CEPA eprint 3374

Formation, and neural organization, of perceptual spaces

Foerster H. von & Maturana H. R. (2011) Formation, and neural organization, of perceptual spaces. Cybernetics & Human Knowing 18(3–4): 63–71. Available at http://cepa.info/3374
Table of Contents
I. Formation, and Neural Organization, of Perceptual Spaces
II. Proposal
1. PhysiologyNeurophysiological Identification of Perceptual Equivalence
2. PsychologyExperiencing the Fourth Spatial Dimension
I. Formation, and Neural Organization, of Perceptual Spaces
The term “space” in the title as well as in the following description of this proposed study refers to the abstract mathematical notion of “any set or accumulation of things, the members being called elements or points and usually assumed to satisfy a set of postulates of some kind” (James &James, 1959), rather than to the usual notion of the three-dimensional Euclidean space, except when explicitly stated. Thus we may speak of the olfactory space: “The olfactory system constructs a space of many dimensions (no more than the number of different cells) but with low resolution in any dimension” (Gesteland, Lettvin, Pitts, & Chung, 1968, pp. 313-322). Similarly, we may speak of the haptic space, the color space, the visual space, and so forth, in order to replace the descriptive distinctions of subjective experiences as, for instance, “smell,” “touch,” “vision,” and so forth, by the appropriate set of postulates which define the points, the relation between these points, and thus the “spaces” which correspond to these experienced sensory modalities.
Henri Poincarè (1960, pp. 631-646) was probably the first to note that an organism’s appreciation of a domain external to itself can only arise through the formation of appropriate relations between the organism’s actions and the ensuing sensations, rather than through a passive reception of the stimuli impinging on the organism’s surface. Without the interaction between the sensorium (Kraepelin, 1887) with the motorium (Foerster, 1970) these sensations remain uninterpretable.
Poincarè’s observation, however, remained unnoticed by his contemporary neurophysiologists who were under the influence of one of their great colleagues, Johannes Müller, whose famous “law of the specificity of nervous energy” (Müller, 1826) is still found in all textbooks on physiology under his name. This law refers to a correct observation, namely, that subjective experience (sensation) elicited by the activity of any arbitrary set of receptors is independent of the physical agent which causes their activity. However, by associating this observation concept (nervousity of nervous energy), a new concept (nervous energy) is introduced which is not contained in the observation and, hence, carries with it implicitly the notion of an explanation.
However, the absence of a rigorous explanation of this observation in terms of physiological evidence rather than of metaphysical “specific energies” becomes even more evident when this observation is expressed in contemporary language:
The physical modality of a stimulus is not encoded into a receptor’s response.
This formulation, of course, immediately begs the question as to the specific neural organization which indeed interprets certain physical modalities (electro-magnetic waves, atmospheric pressure waves, molecular mean energies, etc.) by corresponding sensory modalities (light, sound, heat and cold, etc.).
It appears that in the last decade major steps toward an answer to this question have been made. Experimental neurophysiologists discovered first stimulus configuration specific cells in neural tissue close to the sensory surface (Maturana, Lettvin, McCulloch & Pitts, 1959, 1960, pp. 129-175; Maturana, 1962, pp. 170-178) and later in higher nuclei along the sensory pathways (Hubel & Wiesel, 1967, pp. 1561-1573; 1968, pp. 215-243); simultaneously, and in part independently, theoreticians developed the mathematical formalism of networks whose nodal elements compute an almost unending array of geometrical properties on their input configurations (Inselberg, Löfgren, Foerster, 1960, pp. 2-41; Weston, 1960, pp. 62-80; Foerster, 1962, pp. 43-82, 1967, pp. 47-93) and, finally, engineers utilized these principles of computation by constructing devices which out-perform in speed their physiological counterparts from counting the number of objects independent of their size, shape and position (Weston, 1961, pp. 44-46) to the identification of objects of mildly sophisticated shapes and forms (Russell, 1966; Sutro, 1968, pp. 811-820).
In the appreciation of these physiological findings and in the development of mathematical and engineering models which “anticipated” (Pitts & McCulloch, 1947, pp. 127-147) or “confirmed” (Hoffman, 1967, pp. 423-440) these findings, the concepts of a signal space (manifold) and of transformation groups operable in this space proved to be of considerable heuristic power. Indeed, it appears that cells which respond selectively to a given set of arbitrarily complex configuration (“Gestalten”) in their receptive field can not only be predicted by theory (Hoffman, 1967, pp. 423-440) but may also be found amongst the (~1010) cortical cells, given sufficient time for the search.
On the other hand, the response of such cells which indicate to the experimenter the presence of particular stimulus configurations in their receptive field gives not the slightest clue as to the organism’s subjective experience of an external domain of certain configurations, unless, of course, built into this organism is a “virtual experimenter” with known perceptive powers who compares his experience of the environment with the response of these cells.
From this vicious circle it is clear that postulates which may be sufficient for defining the signal spaces of an organism’s sensorium are insufficient for establishing its perceptual spaces. The additional information in bridging this gap comes, of course, from including into these considerations the role of the motorium and the relations between an organism’s behavior and its sensations.
In two specific instances the functional integration of somato-sensory areas has been successfully demonstrated. In one instance the topological mapping of the (two- dimensional) proprioceptive system for the extremities in primates into the motor system was shown to provide the necessary and sufficient interactions between the two systems to account for a considerable amount of autonomy of the integrated system in elementary behavior patterns (Whitsel, Petrucelli & Werner, 1969, 170- 183). In another instance the center for taste in man was found to, reside not as one would expect in the rhinencephalon, but in the parietal lobe in close proximity to the areas which control mastication and movements of the tongue (Boernstein, 1970, pp. 673-682).
It is the contention of the two senior investigators of this proposed study that these cases of somato-sensory integration are not isolated instances, but are manifestations of a general principle which stipulates the emergence of perceptual spaces from the interaction of the motorium with the sensorium, and it is precisely the rigorous formulation of this principle and its demonstration which is at the core of this proposed study.
II. Proposal
1. PhysiologyNeurophysiological Identification of Perceptual Equivalence
Our purpose is to reveal experimentally the neurophysiology mechanism through which a perceptual space is generated. To this end we propose to consider the nervous system as an homeostatic system, in which a perceptual space arises as a mode of conduct specified by a class of somato sensory relations. Consequently, our concrete experimental aims are:
(i) To find out how different kinds of cells, independently characterized anatomically or functionally, participate in the generation of conduct.To find out if the participation of a given cell or class of cells in the generation of a given behavior is always the same, or changes with experience through changes in the circumstances under which the conduct is enacted.
(ii) To find out if a given cell or class of cells can participate in the generation of more than one kind of conduct.
(iii) To find out if the reenaction of a given conduct that had been lost from the active repertoire of the system implies the participation of the same kinds of cells.
(iv) To find out how the answers to these questions vary for cells at different levels between the sensory and the effector surfaces.
Obviously, the difficulty in this experimental approach lies in finding a criterion for deciding whether a cell is at any moment participating or not in the generation of a given conduct. This difficulty arises mostly, however, from considering that the observer can decide what constitutes the stimulus that generates certain conduct, and, hence, that he can determine the state of the animal by specifying the input. According to what we have said above such an approach is inadequate, and we propose to face the problem in the reverse manner. That is, we think that the animal should define its own state for us, and that this can take place if we consider that the recurrence of a given behavior always implies that the animal has come back to the same functional state (perceptual state), independently of the internal path that it has followed to reach it, Thus, we propose to carry to its legitimate consequences the notion that two situations that generate the same behavior are perceptually equivalent for the animal, regardless of how different they may look in our description of them. Accordingly, we propose the following experimental procedure that should constitute a general tool for this sort of analysis:
A: Train an animal to give a definite “learned” response that can be performed under experimental conditions designed to record from single cells. We can attain this in a first approach by training the animals to give a cardiac response under conditions of color or form discrimination that can also be manipulated in terms of color and form illusions. The appearance of the cardiac response will be considered as an expression that the animal is in a given perceptual state, and will be equivalent to an answer naming a color or a form.
B: Once point A is attained we shall record at various levels of the visual pathways from cells that we can independently characterize, either by their location or by their mode of response to physically defined stimuli (receptive field properties, for example), and correlate their activity under various visual conditions with the cardiac response. In this manner we shall know whether a given cell or class of cells participates or not in the generation of the conduct (enaction of a particular perceptual state defined by the animal and not by the observer) denoted by the cardiac response. Also, by the proper temporal controls and the repetition of the experiments under different visual conditions, and under different intervening experiences, we shall be able to assert if a given cell or class of cells participates or not in the generation of a given conduct, and whether this participation is constant or changes with the history of the animal. We shall also be able to assert if a given cell or class of cells participates in the generation of only one or of many different conducts (perceptual states).
This experimental procedure has general value. It permits to analyze the activity of individual neurons, and classes of neurons, in reference to the states of activity that they contribute to generate, and not in reference to arbitrarily defined stimuli that lie in the domain of descriptions of the observer. As a general tool it should permit us to analyze the participation of any class of cells, which we can independently identify at any level of the nervous system, in the generation of a given conduct or class of conducts, that is, in the specification of a perceptual space.
2. PsychologyExperiencing the Fourth Spatial Dimension
The information theoretical concept of “decoding,” that is, relating a signal to a symbol, and the linguistic concept of “denoting” that is, relating an object to a name, have in common that both are decision activities producing the most likely interpretations of their inputs.
In the context of this study that aims at an explicit representation of a semantic data base in a computer’s architecture and its program structure, it is of paramount significance if it were possible to expose explicitly the activities that interpret the inputs provided at the sensory receptors so as to yield a “picture of the world” with objects and their names.
It has been shown (Foerster, 1970; Maturana, 1970), that postulates which may be sufficient for defining the signal spaces of an organism’s sensorium are insufficient for establishing its perceptual spaces The additional information in bridging this gap comes – as Poincaré anticipated – from including into these consideration sensory- motor interactions.
As Piaget (Piaget & Inhelder, 1956) and others have shown these interpretational sensory-motor “skills” are, however, incorporated in man during his very early stages of infancy, including the period of his language acquisition, and thus are not accessible through “interviews” or by eliciting through other means expressions of introspections.
After a thorough survey of the literature on this topic, and after some careful analysis of the problem at hand it became clear that in the adult the emergence of interpretations of sensations could only be achieved by exposure to signal spaces that in principle could not have been experienced before.
The method ultimately selected was to create a visual four-dimensional environment to which the subject has manipulative access. The experiment consists of having the subject equipped with goggles that permit each eye to view on a small CRT one image of a stereo-pair of a 3-D projection of a 4-D object. The image is drawn by an on-line computer and changes in its apparent 3-D or 4-D position according to the subject’s movements, either through controls attached to the subject’s head or through controls that are manipulated by the subject’s voluntary motor system (Arnold, 1971).
Computer programs for a variety of 4-D objects – from simple 4-D tetrahedra to “Klein-bottles” and other sophisticated objects – have been developed (see figure 1) and are at present studied in a situation in which the subject is seated before a fixed screen, observing through stereoscopic binoculars pre-programmed variations of these objects in 4-D (figure 2). The manipulative consoles are under construction and the first, preliminary, results will be obtained during the summer months (figure 3).
Figure 1: Six Examples of 4D → 3D Projection of a Hypercube Rotated in Four-space
Figure 2: Experimental Arrangement for Viewing and Manipulating Stereoscopic Views of 4-D → 3-D Projections of 4-D Objects Manipulated in 4-D and/or 3-D Space
Figure 3a: Head-mounted Display at the University of Utah. Note: Employs a display processor similar to that used for the aircraft carrier sequence. two miniature cathode ray tubes are built into the goggles. A mechanical linkage tells the computer where the viewer is looking at each instant. The display processor instantaneously supplies the correct image for each head position. The viewer is free to look anywhere in a 360-degree circle and can look up and down through an angle of about 45 degrees. Two samples of what he sees are shown at the right. The objects grow larger or smaller and move with relation to one another as the observer moves around.
Figure 3b: U of I Graphic Design Program for 4-dimensional Representations
Note: Unpublished text, 1970. Heinz von Foerster Archives, Department. of Contemporary History, University of Vienna.
Arnold P. (1971) Experiencing the Fourth Spatial Dimension. In: Accomplishment Summary 1970/71 (B. C. L. Report 71.2) Urbana IL: Biological Computer Laboratory, Department of Electrical Engineering, University of Illinois.
Boernstein W. S. (1970) Perceiving and thinking: Their interrelationships and organismic organization. Ann. N. Y. Acad. Sc. 169(3): 673–682.
Foerster H. v. (1962) Circuitry of clues to Platonic ideation. In: C. C. Muses (ed.) Aspects of the theory of artificial intelligence (pp. 43–82) New York: Plenum Press.
Foerster H. v. (1967) Computation in neural nets. Currents in Mod. Biology 1: 47–93. http://cepa.info/1625
Foerster H. v. (1970) Preface. In: Notation of movement (B. C. L. Report No. 10.0) Urbana IL: Biological Computer Laboratory, Department of Electrical Engineering,. University of Illinois.
Foerster H. v. (1970) Thoughts and notes on cognition. In: P. Garvin (ed.) Cognition: A multiple view (pp. 25–48) New York: Spartan Books. http://cepa.info/1637
Gesteland R. C., Lettvin J. Y., Pitts W. H. & Chung S. H. (1968) A code in the nose. In: H. L. Oestreicher & D. R. Moore (eds.) Cybernetic problems in bionics (pp. 313–322) New York: Gordon and Breach.
Hoffman W. C. (1967) The neuron as a Lee Group germ and a Lee product. Bull. Math. Biophysics 25: 423–440.
Hubel D. H. & Wiesel T. N. (1967) Cortical and callossal connections concerned with the vertical meridian of visual fields in the cat. J. Neurophysiol 30: 1561–1573.
Hubel D. H. & Wiesel T. N. (1968) Receptive fields and functional architecture of monkey striate cortex. J. Physiol. 195: 215–243.
Inselberg A., Löfgren, L. & Foerster H. von (1960) Property detector nets and fields. In: Some principles of preorganization in self-organizing systems (Technical Report No. 2: 2–41) Urbana IL: Electrical Engineering Research Laboratory, University of Illinois.
James G. & James R. (eds.) (1959) Mathematics dictionary. Princeton NJ: D. Van Nostrand.
Kraepelin E. (1887) Psychiatrie (2nd ed.) Leipzig: Abel.
Lettvin J. Y., Maturana H. R., McCulloch W. S. & Pitts W. (1959) What the frog’s eye tells the frog’s brain. Proc. IRE 47: 1940–1951. http://cepa.info/518
Maturana H. R. (1962) Functional organization of the pigeon’s retina. In: R. W. Gerard & J. W. Duyff (eds.) Information processing in the nervous system (pp.170–178) Amsterdam: Excerpta Medica Foundation.
Maturana H. R. (1970) Biology of cognition (B. C. L. Report 9.0) Urbana IL: Biological Computer Laboratory, Department of Electrical Engineering, University of Illinois. http://cepa.info/535
Maturana H. R. (1970) Neurophysiology of cognition. In: P. Garvin (ed.) Cognition: A multiple view (pp. 3–23) New York: Spartan Books. http://cepa.info/536
Maturana R. H., Lettvin J. Y., McCulloch W. S. & Pitts W. (1960) Anatomy and physiology of vision in the frog (Rana pipiens) J. General Physiol. 43: 129–175.
Müller, J. (1826) Zur vergleichenden Physiologie des Gesichtssinnes. Leipzig: Abel.
Piaget J. & Inhelder B. (1956) The child’s conception of space. New York: Norton.
Pitts W. & McCulloch W. S. (1947) How we know universals. Bull. Math. Biophys. 9: 127–147.
Poincaré, H., (1960) L’espace et la géometrie. English translation in A. Danto and S. Morgenbesser (eds.) Philosophy of science (pp. 374–382) Cleveland: World Publishing Company. (Originally published in Revue de métaphysique et de morale: 631–646, 1895)
Russell J., (1966) A visual image processor (BCL Report No. 5.4) Urbana IL: Biological Computer Laboratory, Department of Electrical Engineering, University of Illinois.
Sutro L. L. (1968) Proposed electronics to represent the properties of a frog’s eye. In: H. L. Oestreicher & D. Moore (eds.) Cybernetic problems in bionics (pp. 811–820) New York: Gordon and Breach.
Weston P. (1960) Some notes on the synthesis of property filtration. In: Some principles of preorganization in self- organizing systems (Technical Report No. 2: 62–80) Urbana: Electrical Engineering Research Laboratory, University of Illinois.
Weston P. (1961) Photocell field counts random objects. Electronics 34(38): 44–46.
Whitsel B. L., Petrucelli L. M. & Werner G. (1969) Symmetry and connectivity in the map of the body surface in somatosensory area II of primates. J. Neurophysiol. 32: 170–183.
Found a mistake? Contact corrections/at/cepa.infoDownloaded from http://cepa.info/3374 on 2016-10-09 · Publication curated by Alexander Riegler