Subject, Verb, Object

Embodied Origins of a Linguistic Universal

For many decades, linguists have proposed a wide range of universal principles of language. The term linguistic universal has developed to describe a proposition is true of all languages spoken by humans anywhere. Most of the proposed universals have turned out to be statistical regularities rather than ontological necessities. Part of the problem with many universalist proposals has been their tendency to focus on form, on the grammar or syntax of the language. Human communities have developed numerous strategies for encapsulating their thoughts into language, and there do not appear to be definite limits on the form of this encapsulation.

One linguistic universal that has stood up to scrutiny is the fact that every language makes a distinction between objects and events. This distinction cuts deeply, through semantics and syntax all the way down to the level of morphology in the form of nouns and verbs. Such a pervasive division requires some explanation. It is not enough to assert that objects and events are actually different; after all, many languages have complex syntax that can treat any event as an object (to make nouns out of verbs, e.g. gerunds and infinitives). Less commonly, one can express events based on how an object is used (to make verbs out of nouns, e.g. to table a motion or to chair a meeting). Nonetheless, people experience the division between object and event deeply.

Speaking speculatively, I propose that the distinction arises from functional divisions in the structure of the human mind. I highly doubt I am the first to propose it, on the basis that events pertain to the behavioral system and objects pertain to the perceptual system.

Why do organisms have perceptual systems? Is it not so that we can respond to a changing environment? And what is it in the environment to which we must respond, if it is not the existence of threats and opportunities? What other purpose, then, might the perceptual system serve than the identification of threats and the recognition of potential for satisfying basic needs?

The original purpose of the perceptual system must lie in the satisfaction of basic needs of hunger, thirst, reproduction, and safety. In organisms with a central nervous system, perception centralizes the task of identifying those things which may serve as food, drink, mates, predators, and other risks of bodily harm. That is, a primary function of the perceptual system is to identify a certain class of objects.

In simpler organisms, the class of perceived objects may be small in accordance with the limited ability of the organism to act on the world. When organisms become more complex, they do so out of a need to adapt to their environments, because the environment in which they live has become more complex. In this situation, the class of perceived objects must also grow to represent the complexity of the environment.

The number of object types recognized must also grow as the set of fundamental drives increases. Adding a social drive requires the ability to perceive and manage relationships. A drive towards exploration, that is, curiosity, requires the ability to distinguish what is already known from what is unknown. The evolutionary development of these new drives and the ability to perceive the things that satisfy them creates a feedback loop. In humans, the class of perceivable objects seems to be unbounded, whether it truly is or not. We cascade objects upon objects, abstracting to the point where we admit vague abstracts entirely lacking corporeal form.

I opine that object recognition originated to serve a functional role. We recognize objects to predicate action. We perceive food to consume it. We perceive water to drink it. We perceive predators to either fight them or flee from them. We perceive people to interact with them or copy them. These basic actions in service of basic drives condition and in fact determine the development of our perceptual capabilities.

Every object, then, arises out of action. That is, at a basic level, those things are objects on which we can act. That on which we cannot act is not perceived as an object. The purpose of perceiving objects is to enable action that satisfies our basic drives. Every object is the object of an action.

One may immediately object that humans can discuss many objects that are not the object of any human action. But to raise this objection is to miss the point. In humans, our rich simulation facility allows us to imagine actions undertaken by others, or even actions undertaken by inanimate processes, such as the wind or time. But the underlying reality is that this simulation machinery has coopted the fundamental neural machinery that represents the self as agent undertaking an action with respect to a predicate object.

And just so we have arrived at the headline of this post: Subject, Verb, Object. There is no reason why this ternary pattern should be so fundamental to language, yet it is. I am aware of no language, however remote or exotic, that does not admit a class of sentences in which an agent (the subject) performs an action (the verb) on an object (the direct object). The subject abstracts the self to play the role of of any actor, as was discussed extensively in the previous post. The verb abstracts the actions undertaken by the behavioral system. And the direct object abstracts the output of the perceptual system. The most common sentence pattern is nothing other than a reflection of the basic mammalian mental architecture adapted and updated for a hypersocial context. The capacity to think about anything humans do think about in terms of subjects, verbs, and objects is a reflection of the planning machinery in the prefrontal and premotor cortices as an evolutionary outgrowth from earlier perceptual and behavioral systems, as coordinated by the basic system of drives.

There is more to be said about the subject. Consider how unnatural it is to speak of inanimate, intangible things as though they were physical actors performing physical tasks. Consider sentences such as the fence runs along the hill or the storm battered the ship. Fences do not even move. The storm has no mind or intent, no mens rea on which to found an accusation against its ill intent. These sentences, on their face, are nonsensical. Yet we understand them intuitively by imagining how we might run along the hill, or how we might sink a toy boat in a bathtub by beating the water into a frenzy. The foundation for understanding language rests on our own embodied experience; every abstract utterance arises ultimately out of analogy to our own experience.

We began with the universality of nouns and verbs. Given the layout above, it is curious that we have only two categories instead of three. Why are subjects just nouns in the same way as objects? Some languages make morphological and syntactic distinctions between animate and inanimate nouns, that is, between natural subjects and unnatural ones, showing the obvious fact that humans can tell the difference between agents and non-agents. Yet even these langauges still allow inanimate nouns to serve as subjects. Why shouldn’t there be a unique and coequal category of agents at a foundational level, a category that is functionally distinct from the category of objects?

One possible answer is that subjects other than the self can only be identified through perception. Thus potential subjects must be recognized first as objects and only secondarily as agents. In that case, it should not surprise us that any object could then be coerced to serve as a subject, since subject-ness is simply the choice to view a particular object as an agent. If this perspective is correct, then personification is built into human language at the deepest level. It is a kind of type coercion that naturally coopts mental tools originally intended to support social functions. One implication of this proposal is that primates generally should have some ability to think in personified terms to some extent, based on the existence of mirror neurons that allow them to perceive the actions of other agents.

This post has talked extensively about objects, but less about events. Within the presentation above, events originate from the behavioral system. Yet we can also perceive events as they happen, and thus events also enter through the perceptual system. Actions performed by others enter into the mind exclusively through perception. And thus it makes sense that events could also be coerced into the role of an object, though it is more difficult to see how the perceptual system accomplishes this. We will not explore events and behaviors extensively in this post, as they will soon receive a post of their own.

To summarize, the linguistic universality of nouns and verbs potentially has deep cognitive origins, with nouns arising from the perceptual system and verbs from the behavioral system. The basic sentence that tells who did what to whom reflects the structure of the simulation facility that humans share, which coerces objects detected by the perceptual system to serve as other agents and models their action as a mirror of self-action. Self-action responds to basic drives that guide the perceptual system towards a basic set of categories, and these actions are often oriented towards a particular object, all of which is reflected in the operation of the simulation facility.

To foreshadow a bit, I conjecture that somewhere within the executive system of the brain, there might exist three distinct collections of neurons that correspond directly to the threefold roles of subject, action (verb), and object, and that the activity in each of these collections forms a kind of neural language, so that at some level of analysis the patterns in the subject region are isomorphic to the patterns in the object region and potentially also to the patterns in the action region. These activity patterns would be analogous in function and composition to the representations commonly used within deep neural networks. The neural collections themselves would be analogous to dedicated registers within a digital computer.

The guiding philosophy behind this point of view is that language is a fairly explicit representation of human cognition, at least as it occurs a the highest levels within the prefrontal cortex. That is, once you take cross-linguistic variation into account, the human mental architecture is transparently visible in the parts of language that we take for granted. This perspective contrasts starkly with earlier viewpoints that posit a sort of isolated language module that performs complex operations to transform speech into underlying cognitive representations that differ widely from the constructs of language. In these statements, I agree strongly with Daniel Everett’s claim that language is fully integrated with human cognition; language and cognition cannot be teased apart and analyzed separately.

In the next post, I will look at the role of place within the understanding of a scenario, in particular how the sense of place arises out of our capacity for locating objects and takes on a shape determined by the features of our bodies. After that I will examine the relationship between adjectives and the classic binding problem, how for example we perceive redness and ballness on separate cognitive pathways yet think of them as a unitary red ball. Finally, I will then discuss the connections among events, behaviors, and actions alongside verbal morphology such as tense, mood, and aspect. Having covered the cognitive origins of the major parts of speech, I will then proceed to the concepts of stories, frames, and ontologies. Please let me know in the comments if there are any particular topics that most interest you.