# Language of Thought The Language of Thought hypothesis (LOTH), developed by [[Jerry Fodor]] in his 1975 book *The Language of Thought*, proposes that thinking occurs in a mental language—sometimes called "Mentalese"—that has combinatorial syntax and compositional semantics like natural languages. On this view, mental representations are structured symbol systems: thoughts are built from primitive concepts combined according to rules, just as sentences are built from words combined by grammar. This explains the productivity and systematicity of thought: why thinking infinitely many new thoughts is possible, and why anyone who can think "John loves Mary" can also think "Mary loves John." LOTH is central to the [[Computational Theory of Mind]] and stands against connectionist/neural network approaches that represent knowledge in distributed patterns rather than discrete symbols. [[Steven Pinker]], [[Zenon Pylyshyn]], and others have defended the hypothesis, while critics like [[Paul Churchland]] and [[Patricia Churchland]] argue that brains don't work like symbolic computers. The debate continues in cognitive science: do our minds manipulate language-like symbols, or does cognition emerge from subsymbolic neural computation? Fodor later defended LOTH 2.0, acknowledging that central cognitive processes might work differently than modular peripheral systems. ## Language of Thought Structure ``` ┌─────────────────────────────────────────────────────────────┐ │ LANGUAGE OF THOUGHT HYPOTHESIS │ ├─────────────────────────────────────────────────────────────┤ │ │ │ NATURAL LANGUAGE MENTALESE │ │ ┌─────────────────┐ ┌─────────────────┐ │ │ │ Words │ │ Concepts │ │ │ │ (lexicon) │ │ (primitives) │ │ │ └────────┬────────┘ └────────┬────────┘ │ │ │ │ │ │ ▼ ▼ │ │ ┌─────────────────┐ ┌─────────────────┐ │ │ │ Grammar │ │ Combinatorial │ │ │ │ (syntax) │ │ rules │ │ │ └────────┬────────┘ └────────┬────────┘ │ │ │ │ │ │ ▼ ▼ │ │ ┌─────────────────┐ ┌─────────────────┐ │ │ │ Sentences │ │ Thoughts │ │ │ │ (infinite) │ │ (infinite) │ │ │ └─────────────────┘ └─────────────────┘ │ │ │ │ "John loves Mary" ───────▶ LOVES(JOHN, MARY) │ │ "Mary loves John" ───────▶ LOVES(MARY, JOHN) │ │ │ │ Same parts, different structure = different meaning │ │ │ └─────────────────────────────────────────────────────────────┘ ``` ## Key Properties | Property | Description | Implication | |----------|-------------|-------------| | **Productivity** | Infinite thoughts from finite primitives | Compositional structure needed | | **Systematicity** | Related thoughts come in clusters | If A thinks aRb, can think bRa | | **Compositionality** | Meaning from parts + structure | Thought content is compositional | | **Inferential coherence** | Valid inference preserves truth | Syntax mirrors semantics | ## Arguments For LOTH | Argument | Description | |----------|-------------| | **Productivity** | We can think novel thoughts | | **Systematicity** | Thoughts come in related families | | **Rational inference** | Valid reasoning requires structure | | **Concept acquisition** | Learning concepts requires hypothesis testing | | **Computational success** | Symbol systems work in AI | ## Arguments Against LOTH | Criticism | Source | |-----------|--------| | **Neural implausibility** | Brains don't look like symbol processors | | **Connectionist alternatives** | Distributed representations work | | **Grounding problem** | How do symbols get meaning? | | **Nativism required** | Where do primitive concepts come from? | | **Frame problem** | How does inference know when to stop? | ## LOTH vs Connectionism | Aspect | Language of Thought | Connectionism | |--------|---------------------|---------------| | **Representations** | Discrete symbols | Distributed patterns | | **Processing** | Rule-governed | Parallel constraint satisfaction | | **Learning** | Hypothesis testing | Weight adjustment | | **Structure** | Explicit syntax | Implicit in weights | | **Compositionality** | Built-in | Emergent (if at all) | ## Key Figures | Thinker | Contribution | | ----------------------- | ------------------------------- | | [[Jerry Fodor]] | Formulated LOTH (1975) | | Zenon Pylyshyn | Defended systematicity argument | | [[Steven Pinker]] | Applied to language acquisition | | [[Paul Churchland]] | Connectionist critic | | [[Patricia Churchland]] | Neuroscientific critic | ## Implications | Domain | Implication | |--------|-------------| | **AI** | Classical symbolic AI vs neural networks | | **Linguistics** | Relation to [[Universal Grammar]] | | **Psychology** | Concept structure, reasoning | | **Philosophy** | Nature of intentionality | | **Neuroscience** | What to look for in the brain | ## References - Fodor, Jerry. *The Language of Thought* (1975) - Fodor, Jerry. *LOT 2: The Language of Thought Revisited* (2008) - Pinker, Steven. "The Language of Thought" in *How the Mind Works* - https://plato.stanford.edu/entries/language-thought/ ## Related - [[Jerry Fodor]] - [[Computational Theory of Mind]] - [[Steven Pinker]] - [[Modularity of Mind]] - [[Philosophy of Mind]] - [[Cognitive Science]] - [[Noam Chomsky]] - [[Connectionism]] - [[Symbol Grounding Problem]]