Cognitive Psychology Series Part 5: Language & Communication

1. The Structure of Language

Linguists analyze language as a hierarchy of levels, each with its own rules and regularities. Understanding this structure is essential for understanding how the mind processes language.

1.1 Phonology: The Sound System

Phonology studies the sound patterns of language. The basic unit is the phoneme — the smallest sound that distinguishes meaning. English has approximately 44 phonemes; Hawaiian has 13; some languages of the Khoisan family have over 100 (including click consonants).

1.2 Morphology: The Structure of Words

Morphology studies how words are built from morphemes — the smallest meaningful units. The word "unhappiness" contains three morphemes: un- (negation) + happy (root) + -ness (nominalization).

1.3 Syntax: The Rules of Sentence Structure

Syntax governs how words combine into phrases and sentences. It is what allows you to understand that "Dog bites man" means something very different from "Man bites dog" — even though they contain the same three words.

Chomsky proposed that syntax is governed by a system of phrase structure rules and transformations that can generate an infinite number of grammatical sentences from a finite set of rules — what he called a generative grammar.

# Simple Phrase Structure Grammar Parser
# Demonstrates Chomsky's context-free grammar concepts

class SimpleGrammar:
    """
    A basic context-free grammar (CFG) demonstrating
    how finite rules generate infinite sentences.
    S -> NP VP
    NP -> Det N | Det Adj N
    VP -> V NP | V PP | V NP PP
    PP -> P NP
    """

    def __init__(self):
        self.rules = {
            'S': [['NP', 'VP']],
            'NP': [['Det', 'N'], ['Det', 'Adj', 'N']],
            'VP': [['V', 'NP'], ['V', 'NP', 'PP']],
            'PP': [['P', 'NP']]
        }
        self.lexicon = {
            'Det': ['the', 'a', 'every'],
            'N': ['cat', 'dog', 'student', 'professor', 'book'],
            'V': ['chased', 'read', 'saw', 'gave'],
            'Adj': ['big', 'small', 'clever', 'old'],
            'P': ['in', 'on', 'with', 'near']
        }

    def generate(self, symbol='S'):
        """Recursively generate a sentence from the grammar."""
        import random

        if symbol in self.lexicon:
            return [random.choice(self.lexicon[symbol])]

        if symbol in self.rules:
            expansion = random.choice(self.rules[symbol])
            result = []
            for s in expansion:
                result.extend(self.generate(s))
            return result
        return [symbol]

    def demonstrate_recursion(self):
        """Show how recursion creates infinite structures."""
        print("=== Generative Grammar Demonstration ===\n")

        # Generate several random sentences
        for i in range(5):
            sentence = ' '.join(self.generate())
            print(f"  {i+1}. {sentence.capitalize()}")

        # Show structural ambiguity
        print("\n=== Structural Ambiguity ===")
        print('  "I saw the man with the telescope"')
        print("  Parse 1: I [saw] [the man] [with the telescope]")
        print("    => I used a telescope to see the man")
        print("  Parse 2: I [saw] [the man with the telescope]")
        print("    => The man had a telescope; I saw him")
        print("\n  Same surface string, two different deep structures!")

        # Show garden-path sentences
        print("\n=== Garden-Path Sentences ===")
        garden_paths = [
            ("The horse raced past the barn fell.",
             "The horse [that was] raced past the barn fell."),
            ("The old man the boats.",
             "The old [people] man [=operate] the boats."),
            ("The cotton clothing is made of grows in Mississippi.",
             "The cotton [that] clothing is made of grows in MS."),
        ]
        for gp, explanation in garden_paths:
            print(f'  "{gp}"')
            print(f'    Parsed: {explanation}\n')

grammar = SimpleGrammar()
grammar.demonstrate_recursion()

1.4 Semantics & Pragmatics

Semantics deals with literal meaning — what words and sentences denote. Pragmatics deals with meaning in context — what speakers intend to communicate, which often differs from literal meaning.

2. Language Acquisition

How do children go from babbling infants to fluent speakers in just a few years? The answer to this question has been one of the most debated topics in all of cognitive science.

2.1 Nativism: Chomsky's Language Acquisition Device

Chomsky proposed that children are born with a Language Acquisition Device (LAD) — an innate module containing Universal Grammar that constrains the hypothesis space for language learning. The child doesn't learn grammar from scratch but rather sets parameters based on the input language.

Evidence for innateness:

• Universal milestones: All children follow the same developmental sequence (babbling at 6 months, first words at 12 months, two-word combinations at 18-24 months) regardless of language or culture
• Creolization: When children grow up hearing a pidgin (simplified contact language), they spontaneously create a full grammar — a creole — with complex structures not present in the input
• Nicaraguan Sign Language: Deaf children in Nicaragua, exposed only to a rudimentary sign system, collectively created a full, grammatically complex sign language within one generation
• FOXP2 gene: Mutations in this gene cause specific language impairments, suggesting a genetic component to language ability

2.2 Learning-Based Approaches

Alternative theories emphasize the role of input and general learning mechanisms:

2.3 The Critical Period Hypothesis

Eric Lenneberg (1967) proposed that there is a critical period for language acquisition — roughly from birth to puberty — after which full native proficiency becomes extremely difficult or impossible to achieve.

3. Language Processing

Understanding a sentence requires real-time integration of phonological, syntactic, semantic, and pragmatic information — all within a few hundred milliseconds per word. How does the brain accomplish this?

3.1 Speech Perception

Speech perception is far more complex than it appears. The acoustic signal is continuous (no pauses between words), variable (every speaker sounds different), and ambiguous (the same sound can map to different phonemes depending on context).

3.2 Parsing & Garden-Path Sentences

Parsing is the process of assigning syntactic structure to a sentence in real-time. The parser must make commitments about structure before the sentence is complete, which sometimes leads to errors:

Two major theories explain how the parser handles ambiguity:

3.3 Lexical Access

Lexical access — retrieving a word's meaning from memory — occurs in approximately 200 milliseconds. The mental lexicon contains roughly 50,000-100,000 words for a typical adult, yet retrieval is nearly instantaneous.

# Modeling Lexical Access: Cohort Model vs Interactive Activation

class CohortModel:
    """
    Marslen-Wilson's Cohort Model (1987):
    As each phoneme arrives, a 'cohort' of matching words
    is activated, then narrowed until only one candidate remains.
    """

    def __init__(self):
        self.lexicon = [
            'captain', 'capture', 'capsule', 'capable', 'capital',
            'catalog', 'category', 'catheter', 'castle', 'casual',
            'elephant', 'elegant', 'element', 'eleven', 'elevator',
            'electric', 'election', 'eliminate', 'elaborate', 'elastic'
        ]

    def process_word(self, spoken_input):
        """Simulate incremental word recognition."""
        print(f"=== Cohort Model: Recognizing '{spoken_input}' ===\n")

        for i in range(1, len(spoken_input) + 1):
            heard_so_far = spoken_input[:i]
            cohort = [w for w in self.lexicon
                     if w.startswith(heard_so_far)]

            status = "SEARCHING" if len(cohort) > 1 else "RECOGNIZED!"
            if len(cohort) == 0:
                status = "NOT FOUND"

            print(f"  Input: '{heard_so_far}' => "
                  f"Cohort size: {len(cohort)} [{status}]")
            if cohort and len(cohort) <= 5:
                print(f"    Candidates: {cohort}")

            if len(cohort) <= 1:
                break

        # Recognition point analysis
        print(f"\n  Recognition point: '{spoken_input[:i]}'")
        print(f"  Uniqueness point reached at phoneme {i}")
        print(f"  (Word recognized before hearing all phonemes!)")

model = CohortModel()
model.process_word('captain')
print()
model.process_word('elephant')

4. Language & Thought

Does language shape how we think? Or does thought shape language? This question — one of the oldest in philosophy and psychology — has produced some of the most fascinating research in cognitive science.

4.1 The Sapir-Whorf Hypothesis

Benjamin Lee Whorf, building on ideas from his teacher Edward Sapir, proposed that the structure of a language influences the habitual thought patterns of its speakers. This idea comes in two versions:

4.2 Inner Speech & Vygotsky

Lev Vygotsky proposed that language and thought have independent origins but become intertwined during development. Young children (ages 3-7) talk aloud to themselves while solving problems — what Vygotsky called private speech. This gradually becomes internalized as inner speech, the silent verbal stream that accompanies much of adult thought.

5. Neurolinguistics

The study of how language is implemented in the brain — neurolinguistics — has a history stretching back over 150 years, beginning with clinical observations of patients who lost specific language abilities after brain damage.

5.1 Broca's & Wernicke's Areas

5.2 Aphasia Types

5.3 Bilingualism & Cognition

Over half the world's population speaks two or more languages. Research has revealed that bilingualism has profound effects on cognitive architecture:

Both languages are always active in a bilingual's brain, even when using only one. Evidence includes cross-language priming (hearing a word in one language activates related words in the other) and language switching costs (brief delays when switching between languages). This constant juggling strengthens domain-general executive control.

6. Communication & Persuasion

6.1 Non-Verbal Communication

Albert Mehrabian's often-cited (and often misquoted) research suggested that in communicating feelings and attitudes, words account for only 7% of the message, tone of voice 38%, and body language 55%. While these specific percentages apply only to ambiguous situations, the broader point stands: non-verbal channels carry enormous communicative weight.

Robert Cialdini's research on persuasion identified six principles that exploit cognitive shortcuts in communication:

6.2 AI & Natural Language Processing

The quest to build machines that understand and generate human language has been central to AI since its inception. The history of NLP mirrors the evolution of cognitive theories of language:

# Evolution of NLP Approaches
# From rule-based to neural language models

class NLPEvolution:
    """
    Historical progression of natural language processing,
    showing how AI approaches mirror cognitive theories.
    """

    def __init__(self):
        self.timeline = [
            {
                'era': '1950s-1960s',
                'approach': 'Rule-Based / Symbolic',
                'cognitive_parallel': "Chomsky's generative grammar",
                'example': 'ELIZA (Weizenbaum, 1966) - pattern matching',
                'limitation': 'Brittle; cannot handle ambiguity or context'
            },
            {
                'era': '1980s-1990s',
                'approach': 'Statistical NLP',
                'cognitive_parallel': 'Connectionist / frequency-based models',
                'example': 'Hidden Markov Models for speech recognition',
                'limitation': 'Data-hungry; limited understanding of meaning'
            },
            {
                'era': '2010s',
                'approach': 'Deep Learning (RNNs, LSTMs)',
                'cognitive_parallel': 'Sequential processing models',
                'example': 'Google Neural Machine Translation (2016)',
                'limitation': 'Long-range dependencies; interpretability'
            },
            {
                'era': '2017-Present',
                'approach': 'Transformers / Large Language Models',
                'cognitive_parallel': 'Parallel processing with attention',
                'example': 'GPT, BERT, Claude - contextual understanding',
                'limitation': 'Grounding problem; no embodied experience'
            }
        ]

    def display_evolution(self):
        print("=== Evolution of NLP ===\n")
        for entry in self.timeline:
            print(f"  [{entry['era']}] {entry['approach']}")
            print(f"    Cognitive parallel: {entry['cognitive_parallel']}")
            print(f"    Example: {entry['example']}")
            print(f"    Limitation: {entry['limitation']}")
            print()

        print("  Key Question: Do LLMs truly 'understand' language,")
        print("  or do they merely simulate understanding through")
        print("  statistical pattern matching at massive scale?")
        print("  This remains one of the central debates in both")
        print("  AI and cognitive science.")

nlp = NLPEvolution()
nlp.display_evolution()

Exercises & Self-Assessment

Conclusion & Next Steps

In this exploration of language and communication, we've traced the cognitive architecture of humanity's most remarkable ability — from the physics of phonemes to the pragmatics of persuasion. Here are the key takeaways:

• Language is hierarchically structured: Phonology, morphology, syntax, semantics, and pragmatics form nested levels, each with its own rules and regularities
• Language acquisition is remarkable: Children master complex grammar by age 5 without formal instruction, prompting debates between nativist (Chomsky) and empiricist (Skinner, Tomasello) accounts
• The critical period hypothesis is supported by cases like Genie and deaf children acquiring sign language late — suggesting a biological window for grammar acquisition
• Language processing is incremental: The parser builds structure word by word, sometimes committing to wrong analyses (garden-path sentences) that require costly reanalysis
• Language influences thought: Weak linguistic relativity is well-supported — language shapes habitual patterns of color perception, spatial reasoning, and temporal thinking
• Language has specific neural substrates: Broca's area for production, Wernicke's area for comprehension — but the full picture is a distributed network
• Bilingualism strengthens executive control, demonstrating deep interactions between language and general cognitive ability