Sebastian Seung noted that I had not read his book. Fair enough. I was not trying to single out Seung, but I can see that my post uses his examples enough that it seems like I am reviewing the book without reading the book.
I bought the book and read it.
The timing has turned out to be good, for today, Carl Zimmer and Robert Krulwich are moderating a debate between Seung and Anthony Movshon at Columbia University. The formal title of this debate is, “Does the brain’s wiring make us who we are?”
The Twitter hashtag? #brainbrawl. Much more fun, although I worry it promises more sparks than one will probably get from an academic debate about neuroscience. For example, The description of the #brainbrawl debate ends with the question, “Are brain maps the future of neuroscience or an empty promise?” I doubt anyone will argue that describing connectomes will yield nothing. As I said, even though I think the enterprise will fail, on some level, it will fail in an interesting and productive way.
Additionally, Seung was recently profiled in Discovery magazine by Zimmer and Wired. The other #brainbrawl moderator, Robert Krulwich, examines the issues behind “Jennifer Aniston neurons” here, and examines mapping, asking “To map or not to map the brain?”
Early in Connectome, Seung says:
This book proposes a simple theory: Minds differ because connectomes differ.
I read through, making notes, and was caught off guard in the last chapter. After spending all this time making arguments for his connectome theory, I was surprised that in the last chapter, Seung surrenders the war.
Yet Seung remains optimistic that he can win one battle.
In Chapter 4, Seung argues that the function of a neuron is determined primarily by its connections with other neurons. Take a neuron from one region of the brain, stick in into another part of the brain, hook up the input and output synapses the right way, and away you go.
To put it another way, Seung initially treats neurons as interchangeable widgets.
At this point, I was getting ready to start a big long list of reasons why neurons are not interchangeable. There are not only neurons that generate action potentials, there are neurons with graded potentials, pacemaker potentials, and plateau potentials. There are the the passive electrical properties of the dendritic tree. These are just a couple of examples of the long list of intrinsic properties of neurons that make them distinguishable by far more than their connections to other neurons.
These are all in addition to changes of neuronal physiology introduced by neuromodulation.
Seung addresses the intrinsic properties of neurons in the book’s last chapter:
In principle, every neuron is unique in its behavior, owing to the unique configuration of its ion channels. This is a far cry from the weighted voting model, according to which all neurons are essentially the same. But it sounds like bad news for brain simulation. If neurons were infinitely diverse, how could we ever succeed at modeling them? Measuring the properties of one neuron would tell you nothing about another.
There is one hope for escaping the morass of infinite variation: neuron types.
The concept of neuron “types” is initially introduced in Chapter 10. Seung initially wants to categorize the neurons by their connections.
If two neurons are connected to similar or analogous partners, they should be grouped in the same type.
Seung then makes an extrapolation that is plausible, but I would say not yet proven: neurons with similar anatomical connections will have similar physiology and other intrinsic properties.
(N)eurons of the same type generally exhibit the same electrical behaviors. This is presumably because their ion channels are distributed in the same way. If this is the case, then neural diversity is actually finite.
By this point in the book, Seung has already introduced the only complete connectome of any animal, that of the 302 cell nervous system of the worm, Caenorhabditis elegans. I expected that this would be used as the flagship case of how connectomes can help us understand behaviour.
Instead, Seung declares defeat.
He estimates that of the 302 neurons in C. elegans, there are about 100 different types of neurons. And that proportion of distinct neuronal types, he says, is too high for a purely connection-based model to explain the worm’s behavior.
As you read this section (emphasized sections are mine), recall that he's been repeated the mantra, “You are your connectome,” for the previous 14 chapters. Now, in the very last chapter, an important missing piece gets added.
Thus the earlier claim should be revised to say, “You are your connectome plus models of neuron types.” (Let’s assume that a connectome is defined to specify the type of each neuron.) But the models of neuron types are likely to contain much less information than the connectome, as most scientists agree that there are far fewer neuron types than neurons. In this sense, “You are your connectome” would remain a very good approximation.
So “You are your connectome” would be a terrible approximation for a worm, even though it might be almost perfect for us.
If Seung is not convinced that the connectome theory yields meaningful information for C. elegans, I am skeptical that the ratio of neuronal types to the total number of neurons in a human brain* is going to be more favourable to connectomes explaining the mind.
At the very least, we have to acknowledge that there is a continuum.
One one end, we have circuits in which activity is dominated by few types of neurons, in which understanding the synaptic connections explains a large proportion of the function. Escape circuits are excellent examples, such as those in crayfish and fishes.
On the other end, we have circuits that are composed of many personalities, and where the intrinsic physiological state of the neuron matters a great deal to the function. Seung suggests C. elegans falls into this category, and I’ll suggest the crustacean stomatogastric nervous system as another.
It is not clear to me where regions of human brains may fall on this continuum. When I visited the Neuroscience department at the University of San Antonio recently, and was discussing my previous post, several people indicated to me that those interested in connectomes placed special emphasis on the cerebral cortex, apparently because those neurons are a little closer to the “interchangeable widget” end of the scale than other brain regions.
Losing the war - the ability of connectome theory to explain large parts of behaviour in many species - may not matter for Seung, however. For Seung, like many neuroscientists, there is only one battle worth fighting, the one for human minds. Near the end of the book, he writes:
(T)here is only one truly serious problem in science and technology, and that is immortality.
The last sections of the book deal with topics like transhumanism, immortality, and putting consciousness into computers. These are a very distinct group of pre-occupations, all revolving around human consciousness.
As a neurobiologist, I want a theory that can explain how nervous systems generate all the behaviours of all the diversity in the animal kingdom. A theory that only explains the human mind is a small and paltry thing.
Connectome theory reminds me of Anne Elk's theory:
In one sense, the theory is trivially true. But in another sense, the theory omits so much that it doesn't end up telling us anything we wanted to know.
Seung S. 2012. Connectome: How the Brain's Wiring Makes Us Who We Are. New York: Houghton Mifflin Harcourt. 384 pgs. ISBN: 978-0-547-50818-4
* Seung uses an estimate that the human brain contains 100 billion neurons. More recent estimates put the count closer to 85 billion neurons, which he could not have had given the lead time of publishing a book.