Status Report: Quantum Holography and the Brain

Pribram, Karl

In 1951, reviewing the state of our knowledge of auditory processes for Steven's Handbook of Experimental Psychology, Licklider ended with: "If we could find a convenient way of showing not merely the amplitudes of the envelopes but the actual oscillations of the array of resonators, we would have a notation (cf. Gabor 1946) of even greater generality and flexibility, one that would reduce under certain idealizing assumptions to the spectrum and under others to the wave form. .... The analogy ... [to] the position-momentum and energy-time problems that led Heisenberg in 1927 to state his uncertainty principle ...has led Gabor to suggest that we may find the solution [to the problem of sensory processing] in quantum mechanics."

During the 1970s it became apparent that Gabor's notation also applied to the cerebral cortical aspect of visual and somatic sensory processing. The most elegant work was done with regards to the visual system. A recent review by Tai Sing Lee (1996) in the IEEE casts these advances in terms of 2D Gabor wavelets and indicates the importance of frames and specifies them for different sampling schemes. For the monkey, the physiological evidence indicates that the sampling density of the visual cortical receptive fields for orientation and frequency provides an almost tight frame representation through over sampling.

Evidence from my laboratory indicates that the Gabor wavelet as recorded from the visual cortex will reduce to the spectrum and to the wave form under certain idealized conditions: electrical stimulation of the temporal lobe and frontal lobe cortices and the related basal ganglia. The 2D Gabor function achieves the resolution limit only in its complex form. Pollen and Ronner did find quadriture phase (even-symmetric cosine and odd-symmetric sine) pairs of visual receptive fields. However, to my knowledge, this is the only such report. In large part this is due to the lack of available techniques that has existed up to now. Currently, recordings made with multiple microelectrodes and data analysis with sufficiently powerful computers has remedied this situation and I hope to have some preliminary data to report on the conditions under which phase encoding might occur.

Another issue concerns the linearity of the sensory process. Suggestions have been made that the process is fractal rather than strictly linear. It may be that under some conditions non-linearity occurs. How pervasive are these conditions?

The neurophysiological community has come to terms with the distributed nature of what I have called the "deep structure" of cortical processing. The accepted view is that distribution entails the necessity of binding together the disparate sites of processing. Binding is accomplished by temporal synchronization and the emphasis has been that under the conditions which produce binding, no phase lead or lag is present. If, as I believe the evidence shows, the elements of the features of an image are already conjoined in a haphazard fashion in the receptive fields of sensory cortical cells, the issue of binding disappears. Instead, an active filter, a frame, acts to "capture" the relevant feature or combination of features. Capture can be implemented by the interference effects among wavelets. Should this be shown to be correct, Gabor's prediction that we might find the solution to sensory (image) processing in the formalism (and perhaps even in the neural implementation)