The construction of concepts has always been a topic of research interest for the Cognitive Sciences and Artificial Intelligence. In this sense, some theoretical approaches to cognition advocate the need to define sets of pre-conceptual structures as the necessary basis for the formation of concepts of greater complexity or level of abstraction. When considering the internal representation of concrete elements as a starting point, the Recognition-byComponent theory postulates that the base blocks for the mental representation of objects would be a set of volumetric primitives such as spheres, cylinders, parallelepipeds called geons. Such theory describes the properties of these primitives, but without considering the leading role that signals to come from sensorimotor areas independent of vision could have in the formation of these mental representations. This research postulates the analogy of a sensorimotor signal mixer as a possible origin for internal representations of volumetric figures, deepening its genesis. For doing it, we carried out a case study on a set of 48 objects from everyday context, used in recognition of 3 volumetric primitives (sphere, cylinder, and parallelepiped). The results highlight the leading role of 3 sensorimotor signals: (i) the recognition of a second geometric contour, (ii) the texture of the object, and (iii) the hand proprioception required to interact.
Language has historically been considered a central element in the origin of human civilization,
understood as scaffolding for the social organization of forms, laws, and the content of thought [
The hypothesis here conceives the geon as the output signal of a mixture of multiple interacting sensorimotor stimuli. This output signal is hosted by a neural structure (signal mixer) that, due to the refinement of everyday experience and presumably by a Hebbian learning mechanism [16], produces an element consistent enough to be valid as an internal model of representation [17]. As for the research approach, we propose the selection of a subset of representative sensorimotor stimuli in the interaction with elements of the environment. Then, through a case study, we weigh their relevance in the emergence of volumetric primitives (sphere, cylinder, or parallelepiped type).
In daily experience, visual apparatus projects flat and volumetric objects. Meanwhile, the perceived shape will change gradually according to the relative movement between observer and object. However, interpreting different stimuli from the same object as separate entities does not seem to be a privileged characteristic by evolution [18].
In perceptual object recognition, the RBC (Recognition-By-Components) theory [19] asserts that input images are segmented into a set of simple geometric components called geons (for "geometrical ions"), such as bricks, cylinders, wedges, cones, and squashes, among others. Furthermore, these volumetric primitives are expected to have high viewpoint invariance, thus facilitating their recognition [20].
• Approaches that conceive a coordinate space that preserves retinal proximities for the matching of input against stored representation [23].
Taking as reference the geon theory, it supports the idea of representation as a structural description, consisting of elements (such as parts) [24], and this paper seeks to propose a functional scheme suitable for such internal representations.
According to Marr [25], the previous and better-understood stage of recognizing contours comes before recognizing volumes. Marr stated that questions of psychological interest could be illuminated and perhaps even explained by neurophysiological terms. Contour recognition is a neural functionality of high evolutionary value because it is part of the vision as one of the senses (the other one is hearing) with the minor proximity requirement for categorizing environmental elements. Although there are other strategic stimuli for human survival, such as sound and color in the auditory and visual channels, respectively, the identification of contours plays a leading role in the recognition of food, predators, containers, sexual partners, and shelter.
When considering contour recognition as a privileged information channel, it would be expected that this fact will materialize, either in the spatial distribution of this function throughout the neural tissue or in the density of connections relative to its operation. In either of these two scenarios, the probability of interaction with other stimuli coming from different channels increased [20]. The present hypothesis formulates that the interaction of these multiple signals led to a neural structure of composition (or mixture) subordinated to contour recognition (as the main signal) but integrative (by bringing together the effect of contingent signals). This neural structure of composition would be the material substrate of geon.
According to Barlow [26], we can provide arguments for promoting the possibility that such a signal composition mechanism takes place in the neural tissue. Barlows expresses it as follows: Neurons do not loosely and unreliably remap the luminous intensities of the visual image onto our sensorium. Instead, they detect pattern elements, discriminate the depth of objects, ignore irrelevant causes of variation, and are arranged in an intriguing hierarchy. Furthermore, there is evidence that they give prominence to what is informationally important, can respond with outstanding reliability, and can have their pattern selectivity permanently modified by early visual experience. These patterns amount to a revolution in our outlook. It is now entirely inappropriate to regard unit acts as a noisy indication of more fundamental and reliable processes involved in mental operations. Instead, we must consider single neurons as the prime movers of these mechanisms. Neurons bring about thinking, and we should not use phrases like “unit activity reflects, reveals, or monitors thought processes” because the activities of neurons, quite simply, are thought processes. (p. 380).
We propose to think of the formation of geon as an underlying Hebbian learning process, in which simultaneous activation of cells leads to pronounced increases in synaptic strength between those cells, what Hebb [16] stated as follows:
Let us assume that the persistence or repetition of a reverberatory activity (or “trace”) tends to induce lasting cellular changes that add to its stability. ... When an axon of cell A is near enough to excite a cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A’s efficiency, as one of the cells firing B, is increased. (p. 62)
Figure 2 presents the analogy of geon as a composite signal. This figure shows the interaction of ten sensorimotor channels that compose an output signal (tracks initially assumed to have equal synaptic weight). The output signal (or geon) will be continuously subjected to a refinement process given by the subject’s access to new experiences. So, the most consistent channels (channels with fewer novel signals) will take synaptic prominence in geon formation.
As a case study, the interaction of a single observer (Male, 32 years old, right-handed) with 48 random everyday objects, 16 of a spheroidal nature (Figure 3), 16 cylindrical, and 16 parallelepipeds was documented. contour (circular) to a 2nd contour (circular again), while in the cylinder it will go from a 1st contour (circle) to a 2nd contour (rectangular)) 3. TASTE: Material or food-specific plus non-specific cases 4. ODOR: Material or food-specific plus non-specific cases 5. SOUND: Material or functionality specific plus non-specific cases 6. TEXTURE: 6 texture values [27] plus non-specific cases 7. HARDNESS: 21 hardness values [28] plus non-specific cases 8. THERMAL STATE: Material or context-specific plus non-specific cases 9. HAND PROPRIOCEPTION: 36 hand postures [29] plus non-specific cases 10.BODY PROPRIOCEPTION: Functionality specific plus non-specific cases Hypothetical channels and their possible values
Proceeding to count the number of Novel Signals per Channel (NSC) and defining a Low Plurality Estimator (LPE) as follows: = (1)
In Table 2, it is possible to evidence the channels of greater consistency (Higher LPE), those of less relevance for the formation of a volumetric primitive, and the Most Repetitive Signal (MRS) resulting from the case study.
The most relevant channels for geon formation in descending order of importance were: 2nd contour, texture, and hand proprioception.
Firstly, the second contour rose in all three cases as the most representative channel in the formation of geon. Someone could argue that the novelty in the signal would tend to be elevated since each variation in the angle would offer a different signal for the second contour. However, it is here where the RBC theory provides theoretical support for invariant detection of properties of edges such as curvature, collinearity, symmetry, parallelism, and cotermination in a two-dimensional image [19].
Second, the texture channel becomes the next in importance, giving clues regarding the common association humans establish between smooth (or slippery) surfaces and perfect volumetric shapes.
Finally, hand proprioception appeared as the third channel in relevance. In each case, it associates with the most common signal for manipulation of such shapes in a daily environment: Power sphere grasp (Sphere), Large diameter grasp (Cylinder), and Two-hand palmar pressure grasp (Parallelepiped). Moreover, it showed a particular link with the gestural language used by a human trying to describe a given volumetric primitive.
This paper shows the hypothesis of a sensorimotor signal mixer as a formal explanation of the processing of pre-conceptual structures. In the Cognitive Sciences of the last 40 years, this proposal is relevant because the evidence about the role of sensorimotor information in processing concepts [12, 14], and the emergency of approaches into the frame of embodied cognition [10, 11]. The theoretical explanation was developed in three stages: (i) the main role of concrete objects in the construction of concepts, (ii) the RBC as alternative for explaining the perceptual object recognition, and (iii) a hypothetical signal mixer capable of composing geons, as an output signal from channels of cognitive relevance in the experience of the conceptualizer.
The paper exposes a case study for illustrating and testing how the signal mixer might work. As a results, a set of sensorimotor channels related to tactile experience add to contour recognition as prominent channels in preconceptual processing. The results suggests that a signal mixer model might provide an adequate formal explanation for being more deeply explored and elaborated. [3] J. A. Fodor, The Language of Thought, Harvard University Press, Cambridge, 1975. [4] R. C. Berwick, N. Chomsky, Why only us: Language and evolution, MIT Press, 2016. [5] N. Chomsky, Syntactic structures, Mouton, The Hague, 1975. [6] D. Bickerton, Language and species, Chicago University Press, Chicago, 1990. [7] J. Piaget, Intellectual evolution from adolescence to adulthood, Human development 15.1 (1972) 1-12. [8] S. Carey, E. M. Markman, Cognitive development, in: Cognitive Science, 201-254, 1999. [9] M. Tomasello, Culture and cognitive development. Current Directions in Psychological Science 9.2 (2000) 37-40. [10] M. M. Hedblom, Image Schemas and Concept Invention Cognitive, Logical, and Linguistic
Investigations, Springer Nature, Switzerland, 2020. [11] L. W. Barsalou, Perceptual symbol systems. Behavioral and Brain Sciences 22.4 (1999) 577-660. [12] A. M. Borghi, F. Binkofski, C. Castelfranchi, F. Cimatti, F. C. Scorolli,, L. Tummolini, The challenge of abstract concepts. Psychological Bulletin, 143.3 (2017) 263. [13] D. Casasanto, R. Bottini, Spatial language and abstract concepts. Wiley Interdisciplinary Reviews:
Cognitive Science 5.2 (2014) 139-149. [14] A. M. Borghi, A future of words: Language and the challenge of abstract concepts, Journal of
Cognition 3.1 (2020). [15] G. Jakus, V. Milutinović, S. Omerović, S. Tomažič, Concepts, Ontologies, and Knowledge
Representation, Springer Briefs in Computer Science, 2013. [16] D. O. Hebb, The Organization of Behavior, Wiley & Sons, New York, 1949. [17] J. J. Koenderink, A. J. van Doom, The internal representation of solid shape with respect to vision,
Biological Cybernetics 32 (1979) 211-216. [18] K. Kirkpatrick-Steger, E. A. Wasserman, I. Biederman, Effects of geon deletion, scrambling, and movement on picture recognition in pigeons, Journal of Experimental Psychology: Animal Behavior Processes 24.1 (1998) 34-46. doi:10.1037/0097-7403.24.1.34. [19] I. Biederman, Recognition-by-Components: A Theory of Human Image Understanding
Psychological Review 94.2 (1987) 115-147. [20] I. Biederman, Geon theory as an account of shape recognition in mind and brain, The Irish Journal of Psychology 14.3, 314-327, (1993). doi: 10.1080/03033910.1993.10557936. [21] J. Buhmann, J. Lange, C. von der Malsburg, J. C. Vorbruggen, R. P. Wurtz, Object recognition in the dynamic link architecture: Parallel implementation of a transputer network, in: Neural Networks for Signal Processing, 121-159, 1991 [22] D. G. Lowe, The viewpoint consistency constraint, International Journal of Computer Vision 1 (1987) 57-72. [23] T. Poggio, S. Edelman, A network that learns to recognize 3-dimensional objects, Nature 343 (1990) 263-266. [24] J. E. Hummel, I. Biederman, Dynamic binding in a neural network for shape recognition,
Psychological Review 99 (1992) 480-517. [25] D. Marr, Vision, San Francisco, 1982. [26] H. B. Barlow, Single Units and Sensation: A Neuron Doctrine for Perceptual Psychology?,
Perception 1.4 (1972) 371–394. doi:10.1068/p010371. [27] J. Faucheu, B. Weiland, M. Mathieu, A. Witt, P. Cornuault, Tactile aesthetics : textures that we like or hate to touch, Acta Psychologica, Elsevier, 2019. [28] B. Schwartz, Hardness of Rubber Rollers, Imperial Rubber Products INC, 2012. [29] T. Feix, J. Romero, H. B. Schmiedmayer, A. M. Dollar, D. Kragic, The GRASP Taxonomy of Human Grasp Types, Human-Machine Systems, IEEE Transactions, 2015.