Low-dimensional chaos in event-related brain potentials

Abstract

The quantification of a chaotic system, such as the nervous system, can be made by calculating the correlation dimension (D2) from a sample of the data it generates. The encephalogram was recorded from the vertex during an auditory "odd-ball" paradigm and was signal-averaged to reveal the event-related potentials (ERPs). A new method for continuously estimating D2, the "Point-D2" (PD2), was determined from the same data, and it also was signal-averaged. The PD2 method was found to be more accurate than others currently used for investigating finite data; it also was found to track nonstationarities that arise within the data. A significant (p < .001) PD2-decrease accompanied the ERPs evoked by target stimuli; the PD2 onset-latency and peak did not correlate with any ERPs. The very short latency suggests that the PD2-decrease may be associated with "selective stimulus set," an interpretation that has been related to early cortical ERP components.