10 The Future of an Illusion
In 1927, Sigmund Freud published The Future of an Illusion, his psychoanalytic diagnosis and history of religion, which he characterized as an illusion he attributed largely to wish fulfillment (Freud, 1989 [1927]). Acknowledging religion’s universality, Freud recognized its almost unbreakable hold on us, even among atheists, the pervasive way it shapes human relations, culture, art, politics, even science.
Much of what we’ve now uncovered about narrative explanation suggests that the theory of mind on which it’s based may also be a universal illusion, but one more basic and pervasive—and one harder to surrender. And its natural history suggests that the theory was an important source for the illusion that religion foists upon us.
Though it was the best solution that natural selection could contrive for the design problem of securing collaboration among our early ancestors on the African savanna, like most such “solutions,” the theory of mind was a quick and dirty contrivance, cobbled together with what was available, by a process that couldn’t wait around for better, more-adaptive components. The result was an imperfect theory, one that badly overshoots and that, once adopted, turned our ancestors and us into hyperactive agency detectors. Ever since the Pleistocene, humans have sought the causes for almost all natural processes in the desires and beliefs of an all-powerful deity to which the theory of mind could be applied, in explanation, supplication, propitiation, exculpation, and celebration. There is a well-known label for this phenomenon: “anthropomorphism.” Indeed, the “overshooting” of this quick and dirty solution to the problem of getting our ancestors to collaborate, by leading people to believe in an all-powerful deity that could enforce morality among them far better than any human could, is likely to have helped the theory of mind not just to forge but also to strengthen and preserve that collaboration.
Ultimately, science would unravel at least the theory of mind’s “design” argument for the existence of God, with crucial implications for the theory itself. It took 2,000 years of science, from Aristotle’s time to Isaac Newton’s, to rid physical nature of the all-powerful agency that the theory of mind had foisted upon it. The theory of mind was banished from the physical domain when Newton showed that motion did not reflect the ends, goals, needs, or desires of anything or anyone. After Newton, physicists recognized that there was no scope for the theory of mind in the explanation of anything that happened in the physical world. Indeed, it’s reported that when Napoléon asked the great French physicist Pierre-Simon Laplace what role God had in Newtonian physics, Laplace replied, “Your majesty, I have no need of that hypothesis.”
It took a further 150 years after Newton for Charles Darwin to do the same for the biological domain and another twenty years for him to finally publish this achievement, hesitating largely out of fear of its consequences for religion. Before Darwin, the only reasonable explanation for the beautiful and universal means-ends economy in the biological realm was the design of an all-powerful deity in accordance with the deity’s benevolent desires and unerring beliefs.
After Darwin, hardly any scientifically educated person could take the theory of mind’s explanations of adaptation seriously, which effectively spelled the end of religion’s most convincing argument for the existence of God.
Notice the trajectory: first, the theory of mind was banished from the physical, then from the biological domain; now neuroscience is completing this demythologizing process for the domain of human psychology. It’s casting the theory of mind out of its last redoubt in science, the human brain.
To see how natural selection has fine-tuned anatomical structure into function, but without leaving any role for purpose, let’s go back to the rat’s brain and to the grid cells, speed cells, direction cells, and barrier cells of its medial entorhinal cortex and the place cells of its hippocampus that we discussed in chapters 7 and 8 (figure 10.1, plate 10).
Place cells in the hippocampus and grid cells in the medial entorhinal cortex of the rat’s brain. From http://
What makes the neurons in the hippocampus and the medial entorhinal cortex of the rat into grid cells and place cells—cells for location and direction? Why do they have that function, given that structurally they are pretty much like many other neurons throughout both the rat and the human brain?
From as early in evolution as the emergence of single-cell creatures, there was selection for any mechanism that just happened to produce environmentally appropriate behavior, such as being in the right place at the right time. In single-cell creatures, there are “organelles” that “detect” gradients in various chemicals or environmental factors (sugars, salts, heat, cold, even magnetic fields). “Detection” here simply means that, as these gradients strengthen or weaken, the organelles change shape in ways that cause their respective cells to move toward or away from the chemicals or factors as the result of some quite simple chemical reactions. Cells with organelles that happened to drive them toward sugars or away from salts survived and reproduced, carrying along these adaptive organelles. The cells whose organelles didn’t respond this way didn’t survive. Random variations in the organelles of other cells that just happened to convey benefits or advantages or to meet those cells’ survival or reproductive needs were selected for.
The primitive organelles’ detection of sugars or salts consisted in nothing more than certain protein molecules inside them changing shape or direction of motion in a chemical response to the presence of salt or sugar molecules. If enough of these protein molecules did this, the shape of the whole cell, its direction, or both would change, too. If cells contained organelles with iron atoms in them, the motion of the organelles and the cells themselves would change as soon as the cells entered a magnetic field. If this behavior enhanced the survival of the cells, the organelles responsible for the behavior would be called “magnetic field detectors.” There’d be nothing particularly “detecting” about these organelles, however, or the cells they were part of. The organelles and cells would just change shape or direction in the presence of a magnetic field in accordance with the laws of physics and chemistry.
The iterative process of evolution that Darwin discovered led from those cells all the way to the ones we now identify as place and grid cells in the rat’s brain. The ancestors to these cells—the earliest place and grid cells in mammals—just happened to be wired to the rest of the rat’s ancestors’ neurology, in ways that just happened to produce increasingly adaptive responses to the rat’s ancestors’ location and direction. In other mammals, these same types of cells happened to be wired to the rest of the neurology in a different way, one that moved the evolution of the animal in a less-adaptive direction. Mammals wired up in less-adaptive ways lost out in the struggle for survival. Iteration (repetition) of this process produced descendants with neurons that cause behavior that is beautifully appropriate to the rat’s immediate environment. So beautifully appropriate, that causing the behavior is their function.
The function of a bit of anatomy is fixed by the particular adaptation that natural selection shaped it to deliver. The process is one in which purpose, goal, end, or aim has no role. The process is a purely “mechanical” one in which there are endlessly repeated rounds of random or blind variation followed by a passive process of environmental filtration (usually by means of competition to leave more offspring). The variation is blind to need, benefit, or advantage; it’s the result of a perpetual throwing of the dice in mixing genes during sex and mutation in the genetic code that shapes the bits of anatomy. The purely causal process that produces functions reveals how Darwin’s theory of natural selection banishes purpose even as it produces the appearance of purpose; the environmental appropriateness of traits with functions tempts us to confer purpose on them.
What makes a particular neuron a grid cell or a place cell? There’s nothing especially “place-like” or “grid-like” about these cells. They’re no different from cells elsewhere in the brain. The same goes for the neural circuits in which they are combined. What makes them grid cells and place cells are the inputs and outputs that natural selection linked them to. It is one that over millions of years wired up generations of neurons in their location in ways that resulted in ever more appropriate responses for given sensory inputs from the rat’s location and direction.
Evolutionary biology identifies the function of the grid and place cells in the species Rattus rattus by tracing the ways in which environments shaped cells in the hippocampus and entorhinal cortex of mammalian nervous systems to respond appropriately (for the organism) to location and direction. Their having that function consists in their being shaped by a particular Darwinian evolutionary process.
But what were the “developmental” details of how these cells were wired up to do this job in each individual rat’s brain? After all, rats aren’t born with all their grid and place cells in place (Manns and Eichenbaum, 2006). So how do they get “tuned” up to carry continually updated environmentally appropriate information about exactly where the rat is and which way the rat needs to go for food or to avoid cats? Well, this is also a matter of variation and selection by operant conditioning in the rat brain, one in which there is no room for according these cells “purpose” (except as a figure of speech, like the words “design problem” and “selection” that are used as matters of convenience in biology even though there is no design and no active process of selection in operation at all).
Like everything else in the newborn rat’s anatomy, neurons are produced in a sequence and quantity determined by the somatic genes in the rat fetus. Once they multiply, the neurons in the hippocampus and the entorhinal cortex, and many other neurons in the rat’s brain as well, make and unmake synaptic connections with each other. Synaptic connections that lead to behavior rewarded by the environment, such as finding the mother’s teat, are repeated and thus strengthened physically (by the process Eric Kandel discovered; Kandel, 2000). Among the connections made, many are then unmade because they lead to behaviors that are not rewarded by feedback processes that strengthen the synaptic connections physically. Some are even “punished” by processes that interrupt them. In the infant rat, the place cells make contact with the grid cells by just such a process in the first three weeks of life, enabling the rat’s brain to respond so appropriately to its environment that these cells are now called “place” and “grid” cells (O’Keefe and Dostrovsky, 1979). Just as in the evolution of grid and place cells over millions of years, so also in their development in the brain of a rat pup, there is no room whatever for purpose. It’s all blind variation, random chance, and the passive filtering of natural selection.
These details about how the place cells and the grid cells got their functions are important here for two reasons. First, they reflect the way that ’natural selection drives any role for a theory of mind completely out of the domain of biology, completing what Newton started for the domain of physics and chemistry. They show how the appearance of design by some all-powerful intelligence is produced mindlessly by purely mechanical processes (Dennett, 1995). And they make manifest that the next stage in the research program that began with Newton is the banishment of the theory of mind from its last bastion—the domain of human psychology.
Second, these details help answer a natural question to which there is a tempting but deeply mistaken answer. If the grid cells and the place cells function to locate the rat’s position and direction of travel, why don’t they contain or represent its location and direction? If they did, wouldn’t that provide the very basis for reconciling the theory of mind with neuroscience after all? This line of reasoning is so natural that it serves in part to explain the temptation to accord content to the brain in just the way that makes the theory of mind hard to shake. By now, however, it’s easy to see why this reasoning is mistaken. For one thing, if the function of the place and grid cells really makes them representations of direction and location, then every organ, tissue, and structure of an organism with a function would have the same claim on representing facts about the world.
Consider the long neck of the giraffe, whose function is to reach the tasty leaves high up in the trees that shorter herbivores can’t reach, or the white coat of the polar bear whose function is to camouflage the bear from its keen-eyed seal prey in the arctic whiteness. Each has a function because both are the result of the same process of random or blind variation and natural selection that evolved the grid cells in the rat. Does the giraffe’s neck being long represent the fact that the leaves it lets the giraffe reach are particularly tasty? Is the coat of the polar bear about the whiteness of its arctic environment or about the keen eyesight of the seals on which the bear preys? Is there something about the way the giraffe’s neck is arranged that says, “There are tasty leaves high up in the trees that shorter herbivores can’t reach”? Is there something about the white coat of the polar bear that expresses the fact that it well camouflages the bear from its natural prey, seals? Of course not.
But even though they don’t represent anything, the long neck of the giraffe and the white coat of the polar bear are signs: the long neck is a sign that there are tasty leaves high in the trees on the savanna, and the white coat is a sign that the bear needs to camouflage itself from its prey in the whiteness of the arctic, the way clouds are signs that it may rain. But for the neck and coat to also be symbols, to represent, to have the sort of content the theory of mind requires, there’d have to be someone or something to interpret them as meaning tasty leaves or a snowy environment. Think back to why red octagon street signs are symbols of the need to stop—symbols we interpret as such—and not merely signs of that need.
The sign versus symbol distinction is tricky enough to have eluded most neuroscientists. The firing of a grid cell is a good sign of where the rat is. It allows the neuroscientist to make a map of the rat’s space, plot where it is and where it’s heading. John O’Keefe called this a “cognitive map,” following Edward Tolman (1948). The “map,” however, is the neuroscientist’s representation. The rat isn’t carrying a map around with it, to consult about where it is and where it’s heading. Almost all neuroscientists use the word “representation,” which in more general usage means “interpreted symbol,” in this careless way—to describe what is actually only a reliable sign. (See Moser et al., 2014 for a nice example.) The mistake is usually harmless since neuroscientists aren’t misled into searching for some other part of the brain that interprets the neural circuit firing and turns it into a representation. In fact, most neuroscientists have implicitly redefined “representation” to refer to any neural state that is systematically affected by changes in sensory input and results in environmentally appropriate output, in effect, cutting the term “representation” free from the theory of mind, roughly the way evolutionary biologists have redefined “design problem” to cut it free from the same theory.
The same process of random or blind variation and natural selection that led to the polar bear evolving its white coat is what led to us evolving our mind-reading ability. Mind reading is something the neural circuits of our brains were tuned up to do as the result of a purposeless process selecting for adaptations in an environment of predators and prey in which first vertebrates, then mammals, apes, primates, hominins, and eventually Homo sapiens evolved.
But how did hominins’ mind-reading ability—their rather powerful, refined, and sophisticated ability to cope with predators, prey, and other hominins—give rise to the theory of mind? Together with evolutionary anthropology, paleoarchaeology, cognitive linguistics, and even philosophy of language to some extent, neuroscience has reached the point where it can give a pretty clear sketch of how the useful skill of mind reading got transformed into the theory of mind and how that theory foisted itself upon us.
Once hominins appeared and began to claw their way up the food chain on the African savanna, their ability to collaborate with one another became indispensable. This is the point at which mind reading begins to exploit another of our brains’ abilities: the ability that gave rise to spoken language. What ability is that and where did it come from in evolution? We don’t know the answers to either of those questions. But there are some powerful theories about what the ability consists in, due mainly to Noam Chomsky (Hauser, Chomsky, and Fitch, 2002).
Although Chomsky himself has despaired of our finding the evolutionary origins of this ability to mind read, he thinks that another of our abilities is at the root of language—“recursion,” the ability to build nested structures in thought (Hauser, Chomsky, and Fitch, 2002). And, indeed, recursion is ever present in language. Thus we can talk about our mothers, our mother’s mothers, our mother’s mother’s mothers, and so on. It’s an interesting and important question how many recursions the normal human brain can deal with. If you were to hear someone say, “I couldn’t not help but not disagree with you more,” could you tell right off whether that someone was agreeing or disagreeing with you? There are four negative recursions embedded in the speaker’s sentence, so what the speaker would really be saying is “I couldn’t disagree with you more,” right? But you had to consciously puzzle that out. With just two or three negative recursions, you could do it unconsciously. So our ability to produce and process recursive thoughts unconsciously has limits, and this ability may vary from person to person, or even from culture to culture. The Pirahã people of northwest Brazil, for example, are reported by some cultural linguists to have a language that lacks recursion (Everett, 2005).
But in Chomsky’s view, recursion is a necessary condition for language. True languages must have grammars. Grammars are rules of recursion. Recursion had to have emerged prior to language and get harnessed together with other abilities to produce language. Chomsky holds that recursion, and thus also true language is only to be found among humans. Other species communicate, of course. But they lack recursion, so they don’t have true language. Our brains acquired this ability sometime after humans split off from other primates 6 million years ago. How and when our brains’ acquiring this ability was selected for no one can explain. Luckily, that’s not our task here. All we need to do is sketch how our ability to mind read (an ability, not a theory) coevolved with our language ability to eventually generate the platitudes about belief, desire, decision, and action, the “folk psychology” in which the theory of mind consists.
The evolution of language among mind readers almost certainly had to have started with gestures and calls—especially warning gestures and calls that might have first emerged as purely emotional expressions of alarm. Although primatologists have helped flesh out this claim in their study of chimpanzees, we would have to assume that hominins shared alarm gestures and calls with the chimps and our last common ancestor. In any case, it’s safe to assume that hominin protolanguage started out this way and moved from gestures and calls to grunts to something like pidgins, to full-blown languages. It’s well known that mind reading isn’t a prerequisite of signaling. Lots of organisms with hardly any neurology at all exchange signals. Even slime molds “decide” whether to aggregate from free-living cells into multicellular reproductive structures by exchanging information—signals—in the form of secreted macromolecules. Game theorists and philosophers of language have shown that signaling between mindless organisms can emerge by Darwinian processes alone (Skyrms, 1996, 2003). But even just a little mind reading would make the emergence of language easier and its evolution faster when cooperation was adaptive. In the long run, language and mind reading clearly coevolved, with each enhancing the power of the other through a cycle of iterated selection.
Only when language finally appeared would the mind readers be ready to turn mind reading into a theory of mind—a set of statements describing how beliefs and desires produced behavior. At that point, of course, language had already long colonized consciousness, turning introspection into a theater of silent speech and visual imagery.
There is substantial evidence that anatomically modern Homo sapiens emerged in Africa some 300,000 years ago (Hublin et al., 2017). Once evolutionary anthropology established this, dating a dominant mystery, the major puzzle of the disciplines of evolutionary anthropology and paleoarchaeology, arose. Why is there no sign of symbolic behavior for the next couple of hundred thousand years or more and no unambiguous evidence of language until between 50,000 and 75,000 years ago? Of course, spoken words don’t fossilize so it’s hard to date the emergence of language. Still, scientists are puzzled by the fact that, even though one-piece tools date back a million years or so, and complex tools are at least 75,000 years old, there’s no evidence of symbolic adornment, burial markings, or anything else that bears an interpretation, that has content, that seems to be about something else, that represents, until the appearance of complex tools. This fact is part of an argument that language came late to Homo sapiens and emerged certainly long after our species became world-class mind readers. If this chronology is right, we can at least roughly estimate the earliest date for the advent of the theory of mind. Formulating, understanding, and using the full-blown theory of mind would require a prior grasp of the words “belief” and “desire” (in one or another human language), and since beliefs and desires are representations, are about stuff, have content, the theory couldn’t have emerged earlier than the advent of language, say, between 50,000 and 75,000 years ago.
Of course, our ancestors were mind reading long before then, long before any of them knew anything about nouns, verbs, adjectives, adverbs, grammar, sentences, statements, and meanings in general. The ability to mind read is not a matter of deploying a consciously articulated theory of mind, expressed in language that identifies beliefs and desires and that knits them into the causes of actions by their recognized relationship as representing with directions of fit that coordinate as ends and means. Mind reading couldn’t have been anything much like that, at least not at first, if more than 100,000 years was required for it to give rise to language, working with recursion, solving Darwinian survival problems on the savanna.
As we saw in chapters 4 and 5, natural selection selected for the theory of mind, laying down in our brains as innate or nearly so this quick and dirty solution to a critical evolutionary design problem, presumably in response to life or death environmental circumstances. There were probably enough obvious adaptive payoffs to small increments in acquiring the theory of mind to make its gradual emergence unsurprising.
There was at least one obvious and another less obvious design problem the solution to which would have selected for acquiring the spoken words eventually required to express the theory of mind: the need for high-quality imitation to make teaching and learning possible and the need to express and resolve conflict over behaviors with different adaptive payoffs.
In discussing two of our ancestors’ profound differences from other primates in chapter 5, we saw how natural selection for the ability to imitate behaviors turned a significant maladaptation—long childhood dependency—into a significant adaptation. The need to solve the design problem of preserving survival technologies—tool making, effective edible food gathering, and prey tracking—would have selected for combinations of signaling and mind reading that would have made it possible for adults to teach and for dependent children to learn complicated tasks and processes (Sterelny, 2012). Once spoken words emerged, they could be used and combined to describe objects and actions, the next evolutionary step would be the emergence of the means to correct, improve, and communicate processes and features (Cloud, 2014). At some point, there would be selection for the words that would eventually be put together to form the theory of mind, words that would figure in sentences like “What I desire is for you to sharpen your spear this way, not that way” or “I don’t believe you can catch a rabbit that way.” Consider the adaptive payoffs to our ancestors in learning and teaching recursive, iterated, nested sentences that could be expressed to themselves in silent speech.
Why would the expression and resolution of conflict over behaviors with different adaptive payoffs put a premium on the emergence of words like “belief” and “desire”? Among our evolutionary forbears well back into the Pleistocene, mind reading would have been a pretty sharp tool for predicting the behavior of other animals they could see in their immediate vicinity over the immediate future, but it would have been a blunter one when they tried to use it to predict the immediate behavior of animals they couldn’t see over a more distant future, and blunter still the more distant that future became. Its bluntness in these uses would have resulted sometimes from their failure to detect their predictive target’s ends, toward which its behavior was directed, and sometimes from their failure to detect their target’s means, the features of their immediate environment salient to the target’s behavior. And sometimes, too, they’d have failed to detect both. Under these conditions, cooperating mind readers using different inputs about the means-ends behavior of their predictive targets would have had different behavioral responses. One response would usually have been more adaptive than another, and there would have been selective pressure on any mechanism that would have enabled cooperators to coordinate on that response—and thus also on the emergence of words that would have enabled cooperators to identify and resolve their disagreements. That’s where a theory of mind would have been of significant help: “We both want to trap that wildebeest. You believe it’s going left. But I believe it’s going right. Here’s why…”
Solving the critical design problem/opportunity our ancestors faced in order to survive and move up the food chain on the African savanna selected for adaptive advances in communication among hominins, advances that, by 50,000 years ago or perhaps earlier, resulted in Homo sapiens developing language. Sometime probably pretty late in the process, our ancestors developed the words that eventually were used to form and express a theory of mind, which then, through the workings of some Baldwin-effect mechanism, became (almost) hardwired into our brains. For maximum adaptive payoffs from the very emergence of the theory of mind, however, its expressions couldn’t have operated just in spoken conversations between cooperating partners. For, even to do that, these expressions would have had to invade the mind, usually in the form of silent (subvocal) speech and even images moving through consciousness (Dennett, 1995).
As it developed over evolutionary time, spoken language gave rise to significant improvements in the abilities of our ancestors to cooperate, coordinate, and collaborate with one another in their material and social culture. One of the ways it did so was by enabling them to “coin” the terms the theory of mind now uses and eventually to frame the theory itself. And, of course, since speaking was something our ancestors had done for thousands of years, it was natural for their (and now our) general theory of why people do things, the theory of mind, to use speech to explain not only human actions but also speech itself and language in general—as we have explained meaning and reference, linguistic representation and content in all our languages.1
The way the theory of mind does this job is pretty simple. It just takes the features of the words and sentences of spoken language that we seek to explain and explains them by replicating them inside our minds, in beliefs and desires, with the same labels— “content,” “aboutness,” “representation,” “meaning,” “symbol”—that we use when we speak or write about them.
Science long ago saw the problem with such explanations. They leave completely unexplained what we originally sought to explain. At some point, content, aboutness, representation, meaning, and the like have to be explained in other terms. That’s where neuroscience comes in. If neuroscientists could apply the theory of mind to improve on the treatments of psychopathology, surgery, and pharmacology, as well as on our predictions of normal behavior that are based on the theory, we could all have confidence that the theory of mind was at least on the right track.
But instead of vindicating the theory of mind by unpacking beliefs and desires into their component parts in the brain’s neural circuits, neuroscientists have found that the brain just doesn’t work that way at all. The theory of mind is based on an illusion, one that’s extremely hard to shake owing to its Darwinian pedigree. But even though it was adaptive in the past and is useful, even indispensable in everyday life, and crucial to our social institutions, culture, and the arts, the theory of mind is still fundamentally mistaken as a description of what makes us tick.
Can any part of the theory be reconciled with science? Can we identify the part that was adaptive for our ancestors—that helped them survive and climb from the bottom of the food chain on the African savanna to the top in a few hundred thousand years? Our ability to mind read has been around for much longer than that (Tomasello, 2014), and it still seems to work in our everyday interactions with one another. Surely, there’s something going on that the theory of mind latches on to when we exercise this ability? If we can find out what that something is, perhaps we can reformulate the theory of mind, or construct a successor that will improve on it?
We’ve noted several times that the core application of the theory of mind, the one we use it for most and in which it works best is in coping with the behavior of other people in immediate contact with us over relatively short periods of time. The more people are hidden from our direct observation, and the longer into the future we attempt to use the theory, the more ambiguous and unreliable its predictions become. Within the domain of its optimal application, the theory works well because it treats its subjects—people and other animals, as ends-means systems and combines this assumption with information about their immediate circumstances.
The theory of mind works best when it doesn’t stray very far from the mind-reading ability we share with a few other mammals. Since mind readers share their target animals’ environments, they have some sensory access to what the target animals see, hear, smell, taste, and so on. Mind readers also have sensory access to their target animals’ current behavior and perhaps memory access (somewhere in the hippocampus or the neocortex) to the past behavior of those and similar animals in the local environment. Mind reading, whether in predators, prey, or cooperators, is just a matter of how the brain makes a means-ends calculation from a target animal’s current behavior in its environment to its behavior in the near future. Think of a lion tracking a gazelle. The lion factors the gazelle’s speed and endurance and the terrain to close in and make the kill. Think of the gazelle trying to escape the lion. The gazelle factors the lion’s speed and endurance and the terrain to take evasive action, which it adjusts to match the lion’s pursuit. The behaviors of both lion and gazelle reflect the means-ends “calculation” that mind reading consists in.
The human mind-reading ability comes to the same thing. But we’re even smarter than lions and gazelles. We have a mind-reading ability that can factor more of the shared environment into its “calculation,” our memories are probably better than most other animals, at least for environmental threats and opportunities important to us, and we can pick up subtler cues than most other animals can. For example, only humans and dogs show enough white in their eyes that they can track the others’ gaze and coordinate on what the others are detecting in the environment (Kaminski et al., 2009). The importance of this small difference between dogs and us, on the one hand, and between most other animals and us, on the other, was probably a critical variation that natural selection strongly selected for to enhance cooperation and coordination first among humans, and then between dogs and humans.
Take animals with a highly developed mind-reading ability to track the means-ends behavior of other animals—conspecifics, predators, prey, and coevolved animal companions—early Homo sapiens, for example. Add strong selection for developing anything that will enhance cooperation among such animals, like gesturing and grunting to coordinate action. With any luck and enough time in the case of Homo sapiens, you’ll get language, first spoken, and then even written language or at least symbolic ornamentation 125,000 years later.
Thus the development of language creates a theory of mind out of the means-ends calculation that the human brain’s mind-reading ability consists in. Then that theory provides language use with a ready-made, easily understood explanation of how language comes by the meanings of its words. Meanwhile, the theory also helps Homo sapiens climb to the top of the African food chain and continues to work well enough in interactions both between cooperating people and between them and their predators and prey (human and nonhuman). Eventually, this adaptation overshoots and humans start to use the theory of mind beyond the immediate circumstances that selected for its adaptive improvements on their mind-reading ability. Until, finally, we begin to speculate, invent conspiracies, to narrate, tell stories and write histories.
By this point, natural selection is no longer able to do much about such overshooting and our use of mind reading outside its effective range of application. Having selected for the theory of mind’s placement in the mind-reading module of our brains and the reward system for using it, it can’t easily fine-tune the theory further. Nor can natural selection select for turning the theory off unless and until it becomes clearly harmful for our survival and reproduction.
So now we can see what the theory of mind gets right: it tells us that humans and the other animals it applies to are means-ends systems. But that’s only a small part of the theory, and the rest of the theory is quite wrong. The rest of the theory tells us that humans and other animals produce their means-ends behavior via the mechanism of beliefs, which provide representation of the means, and desires, which provide representations of the ends. If this were actually the way humans and nonhuman animals produce their means-ends behavior, then by narrowing down exactly what these representations were, we’d be able to sort out good (true) explanations of what people did from bad (false) explanations. But, of course, we aren’t able do this, and never have been. And now we can see why we can’t.
Neuroscience shows us that there are no representations in our brains. So there’s nothing there that would let us, even in principle, narrow down exactly what people believed or desired—and no way for us to filter true from false narrative explanations. Thus we end up relying on false explanations, which at best serve only to satisfy our curiosity, and the satisfaction of our curiosity becomes a criterion of explanatory success. Working backward from the means-end behavior of people, we manufacture a pairings of beliefs and desires that would be obvious in the light of the means and ends of people’s actual behavior. The difference between history and biography, on the one hand, and historical fiction, on the other, is simply that in history and biography—at least if carefully done with sufficient reliable information—the actual behavior of real persons constrains the belief-desire pairings we manufacture, whereas in historical fiction, authors are permitted to take liberties with that behavior and to invent persons and events that are not real. It’s no surprise that a good narrative explanation of what historical agents did satisfies our curiosity: in many cases, it’s pretty much a matter of the narrative explainers, when describing what the agents actually did and how that turned out, producing pairings of beliefs and desires to explain the behaviors of those agents, which readers (or listeners) then readily accept. But those pairings, whatever they are, can’t be the right ones, the ones that correctly explain the agents’ behavior, because there are no right belief-desire pairings. In fact, there are no belief-desire pairings at all. There are only confabulations driven by the application of means-ends calculation—means-ends pairings—from the domain of immediate environmental and biological needs, where such calculation works well, to the domain of human affairs, where it doesn’t really work at all. Which means-ends pairings are those? Well, our mind-reading ability can’t discriminate very finely between means-ends pairings when they differ only slightly in their impact on immediate behavior. And the ability to mind read gets worse when it comes to predicting faraway or long-term future trajectories of means-ends behavior, which is the area where the kaiser and Talleyrand operated, the area that interests historians and biographers.
Why did the kaiser issue his “blank check” to Austria-Hungary? Well, there’s no way of telling now by what ends-means processes the words that came out of Wilhelm’s mouth, the documents he signed his name to, the gestures he made as his courtiers waited on his pleasure, were all held together in the kaiser’s mind. The kaiser’s brain, having writ, moved on quickly to some other means-ends processes. But one thing we can say with confidence is that there were no belief-desire pairings in his mind from among which anyone could have chosen the right answer to the question of why Wilhelm issued his “blank check.” The whole dispute about the kaiser’s actual intentions is based on a mistaken theory—the theory of mind.
So now you can see that the theory of mind is a scientific dead end and that it should go the way of other scientific dead ends. But even if you can, it’ll be hard for me to convince you that a theory so useful in everyday life, so crucial to human culture and institutions, and one that’s been around far longer than any other theory humans have ever contrived, roughly as long as language has been, between 50,000 and 75,000 years or so, is a dead end and that we have to let go of if we’re to truly understand ourselves. But a little history of science may help.
There’ve been several dead ends in science: theories that were considered more or less valid and held sway for a certain period, but that were finally given up as not being even on the right track to explaining what they purported to explain. Three famous examples are alchemy, the phlogiston theory of combustion, and the Ptolemaic theory of planetary movement.
Holding sway for at least 1,500 years, alchemy explained the physical world as consisting only of the four elements of earth, air, fire, and water, and the workings of human body as actions of three bodily humors, all of which governed by the spiritual action of an all-powerful deity; it is also now perhaps best known for positing that base metals could be “transmuted” into precious metals. It would finally give way to modern atomic theory and chemistry and to the discovery and scientific description of the many natural elements. On the spectrum of these three scientific dead ends, alchemy is probably the most completely mistaken of the three, though to this day some may think it vindicated at least in part both by the success of nuclear scientists in “transmuting” certain radioactive elements into other, heavier radioactive elements through neutron bombardment and by the natural process of radioactive decay, in which radioactive elements “transmute” into other, lighter elements, both radioactive and not.
The phlogiston theory explained combustion and its results by hypothesizing the existence of a substance called “phlogiston,” released during burning. On careful measurement, phlogiston turned out to have an (impossible) negative mass, couldn’t be unambiguously isolated from other substances, and didn’t combine with other “elements” in ways that made any sense. The phlogiston theory gave way by the end of the seventeenth century to the scientifically proven theory of combustion formulated by Antoine Lavoisier when he discovered and described the properties of oxygen. Although “phlogiston” is a now a byword for the worst explanatory hypotheses of physical science, the phlogiston theory drove the leading research program of seventeenth- and eighteenth-century chemistry, resulting in the invention of some reliable measuring devices, and some of the results of its most important experiments were preserved in the new theory of combustion, even though the phlogiston theorists deeply misunderstood their own experiments.
Dating from about 200 AD, the Ptolemaic theory of planetary motion posited an extremely complicated system to accurately account for the perceived motion of the planets and Sun around the Earth. (As we’ll see, since it has few of the positive and all of the negative features of the Ptolemaic theory of planetary motion, the theory of mind needs to go the way of the Ptolemaic theory, into scientific eclipse.) The theory held that the planets traveled on quite specific paths that could be inferred from their perceived motions in the night sky. These heavenly bodies, which included the Sun, had been long distinguished from other such bodies because they were perceived to “wander” across the night sky, changing their speed and direction as they did. For this reason, they were called “planetai” by the Greeks, meaning “wanderers.” For at least a thousand years, Ptolemaic theory enabled astronomers to predict with increasing accuracy the complicated paths of the planets and the Sun as they moved around the Earth on “deferents” and “epicycles.”
The “deferent” was a circular orbit whose center (the black dot in figure 10.2) was near the Earth. The “epicycle” was another circular orbit whose center moved along the deferent. The combination produced a planetary path, showing how each planet moved across the night sky as observed from the motionless Earth. A planet’s apparent speed was originally supposed to be determined by its epicycle’s distance from the “aequant” (figure 10.2), which was exactly as far from the center of the planet’s orbit (the black dot) as the Earth was. As the theory developed in the millennium after Ptolemy, it was “improved,” as a result of better data. The aequant was replaced by more epicycles, which had “piled up” to many dozen by the time of Nicolaus Copernicus (1473–1545). It was a pretty good theory, the best for its time, at least in its predictive power. No surprise there. The data to be explained were getting more refined and precise all the time and in fact guided the “piling up” of the epicycles.
Ptolemaic model of planetary motion. From https://
But as an explanation of the actual motion of the planets, the Ptolemaic theory is completely wrong. There are no epicycles, no deferents, no aequants in the planets’ actual paths. The truth—that the planets, including the Earth, all revolve around the sun is simple ellipses—is quite different, as Johannes Kepler, Galileo, Isaac Newton, and eventually Albert Einstein discovered. Thus a theory can be firmly believed for a very long time, can provide satisfactory predictions, indeed increasingly accurate predictions, and still be all wrong.
Like the Ptolemaic theory of planetary motion, the theory of mind finds itself having to “pile up the epicycles”: adding and subtracting beliefs, revising and substituting desires, as the evidence mounts of what exactly people did and said or wrote about what they did or were going to do. The trouble for the Ptolemaic theory was that there was never an independent way of checking whether there were any epicycles and deferents at all, let alone the forty to eighty that were required to predict the paths of the planets by the time of Copernicus.2 And the same goes for the theory of mind: since the theory was first formed thousands of years ago, there’s been no way to tell whether beliefs and desires were “inscribed” on the neural circuits of the brain, and if they were, how. So there’s been no independent way of checking whether any particular application of the theory of mind was right, or even on the right track. There’s been no way, independent of the theory of mind, to tell what people believed and desired. The only way to infer desires and beliefs from action, speech, writing, was to use the theory of mind to reason backward from these to the belief-desire pairings that were supposed to cause them. The only way, that is, until modern neuroscience. But instead of vindicating, improving, correcting, updating, refining the theory of mind, the findings of neuroscience have shown that it’s completely wrong.
But the theory of mind is actually worse off than the Ptolemaic theory, which at least could be used to reliably predict the movement of the planets and even to construct devices for calculating calendars, the timing of Olympiads and astronomical events. As I suggested above, the theory of mind is more like phlogiston theory: its set of hypotheses leads us further and further away from the right approach to explaining the phenomena in its domain. Phlogiston theory was advanced in the eighteenth century to explain the causes and effects of combustion—in particular, the observations that substances lose weight when burned. that burning quickly ceases in enclosed spaces, that animals cannot breathe the air in enclosed spaces after burning has occurred, and that metal ores can be converted to pure metals by burning them with charcoal, which burns up almost completely. Each of these observations was explained by invoking “phlogiston,” a substance that when released in burning would reduce the weight of burned substances. Burning things inside enclosed containers would fill them with phlogiston to their capacity, causing the fire to go out. Phlogiston in a closed container would kill animals placed there because it couldn’t be breathed. And since it was held to be the sole ingredient of charcoal and a major ingredient of pure metals, when charcoal and metal ores were heated together, phlogiston would leave the charcoal and convert the ores into pure metals. But, with the discovery of oxygen and the role it plays in combustion and for the other reasons noted above, Lavoisier and other chemists were able to disprove all the claims of the phlogiston theory and to replace it with a correct theory of combustion.
Neuroscience has shown that, despite their appearance, human behaviors aren’t really driven by purposes, ends, or goals. As in all the rest of the biological domain, there are no purposes, just a convincing illusion of purpose. Every behavior that looks like it’s driven by a purpose is just the result of physical processes, like those of blind variation and natural selection uncovered by Darwin, processes that work across geological epochs to select for new species; that work in fetal development to shape tissues, organs, and innate capacities; that work to shape adaptive behavior during growth and maturation. And as neuroscience now has shown, such processes operate even on the neural circuits of our brains to drive real-time behavior.
Neuroscience is completing the scientific revolution by banishing purpose from the last domain where it’s still invoked to explain and predict. Newton banished purpose from physics, and Darwin banished it from biology. Eric Kandel, John O’Keefe, May-Britt and Edvard Moser, and their fellow neuroscientists have between them begun the process of banishing purpose from psychology. John Watson, Ivan Pavlov, B. F. Skinner, and the other behaviorists tried to do this a half century or more before them. They saw that purpose has no place in the physical world. But they had no acceptable substitute for goals, ends, purposes inscribed in our brains. So, they used bluster, accusing the theory of mind of being empty, unfalsifiable, without empirical content. They were mistaken about that theory, of course. Indeed, the Ptolemaic theory of planetary of motion, the phlogiston theory of combustion, and the theory of mind all have enough empirical content to enable scientists to enlist or devise the necessary equipment to test their claims: Galileo’s telescope, Lavoisier’s pan balance, O’Keefe’s microelectrode arrays. And because all three theories are false, science has managed to falsify them.
Notes
1. A fine example of how the theory of mind does this is to be found in the work of the well-known twentieth-century philosopher of language Paul Grice, who formalized the role the theory of mind is supposed to play in establishing the meaning of what people say when they speak (see, for example, Grice, 1991).
Think back to the dawn of language in the gestures and grunts of the leader of a hunt. What makes his grunt mean “Now is the time to attack.” According to Grice’s widely accepted model of what gives vocalizations meaning, for the leader’s grunt to mean “Now is the time to attack,” three iterated, nested, and recursive conditions of the grunter’s state of mind had to be met: he had to want the other hunters to believe (1) that it was time to attack; (2) that he grunted because he wanted them to believe it was time to attack; and perhaps most important, (3) that it was time to attack because they believed that he grunted to get them to believe that it was time to attack. You probably have to read through these three conditions several times even to grasp them. It’s also worth working out why the theory of mind holds each of these condition to be necessary for the leader’s grunt to really mean “Now is the time to attack.”
Imagine our lead hunter had the nervous habit of grunting involuntarily every time he thought it was the moment to attack a prey, and that he was a very successful hunter that others were prepared to follow and emulate. Well, after a while, the other hunters might come to see the leader’s grunt as a sign (the way clouds are a sign that it may rain) that it was time to attack. But, in this scenario, according to the theory of mind, the leader’s grunt wouldn’t carry the meaning that it was time to attack any more than clouds carry the meaning that it may rain. Why wouldn’t the grunt mean “Now is the time to attack” in this case? Well, let’s work through the three conditions. The hunter didn’t grunt because he wanted the others to believe that it was time to attack. His grunt was “involuntary,” that is, not caused by his beliefs and desires. It’s safe to assume he wanted the other hunters to believe that it was time to attack, but that’s not why he grunted, so the second condition is not met. And since his grunt was not caused by his beliefs and desires, it wouldn’t meet the third condition either—that he grunted because he wanted them to believe that it was time to attack because they believed that he grunted to get them to believe that it was time to attack.
Now it’s pretty obvious that, at the outset of hominins’ or even Homo sapiens’ earliest attempts at some sort of vocal language, they’d need to have a full-blown theory of mind in order to accord the sounds they made the kind of meaning the theory requires. But since the theory already requires that they have a well-developed language, it’s pretty clear that early humans’ vocalizations couldn’t meet these three conditions and so the sounds they made couldn’t have the sort of meaning the theory of mind required them to have.
2. That is, there was no way until Galileo first trained his telescope on the moons of Jupiter in 1610. One look was all he needed. The moons traveled around Jupiter in simple ellipses, not in epicycles.