[>] cannot typically relay a spike: Even if synapses are weak, it’s possible for a single neuron to drive another neuron to spike. The neurons need only be connected by a large number of synapses. However, this situation is apparently rare in practice.
[>] synapses made by the axon: Actually, synapses behave stochastically. With every spike, some randomly fail to secrete neurotransmitter.
[>] all possible pathways: For the snake, your eye communicates with your legs and not your salivary glands. For the steak, it’s the other way around. In telecom networks, such selectivity is achieved through the operation of routing. Every message has an address, which is separate from the content of the message. This is most obvious when you mail a letter. You write the address on the outside of an envelope, the content on the paper within. Similarly, you enter the address of a telephone by punching in its number to request a call, but it’s the ensuing conversation that contains the content. A node in the network routes an incoming message by looking at its address and relaying it to another node that is closer to the destination specified by the address. A message takes a pathway through the network determined by these routing decisions. These are made by human workers in the post office, and by devices called switches in the telephone network. Even if a single pathway could relay spikes, it’s not obvious how the nervous system could route spikes through the right pathway to reach a specific destination. Axons aren’t doing any routing; they just send spikes indiscriminately to all their synapses. Perhaps routing could be found elsewhere in the neuron, but there is a fundamental problem with the whole idea. Since a spike is merely a pulse, it’s unclear how it could carry both the content and the address of a message. This is why telecom networks are probably not such a good metaphor for the brain. That being said, this theoretical argument cannot exclude the possibility that messages consist of sequences of spikes, that assemblies of neurons can function as routing devices, and that the brain is like a communication network when examined at a higher level of organization. In fact, some theorists still contend that the routing operation is helpful for understanding brain function (Olshausen, Anderson, and Van Essen 1993).
[>] If dendrites lack spikes: As explained in Häusser et al. 2000 and Stuart et al. 2007, researchers have challenged the traditional conception that dendrites don’t spike. Experiments on neurons kept alive in slices of brain have demonstrated spikes in dendrites. If this phenomenon also occurs in intact brains, it could be that each dendrite of a neuron takes a vote of its synapses, and then the cell body takes a vote of its dendrites. This would be analogous to the American presidential election, in which the people of each state vote in the general election, and then the states vote in the Electoral College. In principle, it’s possible for a candidate to win this two-stage election without winning the popular vote.
[>] quantifies the weight: This is a simplification, as the notion of the “strength” of a synapse is more complex than can be summarized in a single number.
[>] “weighted voting model”: Engineers call this the “linear threshold model” of a neuron, to contrast the summation in voting, which they call a “linear” operation, with thresholding, a “nonlinear” operation: Yet another name for the model is “simple perceptron.”
[>] ranging from milliseconds: This is yet another dimension in which chemical synapses are more versatile than electrical synapses.
[>] Inhibitory synapses: More-direct evidence for the importance of synaptic inhibition comes from studies of movement. Muscles are generally organized in pairs with opposing effects. The biceps and triceps muscles, which are on either side of your upper arm, are one example. The biceps bends your elbow; the triceps extends it. Your nervous system is constantly sending spikes to both the biceps and the triceps. This is why your muscles are not completely relaxed at rest; they have some degree of “muscle tone.” When you bend your elbow, your nervous system sends more spikes to your biceps, causing it to contract, and simultaneously sends fewer spikes to your triceps, causing it to relax. One reason for this reduction is that the motor neurons controlling the triceps receive inhibition from synapses. More-direct evidence for the importance of synaptic inhibition comes from studies of movement. Muscles are generally organized in pairs with opposing effects. The biceps and triceps muscles, which are on either side of your upper arm, are one example. The biceps bends your elbow; the triceps extends it. Your nervous system is constantly sending spikes to both the biceps and the triceps. This is why your muscles are not completely relaxed at rest; they have some degree of “muscle tone.” When you bend your elbow, your nervous system sends more spikes to your biceps, causing it to contract, and simultaneously sends fewer spikes to your triceps, causing it to relax. One reason for this reduction is that the motor neurons controlling the triceps receive inhibition from synapses.
[>] tends to “inhibit” spiking: In a more accurate definition, excitatory versus inhibitory depends on whether the so-called reversal potential for the synapse is above or below the threshold voltage at which a neuron spikes.
[>] another kind of synapse: An electrical synapse, or gap junction, consists of a cluster of molecules, each of which is a tiny tunnel connecting the interior of one neuron to the interior of the other.
[>] other limitations: Electrical synapses are less versatile in many other ways. The duration of synaptic currents is fixed and short. Electrical current generally flows in both directions, though it may flow more readily in one of them. If two-way sounds superior to one-way, you might regard electrical synapses as more powerful than chemical synapses. But two-way communication between neurons can be established by two chemical synapses, one in each direction, while electrical synapses cannot establish one-way communication. Therefore two-way communication is actually a limitation. Electrical synapses are known to play an important role when a population of neurons needs to generate spikes simultaneously. Fast bidirectional communication makes sense for achieving such synchronicity. Electrical synapses exert only electrical effects, while chemical synapses can additionally trigger molecular signals within the receiving neuron. The extra steps in chemical transmission may slow it down, but they also allow for amplification, and modulation by other processes.
[>] how should our voting model be revised: A simpler effect of inhibition on pathways almost goes without mentioning: A single pathway containing a mixture of inhibitory and excitatory synapses can’t relay spikes, however strong the synapses may be.
[>] veto many excitatory synapses: In 1943, the theoretical neuroscientists Warren McCulloch and Walter Pitts introduced the first voting model of a neuron. The McCulloch–Pitts model adhered to the slogan “One synapse, one vote,” but only for excitatory synapses. An inhibitory synapse was allowed to have complete veto power over many excitatory synapses. It can be shown that the McCulloch–Pitts model is a special case of the weighted voting model, just by giving the inhibitory synapse a very large weight.
[>] makes only excitatory synapses: This follows from Dale’s Principle, because a given neurotransmitter generally has the same electrical effect on any neuron, either always excitatory or always inhibitory. (The sign of the electrical current depends on the molecular machinery on the receiving side of the synaptic cleft.)
[>] A similar uniformity: Also, the uniformity does not extend to strength; a neuron can make a strong synapse onto one neuron and a weak synapse onto another.
[>] most neurons are excitatory: The split is 80–20 in the cortex.
[>] increases its selectivity: Here’s another way of thinking about the significance of selective spiking. Nature has gone to the trouble of preventing crosstalk between wires. Why do this when signals are mixed at every neuron by convergence and divergence? The answer is that selectivity is preserved because neurons often fail to spike.
[>] albeit a very different kind: As computers have pervaded our everyday lives, we have lost sight of how strange they really are. A digital computer is a machine like no other, because of its universa
lity. Like an infinitely versatile Swiss Army knife, a computer can perform any kind of computation if equipped with the right software. (This is an informal statement of the Church–Turing thesis, which is formulated for an abstract computing model known as a universal Turing machine. It’s something like a modern digital computer with a hard disk of infinite capacity.) This is very different from your toolbox, which contains a hammer, a screwdriver, a saw, a wrench, and a drill, all of which are specialized for different functions. Since brain regions are specialized for particular functions, the brain is more like your toolbox than like a universal computer. Just as the structures of a saw and a hammer are closely related to their functions in carpentry, the structures of brain regions are likely to be closely related to their functions.
[>] deviate somewhat from the voting model: The weighted voting model is only an approximation to a real neuron, which may be more complex. Bullock et al. 2005 briefly describes inaccuracies of the approximation, and Yuste 2010 is a book-length review of the properties of dendrites.
4. Neurons All the Way Down
[>] make scientific observations: Quiroga et al. 2005.
[>] photo of Julia Roberts: Fried’s experiment was striking because it was done in humans. His results are less surprising if you’re familiar with the work of his predecessors, who did similar experiments in monkeys and other animals. For example, Desimone et al. 1984 reported neurons that responded selectively to faces.
[>] celebrity supercouple: Actually there were a few spikes, though not many. Fried and his colleagues did find another group of neurons in the same person that was selectively (dare I say nostalgically?) activated by Aniston and Pitt together, but not by Aniston alone.
[>] “celebrity neuron”: In a famous paper Horace Barlow called this the “grandmother cell” theory of perception, joking that there is a neuron in his brain that is active if and only if his grandmother is present (Barlow 1972). Gross 2002, however, credits the “grandmother cell” theory to Jerome Lettvin.
[>] small percentage: This “small percentage” model actually fits the data better than the “one and only one” model. Before, I emphasized the neurons that responded to a single celebrity, but these were actually a small minority. Many more neurons responded to no celebrities in the experiment, and even fewer neurons responded to two celebrities. To see that this is consistent with the “small percentage” model, compare the random sampling of celebrities with throwing darts while blindfolded. Finding a celebrity that activates a neuron is like hitting the dartboard; both events have low probability. It’s most likely that no dart will hit the dartboard. If you’re lucky, one dart will make it. It’s very unlikely that two or more darts will. That being said, the experiment cannot rule out the existence of neurons that truly respond to just one celebrity. To identify such neurons, it would be necessary to show patients a huge number of photos.
[>] number of possible patterns: Here we’ve simplistically defined the activity pattern to be binary: Every neuron is either active or inactive. We could refine the definition to include the rates at which the active neurons spike. Then the activity pattern would contain even more information.
[>] Leibniz was wrong: The philosophically sophisticated may disagree with my claim, saying that Leibniz was referring not to perception but to qualia, the subjective feelings that accompany perception. In other words, he was really referring to consciousness, and measurements of spiking haven’t told us much about that.
[>] This kind of mind reading: Can fMRI also be used for mind reading? Recently some researchers have argued that fMRI could be used to detect when a person is lying (Langleben et al. 2002; Kozel et al. 2005). The standard “lie detector” used in criminal prosecution and employment interviews is the polygraph. This measures blood pressure, pulse, respiration, and skin conductivity, which are supposed to reveal the hidden emotional stress that usually accompanies the act of lying. There is widespread skepticism, however, about the accuracy of the polygraph, and because fMRI directly assesses mental state by measuring the activation of the brain, it could potentially be more accurate. In laboratory experiments, some researchers have claimed good results with using a brain scanner to distinguish between lying and truth-telling human subjects. Based on this research, businessmen have founded two new companies seeking to commercialize fMRI lie detection. It’s still not clear whether fMRI will turn out to be superior to the polygraph, but that’s irrelevant to the discussion here. The point is that fMRI researchers are hoping only for the crudest kind of mind reading. None of them would dream of using fMRI to read out a highly specific mental property like the perception of Jennifer Aniston.
[>] “the shoulders of giants”: Recently some revisionist historians have interpreted this remark as sarcasm rather than modesty, as it comes from a letter to rival scientist Robert Hooke, who was a hunchback. Newton and Hooke later became enemies because of a dispute over optics.
[>] “receives excitatory synapses”: You may have noticed something missing from this rule: inhibitory neurons. Most cortical neurons are excitatory, but we should not neglect the inhibitory neurons, as they surely have some function too. Recall that the “Jennifer Aniston neuron” did not spike for photos of Jen with Brad Pitt. We can emulate this behavior by adding to our construction an inhibitory synapse from a neuron that detects Brad. If this synapse is strong enough, then its vote will override the votes from the neurons that detect components of Jen, and keep the neuron silent if Brad is present. More generally, it has been theorized that inhibitory synapses are helpful for making fine distinctions between similar stimuli. Excitatory synapses may enable a neuron to spike for a certain type of nose, while inhibitory synapses enable it to not spike for similar types of noses.
[>] hierarchical organization: Actually the part–whole rule was used to wire up only every other layer of his network. The other half were wired by another rule: A neuron receives excitatory synapses from neurons that detect slightly different versions of the same stimulus. The neuron has a low threshold for spiking and therefore responds to any of the stimulus variations. This rule is required for achieving another important property of perception: its invariance to “irrelevant” differences between stimuli.
[>] perceptron: Some use perceptron to refer only to the case of a single layer of synapses, and specify multilayer perceptron for the more general case. But Rosenblatt originally meant the term to refer to a multilayer network, and I follow his usage here.
[>] the layer just below: The perceptron has a feature that is not consistent with the known connectivity of the brain. Its pathways go only from the bottom of the hierarchy to the top. In real brains, there are also connections going in the opposite direction. What could be the role of these top-down pathways in perception, and how are they likely to be organized? In the “interactive activation” model of McClelland and Rumelhart 1981, a letter-detecting neuron receives bottom-up connections from neurons that detect the strokes of the letter. (Such part-to-whole connections were discussed in the main text.) But this fails to explain a simple phenomenon: How do you know that the middle letter of C–T is likely to be A, O, or U, and not E or I? In the interactive activation model, a letter-detecting neuron also receives top-down connections from neurons that detect words containing the letter. In the above example, an A detector is assumed to receive a connection from a CAT detector. More generally, one can imagine the rule “A neuron that detects a whole sends excitatory synapses to neurons that detect its parts.” This allows a neuron to detect a stimulus by weighing evidence received from both bottom-up and top-down connections.
[>] people who have blue eyes: It’s because many wholes can share a single part that a hierarchical representation is more efficient than a flat one.
[>] connectionism: The term connectionism more commonly refers to a 1980s movement in cognitive science that sought to explain the human mind using model networks of weighted voting neurons. Philosophers of mind argued over its merits relative to the “symbolic�
�� approach of understanding the mind as a digital computer. As this heated debate recedes into history, it’s better to use the word in the broader sense I’ve defined, as an intellectual tradition that dates back to the nineteenth century and is still evolving.
[>] perception or thought: The MTL is regarded by some as the top of the hierarchy hypothesized earlier (see Figure 51). At the bottom are areas of the cortex devoted to perception alone. Thinking does not activate the neurons in these areas, or at least not so much. The dividing line between perception and thinking does not appear to be sharp. Rather, the involvement of neurons in thinking appears graded, increasing gradually as one ascends the hierarchy.
[>] never function perfectly: According to some theorists, inhibitory neurons may be more precise at controlling the spread of activity than neuron thresholds, providing for superior memory recall.
[>] information overload: Inhibitory neurons increase memory capacity by retarding the spread of activity. To serve this dampening function, the connections of the inhibitory neurons don’t need much organization at all. If each receives synapses from a random selection of excitatory neurons, it will be activated whenever the “mob” is active. If it sends synapses back to another random selection of excitatory neurons, it will exert a dampening effect on the crowd. An engineer would say that inhibitory neurons exert “negative feedback” on excitatory neurons. The household thermostat is the classic example of negative feedback. If the temperature of a heated room increases beyond a certain point, the thermostat turns off the heat; if the temperature decreases, the thermostat turns on the heat. In both cases the thermostat acts to oppose the change in temperature, in the same way that inhibitory neurons act to oppose changes in the activity of excitatory neurons. In this view, inhibitory neurons play a supporting role in brain function, so their connections don’t have to be very specific.
Connectome Page 31