Spectral Density Maps of Receptive Fields in the Rat's Somatosensory Cortex

Abstract

To extend fmdings from visual neurophysiology we plotted responses for 48 locations in the somatosensory "barrel cortex" of the rat to spatial and temporal frequency stimulation of their vibrissae. The recordings obtained from bursts of spikes were plotted as response manifolds resembling field potentials such as those recorded with small macroelectrodes. The burst manifolds were shown to be composed of those obtained from single spikes, demonstrating continuity between two levels of analysis (single spikes and bursts). A computer simulation of our results showed that, according to the principles of signal processing, the somatosensotyreceptive fields can be readily described by Gabor-like functions much as in the visual system. Further, changes with respect to direction of whisker stimulation could be described in terms of spatiotemporal (vectorial?) shifts among these functions. As late as the 1950's, the structure of memory storage and the brain processes leading to perception remained enigmatic. Thus Karl Lashley (1950) could exclaim that his lifelong search for an encoded memory trace had been in vain, and Gary Boring (1929) could indicate in his History of Experimental Psychology that little was to be gained, at this stage of knowledge, by psychologists studying brain function. All this was dramatically changed when engineers, in the early 1960's, found ways to produce optical holograms using the mathematical fonnulation proposed by Dennis Gabor (1948). The mathematics of holography and physical properties of holograms provided a palpable instantiation ofdistributed memory and how percepts (images) could be retrieved from such a distributed store. Engineers, (e.g. Van Heerden, 1963) psychophysicists, (e.g. Julez and Pennington, 1965); and neuroscientists, (e.g. Pribram, 1966; and Pollen, Lee and Taylor, 1971) saw the relevance of holography to the hitherto intractable issues of brain function in meqlory and perception (Barrett, 1969; Campbell & Robson, 1968; and Pribram, Nuwer and Barron, 1974). However, this early promise failed, for a variety of reasons, to take hold in the scientific community. The fact that neurophysiologically the holographic spread function is limited to single, albeit overlapping, receptive fields (patches) was not recognized by psychophysicists who, therefore, spent considerable energy in disproving globally conceived distributed functions. However, engineers, e.g. Bracewell (see review, 1969), soon showed that such patch holography could and did produce correlated three-dimensional images when inverse transformed, a technique that became the basis of optical image processing in tomography. The application of this principle to the receptive field structure (Robson 1975) overcame the psychophysical problem. Further, it was unclear just how the principles involved in holography related to ordinary measures of brain physiology. For instance, the brain waves recorded with scalp electrodes are too slow to carry the required amount of infoffilation. Also, there seemed to be little evidence that the quadrature relation basic to perfoffi1ing a Fourier holographic transfoffi1 could be found in the receptive field properties of the cerebral cortex. Finally, there was considerable confusion regarding just what needed to be encoded to provide a neural holographic process. These objections have, to a large measure, been met. The nanocircuitry of neural microtubles provides an adequately high frequency wave form for microprocessing in synaptodendritic receptive fields (e.g. Hammeroff, 1987). Quadriture has been shown in receptive fields within columns of the visual cortex (Pollen and Ronner, 1980). And, encoding of coefficients of intersections among waves, not of waves per se, was shown critical to the process (Pribram, 1991). Despite this evidence, Churchland (1986), reflecting the received opinion of the neuroscience community, noted that: "the brain is like a hologranl inasmuch as information appears to be distributed over a collection of neurons. However, beyond that, the holographic idea did not really manage to explain storage and retrieval phenomena. Although significant effort went into developing the analogy (see, for eXanlple, Longuet-Higgins, 1966) it did not flower into a creditable account of the processes in virtue of which data are stored, retrieved, forgotten, and so forth. Nor does the mathematics of the hologram appear to unlock the door to the mathematics of neural ensembles. The metaphor did. nonetheless. inspire research in parallel modelling of brain function" (pp. 407-408). In the same vein. Arbib (1969) states: " ... we note that the Cambridge school of psychophysics (see Campbell, 1974 for an early review of their work) has psychophysical data showing that the visual cortex has cells that respond not so much to edges as to bars of a particular width or gratings of a certain spatial frequency. The cells of the visual cortex tuned for spatial frequency can be seen as falling into different channels depending on their spatial tuning. This might seem to support the contention that the brain extracts a spatial Fourier transfoffil of the visual image. and then uses this for holographic storage or for position-independent recognition (Pribram. 1971). However, there is no evidence that the neural system has either the fine discrimination of spatial frequencies or the preservation of spatial phase information for such Foqrier transformations to be computed with sufficient accuracy to be useful" (p. 134-135). This view has also made its way into the popular literature on the subject. For example. Crick (1994) states "This analogy between the brain and a hologram has often been enthusiastically embraced by those who know rather little about either subject. It is almost certainly unrewarding. for two reasons. A detailed methematical analysis has shown that neural networks and holograms are mathematically distinct. More to the point. although neural networks are built from units that have some resemblance to real neurons, there is no trace in the brain of the apparatus or processq; required for holograms." (pI8S). =; i( That such statements can be made in view of so much evidence to the contrary -see, for example, the volumes by Devalois and Devalois (1988) and by Pribram (1991) -shows that something basic is at odds between the received view and those who have provided 1he evidence for the alternate view. We believe that the failure of holographic principles to take hold in neurophysiology is due to what is held to be the cerebral processing medium: ensembles of neurons or overlapping (receptive) fields of synaptodendritic arborizations. The distinction is a sUbtlf. one and concerns the level or scale at whic~ 'f>' processing is conceived to take place. Ensembles of neurons operating as systems (the currerit ",J nomenclature is "modules"), communicating via axo~s, indeed have an important role to play: f~ri(, instance, in information retrieval as indicated by loc~lized clinical disabilities. Nonetheles, within modules, processing relies on distributed architectures such as those used in neural network simulations. It is our contention that, at this level of processing, the ensembles consist, not of neurons, but of patches of synaptodendritic networks. \Vhat is needed is a method for mapping the activity of the overlapping synaptodendritic receptive fields in such a way as to convince the scientific community that something like a holographic process is indeed operating at the synaptodendritic level. '1 Kuffler (1953), provided us with a major breakthrough when he showed that he could map pat'1hes of the dendritic field of a retinal gangliq~ ': cell by recording from its axon in the optic nerve. Kuff1~r's is a simple technique for making receptiv~", field maps, which is now standard in neurophysiology. lBy stimulating a receptor or a set of receptors' in a variety ofdimensions and using ,the density of unit:responses recorded from axons, a map of the' .' . .l;·l' functional organizarion of the synaptodentritic receptive field of that axon can be obtained. (See e.g. reviews by Bekesy, 1967 and Connor and Johnson, 1992 for somesthesis; and by Enroth-Kugel and Robson, 1966; and Rodiek and Stone, 1965 for vision). , . Experiments by Barlow (1986) and by Gilbert and Wiesel, (1990) have shown that sensory stimulation beyond the reach of a particular neuron's receptive field can, under certain conditions, change that neuron's axonal response. Synaptodendritic patches are thus subject to changes produced in a more extended field of potentials occuring in neighboring synaptodendritic fields. What is seldom recognized is that the Kuffler1technique maps relations among local field potentials occurring in such extended overlapping dendritic arbors. The axon(s) from which the records are being made, sample a limited patch of this extended domain of overlapping receptive fields. Recently, Varella (1993) called attention to this relationship by demonstrating the correlation between burst activity recorded from an axon and the local field potentials generated in the synaptodendritic receptive field of that axon. I The current study also aims to explore the r~lations among local field potentials by mapping receptive field organization using the Kuffler technique. The specific questions posed and answered in the affIrmative are 1) whether this technique can map the spectral properties of synaptodendritic receptive field potentials, and 2) whether such maps df receptive fields in the somatosensory cortex show properties of patch (quantum) holography (tha~ is, of Gabor elementary functions) similar tb those recorded from the visual cortex. Methods and Results The rat somatosensory system was chosen for convenience and because the relation between whisker stimulation and central neural pathways has been extensively studied (see review by Gustafson and Felbain-Keramidas, 1977). Whiskers were stimulate