This repository contains a nicely-formatted version of Allen Newell's paper "You Can't Play 20 Questions with Nature and Win." If you study behavior—by whatever means: single-unit recordings, surveys, etc.—you should read this paper. Although the paper itself is quite old (Newell wrote it in 1972), it is still relevant today when we can run hundreds of subjects on MTurk or record from 10,000 neurons in mice.
Since the only version I could find was an ugly copy of the scanned original, I cleaned it up, in the vague hopes that more people might read it--and heed--it if they didn't have to squint.