Cumulative Record. Chapter 11: Reinforcement Today. Quote 28

Fortunately, a statistical program is unnecessary. Most of what we know about the effects of complex schedules of reinforcement has been learned in a series of discoveries no one of which could have been proved to the satisfaction of a student in Statistics A. Moreover, a statistical approach is just wrong. The curves we get cannot be averaged or otherwise smoothed without destroying properties which we know to be of first importance. (p. 174)