[>] left to right: Note that this looks like the perceptron shown earlier, but turned on its side. Although a synaptic chain can be viewed as a special case of a perceptron, it’s quite different from the typical perceptron, which is used to model perception. The neurons in one layer of a perceptron typically detect different stimuli, so each is wired to a different subset of neurons in the previous layer. (Or if they are wired to the same neurons, the strengths of the synapses differ.) All the neurons in one layer of a synaptic chain get activated together, so their connections with the previous layer need not be different. The synaptic chain has been formalized in mathematical models by a number of researchers (see, for instance, Amari 1972 and Abeles 1982). The American theoretical physicist John Hopfield developed related models in the 1980s.
[>] theory of connectionism: Donald Hebb proposed and named the cell assembly (Hebb 1949). Early computer simulations of model networks with cell assemblies were performed in the 1950s. The English theorist David Marr and the Japanese theorist Shun-ichi Amari were two prominent researchers who studied the equations of such models using pencil and paper in the 1960s and 1970s (see, for example, Marr 1971 and Amari 1972). But the real heyday of connectionism came in the 1980s, following the seminal papers of John Hopfield (Hopfield 1982; Hopfield and Tank 1986). Using esoteric mathematical techniques from a branch of physics known as spin glass theory, theoretical physicists had a field day calculating memory capacity through a statistical treatment of the effects of overlap between cell assemblies (see Amit 1989; Mezard, Parisi, and Virasoro 1987; and Amit et al. 1985). By the time this flurry of activity petered out in the 1990s, these researchers had discovered many interesting properties of the models. Also around this time, the PDP Research Group, a collective of cognitive scientists, published an influential two-volume manifesto containing many interesting connectionist models (Rumelhart and McClelland 1986).
[>] “Problem of Serial Order”: Lashley attributed the “associative chain model” to the British psychologist Edward Titchener, citing a book from 1909. Actually both authors spoke of chains of psychological associations rather than neural connections. Strangely, Lashley did not use the word synapse in his article, although he was a neuroscientist. Nevertheless, the notion of synaptic chains is implicit in his writing. Lashley attributed the “associative chain model” to the British psychologist Edward Titchener, citing a book from 1909. Actually both authors spoke of chains of psychological associations rather than neural connections. Strangely, Lashley did not use the word synapse in his article, although he was a neuroscientist. Nevertheless, the notion of synaptic chains is implicit in his writing.
[>] huge variety of activity: There would also have to be points where two chains converge into one, or we would quickly run out of neurons.
[>] problem of syntax: In a similar vein of criticism, some computer scientists have argued that relations between ideas are richer than simple associations. To say that the ideas of fish and water are associated does not do justice to their relationship. It’s more richly descriptive to say that a fish “lives in” water. Computer scientists represent such relationships with a “semantic network,” which looks like a connectome except that each arrow is labeled with a type of relation.
[>] addressed Lashley’s second: These connectionist models achieve greater computational power by introducing latent or hidden variables, to augment the variables that are used to represent explicit ideas.
5. The Assembly of Memories
[>] two-and-a-half-ton: The blocks varied in size; this number is an estimate of the average (Petrie 1883). Most blocks were limestone, but some were granite.
[>] 2.3 million: Petrie 1883. Petrie 1883.
[>] one hundred thousand workers: Herodotus wrote that one hundred thousand slaves labored for twenty years to transport the blocks from a distant quarry to the pyramid. Many recent Egyptologists have disagreed, arguing that the main quarry was nearby, the workforce was far smaller, and the workers were not slaves.
[>] “There exists in the mind”: Plato, Theaetetus.
[>] straight-edged instrument: Draaisma 2000.
[>] Artisans and engineers: The term plasticity comes from materials science. A plastic material holds its new shape when deformed; an elastic material bounces back to its original shape. Because wax is plastic, it can hold an impression and hence store information about the past. In this technical usage, plastic is an adjective that refers to the behavior of materials in response to deformation. Plastic is more commonly used as a noun, to refer to any of the synthetic polymer materials widely used in manufactured products. The common usage is related to the technical usage in that these materials can undergo plastic deformations at higher temperatures, a feature that is often used in manufacturing. These materials are usually elastic at room temperature, however. Furthermore, there are other types of materials, such as metals, that can also undergo plastic deformations.
[>] a phenomenon I’ll call reconnection: The reweighting–reconnection distinction is cleanest when there is at most one synapse from neuron A to neuron B, as is assumed generally in this chapter. The distinction is blurred if there are multiple synapses from A to B. Then synapse creation and elimination might leave the neurons connected, and only change the number of synapses that A sends to B. This would alter the weight of A’s vote in B’s spiking, bringing about reweighting rather than reconnection.
[>] In the 1960s: My claim of “most neuroscientists” comes from hearsay and is difficult to document rigorously. One example is the Australian neuroscientist Sir John Eccles, who wrote that learning involves “growth just of bigger and better synapses that are already there, not growth of new connections” (Eccles 1965). Rosenzweig 1996 provides some historical review from the viewpoint of a neuroscientist, but the issue should be examined by a real historian.
[>] Figure 23: This image is based on data from an experiment described in Yang, Pan, and Gan 2009.
[>] created and eliminated: Spines have also been observed to change in size, which suggests that synapses are changing in strength.
[>] counting synapses: Greenough, Black, and Wallace 1987.
[>] neo-phrenological theory: Some researchers looked at the sizes of synapses as well as their numbers. There is evidence that the size of a synapse is correlated with its strength. The researchers found that enriched environments increased the average size of synapses in the rat cortex. However, one should not equate learning with an increase in synaptic size, just as one should not equate it with an increase in synaptic number. Other experiments have demonstrated decreases in the average size of synapses. Which of these changes dominates depends on the particular location in the cortex as well as the layer of the neurons involved.
[>] Hebb: Hebb 1949. The sequential rule was also proposed in the late nineteenth century by the Scottish philosopher Alexander Bain (see Wilkes and Wade 1997), but his theory never took hold. Perhaps Bain had the misfortune of living too early, when so little was known about the brain. He knew about fibers and pathways, and guessed the existence of connections between pathways, but the existence of neurons or synapses had not yet been established.
[>] Hebbian plasticity refers: Less is known about plasticity of synapses involving inhibitory neurons, so that will not be discussed here. According to the conventional wisdom, the connections between excitatory neurons are more specific and more shaped by learning. Those involving inhibitory neurons are relatively indiscriminate and may be less influenced by learning.
[>] isolate the spikes: This method, known as “single-unit” recording, was pioneered by the English scientist Edgar Adrian, who garnered the 1932 Nobel Prize and eventually the title “Lord.”
[>] mouth of a speaker: Synapses onto muscles had already been studied in the 1930s and 1940s. In the 1950s Sir John Eccles and other researchers refined the method of intracellular recording and applied it to synapses in the spinal cord. Eccles went on to share a 1963 Nobel Prize for his efforts.
[>] b
y injecting electrical current: The text describes using two intracellular electrodes to study a specific pair of neurons. This is the most precise method of studying synapses, and is relatively recent. Eccles used a single intracellular electrode to record from one postsynaptic neuron, and stimulated a large number of presynaptic neurons by injecting current through an extracellular wire.
[>] The size of this blip: If there happen to be multiple synapses from neuron A to neuron B, then the size of the blip is the aggregate strength of all the synapses.
[>] Repeated stimulation: Bi and Poo 1998; Markram et al. 1997. Whether sequential or simultaneous stimulation is more effective at inducing plasticity depends on the type of neurons involved. Strictly speaking, these experiments did not demonstrate change in single synapses. There were multiple synapses between the measured pair of neurons, and the experiments demonstrated a change in their aggregate strength. In general, such experiments have trouble distinguishing between reweighting and reconnection. If the interaction between two neurons strengthens, it could be the result of an increase in the number of synapses between them, not just synaptic strengthening. Another interesting issue, which I don’t have space to discuss here, is the mechanism by which a synapse detects simultaneous or sequential spiking. This appears to happen through a special molecule called the NMDA receptor.
[>] Brad and Jen: It’s simplistic to say that you forgot the relationship between Jen and Brad. Although they’re no longer married, you still remember that they used to be married. To represent this knowledge, you could imagine that there is a marriage neuron and a divorce neuron. Initially the cell assembly includes the Brad, Jen, and marriage neurons. Later on, the cell assembly includes the Brad, Jen, and divorce neurons. This solution is still not entirely satisfactory, but a better solution would have to confront Lashley’s critique that connectionism cannot represent syntax, and is outside the scope of this book.
[>] two neurons are weakened: For example, Stent 1973 proposed that the synapse from A to B is weakened if A is repeatedly inactive while B is active. Other variants were proposed by many theorists. The flip side of the sequential version of Hebb’s rule is: If two neurons are repeatedly activated sequentially, the connection from the second to the first is weakened. Empirical evidence was found by Markram et al. 1997 and Bi and Poo 1998. In combination with Hebb’s rule, it’s known as “spike-timing dependent plasticity.”
[>] direct competition: Miller 1996.
[>] “trophic factors”: Purves 1990.
[>] redundantly represented: Here a cell assembly must be redefined as a set of neurons such that every connection between neurons is a strong synapse, provided that it exists. We could revise Locke’s metaphor by imagining writing on white paper in which someone has randomly poked many holes (without trying to avoid the holes). The missing parts of the paper are analogous to the missing synapses in a sparsely connected network. If your handwriting is much bigger than the holes, the information may still be readable. But if your handwriting is too small, information will be lost.
[>] the method of loci: Yates 1966.
[>] “connections are created between them”: It’s this variant of Hebb’s rule that is expressed by a ditty popular among college students learning neuroscience: “Neurons that fire together, wire together.”
[>] Gerald Edelman: Edelman 1987; Changeux 1985. A contrarian view was presented in Purves, White, and Riddle 1996, which was answered by Sporns et al. 1997.
[>] generated “on demand”: The “on demand” theory of synapse creation is analogous to Jean-Baptiste Lamarck’s theory of evolution. Lamarck argued that animals can pass on acquired characteristics to their offspring, so that variation is adaptive rather than random. For example, he believed that a person who grows larger muscles through physical training can pass on larger muscles. Lamarck’s ideas were discredited but have recently been partially revived by research on epigenetics.
[>] Jeff Lichtman: Findings are reviewed in Lichtman and Colman 2000; a readable introduction to fundamental ideas is provided by Purves and Lichtman 1985.
[>] little or no memory loss: Gilbert et al. 2000.
[>] below 18 degrees: PHCA is sometimes used when neurosurgeons remove brain aneurysms. The circulation is stopped to prevent bleeding while the aneurysm is clipped, and the low temperature prevents the brain damage that would otherwise be caused by lack of oxygen during that time. At such low temperatures the heart doesn’t beat properly; it is stopped completely by injecting potassium chloride (one of the drugs used in execution by lethal injection).
[>] two storage systems: Actually it’s more complex than this, because there are additional information stores inside the microprocessor. The RAM and hard drive are just the offboard information stores.
[>] change more slowly: I should mention that synapses also change their strengths more rapidly and temporarily. This is known as short-term plasticity, and could also be a basis of short-term memory.
6. The Forestry of the Genes
[>] different adoptive families: Bouchard et al. 1990.
[>] persons chosen at random: Strictly speaking, the proper comparison would be with two individuals drawn from different pairs of monozygotic twins raised apart.
[>] little room for argument: In a celebrated case, Sir Cyril Burt, a pioneer in the study of twins, was posthumously accused of fabricating his data. This cast doubts on the whole field, which were eventually dispelled by more solid data.
[>] First Law of Behavior Genetics: Turkheimer 2000. The Second Law is “The effect of being raised in the same family is smaller than the effect of genes,” and the Third Law is “A substantial portion of the variation in complex human behavioral traits is not accounted for by the effects of genes or families.”
[>] If one twin has autism: Steffenburg et al. 1989; Bailey et al. 1995. There is a range because the exact numerical value depends on whether autism is defined strictly or, as in the autism spectrum disorders, more inclusively. Also, sample sizes are fairly small, so the numbers are subject to statistical uncertainty.
[>] concordance rate for autism: Hallmayer et al. 2011 revises the concordance rate for DZ twins upward relative to the earlier studies of Steffenburg et al. 1989 and Bailey et al. 1995. According to the newer estimates, genetic influences are important for autism, but not as much as previously thought.
[>] What about schizophrenia?: Cardno and Gottesman 2000.
[>] synthesize proteins: You might have thought that cells don’t have to make proteins because we ingest them from food. But actually the digestive system chops up proteins into amino acids, and our cells reassemble them into different proteins.
[>] contains the same genome: There are some exceptions to this rule, such as certain cells in your immune system, variations arising from errors in DNA replication, and so-called mosaic organisms.
[>] inside and outside of the neuron: Sometimes the door and tunnel are in a nearby molecule rather than in the receptor itself. The receptor can open the door by sending another signal, much as electrically powered doors are opened by pressing a button off to the side. Such a receptor is not an ion channel, and is said to be “metabotropic.” The type of receptor discussed in the text is an ion channel, and is said to be “ionotropic.”
[>] called a “channelopathy”: Kullmann 2010.
[>] clay or metal caps: Miles and Beer 1996; Leroi 2006.
[>] reduced brain size at birth: Brain size of at least two or three standard deviations below the norm is the clinical definition of microcephaly (Mochida and Walsh 2001).
[>] pattern of folds: Mochida and Walsh 2001.
[>] intermarriage between cousins: Leroi 2006; Mochida and Walsh 2001.
[>] severe mental retardation: Mochida and Walsh 2004.
[>] control neuronal migration: Guerrini and Parrini 2010.
[>] growth cone acts like a dog: Kolodkin and Tessier-Lavigne 2011.
[>] 200 million axons: The numerical estimate comes from Tomasch 1954 and Aboitiz e
t al. 1992.
[>] milder than in microcephaly: Paul et al. 2007. “Split-brain” patients, with a corpus callosum severed by epilepsy surgery, also have relatively minor impairments.
[>] half a million per second: The estimate of half a million comes from Huttenlocher 1990, Figure 1, which summarizes data from Huttenlocher et al. 1982.
[>] number of synapses has dropped: Huttenlocher and Dabholkar 1997. Similar observations were made in the monkey cortex by Rakic 1986.
[>] a less than ideal way: Earlier I mentioned channelopathies, defective ion channels that cause electrical signaling of individual neurons and synapses to malfunction. Because neural activity alters connectomes by mechanisms like Hebbian plasticity, a channelopathy is expected to lead to abnormal connectivity. This example shows that connectopathies may be associated with other types of neuropathology.
[>] autistic brain is slightly smaller: Redcay and Courchesne 2005 is a meta-analysis, combining the results of many studies.
[>] schizophrenia, like autism: Lewis and Levitt 2002; Rapoport et al. 2005.
[>] too few connections: Courchesne and Pierce 2005; Geschwind and Levitt 2007.
[>] schizophrenia, too, be caused: Friston 1998. According to Kubicki et al. 2005, Carl Wernicke and the German psychiatrist Emil Kraepelin proposed the connectopathy theory of psychosis at the beginning of the twentieth century.Friston 1998. According to Kubicki et al. 2005, Carl Wernicke and the German psychiatrist Emil Kraepelin proposed the connectopathy theory of psychosis at the beginning of the twentieth century.
[>] rapidly again in adolescence: Huttenlocher and Dabholkar 1997.
[>] over the edge to psychosis: Is the connectopathy theory consistent with the observed effects of schizophrenia medications? Psychotic symptoms are relieved by drugs interfering with synapses that secrete dopamine. The symptoms are induced in normal people by drugs interfering with synapses that secrete glutamate. (Examples are ketamine and phencyclidine or PCP, which temporarily turn recreational users into schizophrenics, as emergency room physicians can attest.) According to the traditional view, the connectivity of the schizophrenic brain is normal, but the synapses don’t work properly. Synaptic malfunction is corrected by the antipsychotic drugs and induced by the psychosis-generating drugs. But another view would be that antipsychotic drugs cause changes in synaptic function that compensate for connectopathy in schizophrenics, while psychosis-generating drugs mimic the effects of connectopathy in normals. This is possible because changes in synaptic function and changes in connectivity may have similar effects. For example, drastically weakening a synapse may be indistinguishable from removing it altogether. There is an even more subtle possibility. It may be wrong to think of abnormal synaptic function and abnormal connectivity as two independent defects. Suppose that synapse elimination is driven by synaptic weakening, which in turn is dependent on activity. If abnormal synaptic function changes activity patterns, it could end up causing the brain’s connectivity to develop abnormally. Any initial abnormality in connectivity might also lead to abnormal activity patterns, which could cause further development of abnormal connectivity. Connectopathy would accompany schizophrenia, but it would be difficult to say which is cause and which is effect.
Connectome Page 32