Jerry Fodor published “The Language of Thought” through Harvard University Press in 1975. The book set out the language-of-thought hypothesis, often nicknamed mentalese, which became one of the defining positions in cognitive science and a recurring reference point in debates about artificial intelligence.
Fodor’s argument is that any system whose behavior we explain in terms of computation must have an internal medium in which that computation is carried out. He concluded that thought itself is conducted in a structured, language-like symbol system that is innate, productive, and systematic: mental representations have parts that combine according to rules, so that grasping a few concepts lets a mind entertain an unbounded number of thoughts. On this view, reasoning is the rule-governed manipulation of these symbolic structures, not anything reducible to mere associations.
The hypothesis directly underwrote the symbolic, or classical, tradition in AI, in which intelligence is treated as computation over explicit symbol structures. It also framed a long-running dispute with connectionism, which holds that cognition emerges from distributed patterns in neuron-like networks rather than from a built-in symbolic code. That dispute is alive again as researchers ask whether large neural networks possess anything like the compositional structure Fodor insisted thought requires.
For a general reader, Fodor’s book is worth knowing because the symbolic-versus-connectionist tension it helped define still shapes how people argue about what today’s AI systems do and do not understand.