Last January I wrote on this blog that our linguistic “understanding is sensual, not an abstract mentalese.” So naturally I am pleased to end this year by discussing a chapter in a new book that defends the same proposition. The book, Embodiment in Evolution and Culture, includes a chapter by Thomas Fuchs entitled, “The Embodied Development of Language,” (available here) and is part of a movement to rid the cognitive social sciences of their dependence on symbols and ethereal meanings. They emphasize the physical body and the things we learn from actions and sensations.
Bodies are unimportant to computers. It does not matter whether I solved a problem on my smartphone or my desktop device. But animals make use of their bodies in many social ways that have no role in computer networks. For example, two animals react to one another on the basis of their bodies. One may stand next to the other, revealing a bigger body size that scares the other away. Or one animal may become excited by the sexual readiness of the other. Activities like these play no role in designing computer networks and are absent from discussions of artificial intelligence. Indeed, the Turing test assumes that artificial intelligence trials can hide the body without affecting the results. Thus, in a Turing test, you cannot have an exchange in which one participant asks, “How much do you love me?” and the other participant answers by spreading arms and saying, “This much.”
Since these examples of body-based knowledge and communication have no place in a computational realm, traditional cognitive scientists have simply ignored them.
Embodiment scientists take as their working hypothesis that all knowledge is grounded in the body’s sensory and motor activity. In other words, our knowledge does not rest on a set of symbolic representations or concepts, nor does our linguistic ability rest on a system of abstract rules. This hypothesis contradicts popular assumptions, but it has some attractive qualities.
- It is simple: The embodiment hypothesis only requires a sensorimotor system, which is known to exist in all primates. Meanwhile, the cognitive hypothesis requires at least two systems; first, a system of rules and concepts for generating strings of symbols, and then a sensorimotor system for expressing what you generate.
- It requires no miracles: Symbolic systems require inborn concepts that are organized into sentences. Earlier this year Berwick and Chomsky conceded there is no known natural explanation for this demand, “In some completely unknown way, our ancestors developed human concepts.” It is tempting to brush aside this appeal to miracles as some bit of madness. Why not get our concepts from the world around us? But if you say that, you are granting that symbols are grounded in bodily actions and giving up the two-systems argument before you get started. Thus, while not everyone goes to the length of Chomsky, anyone who holds that the brain is a computer and that sentences arise from computations is compelled to assert the initial presence of some abstract concepts in the brain.
- It is evolvable: Primates have had a sensorimotor system since they first appeared some 50 million years ago, and their ancestors had similar systems long before that. If language does not require the evolution of a new symbol-processing system, we need spend no energy trying to determine how such a thing could evolve.
All this sounds nice, but is there any evidence to support it? Fuchs rests his argument on research that establishes there is a “wealth of communication and dialogue that already unfolds in the human individual before learning language” [page 108]. In short, well before a baby says mama the mother and child have established a variety of communicative interactions. For example, a mother feeds some goop on a spoon to the baby, and then offers more goop. The baby turns its head away and tightens its lips. This bit of interaction illustrates what he means by embodied. The interaction is not representational or symbolic but it is instructive.
Indeed, in the trees and forests of the world, primate mothers and infants learn about one another through their bodies. The infant learns about the world in terms of actions and sensations, while the mother learns about the infant by observing its actions. And then, at 8 or 9 months, human babies take a step unknown to the rest of the primate order; they begin to use words. Most people have assumed that the change marks a step beyond embodied knowledge to some form of symbolic understanding. Fuchs says no. The change reflects a difference for sure, but it is a difference in sociability and behavior rather than a move beyond embodied knowledge. Human infants begin to engage in joint-attention with adults. Parent and child, for example, look together at a ball. Perhaps the father holds the ball out and the baby reaches for it. This is something new, a three-way relationship (adult, child, object) rather than two-way (adult and child or child and object) relationships found in the rest of the primate world, but the knowledge continues to be based on senses and actions. We do not need to add a symbolic representation at this point.
Fuchs’ paper insists that people never need to switch to a non-embodied system of knowledge. Our first nouns refer to objects we know through interaction. Juice, for example, is often one of a child’s first words, and it is known by sensory interactions. First verbs too refer to actions we know from experience, eat, run, fall, etc. The relation between verbs and nouns is also embodied knowledge. In a subject relationship—e.g., I eat—a person knows the action of eating from experience, while the object relationship—e.g., eat ice cream—is also known directly.
What about more abstract nouns? A noun like the concession does not seem to have an embodied meaning. The history of the word, however, shows it originated in the Latin cedere which has the concrete meaning of to withdraw or give way. A concession in a negotiation remains a metaphorical giving away some ground. Metaphors are a way of expressing something unembodied as though it were part of the perceptible world. Even the name of the immaterial (spirit) is a metaphorical derivation from the Latin spiritus for breath.
I want to devote a further post to this paper, but let me note here that neither the embodiment nor the symbolic representation schools have yet carried the day. Both sides can be challenged with queries that, as of yet, cannot be answered But embodiment can already explain enough to force supporters of symbolic representation to concede ground. As in all of life, you pays your money and takes your chance. I like the odds on the simple, evolvable, non-miraculous explanation.