The present article reviews experimental studies in the past 50 years on random number generation, and recent developments in behavioral variability research. The former approach, in an attempt to have people generate series of numbers as randomly as possible, has 3 problems: (1) the verbal instructions used, such as "be random," are ambiguous, (2) the participants' behavior is not adequately motivated, and (3) the randomness of the numbers generated by each participant is regarded as a stable trait of the participant, rather than as behavior of the participant. Recent behavioral variability research using operant conditioning procedures enables us to avoid these problems, and has produced evidence that random-like behavior can be learned. Results from this line of research may contribute to various other areas of psychology, such as modifying abnormally repetitive behavior, developing divergent thinking, exploring, and problem solving. But behavioral variability has at least two aspects: (1) "molecular" variability, including IRT and response topography, and (2) "molar" variability, including choice behavior, changes in play, going to various places, and building various kinds of forms. It is still unclear whether the same principle can explain both the molecular and molar levels of variability.
View full abstract