Connectome

by Sebastian Seung

  The “meaning of life” includes both universal and personal dimensions. We can ask both “Are we here for a reason?” and “Am I here for a reason?” Transhumanism answers these questions as follows. First, it’s the destiny of humankind to transcend the human condition. This is not merely what will happen, but what should happen. Second, it can be a personal goal to sign up for Alcor, dream about uploading, or use technology to otherwise improve oneself. In both of these ways, transhumanism lends meaning to lives that were robbed of it by science.

  The Bible said that God made man in his own image. The German philosopher Ludwig Feuerbach said that man made God in his own image. The transhumanists say that humanity will make itself into God.

  Epilogue

  It’s time to return to reality. We’ve each got one life to live, and one brain to do it with. In the end, every important goal in life boils down to changing our brains. We are blessed with natural mechanisms for transformation, but we find their limitations frustrating. Beyond appealing to our curiosity and sense of wonder, can neuroscience give us new insights and techniques for changing ourselves?

  I’ve argued that one of the most important ideas of our time is connectionism, the doctrine that emphasizes the importance of connections for mental function. According to this notion, changing our brains is really about changing our connectomes. Connectionism dates back to the nineteenth century, but empirical evaluation of its claims has been difficult. At long last, thanks to the emerging technologies of connectomics, we are poised to test the doctrine. Is it indeed true that minds differ because connectomes differ? If we succeed in answering that question, we will also be able to identify desirable changes in the brain’s wiring.

  The next step will be to devise new methods of promoting such changes, based on molecular interventions that promote the four R’s: reweighting, reconnection, rewiring, and regeneration. The methods would also utilize training regimens that harness the four R’s to bring about positive changes.

  To realize all these advances, we must continue to develop the necessary technologies. In the history of science, there are many examples of conceptual barriers that could not be surmounted by researchers, however brilliant they were, until the right tools became available. You wouldn’t expect a caveman to figure out the workings of an old-fashioned mechanical clock if he didn’t have a screwdriver. In the same vein, it’s unrealistic to expect neuroscientists to figure out the brain without extremely sophisticated tools. Our technologies are starting to become equal to the task, but we will need to make them many times more powerful.

  We need to create a research environment that fosters these technological advances. One possibility is to undertake “grand challenges,” ambitious projects that stimulate our imagination and mobilize our intellectual efforts. We could set a goal of finding the entire neuronal connectome of a mouse brain using electron microscopy, or the entire regional connectome of a human brain with light microscopy. The projects are of comparable difficulty, because they require the acquisition and analysis of similar amounts of data. I estimate that either would require a decade of intense effort. Both connectomes would be invaluable resources for neuroscientists, just as genomes have become indispensable to biologists.

  These projects would be enormously difficult, but we could simultaneously pursue shortcuts. With the technologies developed, it would be possible to rapidly and cheaply find smaller connectomes. Compared with the grand challenges above, it should be a thousand times faster to find the neuronal connectome of a cubic millimeter of brain, or the regional connectome of a mouse brain. Finding many smaller connectomes would be important for studying individual differences and change.

  Why should we invest in future technologies when we need to find better treatments for mental disorders right now? I think we should do both. Our therapies will surely improve over the next few years, but I expect that it will take decades to find true cures. Since this will be a continuing battle, it’s worth making a reasonable investment today to reap rewards in the long run.

  You may be skeptical that technology will ever progress enough to find connectomes quickly and cheaply. Before the Human Genome Project began, sequencing an entire human genome seemed almost impossible too. Connectomics might look difficult, but there’s a certain sense in which it’s trivial compared with the larger endeavor of neuroscience. Since the goal is well defined, we know exactly what success means, and can quantify progress. In contrast, the broader goal of neuroscience—to understand how the brain works—is only hazily defined. Even the experts don’t agree about what it means. Once a goal is clearly defined, time, money, and effort are likely to yield progress. That’s why I believe that connectomics will achieve its goals, however ambitious they might seem. We just need to rise to the challenge.

  The young boy laughed as he splashed in the water. Returning to land, he asked, “Teacher, why does the stream flow?” The old man gazed silently at the novice and replied, “Earth tells water how to move.” During their journey back to the temple, they crossed a precarious footbridge. The novice clutched the old man’s hand tightly. He looked at the stream far below and asked, “Teacher, why is the canyon so deep?” As they reached the safety of the other side, the old man replied, “Water tells earth how to move.”

  I believe the stream inside our brain works in much the same way. The flow of neural activity through our connectomes drives our experiences of the present and leaves behind impressions that become our memories of the past. Connectomics marks a turning point in human history. As we evolved from apelike ancestors on the African savannah, what distinguished us was our larger brains. We have used our brains to fashion technologies that have given us ever more amazing capabilities. Eventually these technologies will become so powerful that we will use them to know ourselves—and to change ourselves for the better.

  Acknowledgments

  David van Essen planted the seed for this book by inviting me to lecture at the 2007 meeting of the Society for Neuroscience. Speaking before an audience of thousands, I concluded by laying out the challenge of finding connectomes. Upon hearing the buzz that followed, Bob Prior encouraged me to write a book. I took his suggestion but decided to target the general public. Since no knowledge could be assumed, I would have to argue from first principles and question all my beliefs. I was following the prescription “Empty your cup so that it may be filled.”

  When I finished a draft in 2009, Catharine Carlin pointed me to Jim Levine, and Dan Ariely made the introduction. Jim’s enthusiastic offer to serve as my agent was an enormous boost. He recruited the brilliant Amanda Cook, who has repeatedly prodded me with the question “Why should we care?” Beyond editing my writing and improving my storytelling, she has shaped my thinking. I never anticipated how drastically the book would change under her guidance, and I consider myself lucky that it did.

  A life in science comes with a wonderful fringe benefit—opportunities to meet smart and interesting colleagues. Many fascinating discussions with other neuroscientists have enriched this book. The wise counsel of David Tank originally set me on the road to connectomes. The encouragement of Winfried Denk, who critiqued two drafts of the book, kept me writing. Jeff Lichtman patiently educated me about synapse elimination and neural Darwinism. Ken Hayworth explained his cutting machines and passionately argued the case for transhumanism. Daniel Berger contributed many suggestions for improving the book.

  I am grateful to Scott Emmons and David Hall for information on C. elegans, Axel Borst on the fly brain, Kevin O’Hara on the California redwood, Misha Tsodyks and Haim Sompolinsky on associative memory models, Eric Knudsen and Stephen Smith on reconnection and rewiring, Carlos Lois and Fatih Yanik on regeneration, Mitya Chklovskii and Alex Koulakov on wiring economy, Kristen Harris on serial electron microscopy, Guyeon Wei on semiconductor electronics, Dick Masland and Josh Sanes on neuron types, Kathy Rockland and Almut Schüez on cortical anatomy, Harvey Karten and Jerry Schneider on brain evolution, Michale Fee on birdson
g, Li-Huei Tsai and Pavel Osten on brain disorders, Vamsi Mootha on biology, Niko Schiff on neurology, Drazen and Danica Prelec on philosophy and psychology, and Michael Häusser and Arnd Roth on dendritic biophysics.

  Mike Suh and John Shon assisted me with the initial proposal for the book. Along with Janet Choi and Julia Kuhl, they also commented on the final version. Scott Heftler suggested some fun comparisons. Fellow authors Sue Corkin, Mike Gazzaniga, Allan Hobson, and Lisa Randall advised me at critical junctures. The meticulous editing and impeccable logic of Katya Rice polished the prose, to my delight.

  Several public speaking experiences attuned me to the zeitgeist. Ute Meta Bauer invited me to lecture for the Visual Arts Program at MIT, Susan Hockfield brought me to the World Economic Forum, and Sarah Caddick helped me spread the word through a 2010 TED talk.

  Finally, I owe thanks to the Gatsby Charitable Foundation, the Howard Hughes Medical Institute, and the Human Frontiers Science Program for funding my research in connectomics.

  Notes

  Introduction

  page

  [>] “Let man contemplate Nature”: Pensées 72.

  “the eternal silence”: Pensées 206.

  [>] Figure 2: The picture was taken by differential interference contrast (DIC) microscopy, and can be found at wormatlas.org, a wonderful database of information about the worm. The scale bar is 0.1 millimeter. The two ellipsoids are embryonic worms.

  centralized in a single organ: The majority of the worm’s neurons and synapses are found in a structure called the nerve ring. (Actually this is true for hermaphroditic worms, but the nerve ring is less dominant in the much rarer male.) The nerve ring surrounds the worm’s “throat” and is the closest thing to its “brain.” The human brain contains the overwhelming majority of neurons in the human nervous system. The rest are in the spinal cord and scattered in other parts of the body.

  [>] Figure 3: The first map of the entire C. elegans nervous system was published by White et al. 1986. Although their map is generally considered definitive, it is not actually complete. Varshney et al. 2011 updated it with data drawn from other sources but estimated that 10 percent of the worm’s connections were still missing. The diagram shown in Figure 3, summarizing their work, can also be found at wormatlas.org.

  [>] million pages long: To browse the human genome, go to the NCBI Map Viewer (www.ncbi.nlm.nih.gov/projects/mapview). From there you can navigate to a page that displays all the chromosomes in the human genome (look for Homo sapiens, the official name for our species). Clicking on any chromosome will give you a more detailed map showing the locations of genes, and further clicking will display actual DNA sequences. Figure 4 shows the beginning of chromosome 11. To find the sequences of specific genes, you can search for the names of the proteins they encode.

  [>] unique in a way that a worm is not: Worm connectomes, although more similar to one another than human connectomes, are not identical. The topic is explored at greater length in Chapter 12.

  fixed from the moment of conception: It’s an oversimplification to say that your genome is fixed. Each of your cells contains a copy of your genome. (There are exceptions, such as red blood cells, which lack DNA when mature.) The copies are almost identical, but there are slight differences. Some are caused by copying errors as your cells divide, and can lead to cancer. Some differences are important for function, as in certain cells of the immune system. DNA can also be modified in ways that do not change the sequence, which is part of a more general class of phenomena known as epigenetics.

  [>] than your genome has letters: This comparison is based on a figure of one quadrillion (1015) synapses, which was obtained by multiplying 100 billion neurons in the brain with an estimated 10,000 synapses per neuron. This is likely an overestimate, and its exact value should not be taken too seriously. A more reliable enumeration has been performed for a brain structure called the neocortex, and yielded 0.16 quadrillion synapses (Tang et al. 2001).

  [>] and phallic graffiti everywhere: Beard 2008.

  notions of the self: I’m indebted to Ken Hayworth for clarifying this point to me.

  contains 100 billion neurons: A recent study places the average number at 86 billion (Azevedo et al. 2009).

  1. Genius and Madness

  [>] Ivan Turgenev: The brains of Turgenev and other famous Russians are described in Vein and Maat-Schieman 2008.

  [>] Sir Arthur Keith: Keith 1927. Unfortunately for Keith and his reputation, he is remembered less for his scientific discoveries than for his endorsement of the Piltdown Man. These skull fragments, purported to be a “missing link” in the evolution of man from ape, were eventually exposed as fake. Piltdown Man became one of the most famous hoaxes in the history of science.

  [>] French theoretical physicist: Keith resolved his conundrum in a similar way, writing that “a detailed study of Anatole France’s life, so far as it is known, shows us that he was in many senses a primitive man.” He wrapped up his essay by reaffirming his belief that brain size and intelligence are actually related: “In the long run I expect it will be found that there is a close correspondence between brain mass and the degree of function subserved by that organ.”Keith resolved his conundrum in a similar way, writing that “a detailed study of Anatole France’s life, so far as it is known, shows us that he was in many senses a primitive man.” He wrapped up his essay by reaffirming his belief that brain size and intelligence are actually related: “In the long run I expect it will be found that there is a close correspondence between brain mass and the degree of function subserved by that organ.”

  [>] average head size: Galton 1889.

  [>] people with bigger brains: McDaniel 2005.

  [>] with high accuracy: If the correlation coefficient of two variables is r, then knowing one variable reduces the typical prediction error of the other by a factor of √1– r2.

  [>] correlation between IQ and brain volume: McDaniel 2005.

  [>] “Beauty Map”: Galton recounts the story in the last chapter of his memoirs, about “Race Improvement, or Eugenics” (Galton 1908). In a three-volume hagiography, Karl Pearson reminisced about his mentor: “Galton, influenced by his own motto . . . , seldom went for a walk or attended a meeting or lecture without counting something. If it was not yawns or fidgets, it was the colour of hair, of eyes, or of skins” (Pearson 1924, p. 340). Galton.org pays tribute to the man.

  [>] imbecile: Pearson 1906. While Pearson confirmed Galton’s finding that head size and school grades were statistically related, he also noted that head size was a poor predictor of school grades for any particular individual. Even handwriting quality was a better predictor than head size.

  [>] cerebrum, the cerebellum, and the brainstem: Swanson 2000 divides the brain more finely into the cerebral cortex, basal ganglia, thalamus, hypothalamus, tectum, tegmentum, cerebellum, pons, and medulla. Swanson argues that all of the many proposed schemes for coarsely dividing the brain can be regarded as different groupings of these nine basic parts. For example, in the tripartite scheme of Figure 7, the cerebrum is defined as the cortex plus the basal ganglia, and the brainstem as the rest of the parts minus the cerebellum. A book-length exposition of his views can be found in Swanson 2012. Note that some authorities exclude the thalamus and hypothalamus from the brainstem, so its definition is ambiguous.

  [>] spares mental abilities: Although introductory textbooks usually don’t mention it, cerebellar damage does have some effects on emotion and cognition (Strick, Dum, and Fiez 2009; Schmahmann 2010).

  [>] largest of the three parts: The cerebrum is largest by volume, but the cerebellum has the most neurons, with an estimated 70 billion (Azevedo et al. 2009) or 100 billion (Andersen, Korbo, and Pakkenberg 1992). Almost all of these are the so-called granule cells. Because these are very small, the cerebellum takes up only 10 percent of the brain’s volume (Rilling and Insel 1998). The neocortex, the dominant part of the cerebrum, is estimated to contain 20 billion neurons (Pakkenberg and Gundersen 1997).

&n
bsp; [>] into four lobes: The borders of the occipital lobe are defined with additional landmarks but are somewhat arbitrary. The four lobes are named for the four bones of the skull that overlie them. Some authorities define a fifth, limbic lobe. This is visible on the faces of the hemispheres exposed by cutting the cerebrum in half along the longitudinal fissure. Buried inside the Sylvian fissure is a part of the cortex known as the insula, which is large enough that some regard it as another lobe.

  [>] prisons and mental asylums: Micale 1985.

  [>] not confining them in chains: Harris 2003.

  [>] Figure 10: The lesion is centered in the inferior frontal gyrus (fold) of the left cerebral hemisphere. The story of the patient Leborgne, nicknamed Tan, is told in Finger 2005 and Schiller 1963, 1992.

  [>] hemispheres looked so similar: Researchers have also found slight structural asymmetries between the right and left hemispheres, but it’s been difficult to tell whether these have anything to do with lateralization of function (Keller et al. 2009).

  [>] dominant for language: Rasmussen and Milner 1977. In a minority of left-handers and ambidextrous people, the right hemisphere is dominant for language, or both hemispheres are involved.

  [>] Harvey sent specimens: Abraham 2002; Paterniti 2000.

  [>] Sandra Witelson: Witelson, Kigar, and Harvey 1999.

  [>] brains of luminaries: Burrell 2004.

  [>] his 1819 treatise: Gall 1835.

  [>] IQ is correlated: Jung and Haier 2007.Jung and Haier 2007.