Stereotyped human behavior: A nonlinear dynamical analysis

Abstract

Some forms of stereotyped human behavior seem to occur randomly in time. A dynamical analysis of several topographies demonstrates that while such stereotypies have the spectral characteristics of random noise, the rate at which each subject exhibits his/her stereotypy is to some extent predictable and, unlike uncorrelated noise, prediction accuracy declines with increasing prediction interval. Rhythmic stereotypies appear to be more predictable than nonrhythmic topographies but both show a similar decline in prediction accuracy. Furthermore, the distribution of interresponse times exhibits self-similar behavior. These results point to a deterministic, rather than stochastic, origin for the variability of observed rates of stereotyped behaviors.