OK, I am going to stretch some analogies here. No pretense for comprehensive overview or scientific method, just plain hypothesizing about the nature of story writing. The reason – it’s fun. Also, I have found out for myself that relating writing as a process to cognitive theories helps me think more clearly about the former and more enthusiastically about the latter. The post is a bit technical, but I have tried to provide brief explanations and relevant links where needed.
Sitting in a series of lectures on cognitive robotics and spatial cognition this week, I have been bouncing around my brain various sporadic thoughts, only vaguely related to the courses in question. So, here follow, very sketchy and somewhat amorphous, a few ideas about writerly roles.
Writing as astral projection (in your own brain)
What does it take to write good sensory descriptions, to help the reader walk in the shoes of a character, to wrest emotions of fear and disgust out of her, to pump some adrenaline in her bloodstream? Samuel Delany wrote in an essay (either in The Jewel-Hinged Jaw or in About Writing; too bad I don’t have the books to check and provide quotations) that the first step of the writing process is that of imagining a scene as fully as possible. Once that is in place, the writer engages in the somewhat grueling exercise of its cyclic annotation. Suppose I visualize a snippet of action and provide it with the following verbal anchor: “Joe reached for the doorknob.” This is a pretty sparse description and we can safely assume that the mental image it stands for is comparably bare-boned. Then, Delany writes (which he doesn’t actually, because I am just riffing off on memories here, but bear with me), you might ask yourself what color is the doorknob. Is it new or is it much used? “Joe reached for the rusted doorknob.” This sentence might be just barely richer in information than the previous one, but it is nevertheless coupled with a more vivid mental image, in the reader’s and certainly in the writer’s mind. In the essay Delany goes on to describe multiple subsequent annotationsof this kind, but, for the purposes of the post, I believe this example is sufficient.
The important idea to be taken from this discussion is that linguistic output serves as a sort of shorthand to the writer, through which she is able to record mental images and then rerun them in order to edit in more color and verisimilitude. Of course, this process is not as clearly cut into binary stages – shaping the text is often (probably always) influenced by factors that do not pertain to the image in a straightforward manner (such as incorporation of alliterative sound patters, avoiding word repetition, establishing symbolism within the text, inserting intertextual links, etc.). However, since such constraints are mostly relevant to other effects achieved by fiction (see for example Roman Jakobson’s poetic function of language and Hal Duncan’s take on it), let us concentrate on the strictly referential and pretend that the two stages are distinct from each other. What we have up until now is an idealized looped sequence of actions: the writer imagines a scene, then anchors the imagery in text, after which she reruns the mental tape and records over it, meanwhile correcting her notes of the experience, and the same thing goes on and on, until the verbal instructions trigger the construction of a sufficiently well-shaped image (for the purposes of the text, that is). The process might be more or less compressed for different writers and situations, and often times the actual linguistic recording might be taking place as mere subvocalization rather than as writing text on paper/keyboard, but it is always operative.
What Delany was intuitively driving at decades ago, has been recently proven true by the cognitive sciences. Lots of research needs to be done, surely, and the specifics of how exactly it happens in the brain are far from clear, but in essence it is now widely-acknowledged in the field that language comprehension and production are concomitant with active mental simulation. That is, to comprehend language, people “act out” in their brains, subconsciously for most of the time, what they hear or read. But the same holds for the converse direction too – before we know what we are going to say, we first simulate it in our brains. This approach to language processing has come to be known in recent years as Simulation Semantics (also as Embodied Construction Grammar) and it seeks to establish how meaning is grounded in embodied experience.
Very briefly, the embodied cognition thesis holds that human thought and the mind, more generally, are shaped by our bodies and the way we interact with the environment and the cultural structures already in place. Thus, people are able to comprehend a concept like SOUR, because its meaning is embodied in our brains; whenever we hear or think about something that has to do with sourness, neurons that normally code for the perception of that gustatory quality become somewhat active. Imagine vividly that you are putting a lemon slice in your mouth. Unless you just ate a chocolate cake or you are simply not trying very hard, you will probably experience the quale (subjective conscious experience) of “sourness”, albeit to a lesser degree than when you eat an actual lemon. These are your sensory neurons lighting up, not nearly strong enough to make it feel like Christmas, more like watching a Christmas family film. When you use the word “lemon”, some of those neurons will light up again, but this time the quale will be more akin to listening to a retelling of that Christmas family film (what a weird form of cognitive Platonism that sounded like). Because of this crucial dependence on sensorimotor experience, meaning is distributed all over the cortex and dynamically assembled each and every time, not localized and packed neatly in dictionary-like boxes.
This holds not just for words that denote physical objects and qualities but for abstract thought as well. Consider the metaphor ARGUMENT IS WAR, a classic example from Conceptual Metaphor Theory. We say things like: “I demolished his arguments.”, “His criticisms were right on target.”, “He attacked every weak pointin my argument.”, “Your claims are indefensible.”, etc. As argued by CMT researchers and supported by experimental data, when we use language like that we are not just comparing a source domain (in this case WAR) to a target domain (ARGUMENT), we are conceptualizing the second in terms of the first. This means that the embodied meaning of the concept WAR, which is more concrete because it involves more salient physical interaction, is conscripted for the interpretation of the concept ARGUMENT. Thus, when arguing, our adrenaline rises, we often get hot and breathless, some even get physically aggressive; it is like we are in a fight. The same goes for many other metaphors (here is a suggestive, but by no means exhaustive list). Let me briefly look into just one other metaphor: AFFECTION IS WARMTH. According to George Lakoff, the basis for this metaphor is laid down in the mind of the human individual as early as infancy. A child experiences her first moments of being loved when physically close to her mother, where she is safe and warm. Consequently, the neural populations that code for those two states (being warm and being loved), learn to fire jointly (Neurons that fire together wire together!) and a kind of illusory causal relation is established – the person comes to believe on an intuitive level that love is inherently associated with warmth.
The same goes for motor language. Researchers like Ramachandran have studied and popularized the so-called mirror neurons– cells which fire when an animal is carrying out an action, but also when it observes another organism carrying it out. To understand the motor behavior of others, people simulate it in their own minds. Studies in simulation semantics suggest that this is true for language comprehension too. When we read the sentence “John reached for the rusted doorknob.”, our sensorimotor cortex activates schemas that would be needed if we ourselves were to perform the action.
Let us return to Delany’s argument. We can now annotate his description of the imaginative process by recognizing its idealized first stage as mental simulation. The brain invents a virtual reality (VR) of sorts in order to test mental images. It has all the necessary material to do that, as different areas of the cortex store rich distributed multimodal representations of physical experience (see the figure below), which can be evoked to a degree, even in the absence of relevant stimuli in our immediate surroundings.
Cortical brain areas according to their functions
Now some mighty analogy-stretching needs to be done. I have suggested above that a writer engages in a cyclic process of mental simulation and linguistic recording. In a sense, the writer is immersing herself, time and again, in short simulative bursts, into a VR of sorts. Each time the VR is a bit more detailed, but the writer’s embodied sense of it is so too. If we decouple the top-down process in which the writer fleshes out the VR outside of the character’s point of view, as a kind of disembodied godly designer, from the bottom-up process in which she explores the currently coalescing VR as a narrative agent/avatar (NA), we can take a conceptual leap and bring the analogy to the next level. Namely, that a writer projects part of herself into an NA that moves within, manipulates, perceives and learns from the VR environment.
It’s probably a bit loony, as far as suggestions go, but the analogy is actually quite sturdy. One might even go further and say it’s not even an analogy, that it is the literal truth, and one might not be terribly far from being right. The brain has all the necessary resources to simulate a VR that no current technology has ever produced. Just think of dreams and how vivid they can be (their disjointed character is a separate matter, that’s just a different kind of storytelling). Even though we don’t always feel as if we are transported to a different world when imagining (some do), that is exactly what is happening, only the experience is not consciously available, at least not in great detail (which is evolutionarily sensible – a daydreaming antelope wouldn’t fare very well in any predator-inhabited environment).
This remotely operated NA then has two chief functions in the process of writing. The first one links to the top-down designer role (more on that later). The writer reflects on the simulated avatar’s experience in the VR and adjusts the simulated world in order to increase its level of verisimilitude, sort of like integrating feedback from the characters themselves into the product. Actually, I believe there is another simulated participant in the process, that of the idealized reader, whose reactions too are observed by the writer in order to increase the appeal of the VR and the NAs, but let us not get too carried away.
The second function of the NA is to update its own experiential representation of the VR. With each and every take on the simulated mental imagery the writer adds or subtracts various qualia, effectively constructing a behavioral model of the NA. “Shuddering, Joe reached for the rusty doorknob.” Our sensorimotor cortex knows best how emotions and body states are experienced in different situations; this knowledge is rarely accessible as dense and coherent representation to rational thought. So a writer must learn the art of deep immersion through NAs in her own simulated VR. The textual recording of the mental imagery should ideally be a window for the reader to this authorial simulation of the environment and the characters’ psychological reflexes. An effectively captured experience of the NA operating in the VR should induce the reader to simulate an approximation of the episode pre-tested by the writer. The processes of reading and writing are thus very similar. But where the reader spends five seconds to put a sentence into imagery, a writer might take five hours. This role of hers demands that she run the NA as many times as needed, until it achieves a satisfactory internal representation of the VR.
This process is scarily similar to some of the techniques employed in the field of cognitive robotics. In research of that kind robots are run multiple times in specific environments, until they learn by themselves to perform a task in a human-like way. The researchers do not preprogram the robots to execute the tasks; they program the learning rules by which the robots acquire skills and behavior in context. Surely, the robots learn tasks which are a piece of cake for the human brain: recognizing faces, grasping objects, walking, smiling, etc. But the principle is the same – the NA too is learning to experience the VR in a way that achieves effects sought-after in the context of the story. A writer (a decent one, at least) does not decide in advance what the behavioral circuitry of a character will be, she constructs it dynamically through multiple iterations of behavior in simulated contexts, much like developing an artificial brain. In this sense, a writer simulates virtual robots that learn from the environment and update their responses with each iteration. The final linguistic annotation of the simulation process can be viewed as the code that triggers behavioral algorithms in the reader’s brain. It can run on all kinds of OS that are NDHB-compliant (Normally Developed Human Brain), with slightly different results depending on individual differences. But a well-written piece of code will result in readerly simulations with a relatively high level of overlap.
To summarize, within this extended analogy writers project themselves into virtual robots, or narrative agents/avatars, and learn appropriate behavior in the context of a simulated VR. The nature of that appropriate behavior too is controlled by the writer (in cognitive robotics this amounts to adjusting synaptic weights between layers and units within neural networks), so we can say that the writer is the robot and the robot programmer at the same time. Writing out the linguistic anchor/trigger of the simulation is in some ways comparable to the job of a developer working on a piece of software to be run on other platforms (readers). Thus, we also have the writer as editor/debugger.
All this is to be taken with a grain (maybe even a handful) of salt. Analogies are just analogies, and even under an extreme interpretation which literally holds that mental simulation is a VR and character constructs are virtual robots, it should still be born in mind that those cannot be fully equated to their real life counterparts; they are at best in a relation of partial synonymy with one another.
To be continued. Next: Writing as world design.