ABSTRACT
Temporal patterns are sequences of spatial patterns. The effective dimensionality of the dynamics of the network can change up or down as the parameters a, e are increased, and what happens after some pixels have passed the first or later bifurcations is not predictable from the network's much simpler behavior at lower parameter values. A combination of nonlinear pixels and nonlinear networks is a recipe for processes about which the people cannot a priori intuit the outcomes. There is fundamental difference between the approach to psy-chophysics used, and that which is implicit or explicit in the search for the numerical values of the parameters of a specific psychophysical stimulus-response equation. The reader should reflect that environments that are not in general locally autoregressive are incomprehensible to senses, and ones which are sometimes, but not always, abruptly locally unpredictable are highly stressful.
