Windows into the working brain

The man who lost tomorrow

In the summer of 1953, in Hartford, Connecticut, the neurosurgeon William Scoville operated on a twenty-seven-year-old man named Henry Molaison. Molaison had suffered seizures since boyhood, and the drugs of the day could not control them. Scoville, who was trying an experimental operation, removed tissue from the inner surface of both temporal lobes: most of the amygdala, the cortex next to it, and about half of a seahorse-shaped structure called the hippocampus.

In the narrow sense the operation worked: the seizures became less frequent. But Molaison could no longer form new lasting memories of facts and events. His intelligence was untouched (his IQ was around 110 on later tests), and so were his language and his perception. He could hold a number in mind as long as he kept repeating it; when his attention moved, it was gone, and so was the memory of having been asked. For the next fifty-five years every person he met was new. The medical literature knew him only as H.M.

Brenda Milner, a psychologist working with Scoville at the Montreal Neurological Institute, described the case with him in 1957 in a paper that became a founding document of the science of memory. The picture was more careful than “he lost his memory.” His loss of memory for the years before the operation was partial, covering about one to three years in most accounts, and childhood memories were intact. What he lost was the ability to lay down new conscious memories. Then came the discovery that mattered most. Milner gave him a mirror-drawing task: trace the outline of a star while seeing your hand only in a mirror, which is hard at first. Over three days of practice his errors fell steadily, as they would for anyone. Every day he said he had never seen the task before.

H.M. died in December 2008 at the age of eighty-two, and only then was his name made public. In December 2009 researchers at the University of California, San Diego, led by Jacopo Annese, began cutting his brain into 2,401 slices, each 70 micrometers thick, in a procedure of about fifty-three hours that was streamed live on the internet. The slices, described in a 2014 paper, showed that the remnant of the hippocampus was real but damaged, that Scoville’s estimate of how much he removed was too large, and that there was a small lesion nobody had known about, in the left frontal lobe. A single patient had become a standard for a whole field.

How injuries teach

H.M. is the best-known example of a method with a long history: to learn what part of the brain does, see what is lost when it is damaged. In 1861 at the Bicêtre hospital in Paris, the physician Paul Broca examined a man whose only word was “tan.” The patient understood speech but could not produce it, and Broca, finding damage in the left frontal lobe after his death, concluded that this region is needed to produce speech. Thirteen years later Carl Wernicke described patients with damage farther back in the left hemisphere who spoke fluently but made little sense and could not understand. Two kinds of language trouble from two kinds of damage is a double dissociation, much stronger evidence than a single case.

The method has limits. Injuries do not respect the boundaries of a brain region, and they often cut the wiring between regions, so a symptom can mean a disconnection and not a damaged “center.” In 2007 a team scanned Broca’s two preserved brains with modern MRI and found that the damage went deeper into the white matter than the surface lesions he had described. Finding out what a region is needed for is not the same as finding out what it does, and the brain can also reorganize after injury. Lesions tell us what is necessary; scanners, as we will see, tell us what is active.

An iron bar

The most famous brain injury in history happened outside a hospital. On the afternoon of 13 September 1848, in Cavendish, Vermont, Phineas Gage, a twenty-five-year-old foreman on the Rutland and Burlington Railroad, was packing blasting powder into a hole drilled in rock with a tamping iron, a pointed rod three feet seven inches long (1.1 meters) and weighing 13¼ pounds (6 kilograms). The powder went off. The rod entered below his left cheekbone, passed behind his left eye, went through the front of his brain, and left through the top of his skull.

Gage spoke within minutes, rode three-quarters of a mile sitting upright in an ox cart, and walked up the stairs of the hotel. A local doctor, John Martyn Harlow, treated him through an infection that nearly killed him, and by the end of the year he had largely recovered. In 1868 Harlow wrote a report that included the sentence on which the legend rests: the balance between Gage’s “intellectual faculties and animal propensities” seemed to be gone. He had become fitful and irreverent, impatient of restraint; his friends said he was “no longer Gage.” His contractors would not give him back his job.

The modern reconstruction has changed in two ways. First, the brain damage: in 1994 Hanna Damasio and colleagues concluded from the skull that both frontal lobes had been hit, but later CT-based studies, by Peter Ratiu and colleagues in 2004 and John Van Horn and colleagues in 2012, found damage on the left side only, about 4 percent of the cortex and about 11 percent of the white matter. Second, the biography. Malcolm Macmillan, who spent decades on the records, showed that much of the dramatic tale (the circus, the unemployable drifter) has no support, and that Gage held a demanding job for years, driving a stagecoach in Chile, work that required early rising, planning, and dealing with passengers and that may have helped his recovery.

The lecture on which this course draws mixes the Gage story with another famous medical story. The doctor who made experimental observations of digestion through a man’s stomach wound was William Beaumont, an army surgeon, and his patient was Alexis St. Martin, who was shot on 6 June 1822 and lived for another fifty-eight years with a permanent opening into his stomach. Neither story concerned the other man.

What lasts from Gage is not the legend but a question that Antonio Damasio, whom you met in Course 10, took up in 1994 in Descartes’ Error: how can someone who can talk, walk, and reason still lose the ability to manage a life? Patients such as Elliot, and Gage before them, point the same way: the front of the brain does something other than calculate. It helps hold impulses, plans, and feelings together.

A brake still being built

Gage’s injury was to the front of the brain, the prefrontal cortex, and that region has a developmental story of its own. In 2004 Nitin Gogtay and colleagues followed 13 healthy children and teenagers with MRI every two years for eight to ten years, from ages 4 to 21. The thinning of the cortex, which reflects pruning of connections and other changes, started in the strips that handle touch and movement, spread forward and backward, and reached the higher-order association areas last. The orbitofrontal region, just above the eyes, was still changing at 21, the oldest age in the study. The popular claim that the brain “is finished at 25” does not come from this paper, and as Course 10 noted, no single birthday exists: different tissues peak at different ages.

Whether this timetable explains the way a young child will hit a playmate and, a second later, say sorry is harder to know. Self-control does improve rapidly in the preschool years: in a 1996 study of 160 children aged three and a half to seven, Adele Diamond and Cathy Taylor found that most of the improvement on a task that needed holding back a strong response came by age six, though they did not measure the circuits. In Walter Mischel’s marshmallow test, four-year-olds at Stanford were left alone with a treat, with the promise of two if they waited. Early follow-ups suggested that the children who waited did better as teenagers, but a 2018 replication with 918 children found the link about half as strong, and most of what remained was explained by family background and early thinking skills.

Chemistry may also come into the story. PET scans by Harry Chugani and colleagues in the 1980s showed glucose use in the frontal cortex of babies rising at about six to twelve months, and a child’s cortex using roughly twice the adult amount of fuel between the ages of three and ten. Babies also usually become wary of strangers around six to eight months. The two facts coincide in time, and it is tempting to say the first causes the second, but nobody has tested that.

From black box to scanner

For most of the history of psychology, nobody could look at a healthy brain thinking. In 1879 Wilhelm Wundt began running regular experiments in Leipzig, a year usually given as the birth of experimental psychology. In 1913 John B. Watson announced that psychology should be “a purely objective experimental branch of natural science,” which in practice meant studying behavior and treating the mind as a black box. The “cognitive revolution” is often dated to 11 September 1956, a day of a symposium at MIT when a computer program that proved theorems, a new theory of grammar by Noam Chomsky, and experiments on memory convinced the psychologist George Miller that they belonged to one science of mind.

The tools followed. The first CT scan of a patient’s head was made in 1971. Magnetic resonance imaging produced pictures of soft tissue without radiation. In 1990 Seiji Ogawa and colleagues showed that the magnetic signal in the brain depends on the oxygen in the blood, and by 1992 several teams had used that “BOLD” signal to see the brain at work, which is functional MRI, or fMRI. Working neurons need oxygen, and the blood supply overshoots, so a scanner can detect that a patch of brain has taken in more oxygenated blood than before, and such patches are displayed in color.

It is worth knowing what the colors are. The signal change is small, a fraction of a percent to a few percent, and it peaks about five seconds after the event that caused it. Each tiny cube of brain in the image, a voxel, typically holds a few million neurons. The patches are statistical maps: the program compares two conditions, such as looking at a face and looking at a blank screen, and colors the places where the difference was too consistent to be chance. The image does not show that one part of the brain is “on” and the rest is “off.”

What a scan can and cannot tell us

Within one person, in one scanner, with a clean comparison, fMRI has been very useful. In 2000 Andreas Bartels and Semir Zeki scanned 17 people who said they were deeply in love while they looked at a photo of their partner and, for comparison, photos of three friends. Reward-related regions became more active, while regions linked to fear and to judging other people became quieter. In 2006 Gregory Berns and colleagues scanned 32 volunteers who waited for a mild electric shock and let them choose between a stronger shock soon or a weaker one later. Most chose to get it over with, and nine, whom the authors called extreme dreaders, took a bigger shock sooner rather than wait. The dread tracked activity in regions that attend to the expected pain, so it was not simply “fear.” Waiting is part of the cost.

The limits are as important as the findings. In 2009 a group of researchers scanned a dead Atlantic salmon, showed it photographs of people in social situations, and asked it, tongue in cheek, what the people were feeling. With the statistics uncorrected, parts of the dead fish’s brain “lit up.” With the correction applied, nothing did. The poster, later published and awarded an Ig Nobel Prize in 2012, was a joke with a point: thousands of voxels are tested at once, and some will pass any threshold by chance. The activity above lets you see this for yourself.

The problem is not only a joke. In 2016 Anders Eklund, Thomas Nichols, and Hans Knutsson ran about three million fake group analyses on resting data from 499 people and found that a common correction could produce false alarms far more often than the stated 5 percent, up to 70 percent in some settings; their first estimate that this called into question some 40,000 studies was walked back in a correction. Then there is size. In 2022 Scott Marek and colleagues pooled about 50,000 people and found that the typical brain-versus-behavior correlation was tiny, around 0.01, while the median published study had about 25 people. Small studies of links between scans and personality or illness will often not replicate.

A third limit is the step from a patch to a thought. Because the amygdala is active when people are afraid, it is tempting to read activity in it as fear, but it is also active for novelty and for other things. This inference from a region to a mental state, called reverse inference, is weak, as Russell Poldrack explained in 2006: it is only as good as the selectivity of the region, and few regions are very selective.

A last limit comes up in courts and marketing: lie detection. In laboratories, fMRI can separate some lies from truth at rates from about 71 percent to over 90 percent, but real-world error rates are unknown, and in one experiment simple mental countermeasures dropped detection of an individual’s lies from 100 percent to 33 percent. In 2010 a federal magistrate in Tennessee, in United States v. Semrau, kept a defendant’s fMRI test out of evidence, and the Sixth Circuit upheld that in 2012. The idea that the mouth can lie but the brain cannot is not currently true of a scan.

Reason and emotion are not in separate rooms

The old picture of the mind has reason and emotion as two powers in conflict. Plato’s charioteer, in the Phaedrus, steers two unruly horses. A caricature of Descartes, who in fact wrote in 1649 that the passions are good by nature, splits mind from body. In the film The African Queen (1951), Katharine Hepburn’s missionary tells Humphrey Bogart’s riverboat captain, “Nature, Mr. Allnut, is what we are put in this world to rise above.” The line draws a laugh because it states the old view so firmly.

Brain evidence points the other way. As Course 10 described, patients with damage to the ventromedial prefrontal cortex can score well on intelligence tests and still make disastrous choices (Bechara and colleagues, 1994). In 2008 Luiz Pessoa argued that complex behavior comes from coalitions of networks, and that no region is purely “emotional” or purely “cognitive.” The amygdala looks less like a fear organ than a detector of what matters, sensitive to novelty, threat, and reward. Even Joseph LeDoux, who did more than anyone to establish its role in fear, argued in 2012 that the circuits that respond to danger are not the ones that give us the conscious feeling of fear. A woman called S.M., whose amygdalae were destroyed by a rare disease, showed no fear of snakes or a haunted house, but when she breathed air with extra carbon dioxide she, like two other patients with similar damage, panicked. Fear does not live in one place.

The lecture on which this course draws puts it more strongly: decisions are made by emotion, and reason only follows. A version of that view is respectable. Jonathan Haidt’s 2001 article “The emotional dog and its rational tail” proposed that moral judgments usually come from quick intuitions, with reasoning added afterward. It is one influential theory, and it has serious critics, who argue that intuitions are shaped by earlier reasoning and that people do reason in real moral dilemmas. “Emotion decides” and “reason decides” are both too simple. The brain does not keep them in separate rooms.

Borrowed circuits: why children write b for d

The brain handles reading with equipment that was not built for it. Writing is about 5,000 years old, counting from the earliest Sumerian tablets, and the oldest Chinese writing we have, the oracle bone inscriptions, is about 3,250 years old. That is too short for evolution to have shaped a reading organ. Instead, as Stanislas Dehaene and Laurent Cohen argued in 2007, reading “recycles” older circuitry: an area on the underside of the left temporal lobe, in the fusiform gyrus, comes to respond to written words. In a 2010 study of 63 adults, 10 of whom could not read, 22 of whom had learned as adults, and 31 who had read since childhood, reading ability strengthened the response of this area to writing, and most of the change appeared even when reading was learned in adulthood.

Recycling has costs, and the b-and-d problem shows one. For most objects, left and right do not matter: a tiger facing left and a tiger facing right are both tigers, and the visual system learns to treat mirror images as the same thing. Letters break the rule, because b and d are the same shape facing opposite ways, and people who learn to read pay a small “mirror cost” that people who cannot read do not (Pegado and colleagues, 2014). Most young children write some letters backward while they learn, and the reversal fades as reading improves. It is not a sign of dyslexia, which affects somewhere between about 5 and 10 percent of people and comes mainly from trouble with the sounds of language. A child still struggling with reading after the first couple of school years deserves a reading evaluation, because the issue is reading and not the flipped letters.