Measuring all neurons in the brain

Try to contemplate a future device that allows registering in minute detail the behaviors of a cheetah and a gazelle during a chase, including all muscle and skeletal movements. At the same time, we’re capable of recording billions of neurons across the two nervous systems while the entire chase unfolds from before the cheetah initiates the pursuit until its dramatic conclusion.

From: https://www.sulzer.com/-/media/images/about-us/sulzer-technical-review/str/2017-issue-3/nature_speed_cheetah_gazelle_1920.ashx?mw=564&hash=263818FB250C10F71E1F44C9BFE704B965277EDE.jpg

What would we discover? How much of our textbooks would have to be altered?

A radical rethinking might be needed, and a lot would have to be rewritten. An alternative possibility is that many of the experimental paradigms employed to date are quite effective in isolating critical mechanisms that reflect the brain’s functioning in general settings. True, novel findings would be made with new devices and techniques, but they would extend current neuroscience by building naturally upon current knowledge. The first scenario is not idle speculation, however.

So-called naturalistic experimental paradigms are starting to paint a different picture of amygdala function, for example. In one study, a rat was placed at one end of an elongated enclosure and a piece of food was placed midway between the rat and a potential predator, a lego-plus-motor device called a “Robogator.” To successfully obtain the food pellet, the rat had to retrieve it before being caught by the Robogator (don’t worry, capture didn’t occur in practice). The findings[1] were inconsistent with the standard “threat-coding model” that says that amygdala responses reflect fear-like or other related defensive states. During foraging (when approaching the pellet), neurons reduced their firing rate and were nearly silent near the predator. Clearly, responses did not reflect threat per se.

From: Choi, J. S., & Kim, J. J. (2010). Amygdala regulates risk of predation in rats foraging in a dynamic fear environment. Proceedings of the National Academy of Sciences, 107(50), 21773-21777.

Another study recorded from neurons in the amygdala over multiple days, as mice were exposed to different conditions[2]. Mice were exposed to a small open field and were free to explore it. These creatures don’t like to feel exposed, so in the experiment they frequently stayed at the corners of the box a good amount of time. But they also ventured out in the open for periods and navigated around the center of the box with some frequency. The researchers discovered two groups of cells: one that was engaged when the mouse was being more defensive in the corners (these “corner” cells fired more vigorously at these locations), another when the mouse was in an exploratory mode visiting the center of the space (“center” cells fired more strongly when the animal was around the middle of the field). The researchers also managed to record from the exact same cells during more standard paradigms, including fear conditioning and extinction. They then tested the idea that the firing of amygdala neurons tracks “global anxiety”; for instance, they should increase their responses when the animal entered the center of the field in the open-field condition, as well as when they heard the CS tone used in the conditioning part of the experiment. Surprisingly, cells did not respond in this way. Instead, neuronal firing reflected moment-to-moment changes in the exploratory state of the animal, such as during the time window when the animal transitioned from exploratory (for example, navigating in the open field) to non-exploratory behaviors (for example, when starting to freeze).

The above two examples provide tantalizing inklings that there’s a lot to discover – and revise – about the brain. It’s too early to tell, but given the technological advances neuroscience is witnessing, examples are popping up all over the place. For example, a study by Karl Deisseroth and colleagues[3] recorded activity of ~24,000 neurons throughout 34 brain regions (cortical and subcortical). Whereas measuring electrical activity with implanted electrodes typically measures a few cells at a time, or maybe ~100 by using state-of-the-art electrode grids, the study capitalized on new techniques that record calcium fluorescence instead. When cells change their activity, including when they spike, they rely on calcium-dependent mechanisms. In genetically-engineered mice, neurons literally glow based on their calcium concentration. By building specialized microscopes, it is possible to detect neuronal signaling across small patches of gray matter. When mice smelled a “go” stimulus, a licking response produced water as a reward. The animals were highly motivated to perform this simple task as the experimenters kept them in a water-restricted state. Water-predicting sensory stimuli (the “go” odor) elicited activity that rapidly spread throughout the brain of thirsty animals. The wave of activity began in olfactory regions and was disseminated within ~300 ms to neurons in every one of the 34 regions they recorded from! Such propagation of information triggered by the “go” stimulus was not detected in animals allowed to freely consume water. Thus, the initial water-predicting stimulus initiates a cascade of firing throughout the brain only when the animal is in the right state – thirsty.

In another breakthrough study, Kenneth Harris, Mateo Carandini and colleagues[4] used calcium imaging techniques to record from more than 10,000 neurons in the visual cortex of the mouse. At the same time, facial movements were recorded in minute detail. They found that information in visual cortex neurons reflects more than a dozen features of motor information (related to facial movements, including whiskers and other facial features), in line with emerging evidence. These results are remarkable because traditional thinking is that motor and visual signals are only merged later in so-called “higher-order” cortical areas; definitely not in primary visual cortex. But the surprises didn’t stop there. The researchers also recorded signals across the forebrain, including other cortical areas, as well as subcortical regions. Surprisingly, information about the animal’s behavior (at least as conveyed by motor actions visible on the mouse’s face) was observed nearly everywhere they recorded. In considering the benefit of this ubiquitous mixing of sensory and motor information, the investigators ventured that effective behaviors depend on the combination of sensory data, ongoing motor actions, and internal variables such as motivational drives. This seems to be happening pretty much everywhere in the brain, including in primary sensory cortex. The examples above hint that much is to change in neuroscience in the coming decades. And these results come from fairly constrained settings. The amygdala study used a 40 x 40 x 40 plastic box; the thirst study probed mice with their heads fixed in placed; and the facial movement study employed an “air-floating ball” that allowed mice to “run”. Imagine what we’ll discover in the future.


[1] Recordings in the basolateral amygdala. Amir, A., Kyriazi, P., Lee, S. C., Headley, D. B., & Pare, D. (2019). Basolateral amygdala neurons are activated during threat expectation. Journal of Neurophysiology, 121(5), 1761-1777.

[2] Recordings in the basal amygdala: Gründemann, J., Bitterman, Y., Lu, T., Krabbe, S., Grewe, B. F., Schnitzer, M. J., & Lüthi, A. (2019). Amygdala ensembles encode behavioral states. Science, 364(6437), eaav8736.

[3] Allen, W. E., Chen, M. Z., Pichamoorthy, N., Tien, R. H., Pachitariu, M., Luo, L., & Deisseroth, K. (2019). Thirst regulates motivated behavior through modulation of brainwide neural population dynamics. Science, 364(6437), 253-253.

[4] Stringer, C., Pachitariu, M., Steinmetz, N., Reddy, C. B., Carandini, M., & Harris, K. D. (2019). Spontaneous behaviors drive multidimensional, brainwide activity. Science, 364(6437), 255-255.

Fitting behavior inside a 40x40x40 cm box

The central question in neuroscience is to understand the physical basis of behavior. But what kinds of behavior can be studied in a lab? Mice and rats can be placed in chambers and mazes to perform tasks. One can then study the effects of lesions on behavior. But if cell recordings are performed the constraints are much more severe. Until just a few years ago, this required a fair amount of cabling to link the brain to signal amplifiers and other electronics. Experiments in primates are performed in a “monkey chair” that keeps the animal’s body and head in place. Humans, of course, are studied inside MRI tubes that are anything but organic. With the technology available, getting closer to natural behaviors has simply not been possible.

Skinner's Box and Video Games: How to Create Addictive ...
From: https://levelskip.com/how-to/Skinners-Box-and-Video-Games

A type of behavior that fits inside a 40x40x40 cm box is classical conditioning.  Indeed, it has been extensively studied by psychologists since the early twentieth century, and for those interested in studying the biological mechanisms of fear, the paradigm has been a godsend. It has offered a window into this process, while allowing careful control over experimental variables, a fundamental consideration in experimental science. With the paradigm, the neuroscience of fear has been one of the most active areas of inquiry.

The fixation with this paradigm has also incurred nontrivial costs, leading to led to a type of tunnel vision[1]. As Dennis Pare and Gregory Quirk, very prominent “fear” researchers themselves, state:

When a rat is presented with only one threatening stimulus in a testing box that allows for a single reflexive behavioral response, one is bound to find exactly what the experimental situation allows: neuronal responses that appear tightly linked to the CS and seem to obligatorily elicit the conditioned behavior. Paré, D., & Quirk, G. J. (2017, p. 6)

The very success of the approach has led to shortsightedness.

Placed inside a small, enclosed chamber the animal is limited to a sole response: upon detecting the CS, it ceases all overt behavior and freezes in place. It can’t consider other options, such as dashing to a corner to escape; it cannot try to attack the source of threat either, as there isn’t another animal around – the shock comes out of nowhere! Now, when researchers study the rat’s brain under such conditions, a close relationship between brain and behavior is established. But as Pare and Quirk warn, the tight link might be apparent insofar as it would not hold under more general conditions. Neuroscience is experiencing a methodological renaissance. Advances in chemistry and genetics allow precision in targeting regions and circuits in a way that would have sounded like science fiction a decade ago. But if we continue using the paradigms that have been the mainstay of field, we will be cornering ourselves into a scientific cul-de-sac[2]. It’s time to think outside the box.


[1] Text here builds directly from Paré, D., & Quirk, G. J. (2017). When scientific paradigms lead to tunnel vision: lessons from the study of fear. npj Science of Learning, 2(1), 6.

[2] “Cul-de-sac” expression inspired by Kim and Jung (2018).

Evolution and the brain: what is novel?

The geneticist Theodosius Dobhansky famously stated that in biology nothing makes sense unless it’s in light of evolution. The same applies to neuroscience, a biological science. But evolution poses a conundrum. Vertebrates have been evolving for over 500 million years[1]. A telencephalon, a midbrain, and a hindbrain are part of the general plan of their nervous system. Structures like the amygdala and the striatum are found in animals as diverse as a salmon, a crow, and a baboon. Thus, many parts of the brain are “conserved”. But, then, what is novel? Something must be new after all.

From Pessoa, L., Medina, L., Hof, P. R., & Desfilis, E. (2019). Neural architecture of the vertebrate brain: Implications for the interaction between emotion and cognition. Neuroscience & Biobehavioral Reviews, 107, 296-312.

In chapter 9, we described how homology refers to relationships between traits that are shared as a result of common ancestry. The leaves of plants provide a good example[2]. The leaves of a pitcher plant, Venus flytrap, poinsettia, and a cactus look nothing alike, and in fact have distinct functions. In the pitcher plant, the leaves are modified into pitchers to catch insects; in the Venus flytrap they are modified into jaws to catch insects; in the poinsettia bright red leaves resemble flower petals and attract insects and pollinators; cactus’ leaves have become modified into spines, which reduce water loss and can protect the plants from herbivores. Yet, the four are homologous given that they derive from a common ancestor.

A structure adopts new functions during evolution, while its ancestry can be traced to something more fundamental[3]. Take the hippocampus of rodents, monkeys, and humans. There is copious evidence indicating that the area is homologous in the three species, that is, that is a conserved structure. But does this mean that it performs the same function(s) in these species? Does it perform some qualitatively different function(s) in humans, for example? To many neuroscientists this sounds implausible. However, the possibility need not be any more radical than saying that the forelimb does something qualitatively different in birds compared to turtles, say. If common ancestry precluded new functions, no species could ever take flight!

The ongoing discussion is particularly pertinent when we think of emotion and motivation, because researchers invoke “old” structures when studying these mental phenomena. Regions like the amygdala at the base of the forebrain and the periaqueductal gray in the midbrain in the case of emotion; the accumbens (part of the striatum) also at the base of the forebrain and the ventral tegmental area in the midbrain in the case of motivation. Because these regions are deeply conserved across vertebrates, they function in a similar way, or so the reasoning goes. If we entertain these areas in rodents, monkeys, and humans, closer as they are evolutionarily, the expectation would be that they work in largely the same manner. But rodents and primates diverged more than 70 million years ago. Are we to suppose that no qualitative differences have emerged? This seems rather implausible. (In Chapter 9, we briefly reviewed some of the differences in the amygdala of rats, monkeys, and humans.)

The argument made in this book is that we should conceptualize evolution in terms of the reorganization of larger-scale connectional systems. Instead of more cortex sitting atop the subcortex in primates relative to rodents – which presumably allows the “rational” cortex to control the “irrational” subcortex – more varied ways of interactions are possible, supporting more mental latitude.

The brain doesn’t fossilize. Unfortunately, with time, it disintegrates, leaving no trace. So we simply don’t have a way to know exactly what the brain of a common ancestor looked like. Without fossil remains, scientists tend to think of the brain of a common ancestor of rodents, primates, and humans as something like the current brain of a mouse, as this animal is the “most primitive” one. But a mouse encountered today has had 70 million years to evolve from the ancestor in question, and thus specialize to the particular niches it inhabits now.

Evolution is as much about what’s preserved as what’s new. Ever since science was transformed by the independent work of Charles Darwin and Alfred Russell Wallace in the late 1850s, biologists have sought to determine “uniquely human” characteristics. This has led to a near-obsession to identify one-of-a-kind nervous system features, from putative exclusively human brain regions to cell types. The cortex, in particular, has attracted much attention. We described in Chapter 9 how much of the pallium of mammals is structured in a layered fashion, a quality that is not observed in other vertebrates. Well, not exactly, as some reptiles (such as turtles) have a dorsal pallium that is cortex-like, with three bands of cells. Mammals, however, have parts of the cortex that is much more finely layered, with six well-defined zones. In fact, six-layered cortex is often referred to as “neocortex”, with the “neo” part highlighting its sui generis property (in the book, the more neutral terminology “isocortex” was adopted for this type of cortex).

I offer that the concept of reorganization of circuits is a much more promising idea. That is to say, what is unique about humans is the same that is unique about mice, or any other species: their circuits are wired in ways that support survival of the species. This is not to deny that some more punctate differences play a role. But whatever the differences are, at least considering primates with larger body sizes, they are not staring us in the face – they are subtle. For example, all primates exhibit an isocortex that is massively expanded[4]. Primates also have prefrontal cortices with multiple parts, including the lateral component, which neuroscientists often link to “rational” capabilities. More generally, direct evidence for human-specific cortical areas is scant[5].

Let’s go back to Dobhansky’s call to consider biology in light of evolution – always. Biologists would vehemently agree. But evolution is so egregiously complex that the suggestion doesn’t help as much as one would think. Verily, what we observe in practice is that neuroscientists who don’t specialize in studying brain evolution are time and again cavalier, if not outright naïve, about how they apply and think of evolution. By doing so, our explanations run the risk of becoming just-so stories[6].


[1] For a framework on vertebrate evolution, see Pessoa, L., Medina, L., Hof, P. R., & Desfilis, E. (2019). Neural architecture of the vertebrate brain: Implications for the interaction between emotion and cognition. Neuroscience & Biobehavioral Reviews, 107, 296-312.

[2] https://evolution.berkeley.edu/evolibrary/article/0_0_0/lines_04

[3] Sentence closely borrowed from Murray et al. (2016): “a structure adopts new functions during evolution, yet its ancestry can be traced to something more fundamental”. Discussion of the hippocampus until end of the paragraph also from them.

[4] Striedter (2005).

[5] Striedter (2005).

[6] The Wikipedia page on just-so-stories is actually pretty decent: https://en.wikipedia.org/wiki/Just-so_story.