Response Bias and the Generation of Random Sequences

Abstract

30 subjects attempted to produce random series of 300 responses, using the numbers 1 to 10. Digrams of successive responses were analyzed using two measures of randomness, the 1978 random number generation index of Evans and Ginsburg and Goldstein's 1987 cluster ratio. Three forms of bias were studied: repetition avoidance, series responding, and cycling. Only series responding showed a clear relation to measures of randomness. Counting forward and counting backward were both significantly correlated with departure from randomness. It was suggested that the achievement of randomness might be facilitated by using responses which do not involve an ordered set.