Urantia Book

Grupo de Aprendizes da Informação Aberta

Contact

Superior Index    Go to the next: Chapter 4

Print Files: A4 Size.

Book in Text Format (txt).

Chapter 3
Smell


The Nature of Consciousness
A Hypothesis

Susan Pockett
Original Book
Smell
    3.1  Electromagnetic patterns in olfaction
        3.1.1  Insect Olfaction
        3.1.2  Some slightly disturbing implications
        3.1.3  Mammalian Olfaction
    3.2  Summary

     This chapter asks the question "does the brain generate patterns in the electromagnetic field that correlate with olfactory sensations?". The answer is an unequivocal "yes."

     Olfaction, the sense of smell, is probably the most ancient of all the sensory systems. The first cells, arising at the dawn of life billions of years ago, very likely had a rudimentary capacity for sampling the chemical qualities of the environment. Certainly modern bacteria such as Escherichia coli possess finely tuned receptors for various environmental chemicals. They also have the means for transducing the stimuli, decoding, integrating and transmitting information about them and for generating appropriate behavioral responses. The earliest chemoreception systems in multicellular organisms can be found in coelenterates. With further evolution, chemoreception became the function of specialized cells grouped in specific areas of the epithelium. As early in the phylogenetic tree as molluscs and crustaceans, these groups of receptor cells project to structures called olfactory glomeruli, which are clustered in olfactory bulbs not unlike those found in more advanced vertebrates. In fact it has been persuasively argued that olfactory systems quite similar to those of contemporary insects and vertebrates were probably already in place 500 million years ago, conveying exquisite sensitivity to the identities of places, trails, individuals, prey, predators, mates, social groups and food (Hildebrand, 1995 [124]). Thus, if we are interested in the evolution of conscious perceptions, the olfactory system must be an excellent place to start.

3.1  Electromagnetic patterns in olfaction

     Lord Adrian (Adrian, 1950 [6]) was the first to postulate that classes of odors should be correlated with spatial patterns of neural activity in the olfactory receptor layer and olfactory bulb. The inference from Adrian's original hypothesis was that for each kind of odor, there should exist a unique spatial pattern of neural activity in the olfactory bulb. This pattern should be present when and only when a particular odor is present. In fact Adrian's prediction has been born out startlingly well for invertebrates and in line with our current hypothesis the firing patterns that correlate with particular odors turn out to be not merely spatial patterns, but spatiotemporal patterns. In mammals, however, the situation is somewhat more complex (which is perhaps predictable, considering that mammals have more complex minds than insects). The correlation in rabbits, for example, is not between spatiotemporal electromagnetic patterns and the properties of the stimulus, but between spatiotemporal electromagnetic patterns and the meaning the stimulus has for the animal.

     The overall anatomical design of the early processing stages of the olfactory system is remarkably similar in insects and in humans - only the names of the various structures are different. In both systems the point of contact with the environment is the olfactory receptor cells (ORCs). These are located in the epithelium of the antennae in insects, or the nose in humans, where each nostril contains 107 to 108 of them (Freeman, 1972 [91]). Generally speaking, each ORC reacts to a specific kind of odorant (although this specificity is far from absolute) and only a few molecules of odorant may be necessary to elicit a reaction. The ORCs reacting to any one odorant are spread more or less evenly throughout the area of olfactory epithelium (with some general grouping so that broad classes of odorant are more represented in some areas of epithelium than in others), so the overall outcome is that different odorants stimulate different patterns of ORCs in the antenna or nose. Thus, although smell is not a spatial sense in the same way that sight is, the initial representation of an odor stimulus in the olfactory pathway does have a spatial structure.

     In the second stage of the olfactory system, olfactory receptor cells project to regions of the brain called the antennal lobe in insects or the olfactory bulb in mammals. Here projections from the different odor-specific ORCs are collected into odor-specific groupings, in structures called olfactory glomeruli. In humans, each glomerulus receives about 104 receptor axons (Freeman, 1991a [97]). Thus there is a second, differently organized spatial mapping of odor information in the antennal lobe or the olfactory bulb. This mapping is odotopic, rather than somatotopic.

     Finally, there are projections from the insect antennal lobe or the mammalian olfactory bulb to the mushroom body in insects or to a series of primary cortical areas in mammals. Here the mapping takes on a temporal as well as a spatial aspect.

3.1.1  Insect Olfaction

     In locusts, three interacting phenomena have been observed in connection with the representation of odors in the mushroom body (Laurent, 1996 [160]):

     (1) When an odorant is puffed onto the locust's antennae, extracellular recordings from the mushroom body show oscillatory local field potentials that have a frequency of 20-30 Hz. These oscillations occur without a phase gradient over the whole mushroom body calyx and last for the duration of the odor puff or slightly longer. Within any one locust, the oscillations are at least macroscopically the same for all odors. Hence on a large scale they can not be regarded as carrying any information about the nature of the odor, only the fact that an odor is present.

     (2) Within these mass oscillations can be distinguished the second phenomenon. Odors are repeatably and reliably represented by the firing of particular ensembles of the neurons that project from the antennal lobe to the mushroom body. In other words, a particular odor elicits the firing of a particular group of projection neurons. Any given projection neuron can participate in several different odor-specific ensembles, but a certain group of projection neurons always fires in response to presentation of any given odor. This means that there is a spatial pattern of neuronal firing associated with each odor. However, as well as this spatial pattern, there is also a temporal pattern of neuronal firing associated with each odor. This is generated by the fact that not all the projection neurons which fire in response to a particular odor do so at the same time. Some fire mostly at the beginning of the odor presentation, some at the end, some only in the middle, some at the beginning and the end but not in the middle, and so on. These temporal firing patterns are consistent in each neuron for each odor (at the same concentration of odorant). In other words, any particular projection neuron always behaves the same way in response to one particular odor.

     The overall outcome is that any given odorant elicits a very specific spatiotemporal pattern of firing of the neurons projecting to the mushroom body (Laurent, Wehr, & Davidowitz, 1996 [161]). The temporal aspect of the pattern dictates that the spatial aspect of the pattern evolves over the course of the stimulus presentation.

     (3) Additionally, there is a third layer of complexity to the stimulus representation. It is likely that a neuron in the mushroom body will fire in response to synaptic input from a projection neuron only if this synaptic input occurs at a time when the mushroom body cell is at a stage in the mass field potential oscillation when it is relatively depolarised. Since not all of the action potentials in the projection neurons are phase-locked to the macroscopic 20 Hz oscillations in the mushroom body, not all of the action potentials in the projection neurons will induce firing of mushroom body neurons. The situation becomes somewhat circular here, because there is some evidence suggesting that the mushroom body oscillations themselves are likely caused, at least in part, by fluctuating synaptic input from the projection neurons. Cause and effect are difficult to distinguish in this complex situation, but the upshot is that because the effective input from projection neurons to the mushroom body evolves over the course of the stimulus presentation, there is a smaller-scale spatiotemporal structure within the envelope of the macroscopic 20-ish Hz oscillation which does carry repeatable and reliable information about the characteristics of the odor.

     This looks suspiciously as though what we are dealing with is a 20-Hz carrier wave, which could plausibly be identified with olfactory consciousness per se, plus a series of amplitude and frequency modulations of the carrier wave which covary with the contents of olfactory consciousness.

3.1.2  Some slightly disturbing implications

     But wait a moment. All this refers only to the humble locust. At the best of times, one can never be certain even that other humans experience things in the same way as one does oneself (and indeed the vicissitudes and misunderstandings of everyday life tend to suggest that they often do not). So, taking into account the fact that locusts are very small and don't actually look much like humans, it may be regarded as a moot point whether such insects can be allowed to claim even the most rudimentary form of our own highly evolved Consciousness. The relevance of the above results to our quest might reasonably be questioned.

     However, as we will soon see, remarkably similar experimental results have been obtained from rabbits, rats and cats. Such creatures are relatively large, pleasantly furry and much more often the subject of anthropomorphic children's stories than are insects (pace Jiminy Cricket). In short, they are animals to which one is altogether more inclined to accord the benefit of the doubt, consciousness-wise.

     This is mildly disturbing. If the olfactory system of locusts works in the same way as that of cats, and if we are willing to allow that cats possess, if not intellectual brilliance, at least simple consciousness, then we are forced to contemplate the idea that locusts may also have a similar consciousness to our own, at least when it comes to the basic raw feel of smell sensations. Of course because insects are so much smaller, there would be less of it in an absolute sense, this putative consciousness. So maybe we can still regard it as OK to squash mosquitos, because in squashing a mosquito one is only squashing a very tiny aliquot of a very simple sort of consciousness. Well, maybe it's OK to squash mosquitos (or to eat rabbits or cows) anyway, because that's just the natural way of the animal kingdom - everything kills and/or eats anything else that is not a member of its own species, if it can. That mosquito was eating us, after all ... . Before such outrageously unscientific ruminations go any further however, we should also admit the idea that this kind of spatiotemporal coding may be merely the pre-conscious activity of the nervous system. Perhaps there is more to consciousness than low frequency spatiotemporally modulated electromagnetic oscillations.

     Let us examine the evidence from rabbits, rats and cats.

3.1.3  Mammalian Olfaction

     Because mammals have hugely more neurons than insects, it is no longer feasible to measure large-scale patterns of neural firing in mammals by single-cell recording techniques. Extracellular measurements of so-called "field potentials" must be used. Field potentials are just spatial summations of the electromagnetic field changes caused by the more-or-less simultaneous firing of large groups or masses of neurons.

     When field potentials are recorded from an array of chronically implanted electrodes in the olfactory bulb of rabbits, cats, or rats, macroscopic oscillations similar to those described in the locust are measured. The frequency of the oscillations is within the gamma range, but the details vary with the species - in rats the oscillations are at about 52 Hz, in rabbits about 56 Hz and in cats around 38 Hz (Bressler & Freeman, 1980 [35]). Thus, given that the comparable oscillations in locusts are at about 20 Hz, there does not appear to be any consistent correlation between oscillation frequency and size of animal.

     Notably, these oscillations in mammalian olfactory bulbs have a global character. This is shown by a high level of wave form similarity across the extent of the rabbit olfactory bulb and also within a large portion of the prepyriform cortex, to which the olfactory bulb projects (Bressler, 1984 [36]). Such a global quality sits well with requirement 4 in Chapter 1, where it is argued that consciousness per se has a global character and thus that the neural correlate of consciousness should have a global character. Further, if we assume for the moment that these global oscillations are identical with olfactory consciousness, the different frequencies of the oscillation in different species of animal would suggest that in general, what it is like to smell is different for cats than it is for rabbits, and that it is different again for rats. On the other hand, what it is like for one rabbit to smell is probably quite similar to what it is like for another rabbit to smell (with certain provisos which will be seen in the next paragraphs). These ideas are, perhaps, not counterintuitive.

     At this stage however, the similarity between insects and mammals starts to break down. Early experiments involving simple presentation of odors to rabbits (Freeman, 1978 [92]) failed to elicit any invariant odor-specific modulations of the global carrier oscillations. With the benefit of hindsight, it appears likely that the real reason for this is that suitable methods of signal processing and pattern recognition had not yet been developed, but at the time the other theoretical possibility, which was that the rabbits may not actually have been consciously experiencing odor-specific sensations, (either because they were not paying attention or because they were physiologically incapable of smelling certain odors) also seemed compelling. Because of this possibility, a training procedure that would allow unequivocal inference from behavior that the animals had perceived and discriminated particular odors was adopted. Thus the more mathematically sophisticated later experiments in this series measure not only sensation per se, but also memory and recognition processes and presumably motor processes preparatory for the trained movements. This may be seen as unfortunate for our search for electromagnetic correlates of sensation per se, or it may be fortunate in that it simply reflects the real situation, which may be that odors are not actually experienced in any stable and repeatable way until they can be recognized or identified. At any rate, what we now have are some excellent data from experiments in which animals were trained to associate certain odors with the expectation of reward or punishment, and so to behave accordingly when they identified those smells.

     The results of a series of experiments designed to identify odor-specific patterns in the EEG recorded from an array of 64 electrodes implanted over the olfactory bulb of rabbits that had been trained to discriminate particular odors are reported in three classic papers (Freeman & Baird, 1987 [94]; Freeman & Grajski, 1987 [95]; Freeman & Viana Di Prisco, 1986 [93]). A number of mathematically sophisticated pattern recognition techniques were applied to the raw data obtained in these experiments in order to identify odor-specific patterns, and since the results are so important to the overall argument put forward in this book, it is probably valuable to have at least some non-mathematical understanding of these techniques.

     The basic problem facing the experimenters was that there was no clear prior idea of what the odor-specific patterns might look like, or where or when they might exist in the olfactory bulb. Therefore the rationale guiding the development of the EEG processing and pattern recognition procedures was essentially an ad hoc one. First, it was assumed that patterns existed in the EEG which were distinctly different for (a) odors the animals had been conditioned to react to (b) odors they had been conditioned not to react to, and (c) a control condition of no presented odor. Then any mathematical algorithm that helped in correctly classifying EEG segments as having been recorded during one or other of these stimulus conditions was adopted.

     The first step in the analysis was to apply some general signal processing techniques, designed to smooth the data and remove trends that had no connection with the patterns being searched for. This involved replacing data from bad channels (those that produced obvious artifacts) with the average of 2 adjacent channels, temporal smoothing of the data from each channel (taking the mean of each time point plus its 2 adjacent time points weighted by 0.5) and "detrending" of the data to remove the underlying respiratory wave. Then began the serious data reduction.

     First, the average time series for all 64 electrodes (the time ensemble average) was calculated. This completely removed spatial information. A Fast Fourier Transform (FFT) of this time ensemble average1 gave information that allowed fitting of an amplitude and frequency modulated cosine wave to the time ensemble average. The data that were not fitted (the residuals) were fitted by another cosine wave, and so on for the first five fits. Thus the ensemble average time series was decomposed into five components. The one that carried most of the power (usually the first) was called the dominant component. This manipulation identified two kinds of burst occurring when odors were presented. The first had a dominant frequency greater than 55 Hz and one narrow dominant spectral peak (i.e. not much frequency modulation). The second type of burst had a dominant frequency less than 55 Hz and was disorderly, with broad spectral patterns. A behavioral assay showed that the high- and not the low-frequency bursts carried odor-specific information.

     Now the same sort of averaging procedure was performed in the spatial domain. For each of the 64 electrodes, the root mean square of the amplitude of the whole sample of EEG collected for each odor condition was calculated. This gave 64 numbers, each of which represented the average voltage for the whole recording time at one of the 64 points in space where the electrodes were recording. At this very crude level of spatial analysis, it was found (not surprisingly) that there was no correlation between the odor condition and the spatial pattern. Instead, it turned out that each rabbit had a characteristic spatial pattern that was like an individual signature; never exactly the same twice, but still easily recognizable (Freeman & Grajski, 1987 [95]).

     In order to pull odor-specific information out of the spatial patterns, it was found necessary to go back to the original data and perform several further processing steps, involving both temporal and spatial filtering and also a mathematical procedure called spatial deconvolution. First, the two different types of burst (those with a dominant frequency greater than 55 Hz and those with a dominant frequency less than 55 Hz) were processed separately. This was a vital step, because as it turned out, the orderly > 55Hz bursts carried odor-specific spatial information while the disorderly < 55Hz bursts did not. Secondly, it was hypothesized that synaptic currents in the granule cells (the inhibitory cells in the olfactory bulb) constituted the signal in this context while far-field currents from other neurons were noise. Therefore, spatial filters were developed to identify and enhance the contribution of the granule cells. This turned out to be equivalent to applying a low-pass spatial filter. This manipulation showed that while high frequencies in the time series carried the information, low frequencies (0.15 to 0.25 cycles/mm) in the spatial series carried the information.

     Thirdly, a mathematical procedure called spatial deconvolution was used. This was designed to correct the distortion caused by volume conduction (i.e. partially to undo the blurring that inevitably occurred because measurements could only be taken at the surface of the olfactory bulb and not deep in the bulb where the electromagnetic patterns were generated). Finally, the data were normalized to remove the effect of each rabbit's individual "signature".

     Out of all this processing, some very interesting facts emerged.

  1. Odor-specific spatial patterns were detectable in the olfactory bulb of rabbits trained to discriminate between two odors.
  2. Most interestingly, the odor-specific information was not localizable to subsets of channels. On the one hand, deletion of channels decreased the power of the remaining channel data to classify bursts correctly. But on the other hand, subsets of as few as 16 randomly selected channels had the power to classify bursts correctly at better than chance levels. This suggests that the information was broadly distributed over the bulb. Other results suggested that the granule cell activity patterns that contained odor-specific information extended well beyond the limits of the array window of observation. In particular, the phase gradient suggested that they involved the entire main bulb.
    Thus the discriminative output of the bulb apparently involved the entire structure, even though the receptor input was delivered to limited subsets of mitral cells in the bulb.
  3. Overall burst amplitude played a role in separating odor from control bursts, but no role in separating bursts occurring during presentation of different odors.
  4. The frequency and phase properties played no role.
  5. The amplitude properties that determine the characteristic signature of each animal played no role.
  6. In the words of the authors (Freeman & Baird, 1987 [94]) "During learning to identify an odor, it appears that a nerve cell assembly is formed by strengthened connections among the mitral cells that are coactivated by the odor under reinforcement ... . Thereafter, the arrival of the odor may lead to formation of a stereotypic spatial pattern over the entire bulb as a necessary albeit insufficient condition for correct response to that odor. On this interpretation, the bulbar output is not localized but is truly global; every neuron participates in every learned odor response but in differing degrees for different odors."

     These experiments, then, seem to suggest that there do exist electromagnetic field patterns which can be used to classify correctly the olfactory sensations being experienced by the animals from which the patterns were recorded. But how do these results stack up against the ground rules laid out in Chapter 1?

     In the methodological arena, the experiments pass with flying colors the requirements concerning elimination of artefacts and control of general physiological factors. They fare less well when it comes to the necessity to isolate consciousness per se from other brain processes - in particular, they deliberately conflate conscious experience with memory. However, all of the general requirements specified under the rules of evidence for identifying an electromagnetic pattern with a specific subjective experience are indeed fulfilled:

  1. In as far as this can possibly be ascertained using non-verbal subjects, the spatial electromagnetic patterns were certainly present when and only when the subject was conscious of a particular experience. This was shown by the fact that the patterns were only present when the rabbit responded behaviorally to the presence of a particular odorant. There was no odorant-specific spatial pattern in the absence of an odorant, and in cases when the odorant was present but there was no behavioral response (presumably because the rabbit was not paying attention) the dominant burst frequency was less than 55 Hz and there was no odorant-specific spatial pattern of amplitudes.
  2. The electromagnetic patterns certainly did correlate with subjective experience rather than with the physical stimulus. In fact, as the experimenters put it, "the spatial patterns lacked invariance with respect to odorant conditioned stimuli, showing instead a dependence on brain state, behavioral context and training history" (Barrie, Freeman, & Lenhart, 1996 [23]; Freeman, 1991a [97]). In other words, the patterns correlated with the meaning the stimulus had for the animal, not with the stimulus itself. Furthermore, although it was not directly demonstrated that the spatial amplitude pattern in the bulb varied according to sensation rather than odorant concentration, dynamic range compression compatible with Weber's Law (see Appendix A) was shown to occur at the input stage to the olfactory bulb (Freeman, 1991a [97]).
  3. The electromagnetic patterns were certainly generated by a relatively localized area of the brain.
  4. But they had a striking global quality. This was shown by the facts that (a) almost as much information was carried by the pattern recorded using an array of only 16 electrodes as was carried by the full array of 64 electrodes and (b) there were no "edge effects" i.e. channels at the edge of the array were just as useful information-wise as those in the middle. One likely interpretation of these observations is that the overall electromagnetic pattern in the whole olfactory bulb was the important thing, not any smaller-scale, localized part of this. The explanation of the lack of edge effects would then be that the overall size of the important pattern was significantly larger than the electrode array, so that edge effects would not be expected in this sub-sample of the main pattern. We can also say that the spatial resolution of the pattern must have been such that in an array of the size used, the number of sampling points could be reduced to as few as 16 before spatial aliasing became a problem.

3.2  Summary

     In the mammalian olfactory system, the existence of spatiotemporal electromagnetic patterns that obey almost all of the criteria in Chapter 1 has been unequivocally demonstrated.

Bibliography

[6]
Adrian, E.G. (1950). Sensory discrimination, with some recent evidence from the olfactory organ. British Medical Bulletin, 6, 330-332.
[23]
Barrie, J.M., Freeman, W.J., & Lenhart, M.D. (1996). Spatiotemporal analysis of prepyriform, visual, auditory, and somesthetic surface EEGs in trained rabbits. Journal of Neurophysiology, 76(1), 520-39.
[35]
Bressler, S.L., & Freeman, W.J. (1980). Frequency analysis of olfactory system EEG in cat, rabbit, and rat. Electroencephalography & Clinical Neurophysiology, 50(1-2), 19-24.
[36]
Bressler, S.L. (1984). Spatial organization of EEGs from olfactory bulb and cortex. Electroencephalography and Clinical Neurophysiology, 57, 270-276.
[91]
Freeman, W.J. (1972). Linear analysis of the dynamics of neural masses. Annual Review of Biophysics & Bioengineering, 1, 225-56.
[92]
Freeman, W.J. (1978). Spatial properties of an EEG event in the olfactory bulb and cortex. Electroencephalography & Clinical Neurophysiology, 44(5), 586-605.
[93]
Freeman, W.J., & Viana Di Prisco, G. (1986). Relation of olfactory EEG to behavior: time series analysis. Behavioral Neuroscience, 100(5), 753-63.
[94]
Freeman, W.J., & Baird, B. (1987). Relation of olfactory EEG to behavior: spatial analysis. Behavioral Neuroscience, 101(3), 393-408.
[95]
Freeman, W.J., & Grajski, K.A. (1987). Relation of olfactory EEG to behavior: factor analysis. Behavioral Neuroscience, 101(6), 766-77.
[97]
Freeman, W.J. (1991a). Nonlinear dynamics in olfactory information processing. In J.L. Davis & H. Eichenbaum (Eds.), Olfaction: a model system for computational neuroscience. (pp. 225-249). Cambridge MA: MIT Press.
[124]
Hildebrand, J.G. (1995). Analysis of chemical signals by nervous systems. Proceedings of the National Academy of Sciences USA, 92, 67-74.
[160]
Laurent, G. (1996). Dynamical representation of odors by oscillating and evolving neural assemblies. Trends in Neurosciences, 19(11), 489-96.
[161]
Laurent, G., Wehr, M., & Davidowitz, H. (1996). Temporal representations of odors in an olfactory network. Journal of Neuroscience, 16(12), 3837-47.

Footnotes:

1 FFTs break a signal down into a series of simple sine waves of different frequencies and show how much the sine wave at each frequency contributes to the original complex signal.