Analyzing human random time generation behavior: a methodology and a computer program.

Abstract

In the present article, it is argued that in addition to the traditional random generation tasks discussed by Towse and Neil (1998), random time interval generation tasks should be considered as useful alternatives, because they allow a better empirical control of the executive task component in dual-task situations. First, a framework for discussing randomness over time is presented. Then, the article goes on to present three methods for the analysis of such tasks. A first method is based on the correlation between the intervals produced. The second method calculates the approximate entropy, and the third method converts the time sequences into binary sequences and estimates the statistical properties of the sequence on the basis of these binary data. A principal components analysis on 19 different measures based on 1,381 sequences produced in a number of single-task and dual-task experiments shows that the proposed measures form two general clusters, one related to output probability, perseveration, and alternation, and one related to sequential commonalities. The article also briefly describes a computer program that implements these methodologies.