Some Experiments on a Simple Thought-Problem
Eugene Galanter, Walter Smith · The American Journal of Psychology · 1958
The experiments to be reported here support an hypothesis about the relationship between thought and trial-and-error learning.' We conjecture that trial-and-error learning can be characterized by a stochastic process.2 Each response is independent of the sequence of preceding responses; the change in the probability of some response is determined solely by the outcome of the response that immediately precedes it. Thinking, by contrast, is recursive in character. A response on any given trial is determined by some sequence of preceding responses. The components of a responsesequence are considered to be perfectly determined, and are therefore independent at any given stage of the outcome of a particular response. This is not to say that the mechanism that generates the sequence of responses may not be probabilistically determined, but that a particular sequence of responses is normally describable by a recursive function rather than by a stochastic process.