The Physical Symbol System Hypothesis

Definition

A physical symbol system consists of physical patterns ("symbols") that can be combined into larger structures ("expressions"), plus a set of processes that create, modify, reproduce, and destroy these expressions. Newell and Simon state the hypothesis directly: "A physical symbol system has the necessary and sufficient means for general intelligent action." It is offered not as a definition but as an empirical, falsifiable claim about what intelligence requires and what suffices to produce it.

In the Book

Newell and Simon build the hypothesis from two technical notions — designation (an expression can affect or behave differently depending on an object it points to) and interpretation (the system can execute a process that an expression designates) — then list closure requirements: any symbol can designate any expression, every process the machine can do has an expression that designates it, expressions are stable once created, and the system's expressive capacity is essentially unbounded. They trace its lineage through Frege, Whitehead and Russell's formal logic, through Turing's universal machine, and through early AI programs like Logic Theorist, arguing that a system meeting this description is "an instance of a universal machine." The chapter presents this as a law of qualitative structure — the same status as the cell doctrine in biology or plate tectonics in geology — and marshals evidence for both its sufficiency (programs that solve problems by symbolic search) and its necessity (human problem-solving protocols, like those in chess and cryptarithmetic, that show people navigating symbolically-represented "problem spaces" via heuristic search).

Why It Matters

The hypothesis converts "what is intelligence" from a philosophical puzzle into an architectural claim with testable consequences: if true, general intelligence anywhere — human, animal, or machine — must be implementable by symbol manipulation, and any sufficiently large symbol system can in principle be organized to be intelligent. This is the target that every other essay in the book's opening half either builds on (Newell & Simon, early GOFAI) or attacks (Dreyfus, Searle, the connectionists) — making it the load-bearing premise for a much larger argument about whether minds are, at bottom, computers running symbolic programs.