The Language of Thought Hypothesis (LOTH) #
The Language of Thought Hypothesis (LOTH) posits that thinking occurs in a mental language, often called Mentalese. It shares several key features with spoken language:
- Words and sentences: Mentalese contains words that combine into sentences.
- Meaning: These words and sentences have meaning.
- Compositionality: Each sentence's meaning depends systematically on the meaning of its component words and their combination.
LOTH proposes that to believe a proposition is to hold an appropriate psychological relation to a Mentalese sentence expressing that proposition.
Key Core Commitments of LOTH Theorists #
- Representational Theory of Thought (RTT): Propositional attitudes (belief, desire, intention, etc.) are relations to mental representations with semantic properties.
-
"X believes that p iff there is a mental representation S such that X believes* S and S means that p." (Fodor, 1981)
-
- Compositionality of Mental Representations (COMP): Complex mental representations are composed of simpler constituents, and the meaning of a complex representation depends on the meaning of its constituents and their arrangement.
-
"intending that P&Q requires having a sentence in your intention box… one of whose parts is a token of the very same type that’s in the intention box when you intend that P, and another of whose parts is a token of the very same type that’s in the intention box when you intend that Q." (Fodor, 1987)
-
Logical Structure in Mentalese #
- Logically Structured Mental Representations (LOGIC): Some mental representations have logical structure. The compositional semantics for these representations resembles logically structured natural language expressions.
- Medieval theories used syllogistic and propositional logic, while contemporary theories employ predicate calculus.
Scope of LOTH #
- Primarily focuses on high-level cognition (reasoning, decision-making, planning).
- Applies to other mental phenomena like perception and navigation:
- Perception: Unconscious inferences based on mental representations.
- Navigation: Cognitive maps as mental representations of spatial layout.
Mental Computation #
- Computational Theory of Mind (CTM): The mind is a computational system.
- Classical Computational Theory of Mind (CCTM): The mind is a computational system similar to a Turing machine, with core mental processes as computations.
- Formal-Syntactic Conception of Computation (FSC): Mental computation manipulates symbols based on their formal syntactic properties, not their semantic properties.
-
"The mind is a ‘syntactic engine’." (Fodor, 1987)
-
Arguments for LOTH #
- Argument from Cognitive Science Practice: Our best cognitive science theories postulate computations over mental representations.
- Argument from the Productivity of Thought: The ability to entertain an infinite number of thoughts, despite finite cognitive resources, is explained by a finite base of Mentalese symbols combined through compositional operations.
- Argument from the Systematicity of Thought: Systematic relations among thoughts are explained by the fact that the same Mentalese words can be combined in different ways to create different sentences.
- Argument from the Systematicity of Thinking: Systematic relations among inferences are explained by the uniform applicability of mechanical operations over Mentalese symbols based on their logical structure.
The Connectionist Challenge #
- Connectionism: An alternative computational framework using neural networks rather than Turing-style models.
- Eliminative Connectionism: Neural networks are a replacement for classical computation.
- Implementationist Connectionism: Neural networks model how classical computations are implemented in the brain.
- Fodor and Pylyshyn's argument against eliminative connectionism: It does not explain the nomic necessity of productivity and systematicity in cognition.
Regress Objections to LOTH #
- Learning a Language: If Mentalese is learned through hypothesis testing, then we face an infinite regress of meta-languages.
- Solution 1: Concepts are innate, not learned (Fodor, 1975).
- Solution 2: Concepts are learned through means other than hypothesis testing.
- Understanding a Language: Understanding a Mentalese word would require representing its denotation, leading to an infinite regress.
- Solution: Ordinary thinkers do not represent Mentalese words as having denotations. Mentalese is a medium for thought, not an object of interpretation.
Naturalizing the Mind #
- Naturalism: Attempting to depict mental states and processes as denizens of the physical world.
- LOTH as a tool for naturalism: RTT and COMP provide a framework for explaining how mental representations acquire their semantic properties.
- Challenges: Naturalizing intentionality (the aboutness of mental states) remains an open problem.
- Semantically permeated view: Mental words have their denotations essentially, which may conflict with naturalistic ambitions.
Individuation of Mentalese Expressions #
- Neural Individuation: Mentalese types are individuated by their neural properties.
- Problem: Conflicts with multiple realizability (the same mental state can be realized by different physical systems).
- Functional Individuation: Mentalese types are individuated by their functional roles.
- Molecular functional individuation: Based on canonical relations to other symbols.
- Holist functional individuation: Based on total functional role.
- Semantically Permeated Individuation: Mentalese words are individuated partly through their denotations.
- Problem: Inherently meaningful mental representations may seem like suspect entities.
- Possible solutions: Consider mode of presentation, which requires further theoretical development.
Conclusion #
The Language of Thought Hypothesis remains a powerful and influential framework for understanding cognition. However, it faces ongoing debates and challenges concerning the nature of Mentalese, its acquisition, and how it relates to natural language and the physical world.