The language of thought thesis

Date of Award




Degree Name

Doctor of Philosophy (Ph.D.)



First Committee Member

Edward Erwin, Committee Chair


A foundational belief in cognitive psychology and cognitive science is that those mental processes that fall under the general heading of thought processes involve the manipulation of mental representations. Thinking is a computational process defined over mental states that have a combinatorial syntax and semantics. This internal representational system is considered by some to be a language of thought.As developed and promoted by Jerry Fodor and others, the claim that thought occurs in an internal language has been widely debated, but as yet, it remains undefeated by rivals. On Fodor's version of the theory, the inner language has four main characteristics: it is innate and inaccessible, it is sententially structured; propositional attitudes (e.g., beliefs, desires, hopes, thoughts, etc.) are relations to token symbols in the language; and it has the representational capacity of any possible natural language. A successful defense of the language of thought thesis requires a defense of each of these characteristics.This dissertation challenges Fodor's claim that the language of thought thesis is the best explanation of the cognitive capacities of human beings. Specifically, I question the claims that the language of thought is innate and that the proper treatment of propositional attitudes requires an internal language. Also, I argue that Fodor is unable to provide an adequate theory of content for the language of thought. While the inadequacy of Fodor's theory of content does not show that the language of thought thesis is untenable, it does cast doubt on the viability of the thesis.In the concluding chapter, I suggest that connectionist models provide an alternative that, although unproven and in development, is sufficiently plausible to defeat Fodor's claim to have the "only game in town." Rather than taking connectionism and classical cognitivism as rivals, it seems more plausible to hold that we have access to multiple cognitive architectures that employ various representational systems depending on the task at hand.



Link to Full Text