Language

Psycholinguistics

Study of the psychological and neurobiological factors that enable humans to acquire, perceive, produce, use, and comprehend language

Linguistics

language structure and change

Neurolinguistics

links between brain, cognition, and language

Sociolinguistics

links between social behavior and language

Computational linguistics

computational modeling of language

What is Language?

A system of communication composed of symbols that are arranged according to a set of rules (i.e. grammar) to form meaningful expressions

Symbolic

auditory sounds and visual icons are combined to represent objects, ideas, experiences, etc.
• The assignment of a symbol is arbitrar

Semantic

Symbolic combinations carry meaning

Infinitely Generative

A limited number of symbols can be combined an infinite number of ways

• Allows for the creation of new words – fungivorous; sunburstery; normie; on fleek; hangry

Recursive

• New ideas can be nested within a sentence

• Unlimited number of grammatical sentences are possible

Phoneme

Smallest linguistic unit of speech sounds

• Different languages use different phonemes

• English is made up of vowel and consonant sounds

Morpheme

Smallest linguistic unit with meaning

• root words, prefixes, and suffixes

Content Morphemes

Convey the bulk of meaning

Function Morphemes

The Lexicon

Hierarchical

A series of smaller components are combined to create larger units

• phonemes → morphemes → words → phrases → sentences → thoughts / ideas

Syntax

Arbitrary rules that govern how words/sounds are put together (grammar)

Semantics

The meaning of a word, phrase, or sentence

Denotation

Strict, literal definition of a word

Connotation

Emotional content — varies between individuals and cultures

Speech Perception

• Humans can process up to 50 phonemes per second

• Continuous flow of sound perceived as discrete items

• Influenced by experience/expertise

Coarticulation

We pronounce more than one phoneme at a time — each sound influenced by the next

Speech Segmentation

Often no pause between words in natural speech

Template/Feature-Matching

• Speech sounds broken into components

• Analyzed for patterns, matched to prototype/templat

Phonetic Refinement Theory

• Analysis shifts to higher processing levels

• Words identified by progressively narrowing phoneme matches

The TRACE Model

• Three levels: acoustic features, phonemes, words

• Uses spreading activation — lower and higher levels interact bidirectionally

• Cognition and context influence perception

Is Speech Processing Special?

• All models debate whether speech uses general or dedicated cognitive processes

• Evidence points to top-down influences from meaning and contex

Phonemic Restoration Effect

Categorical Perception

The McGurk Effect

Your other senses interact with audio

Auditory Pareidolia

• Hearing illusory patterns (such as words) in random noise

The Word Superiority Effect

Letters are identified more accurately when they appear inside a real word than in isolation or in a non-word string

Classic finding (Reicher 1969; Wheeler 1970):

Sentence superiority effect:

Unrelated words take 2× longer to read than words in a sentence — context speeds processing

Dyslexia

difficulty in phonological processing → impaired word identification

affected processes:

• phonological awareness

• word decoding

• lexical access speed

Brain bases in dyslexia

Reduced activation in left temporo-parietal cortex

General Process Theory

Special Process Theory

Noam Chomsky (1928 – now!)

Proposed the Language Acquisition Device

• An innate (cognitive) mechanism that facilitates the learning of language

• Argued that there is a universal course of language development

• Nature and nurture are necessary for language development

Surface Structure

The visible form of a sentence that can be parsed conventionally

Deep Structure

An underlying form containing the information necessary for meaning

Transformational Rules

The laws governing how surface and deep structures map onto each other

Innateness Claim:

• Essential components of grammar are innate — humans possess an inborn language schema

• Particular grammars are learned, but the capacity for grammar is wired in

Critical Periods in lang acquisition

Developmental period in which a skill is most easily acquired

• Infants can distinguish between and produce phonemes from ALL existing languages

• This ability is lost by 12 Months Old

Approach (Jamieson et al., 2022):

Train model on a large corpus of text (millions of sentences) to approximate human experience

• Model extracts statistical associations about word co-occurrence in context

LSA (Latent Semantic Analysis)

Most recent large language models (e.g., GPT series) perform remarkably well on language tasks using similar principles

The Distributional Hypothesis:

You shall know a word by the company it keeps" — meaning is derived from context

• Feed the model millions of text examples

• Model learns which words co-occur with which other words

• Representations emerge from statistical regularities — no hand-coded grammar

Lexical Ambiguity

Many words have multiple meanings: 'fast,' 'buckle,' 'clip,' 'sanction,' 'cleave'

• Context rapidly disambiguates meaning

Mental Models

• Internal representations of the situation described in text

• Next sentence can shift which mental model is active

Discourse & Macrostructure

• Working memory holds propositions, not just words

• Thematically crucial ideas (macropropositions) retained longer

Broca’s Area

• Language production

damaged and trouble speaking = Broca’s Aphasia

Wernicke’s Area

• Language comprehension

damaged and trouble understanding = Wernicke’s Aphasia