The analytic algorithms derived from non-linear deterministic models may be more sensitive to differences in physiological data than those based on linear stochastic models. Among the non-linear algorithms the time-dependent dimensional ones appear to be the most sensitive discriminators. In the present study dimensional responses were examined in both electronically and mathematically generated data and in high-resolution physiological data. The latter were event-related potentials (ERPs) recorded from the primary auditory cortex of cats during classical conditioning. Two techniques were found to lengthen and stabilize the linear scaling region in the correlation integral of the dimensional algorithms: (1) linking trials to increase data length; and (2) gain reduction to lower integer-values of noise, combined with algorithmic setting of slopes <0.5 to zero. Of the three dimensional algorithms examined, only the time-dependent Point Correlation Dimension (PD2i) showed low error rates when tracking the dimensional shifts in non-stationary generated data. This algorithm also uniquely distinguished between the conditioned and unconditioned physiological responses. The ERPs had corresponding PD2i's that were significantly different from each other as well as from their own randomized-phase surrogates. The brief dimensional reduction that follows a conditioned stimulus is interpreted to be related to 'cooperativity' among the underlying cortical neurons that contribute to its electrogenesis. Copyright (C) 1999 Elsevier Science B.V.
ASJC Scopus subject areas
- Neuropsychology and Physiological Psychology
- Physiology (medical)