Cumulative Record. Chapter 31: The Processes Involved in the Repeated Guessing of Alternatives. Quote 3

The first guess in a series of five, as in the Zenith experiments, is apparently controlled by an abiding preference, by biased preliminary conditions, or by trivial circumstances which cancel out in the long run and are spoken of as “chance.” The second guess raises a different problem, for it is under the additional control of the first. (p. 455)