Abstract
Large neural systems are conveniently analyzed through a continuum model in which the differential element contains many cells and consequently near-neighbor synaptic connections can be neglected. At this level of detail, the constituent neurons can be replaced by figurative cells that are purely excitatory or inhibitory. The simplest continuum configuration that approximates the mammalian cortex consists of two layers each containing both species of cell, with a time delay in the interlayer synapses. If it is variable, the delay acts like a control, determining the wavelengths and frequencies of waves that preferentially grow out of a small disturbance in neural activity level. The amplified mode structure becomes more complex as delay increases and is always characterized by wavelengths significantly larger than typical synaptic connection ranges. As the favored modes grow and propagate they associate, through simultaneous activation, mutually distant continuum elements. The activation patterns are reproducible in wavelength spectrum, but, especially at large delays, their spatiotemporal forms are partly functions of the initial random noise. Since the individual continuum elements contain enough neurons to be capable of storing as much information as artificial neural networks, it can be argued that the delay-controlled waves provide a mechanism by which “memories” can be reproducibly recalled, and “creative thoughts” generated and developed.