POS.
  • -tion
  • -al
  • -ally
  • -ed
• inflectional morphemes: they only attach to certain categories
  • -est has to attach to an adjective already, and doesn’t change it to another POS
    • big; biggest;
    • love; *lovest;

How robust are these POS correlated with open/closedness?

Subcategories

1. (The/A) (teacher/prof) (teaches/helps) that class
2. The teacher *shouts that class.

Hyp 1: Grammar treats words from the same POS similarly.

Hyp 2: Grammar treats words within the same subcategory from the same POS similarly.

Nouns:

• $\pm$PLURAL
  • book/books: $\pm$morphoPL
  • glasses: +morphoPL, -semanticPL
  • committee: -morphoPL, +semanticPL
• $\pm$COUNT
  • much/many
  • correlation with morphoPL
• $\pm$PRONOUN; $\pm$ANAPHOR
  • [+PRONOUN, -ANAPHOR]
  • [+PRONOUN, +ANAPHOR]
  • [-PRONOUN, -ANAPHOR]
  • ???? [-PRONOUN, +ANAPHOR]

Verbs:

• TRANSITIVITY
  • Number of required arguments
  • Ambiguity? “She helps me”; “Helps me”;
    • Type of required arguments
    • “Helps me”: Hidden subject? –> Imperatives have an implicit/understood subject “YOU”