2

FATHOMING UNCONSCIOUS DEPTHS

How deep can an invisible image travel into the brain? Can it reach our higher cortical centers and influence the decisions we make? Answering these questions is crucial to delineating the unique contours of conscious thought. Recent experiments in psychology and brain imaging have tracked the fate of unconscious pictures in the brain. We recognize and categorize masked images unconsciously, and we even decipher and interpret unseen words. Subliminal pictures trigger motivations and rewards in us—all without our awareness. Even complex operations linking perception to action can unfold covertly, demonstrating how frequently we rely on an unconscious “automatic pilot.” Oblivious to this boiling hodgepodge of unconscious processes, we constantly overestimate the power of our consciousness in making decisions—but in truth, our capacity for conscious control is limited.

Time past and time future allow but a little consciousness.

—T. S. Eliot, Burnt Norton (1935)

During the 2000 presidential campaign, a nasty commercial concocted by George W. Bush’s team featured a caricature of Al Gore’s economic plan, accompanied by the word RATS in huge capital letters (figure 8). Although not strictly subliminal, the image went largely unnoticed, for it flew by inconspicuously at the end of the word bureaucrats. The offending epithet stirred a debate: Did the viewer’s brain register the hidden meaning? How far did it travel in the brain? Could it reach the voter’s emotional center and influence an electoral decision?

FIGURE 8. Subliminal images are occasionally used in the media. During the 1988 French presidential campaign, the face of president and candidate François Mitterrand was briefly flashed within the logo of the major public TV program. In 2000, in one of George W. Bush’s commercials, Al Gore’s economic plan was surreptitiously labeled with the word RATS. Are such unconscious images processed by the brain, and do they influence our decisions?

The French elections, twelve years earlier, had been the theater of an even more controversial use of subliminal images. The face of presidential candidate François Mitterrand was briefly flashed within the logo of the main state television program (figure 8). This invisible image appeared daily at the opening of the eight p.m. news broadcast, a popular program for French viewers. Did it bias the votes? Even a very small shift, in a nation of fifty-five million, would mean thousands of ballots.

The mother of all subliminal manipulations is the (in)famous insertion of a frame with the words Drink Coca Cola into a 1957 movie. Everybody knows the story and its outcome: a massive increase in sales of soft drinks. Yet this foundational myth of subliminal research was a complete fabrication. James Vicary made up the story and later admitted that the experiment was a hoax. Only the myth persists, and so does the scientific question: Can unseen images influence our thoughts? This is not just an important issue for freedom and mass manipulation, but also a key interrogation for our scientific understanding of the brain. Do we need to be conscious of an image in order to process it? Or can we perceive, categorize, and decide without awareness?

The issue has become all the more pressing now that a variety of methods exists for presenting information to the brain in an unconscious manner. Binocular images, inattention, masking, and many other situations render us oblivious to many aspects of our surroundings. Are we just blind to them? Whenever we attend to a given object, do we cease to perceive all the unattended surroundings? Or do we continue to process them, but in a subliminal manner? And if we do, how far can they progress into the brain without receiving the beam of consciousness?

Answering those questions is particularly crucial for our scientific goal of identifying the brain signatures of conscious experience. If subliminal processing is deep, and if we can fathom that depth, then we will understand the nature of consciousness much better. Once we know, for instance, that the early stages of perception can operate without awareness, we will be able to exclude them from our search for consciousness. By extending this process of elimination to higher-level operations, we will learn more and more about the specifics of the conscious mind. Delineating the contours of the unconscious will progressively print a negative photograph of the conscious mind.

Pioneers of the Unconscious

The discovery that a dramatic amount of mental processing occurs outside our awareness is generally credited to Sigmund Freud (1856–1939). However, this is a myth, crafted in large part by Freud himself.1 As noted by the historian and philosopher Marcel Gauchet, “When Freud declares, in substance, that prior to psychoanalysis the mind was systematically identified with consciousness, we have to declare this statement rigorously false.”2

In truth, the realization that many of our mental operations occur sub rosa, and that consciousness is only a thin veneer lying atop sundry unconscious processors, predates Freud by decades or even centuries.3 In Roman antiquity, the physician Galen (ca. 129–200) and the philosopher Plotinus (ca. 204–270) had already noticed that some of the body’s operations, such as walking and breathing, occur without attention. Much of their medical knowledge was in fact inherited from Hippocrates (ca. 460–377 BC), a keen observer of diseases whose name remains an emblem of the medical profession. Hippocrates wrote an entire treatise on epilepsy, called The Sacred Disease, in which he noted that the body suddenly misbehaves against its owner’s will. He concluded that the brain constantly controls us and covertly weaves the fabric of our mental life:

Men ought to know that from the brain, and from the brain alone, arise our pleasures, joys, laughter and jests, as well as our sorrows, pains, grieves and tears. Through it, in particular, we think, see, hear and distinguish the ugly from the beautiful, the bad from the good, the pleasant from the unpleasant.

During the Dark Ages, which followed the fall of the Roman Empire, Indian and Arab scholars preserved some of antiquity’s medical wisdom. In the eleventh century, the Arab scientist known as Alhazen (Ibn al-Haytham, 965–1040) discovered the main principles of visual perception. Centuries before Descartes, he understood that the eye operates as a camera obscura, a receiver rather than an emitter of light, and he foresaw that various illusions could fool our conscious perception.4 Consciousness was not always in control, Alhazen concluded. He was the first to postulate an automatic process of unconscious inference: unknown to us, the brain jumps to conclusions beyond the available sense data, sometimes causing us to see things that are not there.5 Eight centuries later the physicist Hermann von Helmholtz, in his 1867 book, Physiological Optics, would use the very same term, unconscious inference, to describe how our vision automatically computes the best interpretation compatible with incoming sense data.

Beyond the issue of unconscious perception lay the greater issue of the origins of our deepest motives and desires. Centuries before Freud, many philosophers—including Augustine (354–430), Thomas Aquinas (1225–74), Descartes (1596–1650), Spinoza (1632–77), and Leibniz (1646–1716)—noted that the course of human actions is driven by a broad array of mechanisms that are inaccessible to introspection, from sensorimotor reflexes to unaware motives and hidden desires. Spinoza cited a hodgepodge of unconscious drives: a child’s desire for milk, an injured person’s will for revenge, a drunkard’s craving for the bottle, and a chatterbox’s uncontrollable speech.

During the eighteenth and nineteenth centuries, the first neurologists discovered proof after proof of the omnipresence of unconscious circuits in the nervous system. Marshall Hall (1790–1857) pioneered the concept of a “reflex arc,” linking specific sensory inputs to particular motor outputs, and he emphasized our lack of voluntary control over basic movements that originate in the spinal cord. Following in his footsteps, John Hughlings Jackson (1835–1911) underscored the hierarchical organization of the nervous system, from the brain stem to the cerebral cortex and from automatic operations to increasingly voluntary and conscious ones. In France, the psychologists and sociologists Théodule Ribot (1839–1916), Gabriel Tarde (1843–1904), and Pierre Janet (1859–1947) stressed the broad range of human automatisms, from practical knowledge stored in our action memory (Ribot) to unconscious imitation (Tarde) and even to subconscious goals that date from early childhood and become defining facets of our personality (Janet).

French scientists were so advanced that when the ambitious Freud published his first claims to fame, Janet protested that he owned the paternity of many of Freud’s ideas. As early as 1868, the British psychiatrist Henry Maudsley (1835–1918) had written that “the most important part of mental action, the essential process on which thinking depends, is unconscious mental activity.”6 Another contemporary neurologist, Sigmund Exner, who was Freud’s colleague in Vienna, had stated in 1899: “We shouldn’t say ‘I think,’ ‘I feel,’ but rather ‘it thinks in me’ [es denkt in mir], ‘it feels in me’ [es fühlt in mir]”—a full twenty years prior to Freud’s reflections in The Ego and the Id (Das Ich und das Es), published in 1923.

At the turn of the century, the ubiquity of unconscious processes was so well accepted that in his major treatise The Principles of Psychology (1890), the great American psychologist and philosopher William James could boldly state: “All these facts, taken together, form unquestionably the beginning of an inquiry which is destined to throw a new light into the very abysses of our nature. . . . They prove one thing conclusively, namely, that we must never take a person’s testimony, however sincere, that he has felt nothing, as proof positive that no feeling has been there.”7 Any human subject, he surmised, “will do all sorts of incongruous things of which he remains quite unaware.”

Relative to this flurry of neurological and psychological observations, clearly demonstrating that unconscious mechanisms drive much of our lives, Freud’s own contribution appears speculative. It would not be a huge exaggeration to say that in his work, the ideas that are solid are not his own, while those that are his own are not solid. In hindsight, it is particularly disappointing that Freud never tried to put his views to an empirical test. The late nineteenth and early twentieth centuries saw the birth of experimental psychology. New empirical methods flourished, including the systematic collection of precise response times and errors. But Freud seemed content with proposing metaphorical models of the mind without seriously testing them. One of my favorite writers, Vladimir Nabokov, had no patience with Freud’s method and nastily barked: “Let the credulous and the vulgar continue to believe that all mental woes can be cured by a daily application of old Greek myths to their private parts. I really do not care.”8

The Seat of Unconscious Operations

In spite of the major medical advances of the nineteenth and twentieth centuries, only twenty years ago, in the 1990s, when my colleagues and I started to apply brain-imaging techniques to subliminal perception, an enormous amount of confusion still surrounded the issue of unseen pictures in the brain. Many conflicting accounts of a division of labor were being proposed. The simplest idea was that the cortex—the folded sheets of neurons that form the surface of our two cerebral hemispheres—was conscious while all the other circuits were not. The cortex, the most evolved part of the brain in mammals, hosts the advanced operations that underlie attending, planning, and speaking. Thus, it was a fairly natural hypothesis to consider that whatever information reached the cortex had to be conscious. Unconscious operations, by contrast, were thought to take place solely within specialized brain nuclei such as the amygdala or the colliculus, which had evolved to perform dedicated functions such as the detection of fearful stimuli or the movement of the eyes. These groups of neurons form “subcortical” circuits, so called because they lie underneath the cortex.

A different but equally naïve proposal introduced a dichotomy between the two hemispheres of the brain. The left hemisphere, which hosts the language circuits, could report on what it was doing. Therefore it would be conscious, while the right wouldn’t.

A third hypothesis was that some cortical circuits were conscious, while others were not. Specifically, whatever visual information is carried through the brain by the ventral route, which recognizes the identity of objects and faces, would necessarily be conscious. Meanwhile information carried by the dorsal visual route, which goes through the parietal cortex and uses object shape and location to guide our actions, would forever lie on the unconscious dark side.

None of these simplistic dichotomies held up to scrutiny. Based on what we now know, virtually all the brain’s regions can participate in both conscious and unconscious thought. To get to this conclusion, however, clever experiments were needed to progressively expand our understanding of the range of the unconscious.

Initially, simple experiments in patients with brain injuries suggested that unconscious operations brooded in the hidden basement of the brain, beneath the cortex. The amygdala, for instance, an almond-shaped group of neurons located beneath the temporal lobe, flags important, emotionally laden situations of everyday life. It is particularly crucial for coding fear; frightening stimuli, such as the sight of a snake, can activate it on a fast track from the retina, well before we register the emotion at a conscious cortical level.9 Many experiments have indicated that such emotional appraisals are made extraordinarily quickly and unconsciously, mediated by the fast circuitry of the amygdala. Early in the 1900s, the Swiss neurologist Édouard Claparède demonstrated an unconscious emotional memory: while he was shaking the hand of an amnesic patient, he pricked her with a pin, and the next day, while her amnesia prevented her from recognizing him, she emphatically refused to shake his hand. Such experiments provided a first proof that complex emotional operations could unfold below the level of awareness, and they always seemed to arise from a set of subcortical nuclei specialized for emotional processing.

Another source of data on subliminal processing was “blindsight” patients, those with lesions of the primary visual cortex, the main source of visual inputs into the cortex. The oxymoronic term blindsight may seem bizarre, but it accurately describes these individuals’ Shakespearean condition: to see, but not to see. A lesion in the primary visual cortex should make a person blind, and it does deprive such patients of their conscious vision—they assure you that they cannot see anything in a specific part of the visual field (which corresponds precisely to the destroyed area of cortex), and they behave as if they were blind. Incredibly enough, however, when an experimenter shows them objects or flashes of light, they accurately point to them.10 In a zombielike manner, they unconsciously guide their hand to locations that they do not see—blindsight indeed.

Which intact anatomical pathways support unconscious vision in blindsight patients? Clearly, in these patients, some visual information still makes it through from the retina to the hand, bypassing the lesion that makes them blind. Because the entry point into the patients’ visual cortex had been destroyed, the researchers initially suspected that their unconscious behavior arose entirely from subcortical circuits. A key suspect was the superior colliculus, a nucleus in the midbrain that specializes in the cross-registration of vision, eye movements, and other spatial responses. Indeed, the first functional MRI study of blindsight demonstrated that unseen targets triggered a strong activation in the superior colliculus.11 But that study also contained evidence that the unseen stimuli evoked activations in the cortex—and sure enough, later research confirmed that invisible stimuli could still activate both the thalamus and higher-level visual areas of the cortex, somehow bypassing the damaged primary visual area.12 Clearly, the brain circuits that take part in our unconscious inner zombie and that guide our eye and hand movements include much more than just old subcortical routes.

Another patient, studied by the Canadian psychologist Melvyn Goodale, strengthened the case for a cortical contribution to unconscious processing. At the age of thirty-four, D.F. suffered carbon monoxide intoxication.13 Lack of oxygen caused dramatic and irreversible damage to her left and right lateral visual cortexes. As a result, she lost some of the most basic aspects of conscious vision, developing what neurologists call “visual form agnosia.” For purposes of shape recognition, D.F. was essentially blind—she could not tell a square from an elongated rectangle. Her deficit was so severe that she failed to recognize the orientation of a slanted line (vertical, horizontal, or oblique). Yet her gesture system was still remarkably functional: when asked to post a card through a slanted slit, whose orientation she consistently failed to perceive, her hand behaved with perfect accuracy. Her motor system always seemed to unconsciously “see” things better than she could consciously. She also adapted the size of her grasp to the objects that she reached for—yet she was utterly unable to do so voluntarily, using the finger-to-thumb distance as a symbolic gesture for perceived size.

D.F.’s unconscious ability to perform motor actions seemed to vastly exceed her capacity for consciously perceiving the same visual shapes. Goodale and his collaborators argued that her performance could not be explained solely by subcortical motor pathways but had to involve the cortex of the parietal lobes as well. Although D.F. was unaware of it, information about the size and orientation of objects was still proceeding unconsciously down her occipital and parietal lobes. There, intact circuits extracted visual information about size, location, and even shape that she could not consciously see.

Since then, severe blindsight and agnosia have been studied in a host of similar patients. Some of them could navigate a busy hallway without bumping into objects, all the while claiming total blindness. Other patients experienced a form of unconsciousness called “spatial neglect.” In this fascinating condition, a lesion to the right hemisphere, typically in the vicinity of the inferior parietal lobe, prevents a patient from attending to the left side of space. As a result, he or she often misses the entire left half of a scene or an object. One patient forcefully complained about not being given enough food: he had eaten all the food on the right side of his plate but failed to notice that the left half was still full.

Spatial neglect patients, while dramatically impaired in their conscious judgments and reports, are not truly blind in the left visual field. Their retinas and early visual cortex are perfectly functional, yet somehow a higher-level lesion prevents them from attending this information and registering it at a conscious level. Is the unattended information totally lost? The answer is no: the cortex still processes the neglected information, but at an unconscious level. John Marshall and Peter Halligan elegantly made this point by showing a spatial neglect patient pictures of two houses, one of which was on fire on the left side (figure 9).14 The patient forcefully denied seeing any difference between them—he claimed that the houses were identical. But when asked to choose which one he would prefer to live in, he consistently avoided picking the one on fire. Obviously, his brain was still processing visual information deeply enough that it could categorize the fire as a danger to be avoided. A few years later, brain-imaging techniques showed that in spatial neglect patients, an unseen stimulus could still activate the regions of the ventral visual cortex that respond to houses and faces.15 Even the meaning of neglected words and numbers invisibly made its way into the patients’ brain.16

FIGURE 9. Patients with brain lesions provided the first solid evidence that unconscious images are processed in the cortex. Following a brain lesion, Goodale and Milner’s (1991) patient D.F. lost all visual recognition ability and became utterly unable to perceive and describe shapes, even one as simple as a slanted slit (above). Nevertheless, she could accurately post a card through it, suggesting that complex hand movements can be guided unconsciously. Marshall and Halligan’s (1988) patient P.S., who suffered from massive neglect of the left side of space, failed to consciously perceive any difference between the two houses below. Yet when asked which one he would prefer to live in, he consistently avoided the house on fire, suggesting that he unconsciously understood the meaning of the drawing.

The Brain’s Dark Side

All this evidence initially arose from patients with severe and often massive brain lesions that had arguably altered the separation between conscious and unconscious operations. Do normal brains, in the absence of a lesion, also process images unconsciously at a deep visual level? Can our cortex operate without our awareness? Might even the sophisticated functions that we acquire at school, such as reading or arithmetic, execute unconsciously? My laboratory was among the first to provide a positive answer to these important questions; we used brain imaging to demonstrate that invisible words and digits reach quite deep in the cortex.

As I explained in Chapter 1, we can flash a picture for several dozen milliseconds, yet keep it unseen. The trick is to mask the critical event that we wish to hide from consciousness with other shapes just before and after it (see figure 7). But how far does such a masked picture travel in the brain? My colleagues and I got an indication by using the clever technique of “subliminal priming.” We briefly flashed a subliminal word or picture (dubbed the prime) and immediately followed it with another visible item (the target). On successive trials, the target might be identical to the prime or different from it. For example, we flashed the prime word house so briefly that the participants did not see it, and then the target word radio long enough to be consciously visible. The participants did not even realize that there had been a hidden word. They focused only on the visible target word, and we measured how much time they needed to recognize it by asking them to press one key if it referred to a living thing and another if it referred to an artifact. (Virtually any task will do.)

The fascinating finding, replicated in dozens of experiments, is that the prior presentation of a word, even unconsciously, speeds up its processing when the same word reappears consciously.17 As long as the two presentations are separated by less than a second, repetition leads to facilitation—even when it goes totally undetected. Thus people respond faster and make fewer errors when radio precedes radio than when an unrelated word such as house is presented. This finding is called “subliminal repetition priming.” Much as one primes a pump by flushing water into it, we can prime the circuit for word processing by an unseen word.

We now know that the priming information that is sent down the brain can be quite abstract. For instance, priming works even when the prime is in lower case (radio) and the target in upper case (RADIO). Visually, these shapes are radically different. The lowercase a looks nothing like the uppercase A. Only a cultural convention attaches those two shapes to the same letter. Amazingly, experiments show that, in expert readers, this knowledge has become totally unconscious and is compiled in the early visual system: subliminal priming is just as powerful when the same physical word is repeated (radio-radio) as when the case is changed (radio-RADIO).18 Hence unconscious information proceeds all the way up to an abstract representation of letter strings. From the mere glimpse of a word, the brain manages to quickly identify the letters independently of superficial changes in letter shapes.

The next step was to understand where this operation occurs. As my colleagues and I proved, brain imaging is sensitive enough to identify the small activation elicited by an unconscious word.19 Using functional magnetic resonance imaging (fMRI), we made whole-brain pictures of areas that were affected by subliminal priming. The results showed that a large chunk of the ventral visual cortex could be activated unconsciously. The circuit included a region called the fusiform gyrus, which houses advanced mechanisms of shape recognition and implements the early stages of reading.20 Here priming did not depend on the shape of the word: this brain area was clearly able to process the abstract identity of a word without caring whether it was in upper or lower case.21

Prior to these experiments, some researchers had postulated that the fusiform gyrus always took part in conscious processing. It formed the so-called ventral visual route that allowed us to see shapes. Only the “dorsal route,” they thought, linking the occipital visual cortex with the action systems of the parietal cortex, was the seat of unconscious operations.22 By demonstrating that the ventral route, which cares about the identity of pictures and words, could also operate in an unconscious mode, our experiments and others helped dispel the simplistic idea that the ventral route was conscious while the dorsal route was not.23 Both circuits, although they are seated high up in the cortex, appeared to be capable of operating below the level of conscious experience.

Binding Without Consciousness

Year after year research on subliminal priming has dispelled many myths about the role of consciousness in our vision. One now-discarded idea was that, although the individual elements of a visual scene could be processed without awareness, consciousness was needed to bind them together. Without conscious attention, features such as motion and color floated freely around and were not bound together into the appropriate objects.24 The various sites of the brain had to piece the information together into a single “binder” or “object file” before a global percept could arise. Some researchers postulated that this binding process, made possible by neuronal synchrony25 or reentry,26 was the hallmark of conscious processing.

We now know that they were wrong: some visual bindings can occur without consciousness. Consider the binding of letters into a word. The letters must clearly be attached together in a precise left-to-right arrangement, so as not to confuse words like RANGE and ANGER, where the movement of a single letter makes a huge difference. Our experiments demonstrated that such binding is achieved unconsciously.27 We found that subliminal repetition priming occurred when the word RANGE was preceded by range, but not when RANGE was preceded by anger—indicating that subliminal processing is highly sensitive, not just to the presence of letters but also to how they are arranged. In fact, responses to RANGE preceded by anger were no faster than responses to RANGE preceded by an unrelated word such as tulip. Subliminal perception is not fooled by words that have 80 percent of their letters in common: a single letter can radically alter the pattern of subliminal priming.

In the past ten years, such demonstrations of subliminal perception have been replicated hundreds of times—not just for written words but also for faces, pictures, and drawings.28 They led to the conclusion that what we experience as a conscious visual scene is a highly processed image, quite different from the raw input that we receive from the eyes. We never see the world as our retina sees it. In fact, it would be a pretty horrible sight: a highly distorted set of light and dark pixels, blown up toward the center of the retina, masked by blood vessels, with a massive hole at the location of the “blind spot” where cables leave for the brain; the image would constantly blur and change as our gaze moved around. What we see, instead, is a three-dimensional scene, corrected for retinal defects, mended at the blind spot, stabilized for our eye and head movements, and massively reinterpreted based on our previous experience of similar visual scenes. All these operations unfold unconsciously—although many of them are so complicated that they resist computer modeling. For instance, our visual system detects the presence of shadows in the image and removes them (figure 10). At a glance, our brain unconsciously infers the sources of lights and deduces the shape, opacity, reflectance, and luminance of the objects.

FIGURE 10. Powerful unconscious computations lie beneath our vision. Glance at this image, and you see a normal-looking checkerboard. You have no doubt that square A is dark and square B is light. But amazingly, they are printed in the same exact shade of gray. (Check this by masking the image with a sheet of paper.) How can we explain this illusion? In a fraction of a second, your brain unconsciously parses the scene into objects, decides that the light comes from the top right, detects that the cylinder casts a shadow on the board, and subtracts this shadow from the image, letting you see what it infers are the true colors of the checkerboard beneath it. Only the final result of these complex computations enters your conscious awareness.

Whenever we open our eyes, a massively parallel operation takes place in our visual cortex—but we are unaware of it. Uninformed of the inner workings of our vision, we believe that the brain works hard only when we feel that we are working hard—for instance, when we’re doing math or playing chess. We have no idea how hard it is also working behind the scenes to create this simple impression of a seamless visual world.

Playing Chess Unconsciously

For another demonstration of the power of our unconscious vision, consider chess playing. When grand master Garry Kasparov concentrates on a chess game, does he have to consciously attend to the configuration of pieces in order to notice that, say, a black rook is threatening the white queen? Or can he focus on the master plan, while his visual system automatically processes those relatively trivial relations among pieces?

Our intuition is that in chess experts, the parsing of board games becomes a reflex. Indeed, research proves that a single glance is enough for any grand master to evaluate a chessboard and to remember its configuration in full detail, because he automatically parses it into meaningful chunks.29 Furthermore, a recent experiment indicates that this segmenting process is truly unconscious: a simplified game can be flashed for 20 milliseconds, sandwiched between masks that make it invisible, and still influence a chess master’s decision.30 The experiment works only on expert chess players, and only if they are solving a meaningful problem, such as determining if the king is under check or not. It implies that the visual system takes into account the identity of the pieces (rook or knight) and their locations, then quickly binds together this information into a meaningful chunk (“black king under check”). These sophisticated operations occur entirely outside conscious awareness.

Seeing Voices

All our examples so far have come from vision. Could consciousness be the glue that binds our distinct sensory modalities into a coherent whole? Do we need to be conscious in order to fuse together visual and auditory signals, as when we enjoy a movie? Again, the surprising answer is no. Even multisensory information can be bound together unconsciously—we become aware only of the result. We owe this conclusion to a remarkable illusion called the “McGurk effect,” first described by Harry McGurk and John MacDonald in 1976.31 The video, which can be found on the Internet,32 shows a person speaking, and it seems obvious that she is saying da da da da. Nothing puzzling—until you close your eyes and realize that the true auditory stimulus is the syllable ba ba ba! How does the illusion work? Visually, the mouth of the person moves to say ga—but because your ears receive the syllable ba, your brain is confronted with a conflict. It solves it, unconsciously, by fusing the two pieces of information. If the two inputs are well synchronized, it binds the information together into a single intermediate percept: the syllable da, a compromise between the auditory ba and the visual ga.

This auditory illusion shows us again how late and reconstructed our conscious experience is. As surprising as it seems, we do not hear the sound waves that reach our ears; nor do we see the photons entering our eyes. What we gain access to is not a raw sensation but an expert reconstruction of the outside world. Behind the scenes, our brain acts as a clever sleuth that ponders all the separate pieces of sensory information we receive, weighs them according to their reliability, and binds them into a coherent whole. Subjectively, it does not feel like any of it is reconstructed. We do not have the impression of inferring the identity of the fused sound da—we just hear it. Nevertheless, during the McGurk effect, what we hear demonstrably arises from sight just as much as from sound.

Where in the brain is this conscious multisensory brew concocted? Brain imaging suggests that it is in the frontal cortex, rather than in the early auditory or visual sensory areas, that the conscious outcome of the McGurk illusion is finally represented.33 The content of our conscious perception is first distilled within our higher areas, then is sent back to early sensory regions. Clearly, many complex sensory operations unfold sub rosa to assemble the scene that eventually plays out seamlessly in our mind’s eye, as if coming straight from our sensory organs.

Can just any information be assembled unconsciously? Probably not. Vision, speech recognition, and expert chess have something in common—they are all extremely automatic and overlearned. This is presumably why their information can be bound without awareness. The neurophysiologist Wolf Singer has suggested that we should perhaps distinguish two types of bindings.34 Routine bindings would be those that are coded by dedicated neurons committed to specific combinations of sensory inputs. Nonroutine bindings, by contrast, are those that require the de novo creation of unforeseen combinations—and they may be mediated by a more conscious state of brain synchrony.

This more nuanced view of how our cortex synthesizes our perceptions seems much more likely to be correct. From birth on, the brain receives intensive training in what the world looks like. Years of interaction with the environment allow it to compile detailed statistics of which parts of objects tend to frequently co-occur. With intensive experience, visual neurons become dedicated to the specific combination of parts that characterizes a familiar object.35 After learning, they continue to respond to the appropriate combination even during anesthesia—a clear proof that this form of binding does not require consciousness. Our capacity to recognize written words probably owes much to such unconscious statistical learning: by adulthood, the average reader has seen millions of words, and his or her visual cortex is likely to contain neurons committed to identifying frequent letter strings such as the, un, and tion.³⁶ In expert chess players, likewise, a fraction of neurons may become attuned to chessboard configurations. This sort of automatic binding, compiled into dedicated brain circuits, is quite different from, say, the binding of new words into a sentence. When you smile at Groucho Marx’s sentence “Time flies like an arrow; fruit flies like a banana,” these words bind for the first time in your brain—and part of that combination, at least, seems to require consciousness. Indeed, brain-imaging experiments show that during anesthesia, our brain’s capacity to integrate words into sentences is strongly reduced.37

Unconscious Meaning?

Our visual system is clever enough to unconsciously assemble several letters into a word—but can the word’s meaning also be processed without awareness? Or is consciousness needed to understand even a single word? This deceptively simple question has turned out to be fiendishly difficult to answer. Two generations of scientists have fought over it like mad dogs—each camp persuaded that its answer was obvious.

How could word comprehension not require a conscious mind? If one defines consciousness as “the perception of what passes in a man’s own mind,” as John Locke did in his celebrated Essay Concerning Human Understanding (1690), then it is hard to see how the mind could grasp a word’s meaning without, at the same time, becoming aware of it. Comprehension (etymologically, “together-catching,” the assembling of fragments of meaning in “common sense”) and consciousness (“together-knowing”) are so closely connected in our mind as to be virtually synonymous.

And yet how could language operate if the elementary process of word comprehension required consciousness? As you read this sentence, do you consciously work out each word’s meaning before assembling the words together into a coherent message? No: your conscious mind focuses on the overall gist, the logic of the argument. A glance at each word is enough to place it within the overall structure of discourse. We have no introspection of how a sign evokes a meaning.

So who is right? Thirty years of research in psychology and brain imaging have finally settled the issue. The story of how it was done is interesting, a wild waltz of conjectures and refutations progressively converging toward a stable truth.

It all started in the 1950s with studies of the “cocktail party” effect.38 Picture yourself at a noisy party. Dozens of conversations around you mix up, but you manage to concentrate on just one of them. Your attention operates as a filter that selects one voice and thwarts all others. Or does it? The British psychologist Donald Broadbent postulated that attention acts as an early filter that interrupts processing at a low level: unattended voices are blocked at a perceptual level, he surmised, before they can have any influence on comprehension.39 But this view does not survive scrutiny. Imagine that suddenly one of the party’s guests, standing behind you, casually calls your name, even in a low voice. Immediately your attention switches to that speaker. This implies that your brain did indeed process the unattended word, all the way up to a representation of its meaning as a proper name.40 Careful experimentation confirms this effect and even shows that unattended words can bias a listener’s judgment of the conversation that he or she focuses on.41

Cocktail party and other divided-attention experiments suggest an unconscious comprehension process, but do they offer watertight evidence? No. In those experiments, listeners deny splitting their attention and swear that they could not hear the unattended stream (that is, before their name was called), but how can we be sure? Skeptics easily destroy such experiments by denying that the unattended stream is truly unconscious. Perhaps the listener’s attention switches very quickly from one stream to the other, or perhaps one or two words pass through during a blank period. The cocktail party effect, although impressive in a real-life context, was hard to transform into a laboratory test of unconscious processing.

In the 1970s the Cambridge psychologist Anthony Marcel went one step further. He used the masking technique to flash words below the threshold of conscious perception. With this method, he achieved complete invisibility: every participant, on every trial, denied seeing any word. Even when they were told that a hidden word was present, they could not perceive it. When they were asked to venture a response, they remained unable to say whether the hidden string was an English word or just a random string of consonants. Nevertheless, Marcel was able to show that the participants’ brains processed the hidden word unconsciously all the way to its meaning.42 In a key experiment, he flashed a color word such as blue or red. Participants denied seeing the word, but when they were subsequently asked to choose a patch of the corresponding color, they were faster by about one-twentieth of a second than when they had been exposed to another, unrelated word. Thus, an unseen color word could prime them to choose the corresponding color. This seemed to imply that their brains had unconsciously registered the meaning of the hidden word.

Marcel’s experiments uncovered another striking phenomenon: the brain seemed to unconsciously process all possible meanings of words, even when they were ambiguous or irrelevant.43 Imagine that I whisper in your ear the word bank. A financial institution comes to your mind—but on second thought, perhaps I meant the edge of a river. Consciously, we seem to become aware of only one meaning at a time. Which meaning gets selected is clearly biased by the context: seeing the word bank in the context of Robert Redford’s beautiful 1992 movie A River Runs Through It primes the water-related meaning. In the lab, one can show that even a single word, such as river, suffices to make the word bank prime the word water, while seeing save before bank primes the word money.44

Crucially, this adaptation to context seems to occur only at the conscious level. When the prime word was masked down to a subliminal level, Marcel observed a joint activation of both meanings. After flashing the word bank, both money and water were primed—even when a strong context favored the river meaning. Thus our unconscious mind is clever enough to store and retrieve, in parallel, all the possible semantic associations of a word, even when the word is ambiguous and even when only one of its meanings actually fits in the context. The unconscious mind proposes while the conscious mind selects.

The Great Unconscious Wars

Marcel’s semantic priming experiments were very creative. They strongly suggested that sophisticated processing of a word’s meaning could occur unconsciously. But they were not watertight, and the true skeptics remained unmoved.45 Their skepticism launched a massive fight between the champions and the detractors of unconscious semantic processing.

Their disbelief was not entirely unjustified. After all, the subliminal influence that Marcel found was so small that it was close to negligible. Flashing a word facilitated processing by a very small amount, sometimes less than one-hundredth of a second. Perhaps this effect arose from a very small fraction of trials on which the hidden word had, in fact, been seen—albeit so briefly as to leave very little or no trace in memory. Marcel’s primes were not always unconscious, his detractors argued. In their opinion, the participants’ mere verbal report of “I didn’t see any words,” recorded only at the end of the experiment, failed to provide convincing evidence that they had never seen the prime words. Much greater care was needed to measure prime awareness as objectively as possible, in a separate experiment in which subjects were asked, for instance, to venture a name for the hidden word, or to categorize it according to some criterion. Only random performance on this secondary task, the skeptics contended, would indicate that the primes were truly invisible. And this control task had to be run under exactly the same conditions as in the main experiment. In Marcel’s experiments, they argued, either these conditions were not met or, when they were, there was indeed a significant fraction of above-chance responses, suggesting that subjects might have seen a few words.

In response to these critiques, the advocates of unconscious processing tightened up their experimental paradigms. Remarkably, the results still confirmed that words, digits, and even pictures could be unconsciously grasped.46 In 1996 the Seattle psychologist Anthony Greenwald published in the top-ranking journal Science a study that seemed to provide definitive evidence that the emotional meaning of words was processed unconsciously. He had asked participants to classify words as emotionally positive or negative by clicking one of two response keys; unknown to them, each visible target was preceded by a hidden prime. The word pairs were either congruent, reinforcing each other’s meaning (both positive or both negative, as when happy was followed by joy), or incongruent (e.g., rape followed by joy). When participants responded extremely fast, they performed better on congruent than on incongruent trials. The emotional meanings evoked by the two words seemed to pile up unconsciously, helping the final decision when they shared the same emotion, and hindering it when they did not.

Greenwald’s results were strongly replicable. Most subjects not only swore that they could not see the hidden primes but were objectively unable to judge their identity or emotion above chance level. Furthermore, how well they did on such direct guessing tasks was unrelated to how much congruency priming they showed. The priming effect did not seem to be carried by a small set of people who could see the prime words. Here was, at long last, a genuine demonstration that an emotional meaning could be activated unconsciously.

Or was it? Although the strict referees of Science magazine accepted it, Tony Greenwald was a tougher critic of his own work, and a few years later, with his student Richard Abrams, he came up with an alternative interpretation of his own experiment.47 He pointed out that his experiment had used only a small set of repeated words. Perhaps, he surmised, the participants responded to the same words so often, and under such a tough time pressure, that they ended up associating the letters themselves rather than the meanings with the response categories—thus bypassing meaning. The explanation was not absurd because in the Science experiment, subjects repeatedly saw the same words as primes and as targets, and always classified them according to the same rule. After consciously classifying happy twenty times as a positive word, Greenwald realized, perhaps their brains wired up a direct nonsemantic route from the meaningless letters h-a-p-p-y to the “positive” response.48

Alas, this hunch turned out to be correct: in this experiment, priming was indeed subliminal, but it bypassed meaning. First, Greenwald showed that meaningless scrambled primes were just as effective as the real words—hypap was just as powerful a prime as happy. Second, he carefully manipulated the resemblance of the words that people consciously saw with those that served as hidden primes. In a crucial experiment, two of the conscious words were tulip and humor, which participants obviously classified as positive. Greenwald then recombined their letters to create a negative word, tumor, which he presented only unconsciously.

The fascinating finding was that, unconsciously, the negative word tumor primed a positive response. Subliminally, the participants’ brain put tumor together with the words tulip and humor from which it was derived—even though their meaning could not be more different. This was a definite proof that priming depended on a shallow association between specific sets of letters and their corresponding response. Greenwald’s experiment involved unconscious perception but not the words’ deeper meaning. Under these experimental conditions at least, unconscious processing was not smart at all: instead of caring about a word’s meaning, it merely depended on the mapping between letters and responses.

Anthony Greenwald had destroyed the semantic interpretation of his own Science paper.

Unconscious Arithmetic

By 1998, although unconscious semantic processing remained as elusive as ever, my colleagues and I realized that Greenwald’s experiments were perhaps not the final word. An unusual feature of those experiments is that the participants were asked to respond within a strict deadline of 400 milliseconds. This delay seemed too short to compute the meaning of a rare word such as tumor. Given such a tight deadline, the brain had time only to associate letters with responses; perhaps with a more relaxed schedule, it would unconsciously analyze a word’s meaning. So Lionel Naccache and I started some experiments that would definitely prove that a word’s meaning could be unconsciously activated.49

To maximize our chances of obtaining a large unconscious effect, we settled on language’s simplest category of meaningful words: numbers. Numbers below ten are special: they are very short words, frequent, extremely familiar, and overlearned since early childhood; their meaning is transparently simple. They can be conveyed in a remarkably compact form—by a single digit. In our experiment, we therefore flashed the numbers 1, 4, 6, and 9, preceded and followed by a string of random letters that made them entirely invisible. Immediately afterward we displayed a second number, this time clearly visible.

We asked our participants to follow the simplest possible instruction: Please tell us, as fast as you can, whether the number that you see is larger or smaller than 5. They had no idea that there was a hidden number; in a separate test, at the end of the experiment, we showed that even when they knew there was one, they could not see it or classify it as large or small. Still, the invisible numbers caused semantic priming. When they were congruent with the target (e.g., both larger than 5), the participants responded more quickly than when they were incongruent (e.g., one smaller and the other larger). For instance, flashing a subliminal digit 9 accelerated the response to 9 and 6, but slowed the response to 4 and 1.

Using brain imaging, we detected a trace of this effect at the cortical level. We observed a very tiny activation in the motor cortex commanding the hand that would have been an appropriate response to the invisible stimulus. Unconscious votes were traversing the brain, from perception to motor control (figure 11). This effect could arise only from an unconscious categorization of the meaning of invisible words or digits.

FIGURE 11. Our motor cortex can prepare a response to a stimulus that we do not see. Here, a volunteer was asked to classify numbers as larger or smaller than 5. In this example, the visible target was 9. Just before the target, a hidden number was flashed (the word one). Although the hidden number was invisible, it nevertheless sent a small unconscious activation to the motor cortex, commanding the hand that would have been appropriate to respond to it. Thus, an unseen symbol may be identified, processed according to arbitrary instructions, and propagated all the way to the motor cortex.

Subsequent work put the final nail in the skeptics’ coffin. Our subliminal effect was entirely independent of the notation used for the numbers: four primed 4 just as much as an exact repetition of 4 primed 4, suggesting that all the effect arose at the level of abstract meaning. We later showed that priming persisted when the prime was an invisible visual number and the target a conscious spoken number.50

In our initial experiment, the effect might have been caused by a direct association between visual shapes and responses—the same problem that had plagued Greenwald’s experiments with emotional words. However, subliminal number priming avoided this criticism. We proved that hidden numbers that had never been consciously seen in the entire experiment still caused semantic priming.51 By imaging brain activation with functional MRI, we even obtained direct evidence that the “number sense” regions of the brain, in the left and right parietal lobes, were influenced by the unseen number.52 These regions encode the quantity meaning of numbers53 and are thought to house neurons tuned to specific quantifies.54 During subliminal priming, their activity decreased whenever we displayed the same number twice (e.g., nine followed by 9). This is a classical phenomenon called “repetition suppression” or “adaptation,” which indicates that the neurons recognize that the same item is displayed twice. It seemed that the neurons coding for quantity were habituating to seeing the same number twice, even when the first presentation was unconscious. The evidence had mounted: a higher brain area cared about a specific meaning and could be activated without consciousness.

The final knockout came when our colleagues demonstrated that the number priming effect varies as a direct function of the overlap in number meaning.55 The strongest priming was obtained by displaying the same quantity twice (e.g., a subliminal four preceding 4). The priming decreased slightly for nearby numbers (three preceding 4), got even smaller for numbers at a distance of 2 (two preceding 4), and so on. Such a semantic distance effect is a hallmark of number meaning. It can arise only if the subject’s brain encodes that 4 resembles 3 more than 2 or 1—a definite argument in favor of an unconscious extraction of that number’s meaning.

Combining Concepts Without Consciousness

The skeptics’ last resort was to accept our demonstration but to assume that numbers were special. Adults have so much experience with this closed set of words, they argued, that it should be no surprise that we can automatically understand them. Other categories of words, however, would be different—surely their meaning would not be represented without consciousness. But even this last line of resistance collapsed when similar priming techniques revealed semantic congruity effects with unseen words outside the number domain.56 For instance, deciding that the target piano is an object rather than an animal can be facilitated by the subliminal presentation of the congruent word chair, and hindered by the incongruent word cat—even when the primes are never seen throughout the experiment.

Brain-imaging techniques also confirmed the cognitive scientist’s conclusions. Recordings of neural activity provided direct evidence that the brain regions involved in semantic processing could be activated without consciousness. In one study, my colleagues and I took advantage of electrodes that had been implanted deep in the brain, in subcortical regions specialized in emotional processing.57 Naturally, such recordings were performed not in healthy volunteers but in patients with epilepsy. In many hospitals throughout the world, it has become clinical routine to insert electrodes deep inside the patient’s skull, in order to identify the source of epileptic discharges and ultimately excise the impaired tissue. In between the seizures, if the patient agrees, we can use the electrodes for a scientific purpose. They grant us access to the average activity of a small brain region or sometimes to the signal emitted by just one neuron.

In our case, the electrodes reached deep into the amygdala, a brain structure involved in emotion. As I explained earlier, the amygdala responds to all sorts of frightening stuff, from snakes and spiders to scary music and strangers’ faces—even a subliminal snake or face may trigger it.58 Our question was, Would this region activate to an unconscious frightening word? So we flashed subliminal words with a disturbing meaning, such as rape, danger, or poison—and to our great pleasure, an electrical signal appeared, which was absent for neutral words such as fridge or sonata. The amygdala “saw” words that remained invisible to the patients themselves.

This effect was remarkably slow: it took half a second or more before an invisible word caused an unconscious emotional dip. But the activation was completely unconscious: at the same time that his amygdala fired, a participant denied seeing any word and, when asked to guess, had no idea what it was. Thus a written word could slowly make its way down into the brain, be identified, and even be understood, all without consciousness.

The amygdala is not part of the cortex, so perhaps this makes it special and more automatic. Could the language cortex fire to an unconscious meaning? Further experiments gave a positive answer. They relied on a cortical wave that marks the brain’s response to an unexpected meaning. “At breakfast, I like my coffee with cream and socks”: as you read such a silly sentence, the bizarre meaning of the final word generates a particular brain wave called the N400. (The N refers to its shape, which shows a negative voltage on the top of the head, and the 400 to its peak latency, about 400 milliseconds after the word appears.)

The N400 reflects a sophisticated level of operation, which evaluates how a given word fits within a sentence’s context. Its size varies directly with the degree of absurdity: words whose meaning is roughly appropriate cause a very small N400, while utterly unexpected words generate a larger one. Remarkably, this brain event occurs even with words that we do not see—whether they are rendered invisible by masking59 or by inattention.60 Networks of neurons in our temporal lobe automatically process not only the various meanings of invisible words but also their compatibility with the past conscious context.

In recent work, Simon van Gaal and I even showed that the N400 wave could reflect an unconscious combination of words.61 In this experiment, two words appeared in succession, both of them masked below the awareness threshold. They were selected to form unique combinations of positive or negative meanings: “not happy,” “very happy,” “not sad,” and “very sad.” Immediately after this subliminal sequence, the subjects saw a positive or negative word (say, war or love). The N400 wave emitted by this conscious word was modulated by the global unconscious context. Not only did war evoke a large N400 when preceded by the incongruous word happy, but this effect was strongly modulated, up or down, by the intensifier very or the negation not. Unconsciously, the brain registered the incongruity of a “very happy war,” and judged “not happy war” or “very sad war” as better fits. That experiment is as close as one gets to proving that the brain can unconsciously process the syntax and meaning of a well-formed word phrase.62

Perhaps the most remarkable aspect of these experiments is that the N400 wave has exactly the same size whether the words are conscious or invisible. This finding is rife with implications. It means that, in some respects, consciousness is irrelevant to semantics—our brain sometimes performs the same exact operations, all the way up to the meaning level, whether or not we are aware of them. It also means that unconscious stimuli do not always generate minuscule events in the brain. Brain activity can be intense even though the stimulus that causes it remains invisible.

We conclude that an invisible word is fully capable of eliciting a large-scale activation in the brain’s meaning networks. However, an important caveat is in order. Accurate reconstruction of the sources of semantic brain waves shows that the unconscious activity is confined to a narrow and specialized brain circuit. During unconscious processing, brain activity remains within the boundaries of the left temporal lobe, which is the primary site of the language networks that process meaning.63 Later we shall see that conscious words, conversely, gain the upper hand over much larger brain networks that invade the frontal lobes and that underlie the special subjective sense of having the word “in mind.” What this means is that, ultimately, unconscious words are not as influential as conscious ones.

Attentive but Unconscious

The discovery that a word or a digit can travel throughout the brain, bias our decisions, and affect our language networks, all the while remaining unseen, was an eye-opener for many cognitive scientists. We had underestimated the power of the unconscious. Our intuitions, it turned out, could not be trusted: we had no way of knowing what cognitive processes could or could not proceed without awareness. The matter was entirely empirical. We had to submit, one by one, each mental faculty to a thorough inspection of its component processes, and decide which of those faculties did or did not appeal to the conscious mind. Only careful experimentation could decide the matter—but with techniques such as masking and attentional blink in our hands, testing the depth and limits of unconscious processing had never been so easy.

The past ten years have now seen a flurry of novel results challenging our picture of the human unconscious. Consider attention. Nothing seems more closely related to consciousness than the capacity to attend to stimuli. Without attention, we may remain totally unaware of external stimuli—as made clear by Dan Simons’s gorilla movie and a zillion other effects of inattentional blindness. Whenever there are multiple competing stimuli, attention seems to be a necessary gateway to conscious experience.64 In such conditions at least, consciousness requires attention. Amazingly, however, the converse statement turns out to be false: several recent experiments demonstrate that our attention can also be deployed unconsciously.65

It would be strange indeed if attending required the supervision of awareness. The role of attention, as already noted by William James, is to select “one out of several possible objects of thought.” It would be oddly inefficient for our mind to be constantly distracted by dozens or even hundreds of possible thoughts and to examine each of them consciously before deciding which one is worthy of a further look. The determination of which objects are relevant and should be amplified is better left to automatic processes that operate sub rosa, in a massively parallel manner. Unsurprisingly, it turns out that our attentional spotlight is operated by armies of unconscious workers that silently sift through piles of rubble before one of them hits gold and alerts us of its finding.

In recent years, experiment after experiment has revealed the operation of selective attention without consciousness. Suppose we flash a stimulus in the corner of your eye so briefly that you cannot see it. Several experiments have shown that although it remains unconscious, such a flash may still attract your attention: you will become more attentive, and therefore faster and more accurate at attending to other stimuli presented at that same location, although you have no idea that a hidden cue caught your eye.66 Conversely, a hidden picture may slow you down when its content is irrelevant to the task at hand. Interestingly, this effect works better when the distracting stimulus remains unconscious than when it is visible: a conscious distractor can be voluntarily extinguished, while an unconscious one preserves all its nuisance potential because we are unable to learn to control it.67

Loud noises, blinking lights, and other unexpected sensory events, as we all know, can irrepressibly attract our attention. However hard we try to ignore them, they invade our mental privacy. Why? They are, in part, an alerting mechanism, keeping us on the watch for potential dangers. As we concentrate on doing our taxes or on playing a favorite video game, it would be unsafe to tune out completely. Unexpected stimuli, such as a scream or the call of our own name, must remain able to break through our current thoughts—and therefore the filter called “selective attention” must continually operate outside our awareness, in order to decide which incoming inputs call for our mental resources. Unconscious attention acts as a constant watchdog.

Psychologists long thought that such automatic and bottom-up processes of the mind were the only ones that operated unconsciously. Psychologists’ favorite metaphor for unconscious processing was that of a “spreading activation”: a wave that starts from the stimulus and passively spreads through our brain circuits. A hidden prime climbed up the hierarchy of visual areas, progressively contacting processes of recognition, meaning attribution, and motor programming, as it tagged along with, without ever being influenced by, the subject’s conscious will, intention, and attention. Thus, the results of subliminal experiments were thought to be independent of the participants’ strategies and expectations.68

Consider it a major surprise, then, when our experiments shattered this consensus. We proved that subliminal priming is not a passive, bottom-up process, operating independent of attention and instructions. In fact, attention determines whether an unconscious stimulus is or is not processed.69 An unconscious prime that is presented at an unexpected time or place produces virtually no priming onto a subsequent target. Even the mere repetition effect—the accelerated response to radio followed by radio—varies with how much attention is allocated to these stimuli. The act of attending causes a gain that massively amplifies the brain waves evoked by stimuli presented at the attended time and place. Remarkably, unconscious stimuli benefit from this attentional spotlight just as much as conscious ones do. In other words, attention can amplify a visual stimulus and still leave it too weak to break into our awareness.

Conscious intentions can even affect the orientation of our unconscious attention. Imagine that you are shown a set of shapes and are asked to detect only the squares while ignoring the circles. On a critical trial, a square appears on the right and a circle on the left—but both shapes are masked, so that you fail to detect them. You press randomly, not knowing on which side the square was shown. But a marker of parietal lobe activation called the N2pc reveals an unconscious orientation of your attention toward the appropriate side.70 Your visual attention is surreptitiously attracted to the correct target, even on totally invisible trials and even if you eventually select the wrong response side. Similarly, during the attentional blink, within an entire stream of letters, the symbol that is arbitrarily designated as a target evokes noticeably more brain activity, even though it remains undetected.71 On such trials, attention begins to unconsciously sieve the shapes for their relevance, although this process falls short of bringing the target stimulus into participants’ conscious awareness.

The Value of an Invisible Coin

How does our attention decide whether a stimulus is relevant? A key component of the selection process is the assignment of a value to each potential object of thought. In order to survive, animals must have a very quick way of assigning a positive or negative value to every encounter. Should I stay, or should I go? Should I approach, or should I retreat? Is this a valuable treat or a poisonous trap? Valuation is a specialized process that calls upon evolved neural networks within a set of nuclei called the basal ganglia (because they are located near the base of the brain). And as you may have guessed, they too can operate totally outside our conscious awareness. Even a symbolic value such as money can be unconsciously appraised.

In one experiment, a picture of a penny or a pound sterling coin served as a subliminal incentive (figure 12).72 The subjects’ task was to squeeze a handle, and if they managed to exceed a certain amount of force, they would earn money. At the beginning of each trial, the image of a coin indicated how much money was at stake—and some of these pictures were flashed too fast to be consciously perceived. Although the participants denied having any awareness of either coin image, they exerted a stronger force when their potential gain was a pound than when it was a penny. Furthermore, the expectation of gaining one pound made the subjects’ hands sweat in anticipation of this unconscious reward—and the brain’s reward circuits were surreptitiously activated. The subjects remained unaware of the reason their behavior varied from trial to trial: they had no idea that their motivation was being unconsciously manipulated.

FIGURE 12. Unconscious incentives can affect our motivations. In this experiment, participants were asked to squeeze a handle as strongly as they could in order to gain money. When a flashed picture specified that the stake was a pound sterling rather than a penny, people exerted a stronger force. They continued to do so even when the image was masked so that they were unaware which coin was presented. The reward circuits of the brain were unconsciously preactivated, and even the hands sweated in anticipation of gain. Thus, an unconscious image can trigger the circuits for motivation, emotion, and reward.

In another study, the values of the subliminal stimuli were not known in advance but were demonstrably learned during the course of the experiment.73 The subjects, upon seeing a “signal,” had to guess whether to press a button or refrain from pressing it. After each instance, they were told whether they had gained or lost money as a result of pressing or not pressing. Unknown to them, a subliminal shape, flashed within the signal, indicated the correct response; one shape cued the “go” response, another the “withhold” response, and a third was neutral—when it appeared, there was a 50 percent chance that either response would be rewarded.

After playing this game for a few minutes, subjects inexplicably got better at this task. They still could not see the shapes that were hidden within the signal, but they had the “hot hand” and began to earn a significant sum of money. Their unconscious value system had kicked in: the positive “go” shape began to trigger key presses, while the negative “withhold” shape evoked systematic withholding. Brain imaging showed that a specific region of the basal ganglia, called the ventral striatum, had attached the relevant values to each shape. In brief, symbols that the subjects had never seen had nevertheless acquired a meaning: one had become repulsive and the other attractive, thus modulating the competition for attention and action.

The outcome of all these experiments is clear: our brain hosts a set of clever unconscious devices that constantly monitor the world around us and assign it values that guide our attention and shape our thinking. Thanks to these subliminal tags, the amorphous stimuli that bombard us become a landscape of opportunities, carefully sorted according to their relevance to our current goals. Only the most relevant events draw our attention and gain a chance to enter our consciousness. Below the level of our awareness, our unconscious brain ceaselessly evaluates dormant opportunities, testifying that our attention largely operates in a subliminal manner.

Unconscious Mathematics

A return from the over-estimation of the property of consciousness is the indispensable preliminary to any genuine insight into the course of psychic events.

—Sigmund Freud, The Interpretation of Dreams (1900)

Freud was right: consciousness is overrated. Consider this simple truism: we are conscious only of our conscious thoughts. Because our unconscious operations elude us, we constantly overestimate the role that consciousness plays in our physical and mental lives. By forgetting the amazing power of the unconscious, we overattribute our actions to conscious decisions and therefore mischaracterize our consciousness as a major player in our daily lives. In the words of the Princeton psychologist Julian Jaynes, “Consciousness is a much smaller part of our mental life than we are conscious of, because we cannot be conscious of what we are not conscious of.”74 Paraphrasing Douglas Hofstadter’s whimsically circular law of programming (“A project always takes longer than you expect—even when you take into account Hofstadter’s Law”), one might elevate this statement to the level of a universal law:

We constantly overestimate our awareness—even when we are aware of the glaring gaps in our awareness.

The corollary is that we dramatically underestimate how much vision, language, and attention can occur outside our awareness. Might some of the mental activities that we consider hallmarks of the conscious mind actually run unconsciously? Consider mathematics. One of the world’s greatest mathematicians ever, Henri Poincaré, reported several curious incidents in which his unconscious mind seemed to do all the work:

I left Caen, where I was living, to go on a geologic excursion under the auspices of the School of Mines. The incidents of the travel made me forget my mathematical work. Having reached Coutances, we entered an omnibus to go some place or other. At the moment when I put my foot on the step, the idea came to me, without anything in my former thoughts seeming to have paved the way for it, that the transformations I had used to define the Fuchsian functions were identical with those of non-Euclidian geometry. I did not verify the idea; I should not have had time, as, upon taking my seat in the omnibus, I went on with a conversation already commenced, but I felt a perfect certainty. On my return to Caen, for conscience sake, I verified the result at my leisure.

And then again:

I turned my attention to the study of some arithmetical questions apparently without much success and without a suspicion of any connection with my preceding researches. Disgusted with my failure, I went to spend a few days at the seaside and thought of something else. One morning, walking on the bluff, the idea came to me, with just the same characteristics of brevity, suddenness and immediate certainty, that the arithmetic transformations of indefinite ternary quadratic forms were identical with those of non-Euclidian geometry.

These two anecdotes are reported by Jacques Hadamard, a world-class mathematician who dedicated a fascinating book to the mathematician’s mind.75 Hadamard deconstructed the process of mathematical discovery into four successive stages: initiation, incubation, illumination, and verification. Initiation covers all the preparatory work, the deliberate conscious exploration of a problem. This frontal attack, unfortunately, often remains fruitless—but all may not be lost, for it launches the unconscious mind on a quest. The incubation phase—an invisible brewing period during which the mind remains vaguely preoccupied with the problem but shows no conscious sign of working hard on it—can start. Incubation would remain undetected, were it not for its effects. Suddenly, after a good night’s sleep or a relaxing walk, illumination occurs: the solution appears in all its glory and invades the mathematician’s conscious mind. More often than not, it is correct. However, a slow and effortful process of conscious verification is nevertheless required to nail all the details down.

Hadamard’s theory is seductive, but does it stand up to scrutiny? Does unconscious incubation truly exist? Or is it just retrospective storytelling glorified by the elation of discovery? Can we truly solve complex problems unconsciously? Cognitive science has only recently begun to bring these questions to the lab. Antoine Bechara, at the University of Iowa, developed a gambling task that studies people’s protomathematical intuitions of probability and numerical expectation.76 In this test, the subjects are given four decks of cards and a loan of $2,000 (in fake bills—psychologists aren’t that rich). Turning a card over reveals a positive or negative message (e.g., “you win $100” or “you pay $100”). Participants try to optimize their gains by choosing at will from all four decks. What they do not know is that two of the decks are disadvantageous: they initially provide large earnings but quickly give rise to massive costs, and in the long run the outcome is a net loss. The other two decks lead to moderate ups and downs. In the long run, pulling cards from them yields a small but steady gain.

Initially, the players sample randomly from the four decks. Progressively, however, they develop a conscious hunch, and in the end they can easily report which decks are good and which are bad. But Bechara was interested in the “pre-hunch” period. During this phase, which resembles the mathematician’s incubation period, the participants already have a lot of evidence about the four decks but still pull from all of them at random and claim to have no clue as to what they should do. Fascinatingly, just before they choose a card from a bad deck, their hands begin to sweat, thus generating a drop in skin conductance. This physiological marker of the sympathetic nervous system indicates that their brain has already registered the risky decks and is generating a subliminal gut feeling.

The alarm signal probably arises from operations performed in the ventromedial prefrontal cortex—a brain region specializing in unconscious valuation. Brain imaging shows a clear activation of this region, which is predictive of performance, on disadvantageous trials.77 Patients with lesions to this region do not generate the anticipatory skin conductance in advance of unwittingly choosing from the bad-outcome deck; they do so only later on, once the bad outcome is revealed. The ventromedial and orbifrontal cortex contains a whole array of evaluative processes that constantly monitor our actions and compute their potential value. Bechara’s research suggests that these regions often operate outside our conscious awareness. Although we have the impression of making random choices, our behavior may, in fact, be guided by unconscious hunches.

Having a hunch is not exactly the same as resolving a mathematical problem. But an experiment by Ap Dijksterhuis comes closer to Hadamard’s taxonomy and suggests that genuine problem solving may indeed benefit from an unconscious incubation period.78 The Dutch psychologist presented students with a problem in which they were to choose from among four brands of cars, which differed by up to twelve features. The participants read the problem, then half of them were allowed to consciously think about what their choice would be for four minutes; the other half were distracted for the same amount of time (by solving anagrams). Finally, both groups made their choice. Surprisingly, the distracted group picked the best car much more often than the conscious-deliberation group (60 percent versus 22 percent, a remarkably large effect given that choosing at random would result in 25 percent success). The work was replicated in several real-life situations, such as shopping at IKEA: several weeks after a trip there, shoppers who reported putting a lot of conscious effort into their decision were less satisfied with their purchases than the buyers who chose impulsively, without much conscious reflection.

Although this experiment does not quite meet the stringent criteria for a fully unconscious experience (because distraction does not fully ensure that the subjects never thought about the problem), it is very suggestive: some aspects of problem solving are better dealt with at the fringes of unconsciousness rather than with a full-blown conscious effort. We are not entirely wrong when we think that sleeping on a problem or letting our mind wander in the shower can produce brilliant insights.

Can the unconscious solve any type of problem? Or, more likely perhaps, are some categories of puzzles especially conducive to being solved by an unconscious hunch? Interestingly, Bechara’s and Dijksterhuis’s experiments involve similar problems; both require subjects to weigh several parameters. In Bechara’s case, they must carefully weigh the gains and losses incurred with each deck of cards. In Dijksterhuis’s, they must choose a car based on a weighted average of twelve criteria. When made consciously, such decisions put a heavy load on our working memory: the conscious mind, which typically focuses on one or a few possibilities at a time, is easily overwhelmed. This is probably why the conscious thinkers in Dijksterhuis’s experiment did not do so well: they tended to place exaggerated weight on one or two features without seeing the bigger picture. Unconscious processes excel in assigning values to many items and averaging them to reach a decision.

Computing the sum or average of several positive and negative values indeed lies within the normal repertoire of what elementary circuits of neurons can do without consciousness. Even a monkey can learn to make a decision based on the total value brought about by a series of arbitrary shapes, and the firing of parietal neurons keeps track of the sum.79 In my laboratory, we proved that approximate addition is within grasp of the human unconscious. In one experiment, we flashed a series of five arrows and asked subjects whether more arrows were pointing right or pointing left. When the arrows were made invisible by masking, participants were asked to guess, and indeed they thought that they were responding randomly, but in reality they continued to do much better than chance would predict. Signals from their parietal cortex gave evidence that their brain was unconsciously computing the approximate sum of the overall evidence.80 The arrows were subjectively invisible, but they still made their way into the brain’s weighting and decision systems.

In another experiment, we flashed eight numerals; four of them were visible consciously while the other four were invisible. We asked participants to decide if their mean was larger or smaller than five. The responses were quite accurate on average, but remarkably, the participants considered all eight of the available numbers. Thus, if the conscious numbers were larger than five, but the hidden numbers were smaller than five, the subjects were unconsciously biased to respond “smaller.”81 The averaging operation that they were asked to perform with the consciously visible numbers extended to the unconscious ones.

Statistics During Sleep

Clearly, then, some elementary mathematical operations, including averaging and comparing, may unfold unconsciously. But what about genuinely creative operations such as Poincaré’s insight on the omnibus? Can insight really strike us at any time, even when we least expect it and are thinking of something else? The answer seems to be positive. Our brain acts as a sophisticated statistician that detects meaningful regularities hidden in seemingly random sequences. Such statistical learning is constantly running in the background, even as we sleep.

Ullrich Wagner, Jan Born, and their colleagues tested scientists’ claim that they often have a sudden insight upon waking up from a good night’s sleep.82 To bring this idea to the lab, they had subjects participate in a nerdy math experiment: they had to mentally transform a sequence of seven digits into another sequence of seven digits according to an attention-demanding rule. They were asked to name only the last digit of the answer—but finding its value required a long mental calculation. Unknown to them, however, there was a shortcut. The output sequence had a hidden symmetry: the last three digits repeated the previous three in reverse order (e.g., 4 1 4 9 9 4 1), and as a result the last digit was always equal to the second one. Once the participants recognized this shortcut, they could save enormous time and effort by stopping after the second digit. During the initial test, most of the subjects failed to notice the concealed rule. However, a good night’s sleep more than doubled the probability that they would have the insight: many participants woke up with the solution in mind! Controls established that the elapsed time was irrelevant; what mattered was sleep. Falling asleep seemed to enable the consolidation of previous knowledge into a more compact form.

We know from animal studies that neurons in the hippocampus and the cortex are active during sleep. Their firing patterns “replay,” in fast-forward mode, the same sequences of activity that occurred during the previous period of wakefulness.83 For instance, a rat runs through a maze; then upon falling asleep, his brain reactivates his place-coding neurons so precisely that the pattern can be used to decode the locations where he is mentally traveling—but at a much faster speed, and sometimes even in reverse order. Perhaps this temporal compression offers the possibility of treating a sequence of digits as a near-simultaneous spatial pattern, thus permitting the detection of hidden regularities by classical learning mechanisms. Whatever the neurobiological explanation, sleep is clearly a period of boiling unconscious activity that supports much memory consolidation and insight.

A Subliminal Bag of Tricks

These laboratory demonstrations are a far cry from the type of mathematical thinking that Poincaré had in mind when he was unconsciously exploring Fuchsian functions and non-Euclidean geometry. However, that gap is being reduced as innovative experiments study the greater range of operations that can be performed, at least in part, without awareness.

It was long thought that the mind’s “central executive”—a cognitive system that controls our mental operations, avoids automatic responses, switches tasks, and detects our errors—was the sole province of the conscious mind. But recently, sophisticated executive functions have been shown to operate unconsciously, based on invisible stimuli.

One such function is the ability to control ourselves and inhibit our automatic responses. Imagine performing a repetitive task, such as pressing a key whenever a picture appears on screen—except that on rare occasions, the picture depicts a black disk, and then you absolutely have to refrain from clicking. This is called the “stop signal” task, and much research shows that the ability to inhibit a routine response is a marker of the mind’s central executive system. The Dutch psychologist Simon van Gaal asked whether refraining from responding requires consciousness: would subjects still manage to avoid clicking if the “stop” signal was subliminal? Amazingly, the answer was yes. When an unconscious “stop” signal was briefly flashed, the participants’ hands slowed down, and occasionally, they stopped responding altogether.84 They did so without understanding why, because the stimulus that triggered this inhibition remained unseen. These findings indicate that invisible is not synonymous with out of control. Even an invisible stop signal can trigger a wave of activity that spreads deep into the executive networks that allow us to control our actions.85

Similarly, we can detect some of our errors without being conscious. In an eye movement task, when the participants’ eyes deviate from the plan, the error triggers an activation of the executive control centers in the anterior cingulate cortex—even when participants are unaware of the error and deny that their eyes wandered off the target.86 Unconscious signals can even cause a partial switch to another task. When subjects are shown a conscious cue that tells them to change from task one to task two, flashing this cue below the threshold for awareness still has the effect of slowing them down and triggering a partial task switch at the cortical level.87

In a nutshell, psychology has amply demonstrated not only that subliminal perception exists but that a whole array of mental processes can be launched without consciousness (even though, in most cases, they do not run to full completion). Figure 13 summarizes the various brain regions that, in experiments discussed in this chapter, have been shown to activate in the absence of awareness. The unconscious clearly has a large bag of tricks, from word comprehension to numerical addition, and from error detection to problem solving. Because they operate quickly and in parallel across a broad range of stimuli and responses, these tricks often surpass conscious thought.

FIGURE 13. An overview of unconscious operations in the human brain. The figure shows only a subset of the many circuits that can activate without awareness. We now believe that virtually any brain processor can operate unconsciously. For greater readability, each computation is pinned to its dominant brain site, but it should be remembered that such neuronal specialization always rests on an entire brain circuit. Some of our unconscious processors are subcortical: they involve groups of neurons located below the surface of the cortex (denoted by dashed ellipses) and often implement functions that appeared early in our evolution, such as the detection of frightening stimuli that warn us of an impending danger. Other computations recruit various sectors of the cortex. Even high-level cortical areas that encode our acquired cultural knowledge, such as reading or arithmetic, may operate outside our awareness.

Henri Poincaré, in Science and Hypothesis (1902), anticipated the superiority of unconscious brute-force processing over slow conscious thinking:

The subliminal self is in no way inferior to the conscious self; it is not purely automatic; it is capable of discernment; it has tact, delicacy; it knows how to choose, to divine. What do I say? It knows better how to divine than the conscious self, since it succeeds where that has failed. In a word, is not the subliminal self superior to the conscious self?

Contemporary science answers Poincaré’s question with a resounding yes. In many respects, our mind’s subliminal operations exceed its conscious achievements. Our visual system routinely solves problems of shape perception and invariant recognition that boggle the best computer software. And we tap into this amazing computational power of the unconscious mind whenever we ponder mathematical problems.

But we should not get carried away. Some cognitive psychologists go as far as to propose that consciousness is a pure myth, a decorative but powerless feature, like frosting on a cake.88 All the mental operations that underlie our decisions and behavior, they claim, are accomplished unconsciously. In their view, our awareness is a mere bystander, a backseat driver that contemplates the brain’s unconscious accomplishments but lacks effective powers of its own. As in the 1999 movie The Matrix, we are prisoners of an elaborate artifice, and our experience of living a conscious life is illusory; all our decisions are made in absentia by the unconscious processes within us.

The next chapter will refute this zombie theory. Consciousness is an evolved function, I argue—a biological property that emerged from evolution because it was useful. Consciousness must therefore fill a specific cognitive niche and address a problem that the specialized parallel systems of the unconscious mind could not.

Ever insightful, Poincaré noted that in spite of the brain’s subliminal powers, the mathematician’s unconscious cogs did not start clicking unless he had made a massive initial conscious attack on the problem during the initiation phase. And later on, after the “aha” experience, only the conscious mind could carefully verify, step by step, what the unconscious seemed to have discovered. Henry Moore made exactly the same point in The Sculptor Speaks (1937):

Though the non-logical, instinctive, subconscious part of the mind must play its part in [the artist’s] work, he also has a conscious mind which is not inactive. The artist works with a concentration of his whole personality, and the conscious part of it resolves conflicts, organizes memories, and prevents him from trying to walk in two directions at the same time.

We are now ready to walk into the unique realm of the conscious mind.